Upload
lyque
View
220
Download
2
Embed Size (px)
Citation preview
Adhesion of electrolessly deposited Ni(P) on aluminaceramicSeverin, J.W.
DOI:10.6100/IR395680
Published: 01/01/1993
Document VersionPublisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the author's version of the article upon submission and before peer-review. There can be important differencesbetween the submitted version and the official published version of record. People interested in the research are advised to contact theauthor for the final version of the publication, or visit the DOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and page numbers.
Link to publication
Citation for published version (APA):Severin, J. W. (1993). Adhesion of electrolessly deposited Ni(P) on alumina ceramic Eindhoven: TechnischeUniversiteit Eindhoven DOI: 10.6100/IR395680
General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal ?
Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.
Download date: 18. Feb. 2018
Adhesion of Electrolessly Deposited Ni(P)
on Alumina Ceramic
J .W. Severin
The cover shows a cross-section TEM micrograph of the alumina - Ni(P) interface
of a sample which has been heat-treated at 580 oc.
Adhesion of Electrolessly Deposited Ni(P)
on Alumina Ceramic
PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de
Technische Universiteit Eindhoven, op gezag van
de Rector Magnificus, prof. dr. J.H. van Lint,
voor een commissie aangewezen door het College
van Dekanen in het openbaar te verdedigen op
dinsdag 27 april 1993 om 16.00 uur
door
Jan Willem Severin
geboren te Geldrop
Dit proefschrift is goedgekeurd
door de promotoren
prof. dr. G. de With
en
prof. dr. H.H. Brongersma.
The work described in this thesis has been carried out at the Philips Research
Laboratories Eindhoven as part of the Philips Research programme.
ii
Aan Marijke, Dirk en Anne
Aan mijn ouders
iii
IV
Table of Contents
Preface
Chapter 1: Introduction
1.1. l.l.l. 1.1.2. 1.1.3. 1.2. 1.2.1. 1.2.2. 1.2.3. 1.3. 1.4. 1.4.1. 1.4.2. 1.4.3.
Summary Concepts of adhesion
General Fields of application Adhesion theories and practical aspects
Fracture mechanics at interfaces Adhesion strength Peel test Weibull statistics
Electroless metallization Scope of this thesis
Aim Surface analytical techniques Outline of subsequent chapters
Chapter 2: Adhesion of electrolessly deposited Ni(P) on alumina ceramic: an assessment of the current status
2.1. 2.2. 2 3. 2.3.1. 2.3.2. 2.3.3. 2.3.4. 2.4. 2.5.
Summary Introduction General procedures and overview of results Effects of process parameters on the adhesion
Etching conditions Nucleation conditions Metallization conditions Heat treatments
Comparison of Ni(P) with Ni(B) and Cu Final remarks
Chapter 3: A study on changes in surface chemistry during the initial stages of electroless Ni(P) deposition on alumina
3.1. 3.2. 3.3. 3.3.1. 3.3.2. 3.3.3. 3.3.4. 3.4. 3.5.
Summary Introduction Experimental procedures Measurement results
SEM results XRF results Static-SIMS results TEM results
Discussion Conclusions
Page
IX
I 2 2 3 4 9 9 10 11 12 15 15 15 18
21 22 24 28 28 32 34 37 39 40
45 46 47 49 49 49 49 55 58 62
V
Chapter 4: Adhesion and interface characterization of electroless Ni(P) layers on alumina ceramic
4.1. 4.2. 4.2.1. 4.2.2. 4.3. 4.3.1. 4.3.2. 4.4. 4.4.1. 4.4.2. 4.4.3. 4.5. 4.5.1. 4.5.2. 4.5.3. 4.5.4. 4.5.5. 4.6.
Summary Introduction Theory
Adhesion strength Peel test
Experimental procedures Sample preparation Analyses
Results Mechanical properties Interface structure Chemical interface analyses
Discussion Direct pull-off test Peel test Interface microstructure and chemistry Mechanism of adhesion Relation between adhesion strength and fracture energy
Conclusions
Chapter 5: The influence of thermal treatments on the adhesion of electroless Ni(P) on alumina ceramic
5.1. 5.2. 5.2.1. 5.2.2. 5.3. 5.3.1. 5.3.2. 5.3.3. 5.3.4. 5.4. 5.4.1. 5.4.2. 5.5.
Summary In traduction Experimental procedures
Sample preparation Analyses
Results Adhesion measurements Interface and fracture surface structure XPS fracture surface analyses Static-SIMS measurements
Discussion Mechanical behaviour Interface chemistry
Conclusions
Chapter 6: The influence of substrate chemistry on the adhesion of electroless Ni(P) on metal-oxide coated ceramics
6.1. 6.2. 6.2.1. 6.2.2. 6.3. 6.3.1. 6.3.2. 6.3.3. 6.4. 6.5.
vi
Summary Introduction Experimental procedures
Sample preparation Analyses
Results Adhesion measurements Analyses of surface composition Fracture surface analyses
Discussion Final remarks
67 68 69 69 70 72 72 73 78 78 86 91 102 102 104 107 109 Ill 115
121 122 123 123 124 124 124 129 135 136 141 141 144 146
149 150 151 152 153 153 153 154 156 165 170
Chapter 7: The adhesion of electrolessly deposited Ni(P) on alumina ceramic using a vacuum-deposited Ti-Pd nucleation layer
7.1. 7.2. 7.3. 7.3.1. 7.3.2. 7.3.3. 7.3.4. 7.4. 7.4.1. 7.4.2. 7.5.
Summary Introduction Experimental procedures Results
Adhesion measurements Interface chemistry Interface structure Interface formation
Discussion Adhesion Chemical bonding
Conclusions
Chapter 8: Final discussion, conclusions and outlook
Summary 8.1. Aim and status of current knowledge 8.2. New insights 8.3. Suggestions for further work
Summary
Sa men va tting
List of symbols and abbreviations
Dankwoord
Curriculum vitae
173 174 175 177 177 182 182 184 188 188 190 191
193 194 194 196
199
203
207
209
211
vu
viii
Preface
This thesis deals with a study of the adhesion of electrolessly deposited Ni(P) on
alumina ceramic substrates. Electroless metallization is a simple, rapid and cheap
procedure to provide insulating surfaces with a metal layer by successive
immersion of a substrate in a number of aqueous solutions. The process is used in
the production of numerous electronic components such as liquid crystal displays,
IC packages and passive components. The major disadvantage of the electroless
metallization process is the poor adhesion of the metal deposits.
The aim of this work is to obtain insight into the adhesion mechanism, and to
improve the adhesion. Since for many applications etching or abrasion is undesired
for technological or economical reasons, it is the aim to attain strong adhesion on
smooth surfaces. This study is carried out as follows: After a literature review first
a detailed analysis is made of the interface formation. The structure and chemical
composition of the substrate surface is analysed on monolayer scale after each of
the successive process steps. Then, the adhesion is measured by direct pull-off and
peel tests. The results are analysed using fracture mechanics with the Griffith-Irwin
approach and quantitative aspects of the adhesion measurement procedures are
considered. Adhesion strength data are interpreted using Weibull statistics. The
structure and chemical composition of fracture surfaces are analysed both on
micrometer and on nanometer scale in order to determine the fracture path and
to unravel the adhesion mechanism. For the interface- and fracture surface
analyses the following techniques are used: optical microscopy, SEM/EDX,
cross-section TEM, plan-view TEM, AES, XPS, XRF, and static-SIMS.
Subsequently, three procedures are studied to improve the adhesion. Firstly, the
effect of annealing treatments is analysed by adhesion measurements and interface
and fracture surface analyses. Secondly, the composition of the oxidic substrate
surfaces is varied by using various metal-oxide coatings on top of the alumina
ceramic substrates. Thirdly, the effect of an alternative nucleation procedure is
investigated. Instead of the conventional wet-chemical nucleation procedure,
vacuum-deposited Pd was applied, with an underlying Ti base metal layer for
providing strong adhesion between the nucleation layer and the substrate. For this
ix
system the chemical bonding of the Ti base metal with the ceramic substrate is
studied in-situ in a UHV system.
Finally, an assessment is made of the progress with respect to the above described
aims. A summary of the main conclusions is presented, relations between results
obtained in the various studies are discussed, and suggestions for further research
are made.
X
Chapter 1
Introduction
Summary
This introductory chapter first of all provides an introduction to the con
cepts of adhesion. The importance of this field of materials science is il
lustrated with a selection of application areas. The most important
phenomena which influence adhesion are highlighted with the aid of a
number of practical examples. Various types of interfacial interactions,
the occurrence of weak boundary layers, wetting, cleaning and chemical
surface modification are discussed. Subsequently, some aspects of frac
ture mechanics at interfaces are outlined, as this approach is used as a
guideline for the interpretation of all adhesion studies in this work. The
principles of electroless metallization are also presented. The final section
provides a description of the scope of this thesis, including the aim of this
work, an introduction to the surface analytical techniques used and an
outline of the subsequent chapters. The aim of this work is to obtain in
sight into the mechanism of adhesion between electrolessly deposited
Ni(P) and alumina ceramic, and to improve the adhesion.
1.1. Concepts of adhesion
1.1.1. (;eneral
It is sometimes suggested that adhesion is a separate field of materials science and
technology. However, the cohesion within a material is essentially determined by
the same physical and chemical interactions between atoms, ions or molecules as
those which also act across an interface and determine the adhesion. In addition,
the same microstructural aspects are important for the strength of a bulk material
as for the strength of a bonded system. Adhesion should therefore be considered
as just a regular member of the family of Mechanical Properties of Materials. The
characteristic property of this member of the family is the presence of a solid-solid
interface. Very often, this interface is the weakest link in the chain and determines
the strength of the bonded system. This is the reason why investigators working
in the field of adhesion science and technology generally focus their attention on
the state of surfaces, the bonding operation and interface analyses. Nevertheless,
adhesion should not be defined as a property of the interface only. Bulk mechan
ical properties of the adherends, e.g. the substrate and the coating, also play an
important role in adhesion. This will be illustrated in subsequent sections. There
fore, it is important to emphasize that adhesion is a property of a bonded system,
including the interface.
For the sake of clarity a few definitions will first be given of the concepts that are
most important in this thesis.
Adhesion: The interaction which keeps macroscopic parts attached to
each other.
Adhesion strength: The force per unit area required to break apart attached
macroscopic parts.
Fracture energy: The amount of energy required to debond a unit area of the
interface.
2
1.1.2. Fields of application
The importance and the scope of the field of adhesion science and technology can
best be illustrated with a listing of application areas which are encountered in
every-day life.
Polymer coatings are used extensively as glues, paints and inks. The major
function of glues and adhesive tapes is to bring about strong and persistent
adhesion. Paints can only meet requirements for corrosion protection and
decoration when the adhesion withstands high humidity, thermal cycling and
scratching. Similarly, inks on paper and on plastics have to withstand
scratching and friction.
Metal thin films are extensively used on polymers, semiconductors, metals
and insulators for a wide variety of applications such as decoration, resistive
and magnetic films, electrical contacts, light reflection, impermeable films
and anti-corrosion. Well-known examples are chromium on steel and on
plastics for car parts, car lamp reflectors and compact discs, aluminum on
plastic packing foil, zinc on roof gutters, silver on tableware, gold on cheap
jewelry, magnetic films for recording heads and Al, Cu or Ti/W on Si or
Si02 for metallization of integrated circuits (ICs). Proper function of all of
these coatings depends upon reliable adhesion under similar conditions as for
the polymer films. Moreover, for metal thin films additional requirements
are imposed upon the adhesion due to internal stresses which are generally
present in such films (1). The magnitude of these stresses depends upon the
deposition process and deposition parameters and the materials used.
For increasing the life time of steel cutting tools, refractory coatings are used
consisting of metal nitride, boride and carbide compounds deposited by
chemical vapour deposited (CVD) processes. Due to the high mechanical
loads on the films, it is obvious that also for such coatings strong adhesion
is a prerequisite. CVD processes are also used for the deposition of metal
oxide films, e.g. used as optical films in car lamps, on glasses and on TV
screens. Stresses are often present in films deposited at elevated temperatures
3
due to differences in thermal expansion between layer and substrate. Strong
adhesion is required in order to avoid flaking-off during cooling (1).
The reliability of electronic components often depends on the strength of the
metal - metal joint formed by soldering. Other examples of joining techniques
resulting in metal - metal interfaces are conventional welding of steel parts
for large constructions, laser welding, e.g. in the production of electron guns
for TVs, and wire-bonding for ICs.
In vacuum devices, such as TVs and in other gas-tight products such as
lamps, numerous metal- ceramic and metal- glass joints are present e.g. as
feedthroughs and as insulating glass spacers between the grids in electron
guns of TVs, scanning electron microscopes (SEM), transmission electron
microscopes (TEM), X-ray tubes etc. These joints are always prepared at
high temperatures and, therefore, strong adhesion is essential. When cracks
are formed at the metal - oxide interface of a feed through, gas enters into the
lamp or the TV which immediately results in failure of the product.
1.1.3. Adhesion theories and practical aspects
- Adhesion strength
The adhesion strength is generally the most important quantity with respect to
adhesion. According to the Griffith-Irwin theory, the adhesion strength is deter
mined by the fracture energy, the size of flaws (non-bonded areas), the Young's
modulus of the adherends and a geometrical factor. Since we are primarily inter
ested in the relation between interfacial chemistry and adhesion, the fracture en
ergy is further considered. The fracture energy depends on the intensity of
interfacial interactions, the degree of intimate contact and energy dissipation at the
crack tip due to yielding. The presence of stresses in the fJ.lm leads to a lower ap
parent fracture energy. The quantitative relationship between these parameters
will be treated in more detail in section 1.2. A schematic representation of the as
pects which are most important for the adhesion, is given in fJ.g. 1.
4
Interfacial flaws
Weak boundary Interface Stress layers polarization in film
Interface chemistry or Van der Waals'
forces
1111 ---
t Added reactive
layer
Interface morphology, toughness
Film Medium thickness
D
Elastic moduli and yield strength for film and substrate
Fig. 1: Schematic representation of the most important aspects for the adhe
sion of thin films (adapted from a figure in ref. 9).
- Interfacial interactions
Interfacial interactions can be divided into two types: mechanical and non
mechanical interactions. An example of a mechanical interaction is mechanical
interlocking which takes place when film .material penetrates in cavities under the
substrate surface. Another example is the friction which takes place when e.g. fi
bers are pulled out of a polymer matrix when fiber composites are fractured
(2, 3).
The non-mechanical interactions have been subject of many investigations,
discussions and disputes. Firstly, various types of chemical interactions such as
Van der Waals interactions and covalent, ionic or metallic chemical bonding will
be discussed. From a thermodynamical point of view, all of these interactions
lower the free Gibbs energy of the adhering system. As described with the
Young-Dupre equation, the work of adhesion w. is defined as the difference be
tween the sum of the surface energies of both adherends before bonding and the
interfacial energy after bonding. If debonding takes place as a reversible process,
no more than the work of adhesion has to be supplied mechanically. In subsequent
sections it will be shown that debonding is in most cases an irreversible process
and, therefore, the debonding energy is generally a manifold of the above de
scribed interactions due to plastic deformation at the crack tip.
5
When two surfaces are closely spaced, Van der Waals interactions inevitably occur
(4). The work of adhesion due to this type of interactions strongly depends on the
chemical composition of the outermost monolayers of both adherend surfaces and
on the spacing between them. Considerable adhesion due to Van der Waals type
interactions is only possible for spacings smaller than l nm. For adhesion between
apolar surfaces consisting of atoms with a low atomic number, such as those of
polyolefine polymers, Wa is of the order of 0.01 J/m2. This value increases with
increasing polarizability and polarity of the surfaces. It reaches a maximum of
0.5 Jjm2 when hydrogen bonds are formed across the interface. For metallic,
covalent and ionic chemical bonds, w. is between 1 and 5 Jjm2• Fawkes (5) has
presented a theory, the "acid-base theory", for quantitative calculation ofWa. This
theory is based upon the chemical nature of the surfaces of both adherends and
the most likely chemical interfacial interactions. Fawkes used a rather broad de
finition of acids and bases in the theory, ranging from dispersive type Van der
Waals interactions to hydrogen bonds.
For all of these interactions, intimate contact between both adherends is essentiaL
This explains why so much attention is paid to wetting and wettability in all books
and reviews on adhesion. However, strong wetting does not necessarily lead to
strong adhesion. For instance, a low surface energy of a liquid polymer is of pos
itive influence on its wetting of a substrate surface, but for strong adhes(on a high
surface energy of both adherends is more favourable.
An example where adhesion is brought about by electrostatic attraction, is plastic
packing foil, where the attraction is often perceptible at a distance of several
millimeters. In 1955 Deryaguin (6) proposed a theory in which adhesion was ex
plained in terms of electrostatic interactions. At present, it is still uncertain
whether electrostatic interactions significantly contribute to the adhesion between
planar surfaces or not (7). For the adhesion of small particles like dust and
xerographic particles, electrostatic interactions play an important role.
-Weak boundary layers and interphase
A phenomenon which very often plays an important role in adhesion is the oc
currence of a weak boundary layer (WBL). A WBL is an interface layer with a
6
composition different from the composition of each of the adherends and with
weak cohesive interactions. The thicknesses of WBLs may vary from a few
monolayers up to several micrometers. WBLs can originate from unsuitable
cleaning treatments, from the presence of contaminating species during the bond
ing or coating deposition process or from segregation of material out of the bulk
of the adherends after the bonding operation. Examples of WBLs are native oxides
on metal surfaces where a metal - metal bond should be formed such as with
soldering or electrodeposition, pump oil on surfaces which are coated with a vac
uum deposition process and segregation of release agents and plasticizers from
plastic substrates to the interface with a coating. Some glues are designed to dis
solve or displace contaminations by preferential adsorption on surfaces, thus in
herently avoiding the formation of WBLs due to surface contaminations.
The WBL is a specific example of the more general phenomenon "interphase" (8).
Another frequently occuring example of the presence of an interphase is the for
mation of a reaction zone between both adherends. Since this is generally associ
ated with high mutual affinity of both surfaces and with intimate interfacial
contact, such a reaction zone is often found in the case of strong adhesion.
Interfacial reactions between metal layers and ceramic substrates, induced by
high-energy ion beams have been shown to considerably improve the adhesion
(9, 10). For polymer - polymer interfaces the entanglement of polymer chains
across the interface is known to greatly improve the adhesion ( 11 ). This process
is often enhanced by increasing the mobility of polymer chain segments with
thermal treatment or by swelling the polymer surfaces in a pretreatment with a
solvent. However, for some metal - metal joints the adhesion is negatively influ
enced due to the formation of brittle intermetallic compounds and stresses as a
consequence of changes in specific volume by the interfacial reaction.
Chemical surface modification (8, 12) is a popular method for the improvement
of adhesion of polymer films on inorganic surfaces by the introduction of an
interphase. By adsorption or covalent bonding bifunctional molecules, e.g
organosilanes, are attached to the substrate surface prior to the bonding operation.
The remaining functionality of the bonding agent has a high affinity towards the
coating and may form covalent or ionic chemical bonds. In an alternative proce-
7
dure the bonding agent is added in trace amounts to the coating material. In this
case the interphase is formed by segregation.
-Cleaning
The cleaning treatment is one of the most important steps in a bonding operation
(13). As indicated in the above discussion, the adhesion is strongly influenced by
the composition of the interface on a monomolecular level. Each bonding or
coating deposition process has its own specific requirements for the cleaning op
eration. Generally, by the cleaning treatment adsorbed organic contaminations
and macroscopic dirt particles have to be removed (13). Prior to· vacuum
deposition of metal thin films on inorganic or polymer surfaces, a plasma or
sputtering treatment is often applied. Such a treatment not only removes contam
inations, but also chemically modifies the surface. By breaking chemical bonds a
high-energy surface is created. The reactive dangling bonds are saturated with the
metal atoms which results in strong adhesion.
Some practical pitfalls which may occur with cleaning are the following:
When an organic solvent or an aqueous solution of a detergent is used, dis
solved compounds can adsorb and remain on the surface.
When a metal or metal oxide surface is really clean, it generally has a high
surface energy. As a consequence it is completely covered with organic mol
ecules from the ambient atmosphere after just a few seconds exposure.
Fluoride containing solutions and plasmas are found to leave behind fluorine
on the surface, lowering the surface energy.
Oxidizing plasmas or UV-ozone treatment may not only oxidize organic
contaminations, but also give rise to the formation of undesired oxide layers
on metal surfaces. Silver surfaces are black after a UV -ozone cleaning treat
ment due to oxidation.
Oxidizing treatments of polymer surfaces often result in a WBL of low
molecular weight polymer chain fragments on the surface.
These examples illustrate that cleaning is not only an important process step but
also a very difficult one. If reliable experience is not available, careful surface
analysis is required to assess the merits of a cleaning process.
8
In addition to the adhesion aspects discussed above, the performance of a joint is
also strongly influenced by the geometry of the joint design and type of loading.
These aspects however, do not fall within the scope of this thesis and will, there
fore, not be further discussed.
1.2. Fracture mechanics at interfaces
This section deals with theoretical backgrounds of the adhesion strength and
fracture energy measurements. In addition, a statistical method for interpretation
of the adhesion strength data, i.e. Weibull statistics, is discussed.
1.2.1. Adhesion strength
The adhesion strength CTr is determined, among other factors, by the fracture en
ergy Gc and the critical flaw size acr and is usually described by the Griffith-lrwin
relation (2, 14 to 16):
2 CTr = [l]
where K is a geometric factor and E is Young's modulus.
The fracture energy Gc is formed by an intrinsic fracture energy term G; and a
contribution Gp, from plastic deformation of the material at the crack tip:
[2]
The intrinsic fracture energy is the energy required for example to overcome Van
der Waals forces and to break chemical bonds. The order of magnitude of G; is
0.01 to 0.1 J/m2 for Van der Waals interactions and 0.5 to 5 Jfm2 for chemical
bonds. During fracture, stresses are near to the theoretical strength at the crack
tip. This causes plastic deformation during fracture in the adherends near the
interface, represented by Gpl· Since the stresses at the crack tip depend on the
strength of the interfacial bonds, Gp, depends on G; and, therefore, eq. 2 can be
written as (17):
9
[3]
in which f1 is the energy loss factor. For purely brittle fracture, such 3jS with ce
ramics at low temperature, plastic deformation plays a minor role and f, is of the
order of 2 to 10. For metal layers on ceramics Gc values of the order of 100 Jjm2
are found (18), which means that f, is lO to 100. For polymers on rigid substrates
these values are of the order of !000 Jfm2 for Gc (19) and thus 100 to 1000 for f1 •
From eq. l it is clear that in order to evaluate the influence of interface. chemistry
on adhesion strength, the fracture energy Gc must be measured separately. This is
done by the peel test. Conditions under which the peel test can be l!lsed for a
quantitative fracture energy measurement are considered in the next section.
1.2.2. Peel test
The peel test has often been used for measuring adhesion (10, 20, 21), both of
metal films (22) and polymer films (23, 24). In the 90° peel test the peel force is I
measured as a function of displacement. The peel energy Gp is obtained by the
following expression:
[4]
in which Fp is the peel force, ~L is the peeled length, !lA is the peeled area and
W is the width of the peel strip. For this measurement the following energy balance
is valid:
[5]
During peeling energy is consumed by fracture (Gc) and possibly by bulk plastic
deformation of the film (Gct.r), while energy is supplied externally by peeling (Gr)
and internally by relaxation of residual stresses in the film (G.1). All energy terms
are per unit area. Note the difference between Gdef and Gr1 • The first tbrm stands
for bulk plastic deformation in the metal layer, whereas the second term denotes
the plastic deformation in the microscopic crack tip zone. These two terms may
become indistinguishable when the size of the plastic zone is of the order of the
layer thickness. If no energy is lost in bulk plastic deformation of the metal layer
10
and if the residual strain energy in the layer is very small, then the peel energy
equals the fracture energy.
The residual strain energy Ge1 can either be caused by the deposition process as
built-in stresses or by a difference in thermal expansion between layer and
substrate. The amount of elastic strain energy U per unit volume V due to the
difference in thermal expansion is given by:
u leT
= E ede 0
[6]
in which a is the stress, e the strain, eT the thermal strain, E the Young's modulus
of the film, Aa the difference in thermal expansion coefficients and AT the tem
perature difference. This can be expressed in elastic strain energy per unit area if
the volume V is equal to area A times layer thickness D
E(AaAT)2 D
2 [7]
Similarly to eq. 6, with eq. 8 the residual strain energy Gel due to built-in stresses
can be calculated if the amount of internal stress a, is known:
2 (j·
I
2E
1.2.3. Weibull statistics
Weibull (25) suggested that strength data could be fitted with:
(ar au) m Pr = 1 - exp[ - ( ao ) ]
[8]
[9A]
Here Pr is an estimate of the failure probability, ar is the adhesion strength, a0 is
a normalization constant, au is a threshold stress value below which no fracture
occurs, and m is a fit parameter called Weibull modulus. The parameter uu is
usually taken as zero. An estimation of Pr can be made by placing the exper-
11
imental strength values in the order of increasing strength. The failure probability
Pr can then be estimated by (26):
Pr = N +I
[lOA]
or
Pr = 0.5
N [lOB]
in which N is the number of test specimens and i is the rank number of a particular
specimen in the series of measurements. Recent computer simulations have shown
that eq. lOB is more appropriate, i.e. yields the most accurate estimate with the
least bias (26). Initially, Weibull statistics were used for the interpretation of bulk
material strength data, but later they were also used for adhesion strengths (27).
Equation 9A can be rewritten as (uu = 0):
In (-In (I m In u0 [9B]
By plotting In (-In (1 Pr)) versus ln ur a straight line is obtained with slope m,
if a single distribution of flaw types is present. At ur = u0 , failure occurs with
63% probability. Hence the measurement results can be described by two param
eters m and u0 in which m is a measure of scatter and u0 is a measure of location.
For a large value of m, a small variation in strength values is obtained.
1.3. Electroless metallization
The electroless metallization process was discovered and developed in 1946 by
Brenner and Riddell (28). The name "electroless deposition" refers to the more
generally known galvanic process in which an external electric current is required
for the deposition of metal on a conducting surface from a metal ion containing
12
solution. In the electroless process such a current is not required. The metal ions
in the solution are reduced by a reducing agent which is also present in the sol
ution. Therefore, it is possible to deposit metal layers on non-conducting
substrates simply by immersing a sample in the electroless plating solution. In this
solution apart from a metal ion and a reducing compound, one or more stabilizers
are present to prevent the occurrence of a spontaneous, homogeneous redox re
action in the solution. Carboxylic acids and amino acids are used as complexing
agents in the case of electroless Ni. An overview of principles, deposition condi
tions, chemical reactions and deposit properties of electroless Ni is given in a book
by Riedel (29). In the literature the metal deposit is most often termed "electroless
Ni(P)" instead of the gramatically correct expression "electrolessly deposited
Ni(P)". We use both terms throughout the text.
Thermodynamically, the reducing agent is strong enough to reduce the metal ion
but the reaction is kinetically hindered. In order to get selective deposition on a
substrate surface, this surface is provided with a catalyst (nucleation) which locally
lowers the energy barrier for the deposition reaction to take place. Once the de
position has started, the catalyst is covered with a metal layer and the metal itself
acts as a catalyst in a continuous growth process. The catalyst can be deposited
on the substrate also by immersion in a series of aqueous solutions. Therefore, the
electroless deposition process is a relatively cheap and simple alternative for metal
deposition processes by vacuum techniques such as evaporation or sputtering.
Other advantages of this type of processes are its ability to homogeneously coat
substrates with highly irregular shapes where shadow effects occur in vacuum de
position processes, especially in holes, and the fact that it can be done as a con
tinuous process instead of as a batch process.
Electroless deposition also suffers from a number of disadvantages with respect to
the vacuum deposition techniques. It is not possible to deposit very thin films
homogeneously, below a thickness of about 0.2 j.lm. Moreover, there is a limited
choice in the number of metals that can be deposited in this way, i.e. Ni, Cu, Co,
Pd, Au, in the order of decreasing suitability. In contrast, by vacuum deposition
techniques almost every metal and even many multi-component alloys can be de
posited. Process control of electroless deposition is sometimes difficult and trace
13
amounts of compounds in the solutions may drastically affect the deposition
process and properties of the deposited materials. As described above, on non
conducting surfaces as most ceramics, glass, and plastics, electroless deposited
layers show weak adhesion unless use is made of surface roughness (28). The work
described here focusses on the adhesion of an electroless layer on a ceramic
substrate surface.
Of the electroless deposition processes, the one for Ni is most popular. Depending
on the type of reducing agent used for the Ni deposition, borane or
hypophosphite, boron or phosphor is incorporated into the electroless Ni layer.
Consequently, the deposit is denoted by Ni(B) or Ni(P). These as-deposited metal
layers are X-ray amorphous and crystallize upon heating. The P content of the
Ni(P) material may vary between 3 and 15 wt. % (6 and 30 at. %), mainly de
pending on the pH value of the deposition solution. The B content of Ni(B) is
between 0.5 and 5 wt. % (3 and 30 at. %) which also depends on the deposition
conditions and the type of B-containing reducing agent (29).
Although many elements are claimed in the patent literature (e.g. 30), mostly me
tallic Pd is used as a catalyst for the electroless deposition. This metal can be de
posited on an oxidic or plastic surface by first adsorbing Sn ions, e.g. from a
chloride solution. Pd is deposited by a redox reaction with the adsorbed Sn ions
in a subsequent immersion step. As an intermediate step between the Sn and Pd
containing solutions sometimes an Ag containing solution is used (31). One of the
advantages of this step may be that contamination of the electroless solution by
Sn is decreased. This nucleation process with Sn, Ag and Pd will be treated in more
detail in chapter 3. An alternative nucleation procedure is a single dip process in
a solution containing colloidal particles with both Sn and Pd ions (28). On metallic
surfaces Pd can also be deposited by a displacement reaction, which is a redox
reaction between Pd ions in the solution and less noble metal atoms on the
substrate surface.
14
1.4. Scope of this thesis
1.4.1. Aim
As described m previous sections, the electroless metal deposition process is
excellently suited to provide insulating surfaces with a metal coating. In some cases
only the electrolessly deposited metal layer is used, but often a subsequent
electrodeposition step is applied. The major disadvantage of the electrolessly de
posited metal layers on non-metallic substrates is the poor adhesion. A frequently
used solution for this problem is roughening the surface, thus creating possible
sites for mechanical interlocking. However, in many cases this is not possible e.g
due to a very hard or stable substrate material, the presence of thin films, or due
to process requirements. Therefore the aim of the work described in this thesis is
to find procedures to apply strongly adhering metal layers by electroless
metallization, without making use of surface roughness. To achieve this, first in
sight has to be obtained into the mechanism of adhesion.
The system Ni(P) - Ab03 is considered a suitable system for this investigation,
because electroless nickel deposition is one of the most popular electroless
metallization processes and alumina is a good example of a substrate which cannot
easily be roughened. Moreover, various studies on this system have been published
in order to obtain a strongly adhering metallization of IC packages.
1.4.2. Surface analytical techniques
-SEMI EDX
As a routine inspection, fracture surfaces are generally first studied with optical
microscopy (OM). Magnifications up to 600 times are possible, but in may cases
this is not sufficient. Therefore, most fracture surfaces were also inspected with
scanning electron microscopy (SEM). With SEM, surface topography is imaged
by means of a scanning electron beam. The image is formed by backscattered
electrons and electrons originating from secondary emission. Magnifications up to
50.000 times are possible and details of about 20 nm can be observed. Another
important advantage of the SEM is its high depth resolution at high magnifica
tions, compared with optical microscopy. In order to prevent electric charging,
15
insulating surfaces have to be covered first with a conducting layer of about 15
nm thickness. For this purpose generally gold or carbon films are deposited by
sputtering or evaporation.
The high-energy electrons of the scanning beam, with energies up to 30 keV, excite
atoms in the top 1 /liD of the material. These atoms emit characteristic X-rays. By
energy-dispersive analysis of these X-rays (EDX), semi-quantitative information
is obtained on the composition of surfaces. Combined with SEM, EDX is a quick
method for the inspection of surfaces and it provides information on structure and
composition of surfaces on micrometer scale. In special cases, where top layers
of high atom number elements are present, even layers with a thickness of a few
nanometers can be detected with EDX.
-TEM
Further magnification requires the use of transmission electron microscopy
(TEM). For this type of electron microscopy very thin samples are necessary.
Electrons with an energy of about 100 keV are transmitted through a sample with
a thickness of a few hundred nanometers. The intensity pattern of the electrons is
detected on the other side of the sample. Apart from contrast due to differences
in transmittance, also diffraction patterns are generated in crystalline samples.
With TEM atomic resolution can be obtained. The sample preparation, however,
is rather laborious. For imaging a cross-section of an interface, a thin slice is sawn
from the layer - substrate assembly, perpendicular to the interface. Further
thinning is done by polishing and ultimately by ion-milling. For plan-view images
either thin membranes can be used as model substrates, or replicas can be made
by depositing a thin film on the surface to be investigated, and subsequently peel
ing off the film or by dissolving the substrate. With the replica technique it is
sometimes difficult to achieve atomic resolution, and therefore we made use of the
membrane technique. We used silicon nitride as model substrates for these exper
iments, see chapter 3.
- Static-SIMS
For the analysis of the surface composition on monolayer scale, extensive use has
been made of static secondary ion mass spectrometry (static-SIMS). With this
l6
technique a surface is bombarded with low-dose, low-energy primary ions, usually
Ar+. Fragments liberated from the surface are analysed by a high-resolution
time-of-flight (TOF) mass analyser. This analyser is capable of discriminating be
tween different ions of the same nominal mass. The element fragments generally
originate from the outermost few nanometers of the surface, but the larger, mo
lecular fragments generally only originate from the outermost monolayer of the
sample surface. Due to the high sensitivity of the TOF detector, even ppm's of a
monolayer can be analysed. The major disadvantage of this technique is that only
qualitative information can be obtained. For quantitative analyses long-winded
calibration procedures have to be carried out.
- XPS, AES and XRF
With X-ray photoelectron spectroscopy (XPS) the kinetic energy is measured of
electrons which are liberated upon X-ray exposure. This kinetic energy can be
converted into binding energy of electrons in the inner shells of the surface atoms.
With this information the composition of the outermost few nanometers of a
substrate surface can be quantitatively determined. By measuring the kinetic en
ergy within a resolution of about 0.3 eV, it is also possible to identify the chemical
bonding state of the surface species. This often gives insight into the nature of
interfacial interactions and the origin of interface compounds. By alternately
sputtering and analysing, a depth profile can be obtained, although changes due
to sputtering may disturb the analysis. With Auger electron spectroscopy (AES),
similar information can be obtained. However, the major disadvantage of AES is
that it cannot identify chemical bonding states of the surface species. Both with
AES and XPS surface mappings can be made to investigate the distribution of el
ements over the surface. This feature has not been used in this work.
With X-ray fluorescence (XRF) the element composition of a surface is
quantitatively measured by analysing the intensity and peak position of X-ray
fluoresence lines. The analysed area is at least several square millimeters. XRF has
an analysis depth of the order of 100 pm and a detection limit of the order of 1
to 0.01 monolayer, depending on the atomic number.
17
1.4.3. Outline of subsequent chapters
This thesis is organized as follows: In Chapter 2 a literature overview is presented
on the adhesion of electroless Ni(P) to alumina. Attention is paid to the procedures
and techniques used to study the adhesion, the adhesion strengths obtained and
the models proposed to explain the adhesion. In Chapter 3 the formation of the
interface between Ni(P) and alumina is studied. The changes in structure and
chemical composition of the substrate surface is analysed after each successive
process step. In chapter 4 the adhesion is measured with peel tests and direct
pull-off tests. Quantitative aspects of both adhesion measurement procedures are
highlighted. A fracture mechanics approach is followed for the interpretation of
the adhesion measurement results. The metal - ceramic interface and the fracture
surfaces are analysed both on micrometer and on nanometer scale in order to find
a relation between the measured adhesion and interface chemistry.
In chapter 5 the influence of thermal treatments upon the adhesion strength is in
vestigated using a similar approach as in Chapter 4. Special attention is paid to
a distinction between changes in bulk mechanical properties in the metal film and
intrinsic adhesion. In Chapter 6 the influence of substrate surface chemistry upon
the adhesion is investigated. Thin metal oxide layers are deposited on the ceramic
surfaces used in previous chapters. These metal oxide surfaces are metallized and
the adhesion and interface chemistry are characterized. Differences in adhesion are
tentatively explained in terms of surface chemistry and microroughness. In Chap
ter 7 a completely different procedure is followed to apply the Pd catalyst, required
to start the electroless deposition process. Pd layers are evaporated instead of the
usual wet-chemical procedure. A thin titanium layer is applied for improving the
adhesion of the Pd nucleation layer on the substrate. The adhesion and the frac
ture path are determined and a detailed analysis is made of the bonding at the
metal - ceramic interface within the first metal monolayers. This thesis concludes
with a final discussion and conclusion in which relations between the various parts
of this work are emphasized and an overview of major results and conclusions is
presented. In this chapter also remaining questions and suggestions for further
work are indicated.
18
References
1. M. Ohring, The Materials Science of Thin Films, Academic Press Inc., San
Diego, 1992, Ch. 9, p. 403.
2. A.N. Gent, Plast. Rubber Int. 6, (1981), 151.
3. A.N. Gent and C.-W. Lin, J. Adhesion 32, (1990), 113.
4. C.A. Dahlqvist in "Coatings Technology Handbook", D. Satas, ed., Marcel
Dekker, Inc., New York, 1991, Ch. 5, p. 51.
5. F.M. Fowkes, J. Adhesion Sci. Tech. 1, (1987), 7.
6. B.V. Deryaguin, Research 8, (1955), 70.
7. D.A. Hays, in "Fundamentals of adhesion", L.H. Lee, ed., Plenum Press,
New York, 1991, Ch. 8, p. 249.
8. J.D. Miller and H. Ishida in "Fundamentals of adhesion", L.H. Lee, ed.,
Plenum Press, New York, 1991, Ch. 10, p. 291.
9. J.E.E. Baglin, Nucl. Instr. and Meth. B65, (1992), 119.
10. J.E.E. Baglin in "Fundamentals of adhesion", L.-H. Lee, ed., Plenum Press,
New York, 1991, Ch. 13.
11. F. Brochard-Wyart in "Fundamentals of adhesion", L.H. Lee, ed., Plenum
Press, New York, 1991, Ch. 6, p. 181.
12. E.P. Pluedemann, "Silane Coupling Agents", Plenum Press, New York, 1982.
13. K.L. Mittal in "Surface Contamination, Genesis, Detection and Control",
K.L. Mittal, ed., Vol. 1, Plenum Press, New York, 1979.
14. R.J. Good in Adhesion Measurements of Thin Films, Thick films and Bulk
Coatings, K.L. Mittal (ed.), ASTM STP 640, (1978), p. 63.
15. S.J. Bennett, K.L. de Vries and M.L. Williams, lnt. J. Fracture 10, (1974),
33.
16. S.A. Varchenya, A. Simanovskis and S.V. Stolyarova, Thin Solid Films 164,
. (1988), 147.
17. A.J. Kinloch in "Adhesion and Adhesives", Chapman and Hall, London,
1987, Ch. 3.
18. H.F. Fischmeister, G. Elssner, B. Gibbesch and W. Mader, Materials Re
search Society International Meeting on Advanced Materials 8, (1988), 227.
19. Ref. 17, Ch. 4.
20. K.L. Mittal, J. Adhesion Sci. Tech. 1, (1987), 247.
19
21. K-S. Kim, Mat. Res. Soc. Symp. Proc. 119, (1988), 31.
22. J.E.E. Baglin, Mat. Res. Soc. Symp. Proc. 47, (1985), 3.
23. A.N. Gent and J. Schultz, J.Adhesion 3, (1972), 281.
24. G.J. Lake and A. Stevenson in "Adhesion 6", K.W. Alien ed., Applied Sci-
ence Publishers, London, 1982, p. 41.
25. W. Weibull, J. Appl. Mech. 18, (1951), 293.
26. L.J.M.G. Dortmans and G. de With, J. Am. Ceram. Soc. 74, (1991), 2293.
27. J.E. Ritter, L. Rosenfeld, M.R. Lin and T.J. Lardner, Mat. Res. Soc. Symp.
Proc. 130, (1989), 237.
28. A. Brenner and G. Riddell, J. Res. Nat. Bur. Stand. 39, (1946), 385.
29. W. Riedel in "Funktionelle Chemische Vernicklung", E.G. Leuze Verlag,
Saulgau, 1989.
30. R.L. Jackson, U.S. Patent 4.701.351, 1987.
31. C.H. de Minjer and P.F.J. v.d. Boom, J. Electrochem. Soc. 120, (1973), 16.
20
Chapter 2
Adhesion of electrolessly deposited Ni(P) on alumina ceramic:
an assessment of the current status
Summary
Literature data on the adhesion of electrolessly deposited Ni(P) films on
alumina ceramic substrates are reviewed. The influences of conditions of
the successive etching, nucleation and metallization processes on the ad
hesion are discussed as well as the effect of subsequent annealing treat
ments. Also, a comparison is made with the adhesion of electrolessly
deposited Ni(B) and Cu layers. It is noted that in general too limited in
formation is provided by most authors on the adhesion measurement con
ditions and procedures. It is concluded that the etching conditions are of
more importance for adhesion than the nucleation, metallization and
annealing conditions. It is commonly believed that mechanical interlock
ing is the dominant adhesion mechanism. However, bilayer experiments
with electrolessly deposited Ni(P) and Cu suggest that the intimacy of
interfacial contact plays an additional role. This may indicate that van
der Waals or other interfacial interactions significantly contribute to the
adhesion.
In order to obtain further insight into the adhesion mechanism, a fracture
mechanics characterization is suggested. Modern surface analytical tech
niques should be applied to study the interfaces and fracture surfaces.
21
2.1. Introduction
The metallization of alumina ceramic surfaces is widely used in the electronics in
dustry, among other things for integrated circuit (IC) packaging, printed circuit
and sensor applications (l - 6). Alumina is inexpensive ·and offers the advantage
of a relatively high substrate thermal conductivity compared to most other insu
lating materials (3, 7). Often, metallization is performed by applying a paste on the
ceramic surface. This paste contains a mixture of glass and metal powders and an
organic binder. After annealing in air at temperatures of 600 to 800 oc, a well
adhering conducting glass-metal composite layer is obtained (1, 2, 4, 8). Other
frequently used techniques for the deposition of metal layers are evaporation and
sputtering, which are mostly limited to metal films with a thickness less than
1 f.J-m.
An alternative approach is the metallization by electroless deposition of Ni(P).
Since annealing steps or vacuum equipment are not required, electroless
metallization is a relatively rapid and inexpensive process. With electroless
metallization, Ni metal is formed by a chemical reaction of Ni ions with a reducing
agent. Therefore, no external electric current is required and insulating substrates
can be plated by simply immersing the substrates in a series of aqueous solutions.
In refs. (9, 10) backgrounds of the metallization process and properties of the de
posits are described. On top of the conducting Ni(P) layer one or more metal
layers can be deposited by electrodeposition. The major drawback of this proce
dure is that a high adhesion strength of the Ni(P) layer to the alumina ceramic
substrate is difficult to obtain and is generally considered to be too low (11).
Therefore, in the literature a number of experiments have been described in which
the conditions and procedures have been varied to improve the adhesion and to
provide a model for understanding the adhesion effects measured. In this chapter
relevant literature data on this adhesion problem are collected and discussed. In
this literature study the following questions were used as a guideline:
1. What is the adhesion strength between electrolessly deposited Ni(P) and
alumina?
2. What is the influence of experimental conditions on the adhesion strength ?
22
3. How do these conditions influence physical or chemical interactions at the
interface?
4. Which routes should be followed for further optimization of the adhesion
strength?
For electroless metallization, the substrate surface must first be made catalytically
active by a nucleation procedure. In this regard, generally two types of procedures
can be distinguished: a one-step procedure with a colloidal Pd containing solution
(12), and a two-step procedure consisting of a sensitization step with a SnCb
containing solution followed by an activation step with a PdCb containing sol
ution (13, 14). In some cases in the latter procedure an intermediate step is also
used with an AgN03 solution (15, 16). Analogously, this is referred to here as the
three-step procedure. In order to deposit more nucleation material on the surface,
the two-step procedure may also be applied repeatedly. All of these various
nucleation procedures have been used in the adhesion studies as described in the
subsequent sections.
The metallization procedure generally consists of three stages; etching, nucleation
and the actual metallization. In the following sections first the literature data are
presented and discussed for each of these stages. Then, the effect of thermal
treatment is considered and a comparison is made with the adhesion of
electrolessly deposited Cu and Ni(B). In the last section conclusions are drawn and
some recommendations for future work are presented.
2.2. General procedures for sample preparation and adhesion
measurements and overview of results
A summary of the most relevant literature data, such as adhesion strengths and
experimental conditions, is presented in Table 1. For all of the results listed in
Table I, 96 % pure alumina was used as the substrate, unless otherwise stated.
This makes the results from different publications rather well comparable, al
though differences in the microstructure of the ceramics used probably have oc-
23
curred, which may also influence the adhesion strengths, in addition to the various
process conditions.
Unless stated otherwise, the results described in the following text refer to adhesion
strengths obtained by the direct pull-off (DPO) test (1 - 4, 15, 17 - 19). For this
test tin-plated copper wires were soldered to 2 to 4 ,urn thick patches of electro
lessly deposited metal of 2 x 2 mm2 size on the sintered surface of alumina subs
trates. This is schematically shown in fig. 1. In sections 3 to 7 these data will be
described in more detail and discussed. In most references, the number of test
samples used, standard deviations or other information on accuracy or reprodu
cibility of the adhesion strengths were not given. Generally, the standard deviation
in the mean strength s(ar) depends on the Weibull modulus m and the number of
test samples N as given in eq. 1:
[1]
For ceramic- metal joints m is typically :::;; 5 (20). If we take N = 10 and m = 3
then the relative standard deviation in the mean strength is about 10 %. There
fore, differences in adhesion strengths of at least 10 % are not considered to be
significant. Although mostly not indicated in the literature, it is generally assumed
that the strength values reported refer to failure at the metal- ceramic interface.
In addition to the DPO test, Honma et al. ( 4, 19) use another adhesion test in
which an L-shaped wire is soldered onto the electrolessly deposited metal patches
instead of a straight wire as for the DPO test (fig. 1). This test is referred to as a
peel test in ( 4), but as a pull-test in (19). The adhesion values obtained with this
test are expressed in force per unit area, which is at least uncommon for a peel test.
It is not made clear why in some cases this test is chosen instead of the DPO test.
The values obtained with this test cannot be compared with DPO results and can
only be used to show qualitative trends in the adhesion due to variation of process
parameters. In the following, this test is referred to as the L-pull test.
24
Table 1: Adhesion strength (MPa) of electrolessly deposited Ni(P) and Cu on 96 % alumina as measured with the direct pull-off test.
: Experimental conditions DPO strength (MPa) Ref
Etching, (amount) Nucleation Metallizalion Ni(P) Cu
10% HF, (0.15 mg/cm2) Two- step' Ni(P) bath2, Cu batb3 26 IS )
. Pd alkali ion I . 25 14 3
. Pd colloid1 . 25 14 3
", (0.!5 5 mg/cm2) Cucolloidl h 21 12 3
5 % HF, (0.255 mg/cm2) Repeated two step1 Ni(P) and Cu baths as ref. 3 27 14 17
20 % HF, (0.175 mg/eml) Repeated two-step pH 62, as deposited at 90 'C !6 2
Repeated two-step pH 62. I b 300 'C 19 2
i H Repeated two-step pH 62, I h 500 •c 16 2
. Repeated two-step pH 92, as deposited 19 1
! ff Repeated two-step pH 92, I h 300 •c 19 2
. Repeated two-step pH 92, I h 500 'C 22 2
. Repeated two-step pH 102, as deposited 19 2
. Repeated two-step pH J02, I h 300 'C 20 2
. Repeated two-step pH J02, I h 500 •c 21 2
H Repeated two-step Cu bath3 without additive 10 2
H Repeated two-step Additive A4 12 2
. Repeated two-step Additive s4 12 2
. Repeated two-step Additive A+ B l3 2
IOOg/1 NH4F + IOOg{J NaCJ!O Two- step6 Ni(P) citrate bath9 at pH 6, 70 •c 24'5 4
: IOOg{J NH~ + IOOg/1 NaC!lO Catalystfaccelerator7 . 18 4
IOOg{J NH4F + IOOg/1 NaC[IO Activatorfaccelerator8 . 13 4
.No etching Two·step . 21 4
j 10% HF10 Two-step . 14 4
: 10 % HF + 10 % HCilO Two-step . 10 4
100 g{l N~F10 Two-step . 23 4
j IOOg{l N~ + IOOg/1 NaClto Two-step . 2915 4
IOOgfl ~ + IOOg/1 NaCtiO Two-step At 0.1 moi{J NiS04 16 4
10% HF11 Two-step or three-step Maleic acid bath12 at 90 •c 2516 IS
HCI or HNO) etchingll Two-step or three-step . 15 15
80 min I molfl NaOH'3 Activator/acceleratorl4 Citrate bath 14 at pH 4, 90 ·c 30 11
25
1: Two-step nucleation with t g/1 SnCh, L2 g/1 HC1 and 0.1 g/1 PdCh, 0.12 g/1 HCl, both at 25
oc, l min; Pd alkali ion nucleation with Neogant 834, 40"C, 8 min and Neogant W A, 25 oc.
4 min (Schering). Pd colloid nucleation with Cataposit 44 and Accelerator 240. both 40 "C, 4
min (Shipley); Cu colloid nucleation with Ronacat catalyst M (8 min) and Ronacat stripper (1
min) (LeaRonal) both at 25 T. For the repeated two-step procedure, the two-step procedure
described above is carried out twice.
2: Ni(P) bath: 0.1 mol/! nickel sulphate, 0.15 mo!f! hypophosphite, 0.2 molj! sodium citrate, 0.5
mol/1 ammonium sulphate, pH adjusted with NaOH. In refs. 3 and 17 at pH 9 and 90 "C.
3: Cu bath: 0.04 moljl CuS04 .5H20, 0.08 mol/! EDT A.8H20, 0.05 mol/! HCHO, 20 ppm
(C5H4Nh, 50 ppm ~{Fe(CN)6}, pH 12.5, 60 •c, air bubbling.
4: Additive A: ~(Fe(CN)6) 50 mgfl. Additive B: (CsH4Nh 20 mg/1.
5: At optimum etching conditions.
6: Two-step nucleation with 0.05 g/1 SnCI2.2H20 and 0.1 g/1 PdCh both at 40 •c.
7: Precatalyst and catalyst are Cataprep 404 and Cataposit 44, both from Shipley. Acceleration
is done with NaOH 100 gfl.
8: Preactivator, activator and reducer are Neoganth B, Neoganth 834 and Neoganth W A respec
tively, all from Schering.
9: Ni(P) bath: 0.05 molfl nickel sulphate, 0.1 mol/! sodium citrate and 0.2 mol/! sodium hypo-
phosphite.
10: 15 minutes etching at 60 °C.
11: 10 minutes etching, room temperature.
12: 95 % alumina substrate, 15 g/1 NiS04. 6 H20, 24 g/1 NaH2P02. 6 H20, 5 g/1
HOOCCHOHCH2COOH , 5 g/1 C4H40~a2 . 6 H20, 7 g/1 CH3COONa . 3 H 20 and 0.5 ppm
Pb stabiliser in Ni(P) bath.
13: AlN substrates with CaC2 second phase, etching at room temperature.
14: Activator and accelerator HSIOlB and ADP-101 both from Hitachi Chemical. Ni(P) bath same
as in ref. 3, pH 4 adjusted with H2S04 •
15: Different adhesion strength values reported for similar preparation conditions.
16: Fracture in ceramic substrate.
26
Fig. 1:
Fig 2:
Solder
-0 0.8 mm Tin plated copper wire
""" ~ Ni(P) film (4 mm2) ,-;r;rn-rr77:/'77 / /'/ /;
Pu 11 Strength
l-pull Strength
Schematic set-up of direct pull-off test (A) and L-pull test (B), accord
ing to ref. (4).
t 30 ai b) c) d)
., 25 0.
2 .r: & 20 c: ~
"' c: 15 .9 en
"' .c " <t 10
5
0 0.1 0.2 0.3 0 0.1 0.2 0.3 0 0.1 0.2 0.3 0 0.1 0.2 0.3 Amount etched (mg cm-2)-
DPO adhesion strength of electrolessly deposited Ni(P) and Cu versus
degree of etching of the alumina substrate, for various nucleation pro
cedures, according to ref. (3). The nucleation procedures are: two-step
(A), Pd alkali ion (B), Pd colloid (C) and Cu colloid (D), see Table 1.
27
2.3. Effects of process parameters on the adhesion
2.3.1. Etching conditions
The alumina ceramics which were used in the investigations reported in the litera
ture generally contain a few percent glass phase, which is used as a sintering aid
and is present at grain surfaces and grain boundaries. With etching, in principle
two effects can be distinguished. Firstly, by using alkaline or fluoride containing
aqueous solutions, the glass phase is selectively removed and gaps are created be
tween surface ceramic grains, thus providing anchoring sites for adhesion by me
chanical interlocking. This is referred to as low-temperature etching. Secondly, the
alumina grains themselves can be roughened. However, due to the high stability
of a-alumina, this requires severe etching conditions, and thus excludes the use of
aqueous solutions. For this type of etching the use of molten alkali salts at tem
peratures of at least 300 oc is reported. Therefore this is referred to as high
temperature etching. Some examples of both procedures are given below.
Osaka et al. (3) found an increase in adhesion strength of Ni(P) on the sintered
surface of 96 % alumina from 15 MPa to 26 MPa by etching with 10 % HF to a
weight loss of 0.1 mg/cm2 (15), (fig. 2). When polished 96 % alumina substrates
were used, an increase in the adhesion strength from ll MPa to 27 MPa was at
tained after etching to 0.25 mg/cm2 with a 5 % HF solution (17). The adhesion
of the Ni(P) layer on the polished substrates before etching can be explained by
the porosity of the ceramic. The surface pores as shown in SEM micrographs (17)
may have served as anchoring sites. Honma and Mizushima (l) used a SnF2 sol
ution for simultaneous etching and sensitization of the alumina surfaces. This led
to relatively low adhesion strengths of about 10 MPa. Kamijo and Ayuzawa {15)
found an adhesion strength of about 25 MPa on 95 % alumina when HF etching
was used and strengths of the order of 15 MPa when other acids were used for
etching. They found no positive influence of sensitization with SnF2 relative to
SnC}z.
Also Honma and Kanemitsu (4) studied the influence of various etchants upon the
adhesion strength. Without etching, an adhesion strength of 21 MPa was found
which was reduced to 15 MPa by etching with a lO % HF solution. This is sur-
28
prising since other authors (2, 3, 15, 17) all report adhesion improvement by
etching with HF solutions. Moreover, the adhesion strength of 21 MPa obtained
without etching is rather high. By etching with NH4F, mixed HF I NaCl and
NH4F f NaCl solutions adhesion improvements to 25 to 30 MPa were observed.
If these results are correct, they show the importance of the etching procedure.
With the same procedure for nucleation and metallization both the weakest and
the strongest adhesion (Table 1) are found, only by changing the etching proce
dure. In a later publication (19) the adhesion improvement by etching with NH4F
was explained by the observation that cracks and pits had been formed in the glass
phase, creating additional opportunities for mechanical interlocking. This, how
ever, can only be a satisfactory explanation if the glass phase is not removed but
only roughened by this etching procedure.
Etching generally increases the adhesion strength. In the curves of adhesion
strength versus etching time generally an optimum is found (3, 18, 19). The in
crease in adhesion strength is ascribed to an increase in the number and size of
anchoring sites while the decrease is explained by underetching of the surface ce
ramic grains which therefore become weakly adhering themselves (18, 19). This is
schematically shown in fig. 3. This model is supported by the analysis of the metal
fracture surface at various degrees of etching (2). Beyond the optimum adhesion
strength, an increasing amount of ceramic is found to remain on the metal fracture
surface with increasing etching time.
NaOH solutions have been used for etching AlN, resulting in increases in adhesion
strength of Ni(P) from 6 to 30 MPa (11) and from 20 to 30 MPa (7). The difference
between the initial adhesion strengths in the two studies is probably due to a dif
ference in microstructure, resulting from small differences in the fabrication pro
cedure for the AIN substrates (7). Similarly as for the 96 % alumina ceramics,
etching was found to occur mainly at grain boundaries (11) or at grain triple
points (7). This was explained by the selective removal of the CaO second phase,
resulting from the CaC2 sintering aid which was used for the preparation of the
AIN substrates in both studies (7). After prolonged etching a gradual decrease of
the adhesion strength was measured (7) similarly as described above for alumina.
29
The release of top grains from the substrate during prolonged etching was ex
plained by dissolution of AlN at grain boundaries.
Fig. 3:
a)
Schematic representation of unetched (A), etched (B) and over-etched
(C) 96 % alumina surfaces, according to ref. (19).
A comparison was made (11) between etched AlN, mechanically abraded AlN and
mechanically abraded alumina, all with the same center line average roughness (21)
of 0.59 p,m. The adhesion of Ni(P) amounted to 18, 16 and 8 MPa, respectively,
which was explained by the difference in surface morphology as observed with
SEM. A channel-like porosity was formed by etching between the AIN grains. It
is, however, questionable whether the difference in adhesion strength between the
first two samples is significant.
The effects of etching with nitric acid and HF solutions have been compared to the
effect of etching with molten NaOH on adhesion of electrolessly deposited Cu on
96 % alumina (5). The adhesion was measured by a dot-bend test. For this test a
4. 7 mm diameter brass stud is soldered over a 3.8 mm etched Cu dot and pulled
30
at 90° angle to the stud (5). The results are expressed as fracture loads. Due to the
test geometry, the fracture loads cannot be converted into fracture stresses for
comparison with other data. A dot-bend adhesion value of ll N was found after
etching with a boiling 70.6% HN03 solution for 30 min, 14 N when no pretreat
ment was applied, 18 N after etching for 15 minutes in a 48 % HF solution (ul
trasonic) at 50 oc and 23 N after etching with molten NaOH at 420 oc for 15
minutes. From these data it is clear that high-temperature etching yields higher
adhesion than etching in aqueous solutions. The optimum temperature was found
to be 420 oc. However, due to the difference in the adhesion test procedure these
values cannot be compared with direct pull-off values reported in other publica
tions.
The influence of etching with various molten alkali hydroxides on the adhesion
strength of electrolessly deposited Cu on 96 % alumina has been studied (19).
Substrates were first immersed separately in a 10 % solution of NaOH, LiOH,
KOH or combinations of these salts at room temperature. During this treatment
the grain boundary glass phase was dissolved. The optimum immersion time was
found to be 10 minutes. After this step, the samples were not rinsed but imme
diately heated at high temperatures for 15 minutes. Water evaporated and the re
maining alkali hydroxide melted and attacked the uncovered alumina grains. The
three alkali hydroxides gave rise to different types of roughness. With NaOH rel
atively deep channels were etched, with LiOH shallow channels and with KOH an
irregular roughness was obtained. This resulted in a DPO strength of 30 MPa for
etching with NaOH, 20 MPa for LiOH and lO MPa for KOH. As previously de
scribed, by low-temperature etching these values are typically 10 to 15 MPa, see
also in Table 1. The optimum etching temperature of 450 "C agrees well with the
optimum temperature reported in (5).
From the experiments described in this section, it can be concluded that adhesion
improvements of up to 50 % have been realized by optimization of the etching
procedure. This is explained in the various publications by the mechanical inter
locking model. However, also the interfacial area increases by etching. It is difficult
to evaluate the relative contributions of mechanical interlocking and direct inter
facial interactions from these literature data.
31
2.3.2. Nucleation conditions
Many authors have considered the influence of the nucleation on the adhesion.
Schlesinger and Kisel (22) stated that the density of initial Ni(P) sites on the ca-i
talyzed surface determines the adhesion properties of the metal film. They influ-
enced this density by changing the chemistry of the sensitizer solution. However,
they neither gave a reference nor experimental data to support this statement re
garding the adhesion. Also Feldstein et al. (13) considered it reasonable to spec
ulate that the adhesion would be improved, by changing the composition of the
sensitizer solution, which led to a higher catalytic activity of the surface and a
more homogeneous coverage of initial Ni(P) nuclei. They did not give any evidence
to confirm this speculation either. The following authors varied the nucleation
conditions and measured the resulting adhesion strength.
Osaka et al. (3) measured the adhesion of electrolessly deposited Ni(P) and Cu on
96 % alumina after nucleation with the two-step SnCh I PdCb procedure and after
various Pd and Cu colloidal nucleation procedures. They found that the differ
ences in nucleation procedures did not lead to differences in the adhesion strength
for the Pd colloidal nucleation procedures. Only for the Cu colloid nucleation
procedure a weaker adhesion was found, see Table I and fig. 2. On the other hand,
Honma and Kanemitsu (4) found that the one-step Pd colloidal nucleation proce
dures (catalyst I accelerator, and activator I accelerator) led to 25 to 50 % lower
adhesion strengths than with the two-step procedure using similar etching, nu
cleation and metallization conditions as in (3), see Table 1. This may indicate that
nucleation procedures with colloidal solutions can result in equal or lower adhe
sion strengths than with the two-step nucleation, depending on the type of colloi
dal solution. The two-step nucleation in (2) was repeated twice, probably leading
to a higher amount of Sn and Pd nucleation material on the alumina surface.
When, however, the adhesion strengths are compared with those from other ref
erences cited in Table 1, a lower rather than a higher adhesion strength results
from this repeated procedure. Since the data of different authors are not com
pletely comparable, due to, e.g., differences in substrate microstructure, this may
imply that the amount of nucleation material is of minor importance for the ad
hesion. This interpretation is consistent with the fact that the large differences in
Sn and Pd surface coverage due to various nucleation procedures in (1), do not
32
correspond to large differences in adhesion. The differences in adhesion reported
in that paper are probably caused by etching effects.
De Minjer and v.d. Boom (16) introduced an intermediate immersion step with an
AgNOJ solution, in-between the sensitization and the activation step. They claimed
that more homogeneous nucleation occured with this step. This was confirmed by
TEM micrographs after initiation of the metallization, without speculating on the
resulting adhesion. Honma and Mizushima (1) compared the adhesion strength
obtained with and without the AgN03 step and found a 5 to 20 % higher adhesion
strength with the AgN03 step on 96 °/o alumina. A similar effect was observed
using lead zirconate J titanate ceramics. Kamijo and Ayuzawa (15) measured the
adhesion of Ni(P) on 95 % alumina after various etching and nucleation proce
dures. They found that the adhesion was not significantly influenced by the intro
duction of the AgN03 immersion step. From the above described results it can
be concluded that the Ag step has no effect, or, if any, only a small positive effect
on the adhesion strength.
Other conditions varied by Kamija and Ayuzawa (15) were the composition of
etching solutions, Sn and Pd concentrations and the use of fluoride or chloride Sn
sensitizer solutions. They found no significant influence of these variations in nu
cleation conditions on the adhesion. The concentrations used were 1 % for SnF2,
0.1 and 0.5 % for SnCh, 0.1, 0.9 and 1.5 % for AgN03 and 0.1 and 0.5 % for
PdCh. The number of rinsing steps following the sensitization and activation steps
was found not to influence the adhesion either. The major effects they reported
originated from the composition of the etching solution, see Table 1. Also Honma
and Kanemitsu (4) found for the SnCb step that the concentration did not signif
icantly influence the adhesion strength at concentrations between 0.05 and 1 g/l.
In contrast to the above results a lower adhesion strength was found at 5 g/1
(0.5 %), though, with such a large scatter that these results become inconclusive.
Honma and Kanemitsu (4) speculated that the increased adhesion strength which
was observed when NaCl was added to the etching solution (Table 1), can be ex
plained by increased sensitizer adsorption by an ion-exchange mechanism. How
ever, between the etching treatment and the sensitization step a rinsing step in
33
deionized water was always applied and therefore it could be expected that sodium
ions were desorbed before the sensitizer step. Therefore, it is more probable that
NaCl influences the etching action of the solution, for instance by influencing the
ionic strength of the solution, or by increasing the solubility of the silicate etching
products.
De Luca and McCormack (23) also reported the influence of an immersion step
in a halide containing solution upon electroless metallization. Incomplete Cu
coverage was observed on 90 % alumina ceramics, unless an immersion step in an
acid halide solution was applied. This was also ascribed to increased sensitizer
adsorption, again without experimental evidence. However, in this case it seems
more likely that the acid halide dip removes residue from the previous molten al
kali salt etching step, as also reported by Ameen et al. (5).
For alternative nucleation procedures such as with evaporated Pd (24), with an
aminosilane modification replacing the sensitization step (25), or with mixed col
loidal I sol-gel solutions (26) an influence or even a stronger adhesion is claimed,
but no data on the adhesion strength were given. Therefore these procedures are
not further considered.
From the above results it can be concluded that the two-step nucleation procedure
is suitable for obtaining strong adhesion. With this procedure, the adhesion
strength is relatively insensitive to concentrations, the use of an intermediate
AgN03 step and repetition of the nucleation procedure. In the case of one-step
nucleation with colloidal nucleation solutions the type of solution is more critical.
2.3.3. Metallization conditions
The next step generally applied is the actual metallization step. Honma and Ka
nemitsu ( 4) and Osaka et al. (2, 17) studied the dependence of the adhesion
strength on some electroless metallization bath conditions such as pH, type of
complexing agent and Ni and hypophosphite concentrations. Similarly, the dis
solved oxygen content of the bath was reduced (4, 27, 28). As a consequence of
these variations the deposition rate and P content or the deposit also varied.
34
Variation of the pH value of the electroless Ni(P) metallization bath between 3.5
and 6 did not lead to significant changes in the adhesion strength, as it varied
within a range of 10 to 20 % (4). This was found for baths with citrate and glycine
complexing agents. Similar results were found for a citrate bath at pH values of
6, 9 and 10 (2), see Table L
By reducing the NiS04 concentration from 0.05 molfl to 0.01 molfl, an increase in
the adhesion strength from about 3.5 MPa to about 7 MPa was measured by the
L-pull test (fig. 4) by Honma and Kanemitsu (4). With the DPO test an increase
in the adhesion strength of between 20 and 40 % was measured by lowering the
Ni concentration. The Nifhypophosphite ratio was kept at 1/4 in these exper
iments in order to maintain an acceptable deposition rate. The L-pull test exper
iments were carried out with various complexing agents as shown in fig. 4. Most
other studies in this publication were done with citrate baths. The hypophosphite
concentration did not appreciably influence the adhesion strength at concen
trations between 0.03 and 0.07 mol/l.
At the above mentioned low Ni concentrations, the deposition rate dropped by a
factor of 3 to 8 to a rate of 0.5 to 1 p.m per hour, and sometimes a spotty deposit
was observed, or no deposit was formed at all. By bubbling Ar gas through the
solution the 02 content was decreased from 3.2 ppm to 0.3 ppm, and good-quality
deposits could be made at aNi concentration of 0.01 mol/l. At aNi concentration
of 0.1 mol/! a DPO strength of 16 MPa was measured without argon bubbling,
while at Ni concentrations in the range between 0.05 and 0.01 mol/1, DPO
strengths of 28 to 30 MPa were measured with argon bubbling. Since argon bub
bling was not applied with the high-concentration Ni(P) bath, it is not clear
whether the adhesion improvement is caused by the lower oxygen content or by
the lower Ni concentration. For electroless Cu deposition Alpaugh et al. (27, 28)
described procedures in which higher adhesion was obtained by reducing the oxy
gen content of the metallization solution. On the other hand, a low Ni concen
tration and bubbling with Ar is not a prerequisite for obtaining strong adhesion
as has been shown by other studies (2, 3, 17).
35
Fig. 4:
Citric acid bath
Gluconic acid bath
Glycolic acid bath
Rochelle salt bath
Tartaric acid bath
Glycine bath
0
D Low concentration (NiS0 4 = 0.01 moi/L)
E';J High concentration (NiS04 = 0.05 moi/L)
: :
I :
: I
: I
: 2.5 5 7.5
L-pull strength (MPa)-
Influence of various complexing agents on adhesion strength as meas
ured with the L-pull test, according to ref. (4).
As far as indicated, the electroless Ni(P) metallization baths did not contain Ph
stabilizer, except one (15), which contained 0.5 ppm Pb. In commercial electroless
Ni(P) metallization solutions generally a few ppm of Pb is present, because this
enhances the stability and thus facilitates the handling of these solutions under
practical conditions. It has been shown (29) that the Ph concentration at ppm
levels strongly influences the initiation of Ni(P) growth. In fact, it inhibits the
growth of the smallest particles in a similar way as dissolved oxygen does. There
fore, it is possible that Ph negatively influences the adhesion strength, although the
values reported in (15) are not considerably lower, compared with other values in
Table I.
So, it can be concluded that the adhesion strength is not very sensitive to deposi
tion conditions such as pH, temperature, hypophosphite concentration and type
36
of complexing agent. Some data indicate a positive influence on the adhesion
strength at a low Ni concentration and a low concentration of dissolved oxygen.
2.3.4. Heat treatments
In practical applications, the deposited layers are often exposed to elevated tem
peratures. Therefore, some authors studied the influence of temperature on the
adhesion strength. As an example, the adhesion strength as a function of anneal
ing temperature and time is shown in fig. 5 (!). By annealing at temperatures
above 250 oc and for longer than 1 hour, considerable adhesion improvement is
achieved, from an initial value of 2.5 MPa to a maximum value of 15 MPa.
However, it should be noted that the initial adhesion strength in this experiment
is remarkably low. Later Honma and Kanemitsu (4) measured the pull strength
as a function of annealing time at 250 oc and found only a small adhesion im
provement within the first 30 minutes from 23 MPa to 27 MPa (fig. 6). The ad
hesion strength did not change upon longer annealing, up to 24 h. From these
results they concluded that stresses which might occur due to crystallization
shrinkage do not affect the adhesion. The initial change in adhesion strength was
ascribed to desorption of water from the interface.
Osaka et al. (2) measured the adhesion strength of Ni(P) films before and after
heat treatments for 1 hour at 300 and 500 oc. Also the hardness was measured in
order to establish the relation between bulk mechanical properties and adhesion,
which could be expected due to the mechanical interlocking model. They found
adhesion strengths in the range of 16 to 22 MPa, without a significant influence
of the thermal treatment, see Table 1. However, for the Ni(P) films the hardness
considerably increased, which, according to the authors, suggested that besides
mechanical interlocking other factors play a role in the adhesion too. They used
a vacuum atmosphere probably in order to avoid oxidation of the Ni(P) surface.
This oxidation negatively affects the solderability that is required for the adhesion
measurements.
From the above observations on thermal treatments it can be concluded that for
Ni(P) films with a reasonable as-deposited adhesion strength, the effect of thermal
treatments is negligible.
37
Fig. 5:
Fig. 6:
38
t 15
m-a. 2 J:: 0, 10 c: ~ 1i) c: 0 ·;;; Q)
5 J:: "0 <(
R.T 100
Annealing temperature (°C)-
Influence of annealing time and temperature on the adhesion strength
of Ni(P) on alumina, starting with a relatively low adhesion strength,
according to ref. (1 ).
1 30 () ___ -- -- - -- -----~ 25 :/_... o a .o. ............. -a--ss--o ..
0, ;::
~ 20 1i} ;:: 0
·~ 15 .c "0 <(
5
0 2 3 24 Annealing time (h)---
Influence of annealing time at 250 "'C on the adhesion strength of Ni(P)
on alumina, starting with a relatively strong adhesion, according to
ref. (4).
2.4. Comparison Ni(P) with Ni(B) and Cu
Apart from Ni(P), also electrolessly deposited Ni(B) and Cu are frequently used
with the same type of substrates. By comparing the adhesion of these metal layers
with Ni(P) more insight into the adhesion mechanism can be obtained. Kang et
al. (18) measured the adhesion of Ni(P) and Ni(B) on 90 % alumina as a function
of etching time. The adhesion strength ranged between 12 and 15 MPa and was
the same for Ni(P) and Ni(B), within a remarkably small range of a few %. For
both deposits the two-step nucleation procedure was used, after etching with
10% HF + lOO g/1 NaCI at room temperature. The optimum etching time was
2.5 min on this ceramic.
A comparison of the adhesion data summarized in Table 1 shows that the adhe
sion strength of Ni(P) is generally almost two times higher than that of electro
lessly deposited Cu (2, 3, 17). In the curves of adhesion strength versus degree of
etching, the difference in adhesion strength between Ni(P) and Cu remained con
stant in ref. (3), but decreased in ref. (17), where polished 96 % alumina was used
as the substrate. With SEM it was observed that the initial Cu deposits were
rougher, and had a larger particle size than Ni(P) deposits (3, 17). The Ni(P) and
Cu morphology at the initial stage of deposition was investigated also with TEM
(30). For Cu a relatively coarse deposit was observed, with less grains and of a
more angular shape, compared to Ni(P) using identical nucleation procedures. By
using various additives the microstructure of the electrolessly deposited Cu layer
could be influenced (19). The finest structure with the smallest initial Cu particles
was obtained by adding 20 ppm BeS04AH20 to the metallization solution. An
adhesion improvement of 30 % was achieved with this additive, as measured with
the L-pull test. This finer structure of the initial electrolessly deposited Cu particles
may enable a more efficient mechanical interlocking. Similarly, the stronger adhe
sion of the finer Ni(P) deposits can be explained by a more efficient mechanical
interlocking, relative to Cu. If, however, mechanical interlocking determines the
adhesion strength, then the bulk mechanical properties of the metal layer, such as
the tensile strength, can be expected to play an important role, due to metal-metal
fracture at interlocked sites. This was not experimentally confirmed as shown be
low.
39
Osaka et al. (17) first deposited a thin layer of Ni(P) and subsequently continued
metal growth by electroless deposition of Cu. Within a range of 0.05 no 2 pm, in
dependent of the Ni(P) layer thickness, an adhesion strength of 27 MPa was found.
This is equal to the adhesion strength of pure Ni(P) deposits, while tbe adhesion
strength of electrolessly deposited copper films only on the same surfaces
amounted to 14 MPa. The bulk tensile strength of the Ni(P) film, however, was
about 10 times higher than that of Cu. Therefore, with the same degree of me
chanical interlocking, a ten times higher adhesion strength of Ni(P) is expected,
compared to Cu. However, since the mechanical interlocking of Ni(P) is more ef
ficient than for Cu, an even larger difference between the adhesion strengths is
expected, while experimentally a much smaller difference was found. This strongly
suggests that mechanical interlocking is not the only adhesion mechanism for these
systems and that additional, interfacial effects play a role. The fraction of the in
terfacial area which makes intimate contact can be expected to be larger for fine
grained deposits than for coarse deposits. A possible explanation for these
observations and considerations is the occurrence of van der Waals or chemical
interactions at the interface where intimate contact is made.
2.5. Final Remarks
For the adhesion strength of electrolessly deposited Ni(P) on ,..._ 96 % alumina
substrates values ranging from 10 to 30 MPa are found, in most studies close to
about 20 MPa. The etching procedure strongly influences the adhesion strength.
The adhesion strength was relatively insensitive to variation of conditions in the
two-step nucleation, while the one-step procedure was more sensitive to processing
details. Also the metallization conditions were of a minor importance for the ad
hesion strength, as well as the effect of thermal treatments. Most investigators
agree that mechanical interlocking is the dominant adhesion mechanism. However,
experiments with combined electrolessly deposited Ni(P) and Cu layers suggest
that the intimacy of interfacial contact plays an additional role. This may indicate
that van der Waals or other interfacial interactions significantly contribute to the
adhesion.
40
In most studies the adhesion strength was measured as a function of one or more
processing parameters. Generally, the results are explained in terms of interfacial
interactions, although it is well known that strength is also strongly influenced by
the size of interfacial defects which may also vary with processing conditions.
Therefore, in order to obtain information on the influence of intrinsic mechanical
and chemical interactions, both the strength and the fracture energy or fracture
toughness should be measured as a function of processing parameters. The use
of Weibull statistics for the interpretation of adhesion strength data can provide
more insight into the influence of interfacial defects upon the adhesion. Fracture
surface and interface characterization both on micrometre and on molecular scale
with modern surface analytical techniques such as TEM, XPS and static-SIMS are
urgently needed to corroborate conclusions made on the basis of mechanical
measurements.
References
l. H. Honma and S. Mizushima, J. Met. Finish. Soc. Jpn. 33, (1982), 380.
2. T. Osaka, E. Nakajima, Y. Tamiya and I. Koiwa, J. Met. Finish. Soc. Jpn.
40, (1989), 573.
3. T. Osaka, K. Naito, Y. Tamiya, and K. Sakaguchi, J. Jpn. Inst. Printed
Circuit 4, (1989), 285.
4. H. Honma and K. Kanemitsu, Plating Surface Finishing 74(9), (1987), 62.
5. J.G. Ameen, D.G. McBride and G.C. Phillips, J. Electrochem. Soc. 120,
(1973), 1518.
6. S.M. Sze in "VLSI Technology", McGraw-Hill, New York, 1988, p. 596.
7. T. Osaka, T. Asada, E. Nakayima, and I. Koiwa, J. Electrochem. Soc. 135,
(1988), 2578.
8. W.D. Bascom, P.F. Becher, J.L. Bitner and J.S. Murday in "Adhesion Mea
surement of Thin Films, Thick Films and Bulk Coatings", STP 640, K.L.
Mittal, ed., p. 63, American Society for Testing and Materials, Philadelphia,
(1978).
9. W. Riedel in "Funktionelle Chemische Vernicklung", E.G. Leuze Verlag,
Saulgau, 1989.
41
10. W.H. Safranek in "The Properties of Electrodeposited Metals and Alloys",
Elsevier, New York, 1974, Ch. 22.
ll. T. Osaka, H. Nagata, E. Nakajima, I. Koiwa and K. Utsumi, J. Electrochem.
Soc. 133, (1986), 2345.
12. E.J.M. O'Sullivan, J. Horkans, J.R. White and J.M. Roldan, IBM J. Res.
Develop. 32, ( 1988), 591.
13. N. Feldstein, S.L. Chow and M. Schlesinger, J. Electrochem. Soc. 120,
( 1973), 875.
14. R.L. Meek, J. Electrochem. Soc. 122, (1975), 1478.
15. M. Kamijo and N. Ayuzawa, Yamanashi ken, Kogyo Gijutsu Senta
Kenkyu Hokoku 1, 86, (1987). (Research report of the Yamanashi Prefec
tural Industrial Technology Center, adress: Yamanashi- ken, Kogyo Gijutsu
Senta 3-9-4 Satoyoshi, Kofu, Yamanashi- Ken 400 Japan).
16. C.H. de Minjer and P.F.J. v.d. Boom, J. Electrochem. Soc. 120, (1973), 1644.
17. T. Osaka, Y. Tamiya, K. Naito and K. Sakaguchi, J. Surf. Finish. Soc. Jpn.
40, (1989), 835.
18. S.G. Kang, B.S. Jeon and K.J. Park, Yongu Pogu- Kungnip Kongop Si
homwon 39, 413, (1989) (Research Report - National Industrial Research
Laboratory, adress: National Industrial Research Institute, 2 Choongang
Dong, Kwacheon, Kyonggi-Do, S. Korea).
19. H. Honma and Y. Kouchi, Plating Surface Finishing 77(6), 54, (1990).
20. J.T. Klomp and G. de With, accepted for publication in Mater. Manuf.
Proc. (1993).
21. H.C. Ward in "Rough Surfaces", T.R. Thomas, ed., Longman Group Limi-
ted, Harlow, U.K., 1982, Ch. 4, p. 82.
22. M. Schlesinger and J. Kisel, J. Electrochem. Soc. 136, (1989), 1658.
23. M.A. De Luca and J.F. McCormack, U.S. Patent 4,604,299, (1986).
24. T. Osaka, I. Koiwa and L.G. Svendsen, J. Electrochem. Soc. 132, (1985),
2081.
25. T. Hamaya, Y. Kumagai, N. Koshizaki and T. Kanbe, Chemistry Letters
1461, (1989).
26. S.P. Mukherjee, C.J. Sambucetti and D.P. Seraphim, Eur. Patent 0.280.918,
(1988).
42
27. W.A. Alpaugh, W.J. Amelio, V. Markovich and C.J. Sambucetti, Eur. Pat-
ent 0.156.212, (1985).
28. W.A. A1paugh and T.D. Zucconi, U.S. Patent 4,152,467, (1979).
29. A.M.T. van der Putten, J. Electrochem. Soc. (1992), in press.
30. T. Homma, K. Naito, M. Takai, T. Osaka, Y. Yamazaki and T. Namikawa,
J. Electrochem. Soc. 138, (1991), 1269.
43
44
Chapter 3
A study on changes in surface chemistry during the initial stages
of electroless Ni(P) deposition on alumina
Summary
The formation of the interface between electrolessly deposited Ni(P) and
an alumina substrate is investigated. Prior to metallization, the substrate
is cleaned, etched and nucleated with So, Ag, and Pd containing solutions.
With XRF and static-SIMS, changes in surface chemistry due to these
pretreatments are analysed. TEM plan-view micrographs visualize the
changes in surface structure during the pretreatments. The initial stages
of metallization are measured on ShN4 membrane model substrates.
Cross-section TEM micrographs are made of a thin Ni(P) film on the
alumina ceramic, showing a columnar Ni(P) structure, a thin interfacial
layer and an intimate interfacial contact. Possible consequences for ini
tiation and adhesion are discussed.
45
3.1. Introduction
The metallization of alumina ceramic surfaces with electroless Ni(P) is often used
in the electronics industry, among other things for IC packaging, pridted circuit
and sensor applications (I - 5). The use of alumina offers the advantage of a rel
atively high substrate thermal conductivity (6) compared to other insulators,
combined with a low cost price. In order to reach the required properties of the
Ni(P) layer, besides the bulk composition and properties of the Ni(P) material, the
processes that occur at the substrate surface before deposition are also· important
(7). The cleanliness, the chemical composition of the substrate surface and the
nucleation all influence the initiation and the subsequent adhesion both during and
after deposition (1, 8 - 10).
Generally, for the pretreatments three different goals can be distinguished. Firstly,
in a cleaning step adsorbed organic contaminations and particles are removed.
Secondly, by etching the substrate, the surface roughness is increased and possible
sites for mechanical interlocking are created in order to improve adhesion. Thirdly,
by the nucleation procedure the substrate surface is made catalytic for electroless
deposition.
Many aspects of nucleation on the substrate surface have been investigated. For
the nucleation procedures a one-step and a two-step process have been distin
guished (10- 12). For the one-step nucleation, samples are immersed in a SnCh
PdCh colloidal solution (10 - 16). By the two-step procedure, substrates are sensi
tized by immersion in a SnCh containing solution and activated with a PdCb sol
ution (8, 10, 17 - 19). According to Svendsen et al. (18), the one-step nucleation
procedure is not suitable for alumina substrates. This is confirmed by Honma and
Kanemitsu (1) who reported that with a two-step procedure the adhesion of Ni(P)
on alumina is 30 to 50 % stronger than with a one-step procedure, as measured
by the direct pull-ofT technique.
In the present work a two-step procedure is studied including an intermediate
immersion in an Ag-containing solution in between the Sn and Pd steps (2, 17).
Therefore this nucleation procedure is henceforth referred to as three-step
46
nucleation. This three-step procedure on glass substrates has been investigated by
de Minjer and van de Boom (17) using transmission electron microscopy (TEM),
ellipsometry and quantitative analysis with radioactive tracers as analysis tech
niques. They concluded that a more homogeneous nucleation is obtained with the
three-step method than with the two-step method.
In this work the changes in surface chemistry on alumina substrates by cleaning,
etching and nucleation are quantitatively analysed with X-ray fluorescence
spectrometry (XRF) (20). However, this technique does not measure organic
compounds, which may play a role as contaminations since all process steps are
conducted in air. Moreover, for the elements with a low atomic number ( < 23),
the detection limit for XRF measurements is above monolayer coverage. There
fore, more refined additional information is to be obtained otherwise. With static
secondary ion mass spectrometry (static-SIMS) analysis even ppm's of a
monolayer can be measured, both for ion coverages as for organic molecules, al
though static-SIMS has the disadvantage of not being a quantitative technique.
The changes in surface structure during these process steps are analysed on
nanometre scale with TEM. Since plan-view TEM micrographs cannot be made
using ceramic substrates, ShN4 membranes are considered to be the best alterna
tives and are therefore used as model substrates. It is reasonable to expect that the
relevant properties are sufficiently similar for the purposes of this study. Cross
sectional TEM micrographs are made of Ni(P) layers on the alumina ceramic after
deposition of 50 to lOO nm Ni(P).
3.2. Experimental procedures
Polycrystalline 96 % alumina substrates were used with 4 % glass phase, mainly
present at grain boundaries (Hoechst Rubalit 708). As shown by chemical analysis
these substrates contained 441 wt. ppm Na, 231 wt. ppm K, 1.23 wt. % Si,
0.55 wt.% Mg, ± 0.3 wt. % Ca and 0.2 wt. % Fe. The SbN4 model substrates were
prepared as given in the literature (21). Prior to metallization, the substrates were
successively cleaned, etched and nucleated in the aqueous solutions listed in
Table 1. After each step the samples were rinsed in demineralized water. All
47
immersion times were 2 minutes except for the metallization step, where samples
were immersed for 6 and 30 seconds. The Si3N4 model substrates were not etched
because the membranes are attacked by HF solutions. A commercially available
electroless Ni(P) metallization bath (Enlyte 512 from OMI) was used. The condi
tions under which it was operated are listed in Table l. The bath contained
NiClz, NaH2P02 and acetate and lactate complexing agents.
Table 1: Process steps in sample preparation with bath temperature and pH.
Step Function Solution T ("C) pH
l Cleaning Surfactant 1 40 6.5
2 Etching Diluted HF 20 ~I
3 Sensitization SnCll 20 < 1
4 Intermediate AgN033 I 20 ,..., 10
5 Activation PdCh2 20 < 2
6 Metallization Electroless bath 4 65 4.5
l: Amine perfluoralkylsulphonate surfactant 2: pH adjusted with HCI 3: pH adjusted with ammonia 4: Enlyte 512 from OMI
The equipment and experimental conditions for the XRF and static-SIMS analyses
are described in (20) and (22), respectively. A reflectron-type Time-of-Flight
Static-SIMS apparatus (IonTOF Munster) is used for the surface analysis of the
first monomolecular layers of the surfaces. The mass resolution of the spectra,
m/Llm at half peak bight, is that high (3000 - 5000 in the mass range from 20 to
150 amu) that peaks from the metal ions can be separated from those of the
hydrocarbon ions of the same nominal mass.
Plan-view TEM micrographs were taken on a Philips EM 300 transmission
electron microscope at an electron energy of lOO keV. Cross-section TEM micro
graphs of the metal - ceramic interface were taken using a Philips EM 400 trans
mission electron microscope at an electron energy of 120 keV. Samples were
prepared by grinding, polishing and ion milling as described in (23).
48
3.3. Measurement results
3.3.1. SEM results
Before and after etching in the HF solution SEM micrographs were made from the
alumina surfaces, see figs. lA and lB. In these micrographs two effects can be seen
from the etching treatment: the first is that the grain surfaces become less smooth
and the second that gaps appear between the grains. This is caused by the removal
of the glass second phase during etching (1 - 3).
3.3.2. XRF results
With XRF, the surface composition was quantitatively measured after each sub
sequent process step as listed in Table 1. For each step three samples were meas
ured. The analysed surface was about 20 x 30 mm2• The results are given in
Table 2. The coverages are expressed in 1015 at/cm2 which is of the order of a
monolayer of solid material. The detection limits are 0.02 for Sn, 0.1 for Ag, 0.1
for Pd, 0.1 for Cl and I for Na, respectively. The relative accuracy is estimated to
be within 10 % (20). The surface coverage of the glass phase elements Si, K, Ca
and Na in Table 2 was obtained from the difference in the absolute amounts
measured before and after etching in HF solution. The values indicated with step
1 are therefore considered to represent the original surface coverage and in the
subsequent steps these values are taken to be zero, indicated by an asterisk.
The Sn, Ag, Pd and Cl coverages measured with XRF range from 0.5 to
2 . 1015 cm-2 using the present experimental procedures. Cl was found to be present
only after the Pd step. The coverages vary by up to 50 % of the maximum values
in the Table using identical procedures and solutions, with samples prepared im
mediately after one another. A similar spread in coverages was also reported by
Meek (19) using high-energy ion scattering surface analysis.
3.3.3. Static-SIMS results
Static-SIMS was used to analyse the AhOJ samples, also before and after the
treatments listed in Table 1. Figs. 2A and 2B show, as typical examples, positive
ion static-SIMS spectra of the surface before the cleaning step (fig. 2A) and after
49
Fig. 1:
50
SEM micrograph of alumina surface before (A, top) and after (B, bot
tom) etching in HF solution.
Table 2: Coverage (1015 at.fcm2) of alumina substrates after various pretreatment steps for electroless metallization. When the coverage was below the detection limit this is indicated with -, the asterisk is explained in the text.
Step Sn Ag Pd Cl
~· K Na
1 - - - .5 0.5 -1 - - - 14 4.2 0.5 -1 - - - - 19 2.8 0.6 -
2 - - - - * * * * 2 - - - - * * * * 2 - - - - * * * * 3 0.66 - - - * * * *
tf 0.77 - - - * * * * 0.71 - - - * * * *
4 0.54 l.l - - * * * * 4 0.53 1.0 * * * * !
4 0.89 1.7 - - * * * * 5 0.67 1.6 0.5 IL2 * * * * 5 0.63 1.6 0.4 1.4 * * * * 5 0.71 1.9 0.6 1.7 * * I
activation step 5 with a PdCb containing solution (fig. 2B). An overview of the
most relevant static-SIMS results from both the positive and the negative-ion
spectra is given in Table 3.
Due to the high sensitivity of the ~tatic-SIMS technique, large numbers of peaks
are measured of which only the most intense ones are listed in Table 3. It should
be noted that the observed fragments may either originate directly from the sur
face, or be formed during the ion formation process. In the following, a correlation
of the static-SIMS data with the changes in surface composition is made for each
process step.
51
Fig. 2:
52
t i!:' Ui c Q)
.~ Q)
> .cii Qj ([
t -,. :::l
si. ;::-iii c Q)
-~ Q)
.2': a; Qj ([
0 20
0 20
10x
ea+ 118
Si+ ( SiOW
.( ·. ~~--------------~ 40 60 80 100 120 140 160 180
Mass (amu) -----..
AI+ • CxH/
Ag+
·/CH3CQ+ f A IOW
L?":· 10x
40 60 80 100 120 140 160 180
Mass (amu) -----..
Positive-ion static-SIMS spectra of the alumina surface before the
cleaning step (A, top) and after the activation (Pd) step (B, bottom).
A linear intensity scale is used.
Table 3: Overview of most relevant static-SIMS analysis results
Step Function Typical fragments Surface Composition
+ -
0 Blank CH:i C;, C.H- (x=0-4). o-, 02 Hydrocarbons +
CnH:in+l/-1 (n = 2·5) HO-, HC02 , CH,C02 glass phase +
C6H,6NO+(m/z 118), CN-, CNO-, AJO-, SiOz(H)- AhO,
CJHsN+, K', Mg+ Si03(H) , CI-, At Or
CJHt, C6,7H7,9,H,n Si(H)-, C2HO·, N02
Ca+, Na+, AI+, Si+ S03, S04, HS04 !
I Cleaning Na+, AI+, Si+, K+ AIO; (x =0-2), Si02.J, O- Glass phase + AhOJ
y OH-, Si02.3H-, Si-
( x,y 2,3; 3,3; 3,4) SO;; (x 0·4), HS04
2 Etching AI+ AIO; (x = 0-2), Aiz04H- Al20, + F
F-, o-, OH-, AIFx (x=2·4)
3 Sensitization AJ+, Sn+, SnOH+ AIO; (x 0 · 2), CI- As 2 + Sn and Cl
SnOJ(H)-, Sn04, O-, HO-
4 Intermediate AI+, Ag+, Sn+, AIO; (x = 0-2), o-, Ho- As 3 + Ag
SnOH+ Ag-, Sn03(H)-, SnO.r
5 Activation AI+, Ag+ AJO;, Ag,Ciy- (x = 1,2; y = 1,2) As 4 + Pd
Pd+, Sn+, SnOH+ o-, HO-
6 Metallization Ni+, Nit,NiOH+ O-, HO-, P02, P03, 02 Oxidized Ni(P)
NiO+, NizO+, Na+ NiO-, NiOH-, Ni02
NhOH+ HC02, CH,CO:r, NiOzH-
53
The spectra of the surface of the alumina blanks show that they were covered by
various organic compounds with aliphatic, aromatic and aliphatic alkylamine
groups (m/z = 58, 86 and 118, respectively) in (sub)monolayer quantities. Inor
ganic contaminations like sulphates were also found to be present (fig. 2A). Pos
sible sources of such contaminations are ambient air (24) and the plastic packaging
materials in which the substrates were stored and transported.
After cleaning the substrate by immersion in a solution containing an amine
perfluoralkylsulphonate surfactant (step 1), the intensity of the peaks due to the
hydrocarbon contaminations decreased by a factor roughly between 2 and 5. In
addition, the peaks due to Mg+ and Ca+ entirely disappeared. Apparently, the
cleaning step functions very well, serving its purpose even on a (sub)monolayer
scale. Some new or more intense peaks from Na+ and sulphonic acid (see
Table 3), due to the surfactant appeared but with an intensity low enough to be
of no significance.
After etching the sample in an HF solution (step 2) the signals from both Si+ and
Na+ decreased considerably in intensity. The positive-ion spectrum was now
dominated by the signals from the Alz03 substrate, implying that by etching the
glass phase was removed at least up to a submonolayer level over a large fraction
of the surface area. In addition, this measurement shows that after etching the
outermost surface of alumina did not consist of an aluminium silicate like phase
which could be present in this type of ceramic. The relatively strong signal in the
negative-ion spectrum due to F- and the small signals from AlF; (x = 2 to 4) in
dicate that the fluoride was not completely removed by rinsing in distilled water.
However, even small amount of fluoride may give rise to a relatively strong signal
due to its high ionization probability, which means that the coverage may be far
less than a monolayer.
After immersion of the sample in a SnCh containing solution (step 3), the presence
of Sn was revealed by Sn+, SnOH+ and Sn03, Sn03H- and SnO;r fragments. Cl
was also detected. After the sample was immersed in the ammoniacal AgN03
solution (step 4), peaks due to Ag+ of medium intensity were observed in the
positive-ion static-SIMS spectrum. The intensity of the peaks due to Sn+ fragments
54
relative to those of AI+ did not significantly change. This observation will be re
ferred to later.
After activation step 5 (fig. 2B), immersion in a PdCh solution, additional signals
from Pd isotopes became visible at 104 to 110 amu. Complexes of Ag+ with Cl
appeared in the spectra as illustrated by signals e.g. at m/z 177, 179 and 181 from
AgClr fragments. The relative intensities of the Sn and Ag peaks did not signif
icantly change.
After deposition of about 0.1 p.m electroless Ni(P) (step 6), no signals originating
from AI, Sn, Ag or Pd were found due to the static-SIMS information depth of
about 1 to 2 nm. The strongest peaks in these spectra were those from the native
oxide on a closed Ni(P) layer (Ni+, POr and POr).
3.3.4. TEM results
Fig. 3A shows a plan-view TEM micrograph of a Si3N4 membrane after cleaning
and sensitization (step 3). A structure of chains and islands of particles of a few
nanometre in size was observed. The background shows the structure as observed
on a blank sample. A similar structure was observed after subsequent immersion
in an ammoniacal AgN03 solution. As shown in fig. 3B, also after the subsequent
activation with a PdCh solution no significant changes were observed. The larger
particles which appear in fig. 3C (after 6 s metallization) are interpreted as the first
Ni(P) nuclei. As can be seen in this figure, these nuclei started to grow from the
clusters of primary activator particles. This micrograph was made at the earliest
possible stage of growth, since at the bottom right-hand-side the growth had not
yet started. As observed in fig. 3D, larger particles were formed from the nuclei,
covering the whole surface after 30 seconds of metallization.
In fig. 4 a cross-sectional TEM micrograph is shown of a thin Ni(P) film on an
alumina substrate. Between both phases a layer of I to 2 nm thickness with an
amorphous structure is observed. Good interfacial contact is observed on all
micrographs, no voids or interface gaps are discernible within the resolution of
these micrographs (about 0.5 nm). The structure of the material close to the
interface can also be observed. On the micrographs the diffraction lines of the
55
Fig. 3:
56
Plan-view TEM micrographs on Si3N4 substrate surfaces, after
sensitization treatment with SnCiz solution (A, top) and after activation
treatment with PdCh solution (B, bottom).
Fig. 3: Plan-view TEM micrographs on Si3N4 substrate surfaces (continued),
after initiation of Ni(P) growth (6 s) (C, top) and after 30 seconds Ni(P)
growth (D, bottom).
57
Fig. 4: Cross-sectional TEM micrograph of a thin electroless Ni(P) layer on
alumina ceramic. The arrow indicates the interface layer.
crystalline alumina grains are visible. In addition, a branching structure of Ni(P)
columns indicates the coalescence of the initially formed small primary particles
to fewer, broader columns during the growth process. The Ni(P) layer thickness
of this sample was of the order of 50 to 100 nm.
3.4. Discussion
In spite of the large number of studies devoted to the nucleation of surfaces for
electroless deposition, still no consensus has been reached on the chemical proc
esses that take place during this process (17, 25, 26, 28 - 33). This study is not
primarily aimed at providing a decisive explanation for the experimental observa
tions in the nucleation processes. Rather, the experimental results on the alumina
surface will be discussed and compared with data found by other researchers, who
used glass or polymer substrates and employed different experimental conditions.
58
In order to understand the changes in surface chemistry observed with the three
step process as reported in section 3, the two-step process will first be briefly
discussed on the basis of literature data. This process has been investigated much
more extensively than the three-step process. Essentially, the two-step process
consists of the adsorption of Sn2+ ions onto the substrate, followed by deposition
of Pd ions or atoms. The Sn2+ ions may either be adsorbed as single ions or as part
of colloidal particles. De Minjer et al. (17) proposed various types of adsorbed
single Sn2+ ions with glass surface groups, for the reason that the average Sn cov
erage is of the order of a monolayer. Pederson (26) followed this ionic adsorption
model for the interpretation of XPS spectra of sensitized surfaces. However, a real
ionic adsorption is very unlikely at a pH value as low as 2. This can be concluded
from ion adsorption data as a function of pH for various oxidic surfaces described
by Schindler and Stumm (27). For the colloidal adsorption model, however, more
evidence has been supplied in various publications.
On TEM micrographs taken after sensitization of Formvar polymer substrates a
particulate structure has been observed (9), consisting of primary particles of about
2 nm size, which are agglomerated into clumps of about 5 nm size. Similar obser
vations were made by Sard (28) on carbon films. Cohen et al. (29, 30} studied the
chemistry of Sn in the sensitizer solution and after sensitization on substrate sur
faces with Mossbauer spectroscopy and concluded that Sn is deposited onto the
surface by adsorption of colloidal particles which are already present in the
sensitizer solution. They expected that the results obtained on polyimide (Kapton)
surfaces would apply to other insulating surfaces as well (30). They also found that
these particles contained both Sn2+ and Sn4+, in the solution and after the
sensitization step. After the activation step the same amount of Sn was detected,
but in a 4+ state only. Similar results were obtained by Meek (19), who also re
ported that metallic Pd is present after activation as measured with ESCA. These
data support the assumption that reaction I describes the deposition of Pd.
[I]
It is obvious that with an increasing initial Sn4+ content, the apparent efficiency
of reaction I decreases if this is measured by coverage ratio's only. By using aged
59
Sn2+ sensitiser solutions, in which part of the Sn2+ is converted into Sn4+, the effi
ciency was found to be 25 % or less (31).
In several studies the chemical composition of the surface after the sensitization
and activation steps appeared not to be, or at least not completely to be, in
agreement with reaction I, suggesting that different or additional processes take
place (17, 18, 32). For instance, initiation of electroless metal deposition has been
observed when an aged Sn4+ sensitizer was used. This indicates that Pd can also
be adsorbed when a sensitizer is used without addition of Sn2+ (32). However, this
may be explained by the equilibrium between Sn4+ and Sn2+ (26) during aging.
For the nucleation of the substrate surface by the three-step procedure the fol
lowing reactions 11 and Ill can be proposed as a model, in analogy to the simple
reaction I for the two-step Sn-Pd nucleation. In an XPS study for the silver mirror
process, Pederson (26) found experimental evidence that reaction II adequately
describes the deposition of Ag.
[II]
[III]
The Sn coverage on alumina after the sensitization step of about 7 . 1014 cm-·2, see
section 3.2, compares well with the Sn coverage found by de Minjer et al. on glass
slides (5. 1014 cm-2) (17). After the intermediate step, the Ag coverage is two times
the Sn coverage, within the accuracy of the XRF measurements of about 10 %.
This is again in agreement with, but of course no proof for, reaction II. After the
activation step, a higher Ag to Sn ratio is found, which may be caused by partial
dissolution of Sn ions in the acidic activation solution. The dissolution of Sn from
activated surfaces is a well-known phenomenon and sometimes causes problems
due to its poisoning effect when Sn dissolves during initiation in the electroless
bath. After activation, the Pd coverage amounts to about 5 . 1014 cm-2, which is
ten times the coverage minimally required for initiation of electroless deposition
(32). This Pd coverage amounts to about 70 % of the Sn coverage. This may imply
that, if the reaction couple II + Ill describes the Pd deposition, these reactions
have an overall efficiency of 70 % or lower, depending on the amount of Sn ions
60
dissolved in the activator step. Since for reaction II an efficiency of 90 to 100 %
was found, this efficiency decrease must occur after the intermediate step. Proba
bly, this is due to oxidation of Ag, deposited in reaction 11, during the intermediate
step or during rinsing. In that case less Ag0 is available for reaction Ill.
Due to the very low solubility of silver chloride, it is probable that Ag+, formed
by reaction Ill in the PdCh solution, precipitates with Cl- on the activated
substrate surface. Hence, the fact that Ag remains on the surface after activation
should not be used as an indication that Ag does not take part in the reactions II
and III as argued by de Minjer et al. (17). The same holds for remaining Sn (17,
18), which is oxidized to the 4 + state after activation as discussed above.
The species and the intensity changes measured with static-SIMS after the various
process steps correlate well with the quantitative XRF results. In a few cases, the
composition of the fragments may give an indication of surface chemical bonding.
For instance, the SnO and SnOH containing fragments correspond to the
hydrolysed polymeric structures proposed for the sensitiser colloid particles (9).
The occurrence of Ag,Cly fragments (x 1,2; y = 1,2) may support the assump
tion that an AgCl precipitate is formed. Furthermore, it is probable that F- which
is also observed in AlFx (x = 2 to 4) fragments is bonded to AI atoms on the
substrate surface. However, due to the possibility of recombination reactions in the
static-SIMS experiment, the fragments observed should be used as an indication
rather than as a proof for the presence of these surface structures.
The TEM micrographs show that the morphology of the surface which arises
during the Sn sensitization step is maintained during the Ag and Pd activator
steps. The results from the XRF analyses indicate that after the nucleation proce
dure about 3 . 1015 Sn, Ag, Pd and Cl cm-2 are present. From TEM measurements
it is concluded that the activator material is present as small particles, rather than
as a continuous layer. Hence, it is concluded that a large part of the surface area,
that is to say, between the clusters of activator material is not covered by Sn, Ag
or Pd. This is in agreement with the observation that the growth of Ni(P) starts
on these clusters only. The major part of the substrate surface is covered by Ni(P)
through lateral growth of nuclei starting from the activator particles. A similar
61
observation has been made with TEM on various dielectric substrates by Marton
and Schlesinger (33) using a two-step nucleation. They found a linear relation be
tween particle diameter and time of deposition between 15 and 120 nm particle
size. By extrapolation of this relation to the moment of initiation, they concluded
that then the active sites are smaller than 1 nm: this is the size that we found for
the primary particles after activation. The smallest distinguishable Ni(P) particles
are about 2 nm in size. It is interesting to note that Marton and Schlesinger also
observed that growth initiated as a continuous film on Ni and Pd substrates, which
yielded a strong adhesion.
The gradual increase of the particle size is also nicely demonstrated by the TEM
cross-section micrograph shown in fig. 4. The coalescence of many small columns
to fewer thicker columns agrees very well with the plan-view images in figures 3C
and 3D. The interfacial layer with a thickness of about 2 nm cannot be explained
by the presence of the activator material only. This layer may be of importance for
the adhesion between Ni(P) and alumina and will be further investigated.
Homma et al. (10) explained the superior adhesion strength of Ni(P) on alumina
relative to that of electroless Cu by the difference in the initial metal growth
mechanism in the first 30 nm, rather than by the surface nucleation procedures.
They concluded from a plan-view TEM study on carbon films that Ni(P) makes
a more intimate interfacial contact with the substrate than Cu. It is, however, dif
ficult to obtain information on interface morphology by plan-view imaging. Nev
ertheless, our cross-section TEM images confirm that a close interfacial contact
between Ni(P) and alumina is present on nanometre scale.
3.5. Conclusions
The cleaning treatment removes organic contaminants up to the monolayer level.
It is concluded that the subsequent surface preparation steps like etching and
nucleation do not introduce any organic or other contaminations although all
steps are carried out in laboratory air.
62
A glass layer is removed from ceramic grains by etching in an HF solution, also
up to monolayer level. In this step fluoride is deposited onto the surface. Since this
is not completely removed in subsequent steps, it will be present at the interface,
bonded to alumina.
An amount equivalent to about 3 monolayers of activator material is deposited in
small particles of a few nanometre in size. An AgCl coverage of I to 2 . 1Ql5 cm-2
remains on the surface. The Sn, Ag, Pd and Cl coverages can be explained by
simple redox reactions. Though other possibilities cannot be definitively excluded,
for the present discussion we assume that nucleation is satisfactorily described by
these redox reactions.
The activator particles are clustered in islands and chains. Here metallization is
first observed. The structure of the activator material on the surface is determined
by the sensitization step. A large part of the surface area is not covered by acti
vator material and becomes covered with Ni(P) by lateral growth of initial Ni(P)
particles.
The interfacial layer which is observed with cross-section TEM, cannot completely
be explained by the presence of activator material. This layer undoubtedly plays
a crucial role in adhesion for all cases where interfacial failure is observed.
References
l. H. Honma and K. Kanemitsu, Plating and Surface Finishing 74, (1987), 62.
2. H. Honma and S. Mizushima, Kinzoko Hyomen Gijutsu 33, (1982), 380.
3. T. Osaka, E. Nakajima, Y. Tamiya and I. Koiwa, Hyomen Gijutsu 40,
(1989), 67.
4. T. Osaka, Y. Tamiya, K. Naito, and K. Sakaguchi, 40 th ISE Meeting,
Kyoto, (1985).
5. J.G. Ameen, D.G. McBride and G.C. Phillips, J. Electrochem. Soc. 120,
(1973), 1518.
63
6. W. Riedel in "Functionelle Chemische Vernicklung", Eugen G. Leuze Verlag,
(1989), Saulgau I Wi.irttemberg, Ch. 2, p. 17. I
7. T. Osaka, I. Koiwa and L.G. Svendsen, J. Electrochem. Soc. 132, (1985),
2081.
8. M. Schlesinger and J. Kisel, J. Electrochem. Soc. 136, (1989), 1658.
9. N. Feldstein, S.L. Chow and M. Schlesinger, J. Electrochem. Soc 120, (1973),
875.
10. T. Homma, K. Naito, M. Takai, T. Osaka, Y. Yamazaki and T. Namikawa,
J. Electrochem. Soc. 138, (1991), 1269.
11. J. Kim, S.H. Wen, D.Y. Jung and R.W. Johnson, IBM J. Res. Develop. 28,
( 1984), 697.
12. R.L. Cohen, R.L. Meek and K.W. West, Plating and Surface Finishing
1976-5, 52.
13. E.J.M. O'Sullivan, J. Horkans, J.R. White and J.M. Roldan, IBM J. Res.
Develop. 32, (1988), 591.
14. T. Osaka, H. Takematsu and K. Nihei, J. E1ectrochem. Soc. 127, (1980),
1021.
15. T. Osaka, H. Nagasaka and F. Goto, J. Electrochem. Soc. 127, (1980), 2343.
16. R.L. Cohen and R.L. Meek, Plating and Surface Finishing 1976-6, 47.
17. C.H. de Minjer and P.F.J. v.d. Boom, J. E1ectrochem. Soc. 120, (1973), 1644.
18. L.G. Svendsen, T. Osaka and H. Sawai, J. Electrochem. Soc. 130, (1983),
2252.
19. R.L. Meek, J. Electrochem. Soc. 122, (1975), 1478.
20. D.K.G. de Boer, J.J.M. Borstrok, A.J.G. Leenaers, H.A. van Sprang and
P.N. Brouwers, submitted to X-Ray Spectrometry (1992).
21. J.W.M. Jacobs and J.F.C.M. Verhoeven, J. Microscopy 143, (1986), 103.
22. H. van der Wel, P.N.T. van Velzen, U. Jiirgens and A. Benninghoven in
"Analysis of Microelectronic Materials and Devices", M. Grasserbauer and
H.W. Werner ed., 1991, John Wiley & Sons Ltd.,
23. L.C. Feldman en J.W. Mayer in "Fundamentals of surface and thin film
analysis", Elsevier science publishers, New York, 1986.
24. A.H.M. Sondag, M.C. Raas, and P.N.T. van Velzen, Chemical Physics Let
ters 155, (1989), 503.
25. N. Feldstein, J. Weiner and G. Schnable, J. Electrochem. Soc, (1972), 1486.
64
26. L.R. Pederson, Solar Energy Mater. 6, (1982), 221.
27. P.W. Schindler and W. Stumm in "Aquatic Surface Chemistry", W. Stumm,
ed., Wiley Interscience, New York, 1987, Ch. 4, p. 99.
28. R. Sard, J. Electrochem. Soc. 117, (1970), 864.
29. R.L. Cohen and K.W. West, J. Electrochem. Soc. 119, (1972), 433.
30. R.L. Cohen, J.F. D' Amico and K.W. West, J. Electrochem. Soc. 118, (1971),
2042.
31. N. Feldstein and J.A. Weiner, J. Electrochem. Soc. 120, (1973), 475.
32. B.K.W. Baylis, N.E. Hedgecock, M. Schlesinger and A. van Wijngaarden,
J. Electrochem. Soc. 126, (1979), 1671.
33. J.P. Marton and M. Schlesinger, J. Electrochem. Soc. 115, (1968), 16.
65
66
Chapter 4
Adhesion and interface characterization of electroless Ni(P)
layers on alumina ceramic
Summary
The adhesion mechanism of electroless Ni(P) on alumina ceramic substrates has
been investigated. The adhesion was measured by direct pull-off tests and by
90° peel tests, which provided information on adhesion strength and fracture
energy, respectively. The observed mechanical behaviour is rationalized using
the Griffith-Irwin theory. The interface chemistry has been analyzed by
static-SIMS, XPS and AES and the interface microstructure with TEM and
SEM. Two types of alumina substrates with a different roughness were used.
Ni(P) was deposited from two types of electroless Ni(P) solutions, one with
glycine and one with acetate as the complexing agent. Using static-SIMS,
glycine and acetate molecules were found to he present on the interfaces of the
corresponding samples. TEM cross-section micrographs showed a close contact
between the two phases on a nanometre scale for all sample types. In addition,
a 1 to 2 nm thick interfacial layer was observed, probably related to the
nucleation material. Fracture takes place along or through this layer.
The adhesion strength of the glycine-type Ni(P) was much higher and the frac
ture energy was lower than that of the acetate-type Ni(P), for both substrate
types. This implies that the difference in adhesion strength is not caused by
differences in interfacial chemical bonding, but rather by differences in flaw
sizes. Since high adhesion strength was measured on smooth substrates, along
with low peel strength, it is concluded that strong adhesion can be obtained
without making nse of mechanical interlocking. The intrinsic adhesion is as
cribed to van der Waals interactions. Further research should be aimed at
controlling the interfacial flaw sizes.
67
4.1. Introduction
-Aim
The metallization of alumina ceramic surfaces with electroless Ni(P) is often used
in the electronics industry, among other things for IC packaging, printed circuit
and sensor applications (I - 5). Generally, the adhesion between electroless nickel
layers and non-conducting substrates such as polymers, glass and ceramics, is
weak. Various studies have been devoted to the optimization of process parameters
with respect to the adhesion strength of Ni(P) on alumina (l - 6). The adhesion
strength is generally found to be most strongly influenced by etching conditions,
while nucleation and metallization conditions are only of secondary importance.
According to most authors, this suggests that the adhesion is determined by me
chanical interlocking interactions (7). However, for theoretical reasons adhesion
strength data are insufficient for obtaining conclusive information on microscopic
interfacial interactions as described in section 4.2. Moreover, few or no interface
and fracture surface analyses are reported in these literature references.
-Methods
In this work a different approach is followed in order to gain insight into the
backgrounds of the adhesion of both types of Ni(P) on alumina. The mechanical
characterization is supported with interface structure characterization and analysis
of the chemical composition of the interface. In order to vary the contribution of
mechanical interlocking to the adhesion, two types of substrates with different
roughnesses are used, further denoted as rough and smooth-type substrates. In
addition, two types of electroless metallization solutions are used, one with acetate
as the complexing agent and one with glycine. The corresponding deposits are de
noted by acetate and glycine Ni(P}, respectively. Information on adhesion strength,
which is influenced by extrinsic factors such as interfacial flaws (8, 9), is combined
with information on the interfacial fracture energy, which is mainly determined
by intrinsic factors such as interface chemical bonding or mechanical interlocking.
For the adhesion strength measurements the direct pull-off (DPO) test is used
(1 - 6, 10 - 13} and for the fracture energy measurement the peel test is used.
The interface structure is analysed on micrometre and nanometre scale with a
scanning electron microscope (SEM), equipped with energy dispersive analysis of
68
X-rays (EDX) and with transmission electron microscopy (TEM), respectively. For
the analysis of the interface chemical composition, static-SIMS (static secondary
ion mass spectrometry), X-ray photoelectron spectroscopy (XPS) and Auger
electron spectroscopy (AES) are used. Static-SIMS is capable of determining or
ganic structures at a submonolayer coverage. With AES the elemental composition
of a surface can be quantitatively analysed. With XPS inorganic structures and
valencies are quantitatively measured at submonolayer coverage.
-Preface
The following theoretical section deals with adhesion strength and adhesion
measurements. In the third experimental section, first the sample preparation is
described, followed by methods for the characterization of adhesion strength and
fracture energy. Subsequently, the procedures for interface structure analysis by
SEM and TEM and interface chemistry analysis by AES, XPS and by static-SIMS
are explained. In the fourth section the results of this set of analyses are reported.
The last sections deal with the discussion and conclusions. In the Appendix the
meaning of symbols and abbreviations is listed.
4.2. Theory
This section deals with theoretical backgrounds of the adhesion strength and
fracture energy measurements.
4.2.1. Adhesion strength
The adhesion strength ur is determined by, among other factors, the fracture en
ergy Gc and the critical flaw size acr and is usually described by the Griffith-Irwin
relation (14 to 16):
[1]
where K is a geometric factor and E is Young's modulus.
69
The fracture energy Gc is formed by an intrinsic fracture energy term G; and a
contribution from plastic deformation of material at the crack tip Gp~:
[2]
The intrinsic fracture energy is the energy required for example to overcome van
der Waals forces and to break chemical bonds. The order of magnitude of Gi is
0.01 to 0.1 Jjm2 for van der Waals interactions and 0.5 to 5 Jjm2 for chemical
bonds. During fracture, stresses are near to the theoretical strength at the crack tip.
This causes plastic deformation in the metal layer during fracture, represented by
Gpt· Since the stresses at the crack tip depend on the strength of the interfacial
bonds, Gpt depends on G; and therefore eq. 2 can be written as (17):
[3]
in which f1 is the energy loss factor. For purely brittle fracture, such as with ce
ramics at low temperature, plastic deformation does not play a role and f; is
slightly larger than unity. For metal layers on ceramics Gc values of the order of
lOO Jjm2 are found (18), which means that ft is 10 to 100. For polymers on rigid
substrates these values are of the order of 1000 for Gc (19) and thus 100 to 1000
for f1•
From this relation it is clear that in order to evaluate the influence of interface
chemistry on adhesion strength, the fracture energy must be measured separately.
This is done by the peel test. Conditions under which the peel test can be used for
a quantitative fracture energy measurement are considered.
4.2.2. Peel test
The peel test has often been used for measuring adhesion (20, 21), both of metal
films (22)..and polymer films (23, 24). In the 90° peel test the peel force is measured
as a function of displacement, as shown in fig. l. The peel energy GP is obtained
by the following expression:
[4]
70
Fig. I:
Load cell
Layer
Schematic presentation of the peel test. The symbols W, D and RP de
note the width of the peel strip, the layer thickness and the peel radius,
respectively.
in which FP is the peel force, AL is the peeled length, AA is the peeled area and
W is the width of the peel strip. For this measurement the following energy balance
can be proposed:
[5]
During peeling energy is consumed by fracture (Gc) and possibly by bulk plastic
deformation of the film (Gder), while energy is supplied externally by peeling (Gp)
and internally by relaxation of residual stresses in the film (Get). All energy terms
are per unit area. Note the difference between Gder and Gpt . The first term stands
for bulk plastic deformation in the metal layer, whereas the second term denotes
the plastic deformation in the microscopic crack tip zone. These two terms may
become indistinguishable when the size of the plastic zone is of the order of the
layer thickness. If no energy is lost in bulk plastic deformation of the metal layer
and if the residual strain energy in the layer is very small, then the peel energy
equals the fracture energy.
71
The residual strain energy Gel can either be caused by the deposition process as
built-in stresses or by a difference in thermal expansion between layer and
substrate. The amount of elastic strain energy U per unit volume V due to the
difference in thermal expansion is given by:
u leT IBT =
0adf. =E
0ede
2 EeT
2 [6]
in which a is the stress, E is the Young's modulus of the film Aa: is the difference
in thermal expansion coefficients and AT is the temperature difference. This can
be expressed in elastic strain energy per unit area if the volume V is equal to area
A times layer thickness D:
E (AaAT)2 D
2 [7]
In a similar manner as with eq. 6, with eq. 8 the residual strain energy Gel due to
built-in stresses can be calculated if the amount of internal stress a; is known:
[8]
4.3. Experimental Procedures
In this section the experimental procedures are described for the sample prepara
tion, adhesion measurements and interface analyses.
4.3.1. Sample preparation
For the sample preparation two types of alumina were used as substrates. The first
type was a 96 % pure alumina (Hoechst Rubalit 708) with a surface roughness
characterized by an R. value of 0.3 p,m as measured with a Tencor IX-step step
profiler with a tip radius of 2 p,m. These R. values were verified with surface pro
files made with a scanning tunneling microscope (STM), with a tip radius of ea.
50 nm. The second phase was a grain-boundary glass phase, used as a sintering
72
aid. The grain size was of the order of 5 pm as visually estimated. The second
substrate (MRC-996) was a 99.5% pure alumina with an Ra value of 0.06 pm and
a grain size of the order of 1 to 2 pm. The additive in this material was mainly
MgO. The X-ray diffraction pattern of the substrate surfaces showed no prefer
ential orientation of the alumina grains. An impression of the surface topography
is given by the SEM micrographs of the sintered surfaces of both substrate types
in fig. 11. Samples were prepared by first depositing an electroless Ni(P) layer of
about 0.3 pm thickness and subsequently electrodepositing a thicker Ni layer from
a low-stress sulphamate bath (27). For the adhesion strength test samples, a Ni
layer thickness of 2 to 3 pm was used and for the peel test samples this was about
7 .urn.
Prior to the electroless deposition, the alumina plates were first cleaned by
immersion in a fluorinated alkylsulphonate detergent solution, then etched in an
HF solution and subsequently activated by a standard Sn, Ag, Pd procedure as
described in ref. (28). For the 96 % alumina substrates, etching removed the
grain-boundary glass phase from the surface of the alumina grains and from re
gions between surface grains. For the smooth substrates no effect of etching on the
surface structure has been observed. The glycine-containing electroless
metallization solution only contained three compounds: NaH2P02, NiCh. 6 H20
and HOOC- CH2 NH2 in amounts of 10, 30 and 30 g/1 respectively. The
acetate-containing solution was based on the commercially available Enlyte 512
from OMI. Ni layer thicknesses were measured using a Fisherscope X-ray
fluorescence coating thickness meter. The adhesion strength test samples were ob
tained by breaking metallized plates into pieces of about 6 x 6 mm2• For reasons
to be explained later, for a number of these measurements the test samples were
numbered before breaking.
4.3.2. Analyses
- Adhesion measurements: Strength
The adhesion strengths were measured by the direct pull-off (DPO) test
(16, 10 - 13), as schematically depicted in fig. 2. An aluminium pull-stud (QUAD
Sebastian) was bonded with an epoxy adhesive to the metallized ceramic surface
73
Fig. 2:
Fig. 3:
74
Applied Force
Load cell
Sample Holder
Sample
Stud
Schematic presentation of the direct pull-off test.
A pull stud on a metallized rough-type sample.
Fig. 4: Optical micrograph of a cross-section of a pull-stud on a metallized
rough-type alumina sample, magnification 32 x (A, top) and magni
fication 1600 x (B, bottom).
75
by heating for I hour at 160 oc and the force at which fracture occurs at the metal
ceramic interface was measured using a testing machine (ELE 205) at a cross
head speed of 0.5 mm/min in air atmosphere. All adhesion measure~ents were
performed at room temperature (21 ± 2 °C). In fig. 3 a pull-stud bond4d on a test
sample is shown. Cross-sections of such an assembly are shown in figs. 4A and
B. The diameter of the bonded area was 2.5 mm, the height of the studs was 12.5
mm and the angle between nail head and shank of the stud was 140°. The thickness
of the adhesive layer varied between 2 and l 0 11m and no interfacial voids were
observed. At the edge of the stud an adhesive spew fillet of about 0:1 mm was
formed. For each strength measurement about 40 data were fitted using Weibull
statistics (25) with a computer program of Dortmans and de With (26). An outline
of the Weibull statistics is given in Chapter I, section 1.2.3. For estimation of Pr,
eq. lOB was used, while for the strength the nominal value given by the pull-off
force divided by the bonded area under the stud was used.
- Adhesion measurements: Fracture energy
The fracture energy was measured using the 90° peel test at a test rate of 1 mm/min
in air. In fig. l the pe~l-test set-up is schematically depicted. A I N load cell was
used and the overall measuring accuracy was < 1 %. The metal strips were cut
with a razor blade to a length of about 50 mm and a width W of 15 mm. Peel radii
were measured by means of a video camera both during (Rv) and after (R' p) peel
ing. Initially, a frictionless air bearing was used to keep the peel front exactly
below the load cell. However, in later experiments it turned out that the small de
viation from 90° (within 5°) made by peeling from a fixed substrate, did not sig
nificantly influence the peel energy value. In order to obtain comparable results
all peel test samples received the same thermal treatment as the DPO test samples.
- Interface structure: Cross-section TEM micrographs
Cross-section TEM micrographs of the metal - ceramic interface were made using
a Philips EM 400 transmission electron microscope at an electron energy of
120 keV. Samples were prepared by grinding, polishing and ion milling as de
scribed in (29). Apart from TEM interface analyses of the adhesion strength test
samples, samples with a smaller Ni layer thickness were also prepared for these
TEM analyses with about 0.1 11m Ni(P) and without electrodeposited Ni layer.
Such a thinner metal layer facilitates the TEM sample preparation.
76
- Interface chemistry: AES depth profiling
The AES spectra were obtained using a PHI 545 Scanning Auger Microscope
equipped with a cylindrical mirror analyser. The background pressure was about
I0-8 Pa. Sputtering was done using Ar+ ions. A 3 keV ion beam was rastered over
3 x 3 mm1 area, the current density being 90 pA/cm1, from which the sputtering
rate was estimated to be about 13 nm/min. The 3 keV electron beam diameter is
about 5 ,urn and this beam was not rastered. The following Auger electrons were
measured in this analysis, with the electron energy in eV (50) between brackets:
Sn MNN (430), Ag MNN (351), Pd MNN (330), Ni LMM (848), P KLL (120),
S KLL (152), AI LMM (51), 0 KLL (505), C KLL (278). The analysis depth with
this technique is 0.5 to 2 nm, mainly depending on the Auger electron energy (29).
The detection limit is ea. 0.5 atom %. For the AES measurements special samples
with a thinner Ni layer were also prepared in order to minimize the roughening
effect during sputtering. The same layer deposition procedure was followed as for
the special TEM samples.
- Interface chemistry: XPS analysis
Glycine-type Ni(P) layers were peeled from the smooth-type alumina substrates in
a glovebox filled with purified N2. The H20 and 0 2 contents in this atmosphere
were < 0.1 and 2.5 ppm, respectively, though it should be noted that the concen
trations of these contaminants may be considerably higher in the vicinity of the
rubber gloves. From the glovebox the samples are transferred to the vacuum of the
XPS apparatus in a vacuum-tight container.
The XPS measurements were done on a PHI 5400 apparatus equipped with a
hemispherical analyser, using Mg-Ka radiation (1253.6 eV) and an emission volt
age of 13.5 keV. The background pressure was lower than I0-9 Pa. The analyser
was positioned at an angle of 45° relative to the substrate surface. A depth profile
was made by alternatingly measuring and sputtering with 3 kV Ar+ ions with a
rate of 0. 7 nm/min. By rastering the sputter beam, a crater of 7 x 7 mm2 diameter
is formed. The information depth is 0.5 to 2 nm, which corresponds to 2 to 10
atomic layers. The spot size is ea. 2 mm2• Survey spectra, multiscan detail meas
urements and depth profiles were made, the latter only for a Ni fracture surface.
For the depth profiles, the intensities of the following peaks were measured with
77
binding energies in eV between brackets: Ni2p3 (855), P2p (130), Ols (532) and
Cls (285). The exact peak positions (Table 5) are measured using curve fitting. The
relative concentrations (Table 4) are calculated from the measured peak areas, as
suming a homogeneous surface composition, both in depth and laterally. For this
calculation Perkin-Elmer software is used (ESCA series model 8503A version V4.0
Rev. B 19-02-'91). The dependence of information depth upon kinetic energy of
the measured photoelectrons is taken into account in the sensitivity factors in this
software.
• Interface chemistry: Static-SIM...'i analysis
The chemical composition of the outermost mono layers of the nickel and alumina
fracture surfaces was analysed with static-SIMS. After fracture, the nickel and
alumina surfaces were introduced into the vacuum of the apparatus as quickly as
possible, that is within a few minutes. During the measurements, ions were gener
ated from these materials by bombardment of the surface with a primary beam
of 10 keV Ar+ ions of low ion dose, 1012 ions/cm2 • The spot diameter of the pri
mary ion beam was approximately 50 Jim. The secondary ions were accelerated to
2 keV and mass-separated in a Reflectron-type time-of-flight mass analyser de
scribed elsewhere (30). Under these conditions the SIMS instrument operates
within the static limit, i.e. the probability that an ion will hit a previously bom
barded area is negligible. The analysis depth of this technique is of the order of a
few monolayers (about l nm), its sensitivity is in the range of ppm of a monolayer.
More details of the equipment and measuring conditions are given in (30). The
ratio of the integrated signal intensity of a peak characteristic of a surface com
pound and the signal intensity of a peak characteristic for the substrate (e.g. Ni+,
AI+, AlO-, Ni02H-), provides a relative measure of the surface coverage. These
substrate signals can be used as reference intensities for the surface coverage be
cause they originate from the outer 0 to 5 monolayers of the surface, whereas the
organic molecules are in the first monolayer. A linear relationship between the
relative static-SIMS intensities and the absolute coverage has been established by
van der Wel et al. (51). The mass resolution, m/Am, of the spectra is that high
(3000 - 5000 in the mass range 20 to 150 amu) that peaks from the metal ions can
be separated from those of the hydrocarbon ions of the same nominal mass.
78
4.4. Results
In this section the results of the adhesion measurements (mechanical properties),
of the interface structure analysis (SEM, TEM) and of the interface chemical
analysis (XPS, AES, static-SIMS) are reported.
4.4.1. Mechanical properties
- Adhesion strength
The results of the DPO adhesion strength measurements are given below in
Table 1. In fig. SA to D the Weibull plots of these measurement data are shown.
The Weibull modulus m and the Weibull normalization constant o-0 (at which Pr
is 63 %) are also listed in Table I.
Table I: Adhesion strength data measured by the DPO test. (Jr denotes the mean adhesion strength, sn-1 the sample standard deviation, N the number of test samples, m the Weibull modulus and (Jo the Weibull normalization constant
Ah03 Ni(P) (Jr (MPa) Sn-l (MPa) N m (Jo (MPa)
Rough Glycine 20 5.3 45 4.2 22.3 Rough Glycine 17 5.2 46 3.5 18.7 Rough Acetate 10 5.0 45 2.3 11.6
Rough Acetate 13 6.1 45 2.1 15.0 Smooth Glycine 47 12 47 3.9 52.3 Smooth Glycine 31 l3 44 2.2 35.5 Smooth Acetate 3.7 4.9 33 0.9 3.4 Smooth Acetate 6.4 7.9 35 0.7 5.3
The strongest adhesion is measured with the glycine-type Ni(P) on the smooth
substrates, whereas the weakest adhesion is found with the acetate-type Ni(P) on
the same substrates. It is remarkable that the adhesion of the glycine-type Ni(P)
on the smooth substrates is even stronger than on the rough substrates. The dif
ferences in adhesion strength are significant and reproducible. The Weibull moduli
from the samples with glycine Ni(P) are higher than those from the corresponding
acetate-type samples. These observations will be discussed in sections 4.5.4 and
4.5.5.
79
I
I
99
90
50
30
99
90
15.46 19.37 24.27 3Q.42
crt (MPa)---
~ :!- 70
50
30
10
5
1 L---------~--------~--------~--------2~7.22 2.07 3.94 7.51 14.30
crt (MPa)-
Fig. 5: Weibull plots of adhesion strength measurements.
A (top): Sample with rough-type alumina and glycine-type Ni(P).
B (bottom): Sample with rough-type alumina and acetate-type Ni(P).
80
I
t l cL
Fig. 5:
99
90
50
30
10
5
1 ~--------._--------~--------~------~ 17.70 25.01 35.32 49.88 70.44
99
90
70
50
30
10
5
at (MPa)-
X
~~ X X
,;:
7.30 22.93
at(MPa)-
Weibull plots of adhesion strength measurements (continued).
C (top): Sample with smooth-type alumina and glycine-type Ni(P).
D (bottom): Sample with smooth-type alumina and acetate-type Ni(P).
81
In order to identify the original position of the samples on the plates, the test
samples on the metallized plates are numbered before breaking. No systematics
were detected in the distribution of strong and weak samples over the plates.
Moreover, the adhesion strength was found to be independent of the layer thick
ness, within the investigated range of 0.5 to 8 J.lm.
- Fracture energy
In figs. 6A and B two typical peel curves are shown for the acetate and the
glycine-type samples, respectively. The two layer types show a different exper
imental behaviour in the peel test. Despite the constant displacement rate, the
acetate-type Ni(P) always peels in small, rapidly repeating steps smaller than
0.1 mm. This results in the broad line in the peel plot due to a compressed
sawtooth structure of the force-time plot. The glycine-type Ni(P) peels off more
continuously and the variation in peel force is mainly caused by buckling of the
edges of the metal strip. The physical cause of the difference in peel behaviour is
unclear. In Table 2 the results of the peel measurements are listed. In fig. 7 the
peel energy values are plotted against layer thickness with values ranging from 2
to 9 J.lm of the glycine-type Ni(P) on the smooth-type alumina substrates. The
points are mean values of three to four measurements, except for the two points
at the middle, which are both from one measurement. For standard deviations, see
Table 2.
As described in the experimental section, the adhesion measurements were done in
ambient atmosphere. When, however, dry nitrogen was passed over the peel front
of the test sample, the peel energy immediately increased by about 10 20 % for
all sample types. When the nitrogen flow is stopped, the peel force immediately
drops to the original level. A similar decrease is observed when air, saturated with
water is passed over the peel front. These results show that the peel energy de
pends on the humidity of the ambient atmosphere. When peeling is stopped, a re
laxation effect is observed. When, during this relaxation in normal air, humid air
is passed over the sample, the peel force drops to the corresponding level. After
switching back to normal air a small increase in peel load is observed. This latter
effect can only be explained by partial closure of the crack at the peel front since
this whole process occurs at zero peel rate.
82
Fig. 6:
t N 7.5 E 2 >-CJ)
Q; c CD
Qi 5 CD Q.
2.5
2 4 6 8 10 12 14
Displacement (mm) -
t N
30 E 2 >-~ CD c CD
Qi CD 20 Q.
10
0 2 4 6 8 10 12 14 16 18
Displacement (mm) -
Peel curves of (A, top) acetate-type Ni(P)-Ni layer from smooth-type
alumina and (B, bottom) glycine-type Ni(P)-Ni layer from rough-type
alumina.
83
Table 2: Peel test results in which Gp is the peel energy per unit area, Sn-t is the sample standard deviation and N is the number of test samples.
Ah03 Ni(P) GP (J{m2) Sn-1 (J /m2) N
Rough Glycine 24.3 1.63 8
Rough Acetate 40.9 4.39 8
Smooth Glycine 6.09 0.62 8
Smooth Acetate 8.45 0.25 8
I 10
8
"' E + =~--+--r-:1;;
...__ :2 6 =F >-~ Q) c: 4 Q)
Qi Q) a..
2
0 0 2 4 6 8 10
Layer thickness (!lm) -
Fig. 7: Peel energy versus Ni(P) I Ni layer thickness for samples with glycine
type Ni(P) and smooth-type substrates.
In Table 3 the peel radius of the metal film Rv (see also fig. 1) is listed for samples
with various layer thicknesses. This radius gives information on the relative
amount of plastic deformation of the film during peeling. In addition, the radius
of the metal film after peeling and at zero load, Rv', is listed in Table 3. Magnified
images of peel tests were recorded on video tape. The radii were measured by fit
ting the images with circle segments on the monitor screen.
In fig. 8 the load versus time is plotted by the recorder of the peel test equipment
for a sample with acetate Ni(P) and a rough substrate. The peaks represent
loading and unloading cycles with increasing maximum load, measured at a load-
84
Table 3: Metal film radii during peeling (Rp) and after peeling (Rp') for samples with various layer thicknesses D.
AtzO, Ni(P) D (Jl.m) RP (mm) Rp' (mm) Gr (J/m2)
Rough Glycine 11.7 0.9 3.4 23 Rough ~cine 12.5 0.8 3.5 25 Rough etate 0.5 1.5 42
Rough Acetate 0.4 1.2 39 Smooth Glycine S! ') I 0.7 5 6.2 Smooth Glycine 7.4 0.7 4.5 6.0 Smooth Acetate 6.9 0.6 4.5 8.3 Smooth Acteate 9.4 0.5 4.5 8.6
ing and unloading rate of l mm/min. This is done up to the load at which peeling
starts (peak 15). For reasons of clarity only the peaks corresponding to the higher
loads (peaks 10 to 15) are depicted in fig. 8. The surface area under the left-hand
half of the peak is linearly proportional to the amount of energy required for
bending the layer. The right-hand half represents the amount of energy released
elastically from the system upon unloading. If both areas are equal then the system
behaves perfectly elastically.
Fig. 8:
t 15
40 14
E 13 z
'0 30 "' 12 .2 -.;
"' 0.. 20 10
9
Time{min)-
Stress - strain peaks recorded in loading - unloading cycles with in
creasing top load. Peeling starts at peak no. 15.
85
In fig. 9 the difference between both areas (a measure of plastic deformation en
ergy) divided by the loading area, A, is plotted versus maximum load. Within a
range of 5 to 10 % the loading energy equals the unloading energy, up to a peel
energy of about 30 Jjm2• The relative accuracy is smaller at the lower Joads due
to error in the surface area measurements. Only the last point in the plot, corre
sponding to peak 13, shows a significant deviation, which is ascribed to initiation
of peeling at a small part of the peel front. Similar effects are observed at about
10% below the top load of peaks 14 and 15 in fig. 8. Therefore these latter peaks
are not used for the plot in fig. 9. From these measurements it can be concluded
that bulk plastic deformation of the metal film during peeling does not play an
important role up to peel energies of at least 30 Jjm2• This is discussed further in
section 4.5.2.
t 16 13 X
t 12
<1 1 8 X
2 X 10 3 X 4 x4 11
X 78 X xx 9 12
0 X
-4 6 X
5 -8 X
4 8 12 16 20 24 28 32 36
Gp(J/m2)-
Fig. 9: Relation between elastic and plastic deformation versus load. A is the
difference between the left-hand half of the peak areas of fig. 8, minus
the right-hand half, divided by the left-hand area.
4.4.2. Interface structure
- Cross-section TEM
The cross-section TEM micrographs of the interfaces between both types of Ni(P)
and the 96 % and 99.5 % alumina substrates are very similar. As a typical exam
ple, the micrograph of a sample with the rough-type substrate and acetate-type
Ni(P) is shown in fig. lOA. Between both phases a layer of 1 to 2 nm thickness
86
with an amorphous structure is observed. Good interfacial contact is observed on
all micrographs, no voids or interface gaps are observed within the resolution of
about 0.5 nm. The structure of the material close to the interface can also be ob
served. On the micrographs the diffraction lines of the crystalline alumina grains
are visible. In addition, a branching structure of Ni(P) columns indicates the
coalescence of the initially formed small primary particles to fewer, broader col
umns during the growth process. The micrographs show that the Ni(P) layer
thickness of these samples is of the order of 50 to 100 nm.
Fig. lOB shows a TEM micrograph for the same sample type but for a sample that
was used for the strength measurements. The same interface layer is however ob
served. The structure of the Ni(P) material is quite different. Small crystalline
particles are formed in the amorphous Ni(P) layer. The columnar structure has
almost completely disappeared. This is probably due to the fact that this sample
has been heated for 1 hour at 160 oc for bonding the pull stud, as for all other
DPO test samples. However, according to Riedel (31), crystallization of the
amorphous Ni(P) deposit starts at about 260 oc, as analysed by X-ray diffraction.
This apparent discrepancy may be explained by the fact that with TEM smaller
crystals and thus an earlier stage of crystallization can be detected than with X-ray
diffraction.
- SEM fracture surface analysis
After the strength measurements for a number of samples the fracture surfaces are
analysed by SEM and EDX in order to obtain information on the crack initiation
and macroscopic (pm scale) flaws. No significant differences were observed be
tween the fracture surfaces of the samples with acetate-type Ni(P) and with
glycine-type Ni(P). Although some irregularities are observed, no clear indications
of the presence of interfacial flaws are obtained for any of the samples. The metal
side forms an exact replica of the substrate, and no fracture patterns can be dis
tinguished. Samples with a strong adhesion are only fractured in the area under
the stud, whereas for samples showing a weak adhesion large areas of metal film
are peeled off around the pull stud. On the rough-type substrates flat areas of
maximum 15 pm size are observed, where no mechanical interlocking is possible.
87
Fig. 10: Cross-section TEM micrographs of Ni(P)-alumina interfaces for a
sample with rough-type alumina and acetate-type Ni(P) (A, top) and for
the same sample type, after heating for I hour at 160 oc (B, bottom).
88
Fig. 11: SEM fractographs of samples with acetate-type Ni(P).
A (top) shows the alumina side of sample with rough-type alumina.
B (bottom) shows the alumina side of sample with smooth-type
alumina.
89
Fig. 11: SEM fractographs of samples with acetate-type Ni(P) (continued).
C (top) shows the Ni(P) side of sample with rough-type alumina.
D (bottom) shows the Ni(P) side of sample with smooth-type alumina.
90
In fig. ll A D the metal and ceramic fracture surfaces are shown of samples with
acetate-type Ni(P) and with both rough and smooth-type substrates.
With EDX no Ni or P is detected on the ceramic fracture surface of the smooth
type substrates, for both the acetate-type Ni(P) and the glycine-type Ni(P). On the
rough-type substrates small amounts of Ni are detected with EDX at the grain
boundaries. No P is detected on these substrates. On the metal side of the fracture
surfaces some AI, originating from a few detached grains, is observed on layers
from rough substrates but no Al is detected by EDX for the layers which are re
moved from smooth substrates.
4.4.3. Chemical interface analyses
- Auger Electron Spectroscopic depth profiling
The depth profiles obtained with AES from Ni(P) layers on the rough-type
substrates are very different from those obtained from metal layers on smooth-type
substrates, see the schematical representation in fig. 12. In the depth profiles made
from samples with rough substrates a gradual decrease in the intensity of signals
from the metal layer is observed along with a gradual increase in intensity of the
oxygen signal from the substrate. Due to this poor depth resolution, no signals
could be measured of elements of which it is known that they are present only at
the interface (e.g. Sn, Ag, Pd), in monolayer amounts.
From the samples with the smooth substrates the transition from layer to substrate
can be distinguished better in the depth profiles. The elements Sn, Ag and Pd
which are used in the nucleation procedure are detected in the region where the
intensity of the Ni signal decreases and that of the 0 (from the substrate oxide}
signal increases. In the same range and of a similar intensity a signal from C ap
pears as the interface is reach~d and disappears when the interface is passed and
the substrate is measured. In the Ni(P) layer no C or 0 are detected. However, the
signals of these interface species are only slightly stronger than the noise. The first
reason for this is again the surface roughness and variation in layer thickness. The
second reason is the relatively large noise due the short measuring time of 5 min
utes, used in these experiments. This measuring time is kept short in order to
avoid interference by carbon contamination from the vacuum equipment. In a
91
spectrum that was recorded in 5 minutes halfway through the metal layer, no
carbon signal is observed, which means that contamination did not influence this
measurement. As far as possible under these conditions, no differences are ob
served regarding the interfacial carbon between the Ni(P) deposited from the
acetate-containing solution and the glycine-containing solution.
t I Rough substrate
Ni 11 Smooth substrate
p
I I 0 I
c' 0 50 100 150
Sputter depth (nm) ___.
Fig. 12: Schematic representation of Auger depth profiles for rough- and
smooth-type alumina substrates.
- XPS fracture surface analysis
The XPS survey spectra, recorded from both fracture surfaces of a sample with
glycine-type Ni(P) and a smooth-type substrate are shown in fig. 13 A and B, re
spectively. The peaks indicated by "ghost" are due to instrumental effects. In Table
4 the atom concentrations, calculated from the peak areas in the multiscan detail
measurements (32), are listed. Relative accuracies of the values listed in the Table
92
are estimated to be within 10 %, except for the values which are close to the de
tection limit, which is ea. 0.1 % for the elements reported here.
Table 4: Atom concentrations on the Ni(P) and the Ab03 fracture surfaces of a sample with glycine-type Ni(P) on a smooth-type substrate.
Surface C ls 0 Is Ni 2p
Al20l 13 52 4 0.
Ah0r2Dnm 4 59 1.3
27 23 40 0.4 9 0.5 )-20nm 8 4 79 0.1 <0.1 7 <0.1
33 19 41 7
"-20nm": after 20 nm sputtering. "box": after sputtering a subsequent transfer and stay in the glovebox for 30 minutes. "!": An additional amount of about I % N is detected at the Ni(P) surface. "-": not measured.
In Table 5 the assignment is given of exact peak positions to chemical environment
(molecules, ions or compounds) of the same sample as in Table 4. Reference data
are used from (33). These XPS measurements show that 80 % of the Ni in the
outer 2 nm (twice the information depth for the Ni 2p3 signal at 855 eV binding
energy) of the Ni(P) fracture surface is metallic or intermetallic Ni. The other 20
% is oxidized and consists of phosphate, hydroxide or oxide. As a first approxi
mation, this would correspond to an oxidized layer with an average thickness of
0.6 nm, or two atom layers. For the calculation of the relative amounts listed here,
a homogeneous distribution of the various species over the analysis depth is as
sumed. However, the signal intensity of outer surface atoms is higher than those
at about 2 nm below the surface. Since it is reasonable to assume that the oxide
is present in the outermost surface of the Ni(P) layer, it is probably even less than
corresponding to a layer with an average thickness of 0.6 nm, or in other words,
less than 2 monolayers. The depth profiles, which are not shown here, reveal that
most of the carbon and oxide are removed after 1.4 nm sputtering. After exposing
the sputtered surface to the same procedure as used during the sample preparation
(transfer in vessel and stay in glovebox), the same amounts of carbon and oxygen
are found again on this surface, see Table 4. This means that both the carbon and
the oxygen that are detected on the fresh fracture surface may be completely or
partly due to handling.
93
I 10
9
8 !:!:! 7 w z 6 .;::- 5 "(j) c 4 Q)
E 3
2
1
0 1000 800
I 10
9
8
!:!:! 7 w z 6 .;::- 5 "(j) c 4 Q)
E 3
2
1
0 1000 800
600
600
(/)
<.?
400
(/) N q.
200
0. N (L a. ~
<f> "' c
~ ~ ~ N 9
0
---Binding energy (eV)
400 200 0
Binding energy (eV)
Fig. 13: XPS survey spectra of Ni(P) fracture surface (A, top) and of Ah03
fracture surface (B, bottom).
94
Table 5: Assignment of exact XPS peak positions to chemical environment of the same sample as measured in Table 4.
Element Position (eV) Rei. Amount(%) Environment
Ni(P) surface
c 284.8 80 -C-H
286.4 10 -C-O
288.5 lO -0-C 0
0 532.8 25
531.1 75 P04, Ni(OHh or Ni20J
Sn 486.6 100 Sn02 or SnO p 132.3 25 -P5+ (-PO.)
129.5 75 -P- (NiP)
Ni 852.2 70 Ni0 metallic Ni
853.2 10 Ni1+ (NiP)
855.2 8 Ni2+ (Ni(OH)l)
857.0 12 NP+ (NiP04)
Al20> surface
c 284.8 85 -C-H
286.5 15 -C-O
0 533.2 7 -531.0 93 Ah01
Sn 487.6 lOO Sn02 or SnO p 134.2 40 -
132.5 60 P5+, (P04)
Ni 851.7 15 Ni0 (metallic Ni)
852.6 10 Ni1+ (NiP)
I 854.9 60 NP+ (Ni(OH2))
857.7 15 Ni3+ (NiP04}
AI 76.8 100 A!zO,
The amounts of carbon remaining on the Ni(P) and alumina fracture surfaces after
20 nm sputtering (8 % and 4 % respectively) are rather high. The alumina ceramic
does certainly not contain carbon in these amounts and the previously described
AES measurements showed that the Ni(P) layer does not contain carbon either.
A possible explanation can be found in the "shadow effect11, frequently encount
ered on rough surfaces. The sputter beam is positioned at an angle of 54" to the
surface and the analysis beam is at an angle of 40° to the sputter beam. Therefore,
95
both beams do not apply to the same projections on a rough surface. Other phe
nomena which can increase the apparent width of a rough interface is redeposition
of sputtered material and local variations of the sputter rate owing to the varying
angle of the rough surface with the sputter beam.
The amount of nitrogen, detected on the Ni(P) sample surface is very small, of the
order of 1 %. This corresponds to 0.1 monolayer at most (order 10 14 atoms
Nfcm 2). However, as part of an organic molecule, this small nitrogen coverage may
be due to a close packing of glycine molecules on the surface, within the large
uncertainty margin. The 0-C = 0 coverage measured on this surface is in agree
ment with this assignment. Both species might, however, also originate from the
surface contamination. Sn from the surface nucleation is detected both on the
metal and the ceramic side of the interface. Pd is not detected on either side with
XPS, indicating that the amount present at the interface must be less than about
0.1 %. Fluorine, probably originating from the HF etching step, is present on
ceramic, and at most amounts to 5 % of a monolayer. It is not probable that the
Ni and P which are detected on the ceramic surface are present as macroscopic
particles. Firstly, after sputtering of only 20 nm, the signal intensity of both species
decreased more than threefold. Secondly, the smooth substrates used for this ex
periment do not give rise to remaining Ni(P) particles after peeling, as confirmed
by the SEM I EDX analyses. Thirdly, the Ni ions are predominantly in the + 2
and + 3 oxidized state, in contrast to the solid Ni(P) material surface. It is there
fore possible that this material originates from small amounts of metallization
solution which remain at the interface during metallization due to adsorption or
inclusion.
- Static-SIMS fracture surface analysis
In Table 6 a listing is given of the most important ions in the static-SIMS spectra
in decreasing order of signal intensity. The most intense signals originate from the
substrate material on the Ab03 side (AI+ and AIO- and from the metal layer on
the Ni(P) side (Ni+ and Ni02H- ). This probably means that no large amounts of
contaminations (less than several monolayers) are present on any of the samples.
The positive and the negative ion static-SIMS spectra of the blank surfaces of both
96
types after cleaning and etching of the substrates (see section 4.3.1) are almost
identical.
Table 6: Listing of main ions observed in the static-SIMS experiment, in order of decreasing signal intensity. C,Hy fragments originate from aliphatic hydrocarbons with (x, y) = (I, 3), (2, 3), (2, 5), {3, 3), (3, 5), (3, 7), etc.
i Rough-type substrate - Acetate-type Ni(P)
! AlzOJ + : AI, Ni, Na, C,Hy, acetate, (K, Si, Mg) - : 0, HO, H, F, HC02, Si02, AIO, POz , CzH, P01, Cl, CH1COz (acetate}
Ni + : Ni, Na, AI, C,Hy, CH3CO (acetate) - : 0, OH, CH3C02 (acetate), HC02, PO", P03, Ni02H, N02, N03, Cl
i Smooth-type substrate - Acetate-type Ni(P) . Ab01 ! + : Na, AI, Ni, C,Hy, CH3CO (acetate), SiOH, Si . : 0, OH, H, F, P02 , PO), CH1COz, AIO, HC02 , Cl, Si02, S02, S01, NiO,H
Ni + : Ni, Na, C,Hy, 43 CH1CO, 43 C,Hr - : CH1COz (acetate), P01, P02, HCOz, 46, CzH, Oz, NiOzH, Cl, SOz, S01
Rough-type substrate - Glycine-type Ni(P)
Ab01 + :AI, Na, SiOH, CH2NH2 (glycine), Ni, CH1CO (acetate), C,Hy - : 0, OH, H, F, P02, P01, 02, HCOz, Si02, AIO, Cl, AI02, S02 , S01
Ni + : Ni, CH2NH2 (glycine), Na, C.Hy, CH1CO, C,Hr - : 0, OH, H, Cl, POz, PO), HCOz, N02 , glycine, CH1C02 (acetate), NiOH
Smooth-type substrate - Glycine type Ni(P)
Ah01 + : Na, AI, CHzNHz (glycine), C,Hy, CH1CO (acetate), Ni . : 0, OH, H, F, POz, PO), Cl, 02, AIO, HC02, SOz, SO), AI02, CH1COz
(acetate), C2H, S04, HS04
Ni + : Ni, Na, CH2NH2 (glycine), C,Hy, CH1CO , C,Hy, CHzN, C,Hy , CH1 - : 0, OH, H, Cl, POz, P01, HCOz, glycine, NOz, CH1COz (acetate)
Blank Ah01 after cleaning and etching
Rough-type substrate + : Al, C.Hy, Si, SiOH . H, 0, OH, F, C, (x 1,2), C,H (x 1 ,2), AIO, (x 0-2), SiO,H (x = 2,3)
Smooth-type substrate + : Al, Mg, C.Hy, Si, SiOH - H, 0, OH, F, C. (x = 1,2), C,H (x 1,2), AIO, (x = 0-2), SiO,H (x 2,3),
Ah04H, AhOsH Ah06H2
97
For all metallized samples on the Ab03 side, ions are measured which originate
from the Ni(P) layer, such as Ni+, Ni02H- , P02 and P03 . For the samples with
the rough substrates, these fragments may originate from small pieces of Ni(P)
which remain on the substrate after delamination as observed with SEM. How
ever, for the smooth substrates this explanation cannot be valid, since no Ni(P)
pieces remain on these substrates. This will be discussed further in section 4.5.3,
together with the XPS data. On the Ni(P) side of the interface no significant peaks
of fragments characteristic of the smooth-type alumina substrate are found, such
as AI+ and AIO-. The P02 and PO"r that are measured in the negative ion spectra
do not necessarily indicate that the Ni(P) is oxidized at the interface. It is to be
expected that immediately after delamination a natural oxide layer is formed on
the Ni(P) foils before they are introduced in the vacuum equipment.
Relatively strong peaks of Na+ are often found on both sides of the interface of
metallized samples. This means that Na+ from the metallization solution remains
at the interface, since it is not found on the blank samples after cleaning and
etching. In the negative ion spectra of the ceramic surfaces F- is one of the major
peaks. It is also found on the blank alumina surfaces, which are cleaned and etched
in an HF solution. The F ions are therefore assumed to originate from this step.
However, due to the high ionization probability of Na and F it cannot be con
cluded from the relatively high intensities that Na and F are major interface con
stituents.
The activator elements Sn, Ag and Pd, are detected in small amounts (not listed
in Table 6, see fig. 15) on the Ni(P) side of the interfaces. The Ag signal is stronger
than the other ones. On the Ah03 side only a weak signal of Sn is observed. It is
therefore concluded that most of the activator material remains on the Ni side
when fracture takes place.
Fragment ions originating from acetate and glycine are found in the spectra with
considerable intensity, on both the Ni(P) and the alumina fracture surfaces. The
relative intensities of fragments of these compounds are listed in Table 7 to 10.
98
Table 7: Static-SIMS intensities of the negative ion spectra (relative to AJO-, x 100 %) of interface compounds on the Ah03 fracture surfaces of the test samples and blank Ah03 surfaces.
AbOJ Ni(P) CH3C02 (59) H2NCHzC02 (74)
Rough Glycine 32.1 9.6
Rough Acetate 76.8 -Rough Blank 4.3
Smooth Glycine 37.3 17.7
Smooth Acetate 192 -Smooth Blank 8.7 -
Table 8: Static-SIMS intensities of the negative ion spectra (relative to NiOzH-, x 100 %) of interface compounds on the Ni(P) fracture surfaces of the test samples.
AbOJ Ni(P) CH3C02 (59) H2NCH2C02 (74)
Rough Glycine 145 193
Rough Acetate 529 . Smooth I Acetate
210 261
Smooth 679 .
Table 9: Static-SIMS intensities of the positive ion spectra (relative to AI+, x 100 %) of interface compounds on the Ah03 fracture surfaces of the test samples and blank AbO, surfaces.
Ab03 Ni(P) CH2NHt (30) CH3CO+ (43) cm (15)
~' 6.7 2.9 1.4
te 0.1 2.5 2.1
h 0.8 1.6 1.0
Smooth Glycine 15.1 6.9 3.0
Smooth Acetate - 4.5 3.1
Smooth Blank 1.1 3.6 1.6
Table 10: Static-SIMS intensities of the positive ion spectra (relative to Ni+, x 100 %) of interface compounds on the Ni(P) fracture surfaces of the test samples.
AbOJ Ni(P) CH2NH2 (30) CH3CQ+ (43) CH:! (15)
Rough Glycine 638.0 4.0 2.8
Rough Acetate 1.0 1.8 1.6
Smooth Glycine 43.3 5.2 3.8
Smooth Acetate 0.3 9.7 7.1
99
Due to higher peaks from contaminations in the positive ion spectra (see Table 6),
the presence of acetate and glycine can be analysed more accurately from the
negative ion spectra. Peaks from glycine at the mass/charge (m/z) ratio 74
(HzNCHzCOO-) are measured only on samples that are prepared from the
glycine-containing metallization solution, on both the Ab03 and the Ni(lP) side of
the interface, see Tables 7 and 8 respectively. Both on the Ni(P) side and on the
Alz03 side most acetate (m/z 59, CH3COO-) is found for the samples that are
prepared from the acetate-containing electroless solution. On the alumina side an
increase of about 20 times is measured relative to the blank rough and smooth
alumina substrates. However, on the glycine samples an increase in acetate cover
age is also measured, relative to the blank alumina surfaces. This can be explained
by the affinity of acetic acid present in the laboratory ambient, for the basic amino
end groups of the glycine-covered samples.
Also in the positive ion spectra, glycine (m/z 30, CH2 = NH2 ) is mainly found on
the fracture surfaces of samples prepared with the glycine-containing electroless
solution, both on the Ni(P) side and on the Ab03 side. In fact, the intensity of
mfz 30+ is negligible on the other metallized and blank samples. The interpretation
of the acetate coverage of the samples is hampered by the presence of organic
contaminations, which may give rise to the same fragments.
This influence of contaminations is confirmed in a separate experiment for the
glycine-type Ni(P) - smooth-type alumina fracture surface. In this experiment the
sample was peeled in vacuum in the mass spectrometer. On the Ni(P) and the
Ah03 sides no acetate, formiate and hydrocarbon fragments could be found.
However, after placing these fracture surfaces for a few minutes in air, static-SIMS
analyses show the same amounts of acetate, formiate and hydrocarbons as those
observed for the surfaces discussed in Table 6 - 10. In fig. 14 static-SIMS spectra
are shown from the glycine-type fracture surface before and after exposure to air.
100
t + 58Ni+ t 58Ni+ • CxHy CH2=NH2
::; 30 ::i .:i .:i ~ >-
""' (/) "' c: c: .l!l ~ .!: Na+ (I) (I) + > 23 .2: CH~NH2 -~ (ij 30 (j) (j) a:
P0Nt a:
soNi+ NiOH;
Na+ :I:
7"''~~ +Z 23
Ill :I:
I/o . ~ / Ni2o+ lll . il .• ! r . 0 50 100 150 0 50 100 150
Mass(amu) ~ Mass(amu) ~
t 3scr t ::i .:i B "' P03 P02 c: .l!l .s (I)
P02 > 1ii (j) a:
o· 37cr
~H2 I
H"
H- HO" PH2-C02
qN· r v 74
0 50 100 0 50 100
Mass(amu) ~ Mass(amu) ~
Fig. 14; Positive- and negative-ion static-SIMS spectra of glycine Ni(P) surface
after debonding in vacuum (A, top left and C, bottom left) and after
subsequent exposure to air (B, top right and D, bottom right). A linear
intensity scale is used in the static-SIMS spectra.
101
t * sn+ isotopes
• Pd+ isotopes
107 A + 9 109Ag+
x+unknown
l l L ·,~ 't <I X
\
100 105 110
ssNi/
ssNi2W
~
'
I j 1 J1 • ;
I
115
58Ni6oNi+
58Ni60NiW
soNi2+
* soNi2W I . I
l t i
120
Mass(amu)-
Fig. 15: Positive-ion static-SIMS spectrum in the mass range of the activator
elements of the Ni(P) fracture surface of the same sample type as for
fig. 14 A to D. A linear intensity scale is used in the spectrum.
4.5. Discussion
In this section first the advantages and drawbacks of the DPO and peel tests for
the adhesion measurement are discussed. By combining the various interface
analysis results, a good impression of the interface structure and chemical com
position is obtained. The final sections deal with the mechanism of adhesion and
with the relation between adhesion strength and fracture energy.
4.5.1. DPO test
- Literature data
The strength values in the literature of Ni(P) on rough-type 96 % alumina range
from 10 to 30 MPa, but mostly values of about 20 MPa are reported (7). In this
work on similar substrates, values of about 12 and 18 MPa are found for
acetate-type Ni(P) and glycine-type Ni(P), respectively. Since in the literature
102
fracture energies are not reported, it is difficult to explain differences from the
strength values reported here in terms of process conditions and interfacial bond
mg.
- DPO test-sample preparation
The direct pull-off adhesion strength measurement procedure used in this work is
slightly different from the one used in the literature (1 - 6). In the procedure used
in the literature first about 2 Jtm Ni(P) is deposited, then by photolithography flat
patches of 2 x 2 mm2 size are etched and a tin-plated copper wire is soldered. This
procedure is more laborious than using the commercial, adhesive-coated pull studs.
A few other differences from the literature method are the following: In etching
there is always a risk of some degree of underetching, and the shape of a solder
dot is difficult to control. With our studs a more homogeneous stress distribution
is expected since bonding is axisymmetric. With soldering the wire must be accu
rately centered over the Ni(P) patch, while with the pull studs this is not necessary.
Soldering causes a thermal shock to about 250 oc, while the studs are bonded at
160 °C. The amount of elastic strain energy is proportional to the difference in
temperature squared. With the epoxy adhesive on the studs, strength values up to
80 MPa have been measured, while solder is much weaker, thus limiting the max
imum measurable strength. For soldering, The Ni(P) layer thickness must at least
be 2 t-tm because solder reacts with Ni(P). For adhesive bonding any layer thick
ness is suitable, if the layer is closed. In spite of the differences in adhesion
strength measurement procedures, the strength values reported in the literature are
of the same order as the values reported in this chapter.
- Residual strain energy
In the DPO test, pull studs are bonded to the sample with an epoxy adhesive which
is polymerized and solidified at 160 °C. Due to differences in thermal expansion
coefficients elastic strain energy is built up in the adhesive layer during cooling as
given by eq. 7 in section 4.2.2. With an adhesive layer thickness of 10 t-tm and as
suming purely elastic deformation (Young's modulus I GPa, ref. (34)), a worst
case value of the order of 0.5 Jfm2 is obtained, which is small compared to peel
energies of the order of 10 J/m2• Moreover, during or after fracture the adhesive
film cannot freely expand or contract since it is restricted by the aluminium pull
103
stud on which it remains. Therefore, it is more probable that the release of elastic
strain energy is determined by the difference in thermal expansion of the alumin
ium pull stud and the layer- substrate combination and by the degree of relaxation
of this stress by the adhesive layer during cooling. This is, however, a rather
complicated mechanical problem which is not within the scope of this work.
Kinloch (34) states that in most practical cases stresses are relaxed by viscoelastic
deformation of the adhesive during cooling.
- Stress concentration
As described by Kinloch (34), stress concentration takes place at the edge of an
adhesive-bonded butt-joint geometry. Consequently, in such a geometry fracture
starts at the edges. In this study, due to a well-chosen nail-head-shaped pull-stud
geometry (fig. 3 and 4A), the stress at the edge is limited. For these studs the
fracture was observed to start near the middle of the bonded area. This aspect will
be further discussed in section 4.5.5. This is an indication that the stress at the
edge does not exceed the stress at the pull-stud axis. According to Kinloch, the
stress concentration is also reduced by the remaining excess adhesive at the edge,
the spew fillet which can be observed in the optical micrograph in fig. 4A. In
practice the stress is limited by viscoelastic deformation of the adhesive. An upper
boundary value is given by the yield stress. Typical values for the yield stress of
epoxy adhesives are in the range of 30 to 50 MPa (34).
4.5.2. Peel test
- Quantitative aspects
As discussed in the introduction, the peel energy Gr is equal to the fracture energy
Gc only if the contributions of residual strain energy Ge1 and energy of plastic de
formation Gder are negligible (eq. 5). The residual strain energy in the metal layer
is proportional to the layer thickness. If this energy plays a role in the peel test, a
decreasing peel energy with increasing layer thickness should have been found.
However, as shown in fig. 7, the peel energy is found not to depend on layer
thickness. This can be understood by separately considering the contributions of
the electrodeposited Ni layer and the electrolessly deposited Ni(P) layer to the
residual strain energy. The internal stress a; in low-stress sulphamate Ni deposits
104
ranges between 0 and 50 MPa (35). Using the upper boundary value of 50 MPa
the residual strain energy G.1 can be calculated with eq. 8 (section 4.2.2):
2 (J·
I
2E [8]
A residual strain energy Ge~ of 0.007 J/m2 per ttm layer thickness is obtained.
Therefore it can be concluded that the residual strain energy G.1 of the
electrodeposited Ni layer does not play a significant role. Stresses in the electroless
Ni(P) film depend on the phosphor content. For a P content of about 11 wt. %
the stress varies between about 20 MPa tensile and 20 MP a compressive (31 ). Since
both the layer thickness and the internal stress are much smaller for Ni(P) than for
electrodeposited Ni, the contribution of the Ni(P) layer to this residual strain en
ergy Get can be neglected as well.
In order to estimate the contribution of plastic deformation of the metal layer, the
stress - strain curves were measured during loading and unloading cycles up to
various top loads, lower than the peeling load as shown in fig. 9 and as described
in section 4.4.1. Since within the experimental error, the amount of energy stored
in the system during loading was completely released upon unloading, it was con
cluded that plastic deformation is negligible in the peel test for these samples.
Therefore, the radius R' P is caused by deformation in the plastic zone at the crack
tip (see below). This is not a secondary effect, but it forms intrinsic part of Gc, see
section 4.2.2. Therefore, it can be concluded that for this system the peel energy
Gp is equal to the fracture energy Gc.
- Crack tip plasticity
Kinloch (36) found an increasing peel energy value with increasing polymer layer
thickness up to several millimeters, with unaltered intrinsic adhesion. This is as
cribed to the increased volume available for crack-tip plastic deformation. Owing
to the much higher yield strength of metals compared with polymers, the size of
the plastic deformation zone is smaller for metals. The height H of the plastic de
formation zone in the metal side of the present interface (fig. 16) is given by
eq. 11 (36, 37) (plane stress):
105
H [11]
in which ay is the yield strength of the metal phase. Using a fracture energy value
Gc of 10 J/m2, a Young's modulus of 190 GPa and a yield strength Gy of 400 MPa,
a height H of about 2 fJ.m results. In case of plane-strain condition the factor 2n
should read 6rr. In reality, a full 3-D stress situation is present, for which the
constant factor is somewhere between 2n and 6rr. Consequently, the above estimate
is a maximum estimate. It is concluded therefore, that the height of the plastic
deformation zone is not limited by the layer thickness for layer thicknesses greater
than 2 fJ.m. This is in agreement with the observation that the peel energy did not
depend on the layer thickness, between 2 and 9 fJ.m, as shown in fig. 7.
t Deformed Metal
Fig. 16: Schematic representation of plastic deformation at crack tip (ref. 37).
- Types of crack growth
The sawtooth structure that is observed in the peel curves of acetate-type samples
is characterized as unstable brittle crack propagation (36). The unstable, discrete
nature of the crack propagation can either be explained by a strong dependence
of the size of the plastic deformation zone on the crack growth rate or by a dif
ference in intrinsic initiation and propagation fracture energies. Both phenomena
result in a relatively low fracture energy at a high crack propagation rate and a
high fracture energy at a low rate. When during peeling at a constant cross-head
speed the crack tip advances more rapidly than the peel rate, the stress is released
until fracture stops. This corresponds to a decrease in peel load. Subsequently, the
peel load increases up to the higher load, corresponding with slow fracture. These
106
higher and lower fracture energies are interpreted as initiation and arrest fracture
energies, respectively. This mechanism does not, however, explain why a different,
more stable crack growth is found for the glycine-type Ni(P), with fracture energies
in the same range.
4.5.3. Interface microstructure and chemistry
- Fracture path
The cross-sectional TEM micrographs show that an interfacial layer of I to 2 nm
thickness is present for all Ni(P) - alumina samples investigated. In HR-TEM
micrographs of a sputtered Ti layer on the same smooth-type alumina, a sharp
transition from ceramic to metal is observed (38). This means that the interfacial
layer is not a characteristic feature of the substrate material, such as e.g. a
hydrolyzed surface layer. It must be concluded, therefore, that the interface layer
is formed by deposition of Ni(P). Apart from this interfacial layer, with TEM a
good interfacial contact was observed for all samples. With static-SIMS and XPS
it is shown that the outermost monolayers of the fracture surfaces mainly consist
of Ni(P) and alumina, respectively. Therefore, it is concluded that fracture takes
place exactly through or at this interfacial layer which may therefore be regarded
here as the weakest link in the chain. The nature, composition and origin of this
layer are thus of great importance for this investigation. An overview of the most
important information on the composition of this interfacial layer from
static-SIMS, AES and XPS is presented in Table 11. For the sake of completeness,
results of Rutherford backscattering spectrometry (RBS) analyses which were not
previously described, are added.
In the following discussion, a number of possible contributions to the interfacial
layer are considered successively.
- Activator material and complexing agents
The interfacial layer partly consists of activator material. XRF measurements have
shown that the amount of activator material (Sn, Ag and Pd) before deposition
of Ni(P) is about a monolayer (of the order of 1019 metal atoms per m2) with the
present nucleation procedure (39). Also by AES depth profiling and XPS and
static-SIMS fracture surface analysis of Ni(P), these elements are detected at the
107
interface. However, the amount of nucleation material is not enough to account
for the whole layer thickness as observed with TEM. In addition to these activator
elements with static-SIMS also aoetate or glycine, depending on the metallization
solution, is detected on the fracture surfaces. Reliable estimations of coverages
cannot be made on the basis of these static-SIMS measurements. Therefore, it can
only be concluded that the interface layer contains nucleation material and the
organic molecules mentioned above.
Table 11: Overview of most important chemical analysis data on interf~cial composition.
I Technique
Organic
Static-SIMS glycine I acetate XPS Various C AES2 c RBS3
T': Only for glycine-type Ni(P). "2": Depth profile
Constituents Sn, Ag, Pd F N
Sn, Ag, Pd
~ Sn, Ag Sn, Ag, Pd (Sn, Ag, Pd) - -
"3": Analysed after nucleation, before Ni(P) deposition.
- Metallization solution and oxidation of Ni(P)
Cl Other
Cl 0, Ni, P02, PO,
- 0, Ni(OHh, NiP04
- -Cl - !
It is possible that also a thin layer of solution remains at the interface during
metallization. In fact, all of the compounds of the metallization solution can be
recognized in the static-SIMS spectra. The XPS measurements show that on the
alumina fracture surfaoe Ni is present which cannot be explained by the presence
of remaining Ni(P) particles, sinoe this Ni is for the greater part removed by
sputtering only 20 nm. Moreover, the ratio of oxidized Ni versus metallic or
intermetallic Ni is greater on the alumina surface, compared to the Ni signals on
the Ni(P) fracture surfaoe. It is observed that the adhesion of Ni(P) on very
smooth non-conducting substrates such as float glass, is considerably increased
after drying the sample when the first metal layer has been deposited. Even the
slightest stress leads to cracking and buckling of the Ni(P) films during deposition
and the film can be wiped off with a tissue in the wet state. Therefore an inter
mediate drying step is often used after deposition of the first 0.1 Jl.m when using
smooth surfaces. Probably water is bonded to the oxide surfaoe more strongly
than the freshly deposited Ni(P) and capillarity or inclusion effects may play a role
108
as well. After the water is evaporated, remaining components of the metallization
solution may contribute to the formation of an interface layer. Due to the cracking
and buckling effect described above, it is difficult to prepare samples of Ni(P)
layers on perfectly smooth substrates, although the adhesion in the dry state may
be acceptable.
- Carbon at the interface
As described in the previous section, AES depth profiles only gave relevant infor
mation of the interface when recorded from the smooth sample types. Even on
these smooth surfaces the intensity of the interface species was just above the noise.
In a separate experiment smooth, polished alumina was used instead of the
sintered surfaces. On these polished samples the same signals were observed, at a
much higher intensity relative to the noise. This is caused by an increasing depth
resolution with decreasing roughness. Therefore, on the polished samples the ob
servations from the sintered surfaces are confirmed.
It is very probable that the carbon detected at the interface with AES originates
from acetate or glycine detected at the interface with static-SIMS. On the glycine
Ni(P) and alumina fracture surfaces, after debonding in the vacuum of the
static-SIMS apparatus, hardly any other organic compounds (contaminations) are
measured. Therefore, it is unlikely that the carbon signal in the AES spectra at the
interface is caused by organic contaminations. In the case of glycine, N should also
have been detected by AES, which was not the case. It is, however, possible that
the N signal remained below the detection level because only 1 N atom is present
per 2 C atoms, and the signal due to C at the interface is already very weak.
4.5.4. Mechanism of adhesion
- The contribution of mechanical interlocking
The mechanism of adhesion describes the type of intrinsic interfacial (chemical or
mechanical) interactions. Most authors (1 - 6) who studied the adhesion of Ni(P)
using 96 % alumina, observed that the adhesion is strongly influenced by etching
conditions and therefore conclude that the adhesion is determined by mechanical
interlocking (7).
109
Osaka et al. (5) concluded from experiments with electroless Cu using Ni(P)
underlayers, that apart from mechanical interlocking additional, interfacial phe
nomena also play a role in the adhesion. All these literature data are obtained by
adhesion strength measurements which, as explained in the sections 4.1 and 4.2,
are insufficient for drawing conclusions about intrinsic interfacial interactions.
The peel energy values for the rough-type substrates are about five times higher
than those for the smooth substrates, as reported in section 4.3.2. Since small
pieces of Ni or Ni(P) remain between surface grains of the rough ceramic
substrates and local deformations are observed on the metal fracture surfaces, it
is concluded that mechanical interlocking at least contributes to the intrinsic frac
ture energy. This model is illustrated by the cross-section optical micrograph in
fig. 4B.
The influence of surface roughness on the adhesion is more complex than sug
gested by the simple model of rupture of penetrated parts of the film which remain
in substrate pores. Oh et al. (37) illustrated nicely how the interface microstructure
influences the fracture energy with unaltered chemical interactions for
thermocompressed copper foils on glass. By deliberately introducing small
interfacial flaws, bridging ligaments were created behind the advancing crack
front. The additional energy dissipation in these ligaments is larger than the ori
ginal fracture energy. On rough substrates, fracture may take place similarly.
- Van der Waals and other chemical interactions
For both the rough and the smooth-type substrates it is observed that the moisture
content of the atmosphere significantly influences the peel energy value. This can
not be explained by mechanical adhesion, only by chemical adhesion, including
van der Waals interactions. Since the interfacial area increases along with the
roughness during etching, chemical interactions may increase as well as mechanical
interactions. Therefore, with the present results it can be concluded that for the
rough substrates both mechanical and chemical interactions play a role in the ad
hesion. With the present data it is not possible to make a quantitative estimation
of each contribution. For the smooth substrates no evidence is obtained that me
chanical interactions play a role in the adhesion, since no metal remains on the
llO
ceramic fracture surface and on the metal fracture surface no local plastic defor
mations can be distinguished.
Since the fracture always takes place exactly along the interface, except for the
interlocking sites on the rough substrates, it must be concluded that the chemical
bonds in both the metal layer and the substrate are much stronger than at the
interface. Therefore it is probable that interfacial bonding is brought about by van
der Waals interactions. Moreover, in view of the conditions under which both
phases make contact, it is not probable that ionic or covalent bonds are formed
between layer and substrate.
Van der Waals interactions amount to 0.5 J/m2 maximally, but for the interfaces
studied here this is probably less, due to the presence of organic and probably also
inorganic molecules at the interface. Nevertheless, a peel energy of at least 7 J/m2
is measured. This difference may be explained by the plastic deformation processes
described in section 4.5.2. For the thermocompressed Ni - alumina system fracture
energies of about 150 Jfm2 were measured, while an intrinsic fracture energy of a
few J/m2 was calculated (18). This implies that for that system the energy loss
factor f1 (section 4.2.1) is of the order of 100.
4.5.5. Relation between adhesion strength and fracture energy
-Flaw size calculations
Table 12: Critical flaw size, ac, calculated from adhesion strength ar and fracture energy Gc values.
Ah03 Ni(P) IJr (MPa) Gc (J/m2) ac, (pm)
Rough Glycine 22 24.3 120
Rough Acetate 12 40.9 700
Smooth Glycine 45 6.1 6.8
Smooth Acetate 5 8.5 730
As shown by the data in Table 12 there is no proportionality between interfacial
fracture energy and adhesion strength, as might be expected, see eq. 1. Since large
differences in residual stress energy due to built-in stresses or due to differences in
Ill
thermal expansion would become apparent in the peel energy (see eq. 5, section
4.2.2), these phenomena can be excluded as an explanation for the differences in
the adhesion strength. For the calculation of critical flaw sizes a K value (eq. 1)
of Ll3, appropriate for circular flaws, is used (40). Even if the peel energy GP is
not equal to but only proportional to the fracture energy Gc, the trend from
Table 12 is clear. The highest fracture energy is measured on samples with low
strength and the strongest samples have the lowest fracture energy. A wide range
of interfacial flaw sizes is calculated from these values with the Griffith-Irwin
equation. Since the flaw is in the proximity of the interface, an elastic modulus of
the order of a few GPa, typical for the adhesive, rather than a few hundred GPa,
typical for the nickel layer and the alumina substrate has to be used. For the ef
fective elastic modulus a value of 2 GPa is chosen. This is a different value from
the Young's modulus of the epoxy adhesive which is used for the calculation of
residual strain energy in section 4.5.1 (1 GPa). This difference is related to the ge
ometry and the loading conditions. For an adhesive layer which is thin relative to
its lateral dimensions and which is perfectly constrained by relatively rigid
substrates, the relation between the effective modulus of the adhesive E'. and the
Young's modulus E. under normal loading is given by eq. 12 (33):
E' a [12]
in which v. is Poisson's ratio. For a Poisson's ratio of 0.35, which is a typical value
for epoxy adhesives, the effective modulus may be greater than the Young's
modulus by about 50 %, or even more due to the spew fillet.
-Flaw growth during testing
Flaw sizes of about 800 p,m are calculated for the samples prepared with the
acetate-type Ni(P), whereas for the samples with the glycine type Ni(P) much
smaller flaw sizes are obtained. In principle, there are two possibilities: Either these
flaws are present at the interface after sample preparation, or they are introduced
during the strength test. By scanning acoustic microscopy no indications of the
presence of interface flaws are obtained from the metallized samples, within the
resolution of 20 p,m. Also, on the fracture surfaces no features are found which
point to the presence of interface flaws. Since it is not possible to study the for-
112
mation of flaws at the ceramic - metal interface during the strength measurement,
polished glass plates were used as model substrates. After grinding one side was
metallized and during the DPO test the interface was observed through the glass
with a video camera. A sequence of photographs shown in fig. 17, revealed that
during the strength test a flaw appears which grows in discrete steps. At a size of
about 800 p,m, fracture suddenly takes place.
Apparently, during this flaw formation, the Griffith-lrwin energy balance remains
near to equilibrium, otherwise catastrophic failure would have taken place imme
diately. The possibility of slow crack growth at the interface has been considered.
A peel strip with acetate type Ni(P) was loaded with various weights, correspond
ing to 10 to 90 % of its peel load and the advance of the crack front was moni
tored with a camera. The peel front did not move in a week's time, not even at 90
% of the peel force, within the resolution of 20 to 50 p,m. This suggests that slow
crack growth does not play an important role in these systems. Moreover, the flaw
growth that was observed during the DPO test did not take place gradually but in
discrete steps, while for slow crack growth a gradual increase in size is expected.
An alternative explanation for the flaw growth is related to the mechanical be
haviour of the adhesive layer with which the aluminium pull stud is bonded on the
metal layer. The effective elastic modulus for a thin adhesive layer under normal
load, with a layer thickness various orders of magnitude smaller than its lateral
dimensions, is much higher than the Young's modulus, see eq. 12. Only at the
edges, where shear displacement in the adhesive is possible, does the effective
modulus approach the Young's modulus (41). In the vicinity of an interfacial flaw,
no normal load is applied and the effective elastic modulus of the adhesive is lo
cally reduced, limiting the amount of missing elastic strain energy, which causes
catastrophic failure according to the Griffith-Irwin theory. Similar observations
of a debonded area which grows during testing have been made for acrylic
(Plexiglas) plates, bonded with a polyurethane adhesive (42). It should, however,
be stressed that in this case the effect is characteristic of some sample types, rather
than for the DPO test. It is reasonable to assume that with smaller initial flaw
sizes, this flaw growth process does not take place or takes place at higher stresses,
leading more rapidly to catastrophic failure.
113
Fig. 17: Stepwise flaw growth during adhesion strength test on a glass model
substrate.
114
- W eibull statistics
The Weibull moduli found in this work range between 0. 7 and 4.2. This is rela
tively low compared to Weibull moduli generally found for the bulk strength of
ceramics, ranging from 5 to 20 (43). In the few data that are available in the lit
erature on Weibull moduli of adhesion strengths, the Weibull moduli for adhesion
tend to be somewhat lower than for bulk strength (44). For the joint strength of
SbN4 / Ni-Cr systems Weibull moduli ranging from 2.3 to 6.1 were reported (45).
For both the tensile strength and the three-point bending strength of Si3N4 I AI /
Invar joints, Weibull slopes of about 6 were found (46). For the lap shear adhesion
strength of epoxy and acrylate coatings on glass a spread corresponding with a
Weibull modulus of about 2 was reported (9). By adhesion measurement with in
dentation for the same systems Weibull moduli of about 9 were found, which was
explained by the fact that interfacial flaws do not play a role due to the small area
in the indentation tests.
Generally, the distribution of flaw sizes determines the Weibull modulus value.
Apparently, a wide distribution of flaw sizes is present at the interfaces studied in
this work. This width may be partly explained by the stable flaw growth during
testing which is described above. This growth may lead to a decreased Weibull
modulus similar to that found in the case of slow crack growth (47). This is in
agreement with the observation that the lower Weibull moduli were found for
samples with the largest calculated flaw sizes, viz. those with acetate-type Ni(P)
(Table 1). For clarity's sake it is repeated here that slow crack growth probably
does not occnr in our system.
4.6. Conclusions
This contribution clearly demonstrates the complexity of the adhesion of
electroless Ni(P) to alumina ceramics. It has been shown that a fracture mechanical
approach, along with a thorough characterization of chemistry and structure of
the interface is required for obtaining insight in the adhesion.
115
With a direct pull-off adhesion strength measurement, sample standard deviations
of about 30 % are frequently found in these measurements, for which 20 to 30
samples are required. Furthermore, conditions are investigated under which the
90° peel test can be used as a quantitative fracture energy measurement. By using
sulphamate Ni as bulk metal layer, the influence of elastic energy stored in the
layer can be neglected. This is confirmed by the observation that the peel energy
is not in11uenced by the layer thickness. It is tentatively concluded that for the
smooth substrates, the peel energy is a good approximation of the fracture energy.
In the peel measurements with eight samples, standard deviations in the mean of
one or a few percent are obtained.
The Ni(P) - alumina interface structure, studied with cross-section TEM was very
different from that of most other metal - ceramic systems prepared by for example
vacuum-deposition of metal layers or by thermocompression of metal films on
ceramics ( 48). For such systems a sharp transition is generally observed between
metal and ceramic, while for the Ni(P) - alumina system, an interface layer with a
thickness of I to 2 nm is observed for all samples, probably due to nucleation
material and organic molecules detected at the interface with static-SIMS. Fracture
takes place at or in this layer. Apart from the interface layer, a close contact be
tween layer and substrate is observed for all samples. Based on the fracture energy
measurements, it is concluded that differences in adhesion strength of the various
sample types cannot be accounted for by differences in interfacial structure at
nanometer level.
By static-SIMS measurements most of the components in the metallization sol
utions are found on the corresponding fracture surfaces for both sample types.
This was also the case for the glycine and acetate complexing agents, which was
the only significant difference between the spectra of the two sample types.
Nucleation material is also found to be present on both the layer and the substrate
fracture surfaces. In addition, by XPS it was shown that the interfacial layer can
not be completely explained by oxidation of Ni(P) at the interface during or after
deposition.
116
Peel measurements show that for both the rough and the smooth-type substrates
chemical interfacial interactions contribute to the adhesion. In view of the exper
imental conditions it is probable that the chemical interactions are limited to van
der Waals-type interactions, which is in agreement with the order of magnitude
of the measured peel energies. Only for the rough-type substrates evidence has
been obtained that mechanical interlocking contributes to the adhesion. The peel
measurements show that the difference in adhesion strength between the glycine
type Ni(P) and the acetate-type Ni(P) cannot be accounted for by differences in
chemical or mechanical interfacial interactions or differences in residual (built-in
or thermal) stresses. It is therefore concluded that the difference in adhesion
strength is due to differences in interfacial critical flaw sizes. Since strong adhesion
was found for samples with smooth substrates and low peel energies, it can be
concluded that strong adhesion can be obtained, probably by van der Waals
interactions, without making use of mechanical interlocking. For most samples the
adhesion strength is limited by the size of interfacial flaws. The final conclusion
therefore is that further research is required to obtain insight into the origin of
these flaws.
Note added: With a deposit stress analyser (49) an internal stress of 40 MPa was found
in the sulphamate Ni electrodeposits. This confirms the literature data used in the
discussion on p. 105.
Literature
1. H. Honma and S. Mizushima, Kinzoko Hyomen Gijutsu 33, (1982), 380.
2. T. Osaka, E. Nakajima, Y. Tamiya and I. Koiwa, Kinzoko Hyomen Gijutsu
40, (1989), 67.
3. T. Osaka, Y. Tamiya, K. Naito, and K. Sakaguchi, J. Jpn. Inst. Printed
Circuit 4, (1989), 285.
4. H. Honma and K. Kanemitsu, Plating and Surface Finishing 74, (1987), 62.
5. T. Osaka, Y. Tamiya, K. Naito and K. Sakaguchi, J. Surf. Finish. Soc. Jpn.
40, (1989), 835.
117
6. M. Kamijo and N. Aylizawa, Yamanashi - ken, Kogyo Gijutsu Senta
Kenkyu Hokoku l, (1987), p. 86. (Research report of the Yamanashi
Prefectural Industrial Technology Center).
7. J.W. Severin and G. de With, accepted for publication in J. Adhesion Sci.
Technol. (1992).
8. M.D. Thouless, Mat. Res. Soc. Symp. Proc. 119, (1988), 51.
9. J.E. Ritter, L. Rosenfeld, M.R. Lin and T.J. Lardner, Mat. Res. Soc. Symp.
Proc. 130, (1989), 237.
10. R. Jacobsson, Thin Solid Films 34, (1976), 191.
11. R. Jacobsson and B. Kruse, Thin Solid Films 15, (1973), 71.
12. P.C. Hopman, Transactions and Communications, 1984(7), 179.
13. A.G. Dirks and J.J. van den Broek, Thin Solid Films 96, (1982), 257.
14. R.J. Good in Adhesion Measurements of Thin Films, Thick films and Bulk
Coatings, K.L. Mittal (ed.), ASTM STP 640, (1978), p. 63.
15. S.J. Bennett, K.L. de Vries and M.L. Williams, Int. J. Fracture 10, (1974),
33.
16. S.A. Varchenya, A. Simanovskis and S.V. Stolyarova, Thin Solid Films 164,
(1988), 147.
17. A.J. Kinloch in "Adhesion and Adhesives", Chapman and Hall, London,
1987, Ch. 3.
18. H.F. Fischmeister, G. Elssner, B. Gibbesch and W. Mader, Materials Re-
search Society International Meeting on Advanced Materials 8, (1988), 227.
19. Ref. 17, Ch. 4.
20. K.L. Mittal, J. Adhesion Sci. Tech. 1, (1987), 247.
21. K-S. Kim, Mat. Res. Soc. Symp. Proc. 119, (1988), 31.
22. J.E.E. Baglin, Mat. Res. Soc. Symp. Proc. 47, (1985), 3.
23. A.N. Gent and J. Schultz, J. Adhesion 3, (1972), 281.
24. G.J. Lake and A. Stevenson in "Adhesion 6", K.W. Allen ed., Applied Sci-
ence Publishers, London, 1982, p. 41.
25. W. Weibull, J. Appl. Mech. 18, (1951), 293.
26. L.J.M.G. Dortmans and G. de With, J. Am. Ceram. Soc. 74, (1991), 2293.
27. W.H. Safranek, The properties of electrodeposited metals and alloys,
Elsevier, New York, 1974, Ch. 12, p. 260.
118
28. C.H. de Minjer and P.F.J. van der Boom, J. Electrochem. Soc. 120, (1973),
1644.
29. L.C. Feldman en J.W. Mayer in "Fundamentals of surface and thin film
analysis", Elsevier science publishers, New York, 1986.
30. H. van der We1, P.N.T. van Velzen, U. Jiirgens and A. Benninghoven in
"Analysis of Microelectronic Materials and Devices", M. Grasserbauer and
H.W. Werner, eds., 1991, John Wiley & Sons Ltd., Ch. 2.10.
31. W. Riedel in "Funktionelle Chemische Vernicklung", E.G. Leuze Verlag,
Saulgau I Wiirttemberg, 1989, p. Ill.
32. Pers. Communication with F. Vreugdenhil, Philips CFT Eindhoven (1991).
33. K.S. Rajam, S.R. Rajagopalan, M.S. Hegde, B. Viswanathan, Mater. Chem.
Phys. 27, (1991), 141 and National Institute of Standards and Technology
XPS data base version 1.0, october 1989, NIST, Gaithersburg, Maryland,
USA.
34. Ref. 17, Ch. 6.
35. Chin-Min Lin and Ten-Chin Wen, Plating and Surface Finishing 78(9), 70.
36. Ref. 17, Ch. 7.
37. T.S. Oh, R.M. Cannon and R.O. Ritchie, Mat. Res. Soc. Symp. Proc. 130,
(1989), 219.
38. J.W. Severin, R. Hokke, H. van der Wel, M. Johnson and G. de With, ac
cepted for publication in J. Electrochem. Soc., (1993).
39. J.W. Severin, R. Hokke, H. van der Wel and G. de With, accepted for pub-
lication in J. Electrochem. Soc., (1993).
40. G.K. Bansal, J. Am. Ceram. Soc. 59, (1976), 87.
41. P.C.P. Bouten, personal communication (1992).
42. G.P. Anderson and K.L. DeVries in "Treatise on Adhesion", Vol 6, R.L.
Patrick. ed., Marcel Dekker Inc., New York, (1989), Ch. 3, p. 102.
43. R.W. Davidge in #Mechanical Behaviour of Ceramics", Cambridge Univer
sity Press, 1979, Ch. 9, p. 136.
44. J.T. Klomp and G. de With, accepted for publication in Mater. Manuf.
Proc. (1993).
45. S.D. Peteves in "Ceramics Today - Tomorrow's Ceramics" Part B, P.
Vincenzini ed., Elsevier Science Publishers, Amsterdam, ( 1991 ), p. 1469.
119
46. K. Suganuma, T. Okarnoto, M. Koizumi and M. Shirnada, J. Am. Cerarn.
Soc. 69, (1986), p. C-235.
47. J.D. He1finstine, J. Am. Ceram. Soc. 63, (1980), 113.
48. M. Riihle and W. Mader in "Designing interfaces for technological applica
tions", S.D. Peteves ed., Elseviers Science Publishers, London, 1989, p. 145.
49. Deposit stress analyser, Model 683 EC, Electrochemical Co. Inc., 1600
Pennsylvania Avenue, York, P A. 17404.
50. PHI Handbook of Auger electron spectroscopy by L.E. Davis, N.C.
MacDonald, P.W. Palmberg, G.E. Riach and R.E. Weber, published by
Physical Electronics division, Perkin-Elmer Corporation, 6509 Flying Cloud
Drive, Eden Prairie, Minnesota 55343.
51. H. van der Wel, J. Lub, P.N.T. van Velzen and A. Benninghoven,
Microchim. Acta (Wien) 1990 II, p. 3.
120
Chapter 5
The influence of thermal treatments on the adhesion of
electroless Ni(P) layers on alumina ceramic
Summary
The adhesion of electrolessly deposited Ni(P) on 96 % and 99.5 %
alumina was studied as a function of annealing temperature, up to
580 °C. The adhesion was measured with the direct pull--off test and the
peel test. The interface structure was analysed with cross-section TEM.
Fracture surfaces were analysed with SEM / EDX, static-SIMS and
XPS. The optimum annealing temperature was found to be 400 oc, at
which an increase in peel energy and adhesion strength by a factor of 2
to 3 was measured, with respect to the as-deposited value. It was observed
that, upon heating, Ni(P) crystallizes and forms microcracks, mainly
perpendicular to the interface, but not along the interface. The nucleation
material disappeared, the organic molecules decomposed and the amount
of oxygen at the interface decreased, probably by diffusion into the metal
bulk. Since the fracture path remained along the interface, the improve
ment in adhesion properties with annealing temperature can be explained
by the changes in interfacial chemistry.
121
5.1. Introduction
Electroless metallization of oxidic surfaces is frequently used for electronic appli
cations. The thermal behaviour of the metal-ceramic interface is of great impor
tance for these applications. Thermal shocks occur with soldering and thermal
cycling is a standard test procedure for most electronic parts. Retention of strong
adhesion is required since differences in thermal expansion, for example between
electronic components and the printed-circuit board or between metal layers and
substrates cause mechanical stresses. Interfacial fracture may rapidly lead to elec
tronic failure.
Generally, the adhesion strength of metal layers on oxidic substrates increases with
annealing temperature (1, 2). An annealing treatment after deposition might
therefore be a simple method to improve the adhesion of electrolessly deposited
Ni(P) layers. However, for Ni(P) on 96 % alumina diverging results on the effect
of temperature upon adhesion have been reported, as measured with the direct
pull-off (DPO) test. Honma and Mizushima (3) found an increase in adhesion
strength with annealing time and annealing temperature. The greatest effect, an
increase of a factor of 3 to 4, was found after annealing for 1 hour at 250 oc in
air. In contrast, in a later publication than ref. (3) Honma and Kanemitsu (5) did
not measure significant changes in the adhesion strength after annealing at
250 ac in air for between 0.5 and 24 hours, with respect to the as-deposited value.
In addition, Osaka et al. (4) did not find significant differences between the adhe
sion strengths before and after annealing for 1 hour in vacuum at temperatures
of 300 and 500 °C. Since these literature data do not allow a definitive conclusion
to be drawn on the influence of thermal treatments upon the adhesion, more in
sight into this matter is required.
On the basis of adhesion strength data only, as measured by the DPO test in the
references cited above, it is very difficult to explain changes in the adhesion. As
described in more detail in ref. (6), the adhesion strength is determined not only
by interfacial interactions on a molecular scale (intrinsic adhesion) but also by the
size of interfacial flaws due to, for example, pores or foreign particles. For that
reason not only adhesion strength measurements but also fracture energy meas-
122
urements were performed in this study. Moreover, various interfacial analyses were
carried out in order to obtain information on changes of the intrinsic adhesion
with temperature.
In previous investigations (6, 7), cross-section transmission electron microscopy
(TEM) micrographs revealed the presence of an interfacial layer between the
alumina substrate and the Ni(P) layer for as-deposited samples. The thickness of
this interfacial layer was a few nanometres. Static secondary ion mass
spectroscopy (static-SIMS) and X-ray photoelectron spectroscopy (XPS) analyses
of the fracture surfaces showed that this interfacial layer mainly consisted of re
maining components of the metallization solution and nucleation material.
Moreover, these analyses showed that fracture took place through this layer. The
cohesive interactions within this layer are therefore considered to be of decisive
influence for the fracture energy. In this chapter an analysis will be made of how
the interfacial interactions and the fracture path change for samples that have re
ceived a heat treatment.
5.2. Experimental procedures
5.2.1. Sample preparation
For the sample preparation a rough-type 96 % alumina (Maruwa) and a smooth
type 99.5 % alumina (MRC/Coors) were used as the substrates. These ceramic
substrates were cleaned with a detergent, etched with an HF solution and
nucleated with solutions containing Sn, Ag and Pd. By electroless metallization,
Ni(P) layers of about 0.3 Jlm thickness were deposited. Two types of electroless
metallization solutions were ust:d, one with acetate as the complexing agent and
the other one with glycine as the complexing agent. For both solutions the P con
tent of the deposits was ea. 10 wt. %. On top of the electrolessly deposited Ni(P)
layers, galvanic Ni layers were deposited from a low-stress sulphamate bath. The
total metal layer thickness was about 2 Jlm for the DPO test samples and about
7 Jlm for the peel test samples.
123
The galvanic Ni layer was only applied in order to facilitate adhesion measure
ments. For the peel test a metal layer with sufficient strength and stiffness is re
quired. For the DPO test the galvanic Ni layer prevents the penetration of the
epoxy adhesive of the pull studs into cavities of the rough substrate surface during
bonding. However, as described in subsequent sections, in certain cases the
galvanic Ni layer may influence the results of adhesion measurements after thermal
treatments. Therefore, in order to eliminate this possible influence, also a number
of measurements were performed on samples without galvanic nickel. For these
samples Ni(P) layers were electrolessly deposited with a thickness of 2 to 4 pm.
Details of materials, solutions and deposition conditions are given in refs. (6) and
(7). After metal deposition, the samples were annealed at temperatures between
lOO and 600 "C. Details of the annealing conditions are given in section 5.3.1.
5.2.2. Analyses
The adhesion measurements were carried out with DPO and peel tests as described
in ref. (6). The influence of humidity upon the fracture energy was measured by
peeling after applying a drop of water at the peel front. Scanning electron
microscopy and energy dispersive X-ray analysis (SEM I EDX), cross-section
TEM, XPS and static-SIMS analyses were carried out as described in the same
paper (6). For the static-SIMS analyses, samples were peeled in the vacuum
chamber of the apparatus in order to avoid contamination of the fresh fracture
surfaces in the air. The crystallization of the samples with an electrolessly depos
ited Ni(P) layer only, was followed by X-ray diffraction (XRD).
5.3. Results
5.3.1. Adhesion measurements
-Peel tests
The peel test results are shown in fig. 1. The relative accuracy of the peel test re
sults is within 10 % (6). This accuracy is mainly determined by the reproducibility
of the sample preparation. The lower line with open circle symbols represents the
peel energy versus annealing temperature for acetate-type Ni(P) on smooth-type
124
substrates. These samples were annealed for 1 hour in vacuum. From 150 to
580 "Can increase in peel energy is observed of a factor of 2 to 3. Samples without
galvanic Ni(P), onto which electroless Ni(P) layers with a thickness of ea. 3.5 J.lm
was applied, could only be peeled after annealing at temperatures of 200 oc or
lower. For as-deposited samples and samples annealed for 1 hour at 150 and
200 oc in air, peel values of 2.8, 4.3 and 5.1 Jjm2, respectively, were recorded.
Samples annealed at higher temperatures could not be peeled because the peel strip
broke before peeling started, owing to increased brittleness of the Ni(P) material
after annealing.
Fig. 1:
t 120 []
100 N
E [] --.. ::::.
Q.
80 (.9
-V 60
20.~1
-7 ~~30 5,20 .,0 40 ~~~ 20
___.a •0
0 0 100 200 300 400 500 600
T(°C) ...
Peel energy Gp versus annealing temperature T of samples with
smooth-type (circle symbols) and rough-type (solid dot and square
symbols) alumina substrates and acetate-type Ni(P) for samples
annealed for l hour in vacuum. The numbers along the line with solid
dot symbols represent the cumulative annealing time (minutes) in air.
125
The line with the square symbols in fig. I represents the peel energy versus
annealing temperature for samples with rough-type substrates and acetate-type
Ni(P). These samples were heat-treated for I hour each at a different temperature
in vacuum. After an initial decrease in peel energy by about 30 % after annealing
at 150 oc, a gradual increase of up to a factor 3 is observed after annealing at
450 oc.
The line with solid dot symbols in fig. 1 represents the peel energy versus annealing
temperature in the range between room temperature and 300 oc for a sample with
the rough-type alumina substrate and with acetate-type Ni(P). In this experiment
the sample was first peeled and then alternately annealed in air and peeled further.
Annealing was done by placing the sample in a preheated furnace and taking it
out after the annealing time. Every time annealing was done for a longer time or
at a higher temperature, resulting in a cumulative thermal load of the sample. The
numbers along this line in fig. l denote the cumulative annealing time for each
temperature. Due to oxidation of the nickel layer in air, no annealing temperature
higher than 300 oc could be applied in this experiment. Similarly as for the other
measurements with samples with smooth-type and rough-type substrates, adhesion
improvement with temperature was found for this sample with the rough-type
substrate. The initial decrease in peel energy was not observed for this sample.
For fig. 2 the measurement of peel energy versus annealing temperature for sam
ples with rough-type substrates, with square symbols in fig. I, was repeated. Fur
thermore, the relative decrease .1Gp is shown which was found after a drop of
water had been placed at the peel front. This relative decrease is significantly
higher for samples' annealed at temperatures above 250 oc than for samples
annealed at lower temperatures. These results will be discussed in section 5.4.1.
Since plastic deformation may contribute to peel energy values, it was necessary
to estimate relative changes in the yield strength of the metal layer due to the
thermal treatments. This was done by hardness measurements using the Vickers
test. With an indentation load of 0.01 N, Vickers hardness values of 400, 130 and
126
Fig. 2:
t 110 t "' 100
"" .§_ e.. ..., 0.
0. (.? (.? 90 <
80
70
60
50
40 20
30 )· 20 10
10 -· • 0 0
0 100 200 300 400 500 600
T(°C) ...
Relative decrease in peel energies by placing a water drop at the peel
front for samples with rough-type substrates, heat-treated at various
temperatures T. GP denotes the peel energy measured in laboratory air
(circles) and L1Gp the relative decrease in peel energy due to water (solid
dots).
100 MPa were measured for samples annealed at 150, 250 and 450 °C, respectively.
This means that a decreasing trend in the hardness of the metal layer was observed
with increasing temperature. To eliminate the influence of changing mechanical
properties of the galvanic nickel layer, the DPO tests were performed with samples
with only electrolessly deposited nickel.
127
- DPO tests
Since galvanic Ni had not been applied for the DPO test samples, a much thicker
electrolessly deposited Ni(P) layer was applied with a layer thickness of
2 ± 0.4 flm. The DPO strength versus annealing temperature is plotted in fig. 3 for
samples with the rough-type and the smooth-type substrates. The numbers of test
samples and the standard deviations are given in Table l. The samples were
annealed for I hour in an argon atmosphere at the top temperature indicated in
the figure. Since pull-studs were bonded with an epoxy adhesive at 150 oc for the
DPO test, DPO strength values of the as-deposited sample could not be obtained.
At temperatures of 250 oc or lower no systematic trend was observed in the ad
hesion strength, but at 300 and 400 oc a two- to three-fold increase in the adhesion
strength was clearly seen. For both substrate types a remarkable decrease in the
DPO strength was measured after annealing at 500 °C.
1 70
<? 60 0..
6 b
50
40
30
20
10
0
Fig. 3:
128
0
~ 0 J \X ~ \ X 0 0
100 200 300 400
Direct pull-off adhesion strengths (ur) versus annealing temperature for
samples with 2 flm electroless Ni(P) only. The symbols o and x repre
sent samples with smooth-type and rough-type substrates, respectively.
Table 1: Mean adhesion strength ar of electroless Ni(P) on rough- and smoothtype alumina ceramic as a function of annealing temperature T as measured by the DPO test. N is the number of test samples and s;; is the standard deviation in the mean.
TCC) Rough-type substrates Smooth-type substrates
N IJr (MPa) s, (MPa) N ar (MPa) s.- (MPa)
150 21 16.4 0.9 20 28.2 3.3
200 i 21 25.0 1.8 21 19.1 1.4
250 19 20.2 1.6 19 19.2 2.2
300 19 44.8 2.3 19 39.9 5.0
400 21 51.6 2.3 21 53.1 2.9
500 21 27.4 1.8 20 17.9 3.4
The XRD pattern of the DPO samples with rough-type and smooth-type
substrates annealed at temperatures lower than 400 oc only showed a broad band
due to an amorphous Ni(P) phase, apart from peaks originating from the a1umina
substrates. The samples annealed at 400 oc gave rise to an NhP diffraction pat
tern. In addition, a small contribution from amorphous material was still ob
served. The XRD patterns of samples annealed at 500 oc showed peaks
characteristic of NhP and Ni phases. With these samples an indication of the
presence of an amorphous phase was not visible in the XRD patterns anymore.
5.3.2. Interface and fracture surface structure
-SEMI EDX
In order to explain the large difference in peel energies between the two substrate
types, cross-sections were made of the metal-ceramic interfaces. Optical micro
graphs of these cross-sections, shown in fig. 4, show the penetration of the metal
layer into the surface pores of the rough-type substrates. This type of roughness
with narrow structures and cavities is difficult to measure with, for example, a
step-profiler. It is obvious that this roughness gives rise to a much stronger adhe
sion due to mechanical interlocking than on the smooth-type substrate surface
shown in fig. 4B, where such interlocking structures are not present.
129
Fig. 4:
130
Optical micrographs of cross-sections of samples with rough-type
substrate (A, top) and smooth-type substrate (B, bottom).
The SEM micrographs of rough-type alumina fracture surfaces of samples
annealed at 150 and 450 oc, shown in fig. 5, reveal a larger density of remaining
metal particles on the sample annealed at the higher temperature. On the
smooth-type alumina fracture surfaces (not shown), such remaining metal particles
were not found for samples annealed at 150, 320 and 450 oc. With EDX, never
theless, a very small Ni signal was observed for the sample annealed at 320 oc and
a stronger Ni signal for that annealed at 450 oc. When a relatively large area of
about 50 flm diameter was scanned with the electron beam, the same intensities
were found as when a small area of about 1 fim on a smooth alumina grain surface
was irradiated. This means that a very thin, Ni containing layer is present all over
the alumina fracture surfaces of smooth-type samples annealed at 320 and
450 oc. Ni was not detected with EDX on the surface of the sample annealed at
150 oc.
- Cross-section TEM
The TEM images shown in fig. 6 provide information on the material structure
both at the interface and in the bulk of the metal layers after annealing at 150 and
580 oc in vacuum. The columnar structure of the as-deposited Ni(P) material
(fig. 6A) has completely disappeared after annealing at 150 oc (fig. 6B) and
580 oc (fig. 6C). Instead, microcrystalline particles are observed and extensive
microcracking has taken place all over the Ni(P) layer and in all directions
(fig. 6C). The size of these microcrystals is too small to give rise to a crystalline
type XRD pattern. On top of the microcrystalline Ni(P) layer, Ni crystals are vis
ible from the galvanic Ni layer. No cracks along the metal-ceramic interface are
observed. The interfacial layer which is observed for the low-temperature sample
remains present after annealing (fig. 6D). The contrast between the interfacial layer
and the neighbouring phases is much weaker for the annealed sample than for the
as-deposited sample. This may be an indication that the density of the interfacial
layer increases upon annealing.
131
Fig. 5:
132
SEM micrographs of rough-type alumina fracture surfaces from sam
ples annealed at !50 (A, top) and 450 oc (B, bottom). The sample
annealed at the higher temperature shows more remaining metal parti
cles on the substrate surface.
Fig. 6: TEM cross-section images of samples with smooth-type substrate and
acetate type Ni(P) in the as-deposited state (A, top) and after annealing
at 150 oc (B, bottom).
133
Fig. 6:
134
TEM cross-section images of samples with smooth-type substrate and
acetate type Ni(P) (continued). C (top) and D (bottom) both of samples
after annealing at 580 oc.
5.3.3. XPS fracture surface analyses
With XPS the fracture surfaces were analysed of samples with rough-type
substrates, annealed at 150 and 450 "C in vacuum. The values of the peel energies
of these samples are given in fig. 1 (square symbols). The SEM micrographs of
the alumina fracture surfaces of these samples are depicted in fig. SA and B. For
the XPS analyses fresh fracture surfaces were prepared by peeling a small part of
the film in a glove box filled with N2, with less than 0.2 ppm 02 and H20. After
peeling, the Ni(P) and alumina fracture surfaces were transferred in a vacuum-tight
vessel into the XPS apparatus. The surface compositions of the Ni(P) and alumina
fracture surfaces of both samples are listed in Table 2. The relative accuracy of
the XPS relative coverages is within 10 %. The spot area during the XPS meas
urement was ea. 2 mm2, which means that the results are not influenced by inho
mogeneities with the size of a few micrometer.
Table 2: Relative atom concentrations(%) measured with XPS on the Ni(P) and the Ah03 fracture surfaces of samples with acetate-type Ni(P) on rough-type substrates after annealing at 150 oc and at 450 oc.
Surface T (OC) C Is 0 Is Ni 2p Al2p p 2p Sn 3d Ag 3d Pd 3d s 2p
Ah03 150 14.6 53.1 15.6 14.9 1.9 - - - -
Ab03 450 .14.6 52.6 14.5 11.5 6.8 - - - -Ni(P) 150 14.0 41.5 40.7 - 2.6 0.2 0.2 <0.1 0.8
Ni(P) 450 15.2 31.7 43.2 - 6.6 - <0.1 0 3.4
"-": below detection limit
All surfaces show a similar coverage with C, which is probably at least partly due
to organic contaminations in the XPS apparatus or during handling. For that
reason, the coverage with this element will not be discussed further. More re
markable is the relatively high coverage of the alumina fracture surfaces with Ni.
For both annealing temperatures the intensity of the Ni signal is in the same range
as that from AI from the substrate. After annealing at 150 oc the Ni/Al ratio is
1.05 and this changes slightly to 1.25 upon annealing at 450 °C. The P coverage
on alumina is considerably higher after annealing at the higher temperature. This
is also the case for the P coverage on the Ni(P) fracture surface after annealing at
the higher temperature. This points to enrichment of the interface with P, origi-
135
nating from the Ni(P) bulk. The oxygen content on the alumina surface does not
differ for the two temperatures and is probably determined by the oxidic bulk.
The activator material remains almost entirely on the Ni(P) fracture surface and
the coverage is lower after annealing at the higher temperature. Apart from P, S,
too, tends to segregate to the interface at higher temperatures as observed on the
Ni(P) fracture surface. This element probably originates from an additive in the
commercial electroless metallization solution. Another interesting observation is
the significantly lower amount of 0 after annealing at 450 oc. AI was not detected
on the Ni(P) side, which means that few or no alumina grains are detached from
the substrate surface during peeling.
An assignment of the peaks of the elements listed in Table 2 to compounds, ions
or molecules with relative amounts, obtained by multiscan measurements, is listed
in Table 3. The relative coverages are given in atom %. Reference data are used
from ref. (8).
Ni and P which are present on the alumina fracture surface after peeling are en
tirely (Ni) or for the greater part (P) in the oxidized state, for both annealing
temperatures. On the Ni(P) fracture surface Ni and P are for a greater part in the
metallic state after annealing at 450 oc than after annealing at 150 °C. At the
lower temperature the ratios Ni•+fNi0 and pn+fpo are 3.8 and 2.3 while at the higher
temperature these ratios are 1.4 and 1.5, respectively. Nickel- carbon compounds
were not formed at either temperature.
5.3.4. Static-SIMS measurements
The alumina and nickel fracture surfaces of the annealed and as-prepared samples
with smooth-type substrates were analyse.d with static-SIMS. These measurements
were carried out in the first place to see at what temperature the acetate molecules
would decompose and what the decomposition products would be. The charac
teristic acetate peak at mass/charge ratio m/z 59- from CH3COO- did not disap
pear, not even in the spectrum of the sample annealed at 580 oc. Since it is unlikely
that this molecule can withstand such a high temperature, it is concluded that the
136
Table 3: Assignment of exact XPS peak positions to the chemical environment of the same sample as measured in Table 2.
Element Position (eV) Relative amount(%) Environment
150 oc 450 oc
AhO> surface
c 284.8 90 90 -C-H
286.5 10 10 -C-O
0 531.0 lOO lOO Ab03
p 129.5 15 15 Ni(P)
132.5 85 85 P04
Ni 856.2 lOO lOO NhOJ, Ni(OH2), NiP04
AI 73.8 100 lOO AbOJ
Ni(P) surface
c 284.8 85 85 -C-H
288.5 15 15 -OcC 0
0 531.5 100 100 P04, Ni(OH)z or NizOJ
s 162.1 lOO 100 NiS
Sn 486.0 100 - Sn oxide
Ag 367.5 100 lOO Ag oxide
Pd 335.2 lOO - Pd metallic
p 133.5 70 60 -P04
129.5 30 40 NiP
Ni 852.5 21 42 metallic Ni, Ni(P)
856.2 79 58 Ni(OHh, NiP04, Nh03
59- peak originates from contaminations, despite the fact that peeling was done
in the vacuum of the apparatus. By introducing the sample, adsorbed contam
inations on the sample such as acetic acid which is present in the air, can be in
troduced into the vacuum chamber of the spectrometer.
137
The change in relative intensities of various inorganic fragments in the static-SIMS
spectra contained interesting information on the change in the composition of the
fracture surfaces. An overview of the most important results for the alumina
fracture surfaces is given in Table 4. The intensities are normalized to the most
intense peak from the substrate, which is AI+ for the positive-ion spectra and o~
for the negative-ion spectra. The intensity ratios listed in Table 4 are obtained
from two different positions on the fracture surfaces. For each measurement the
analysed area is about 250 pm2• The spread in results represents the spread in
surface composition. The accuracy of the relative static-SIMS intensities is of the
order of 10 %.
Table 4: Relative intensities in static-SIMS spectra from alumina fracture surfaces as a function of annealing temperature for samples with a smooth-type substrate and acetate-type Ni(P).
Rel. Int. Temperature COC) As prep. 200 450 580
Ni+fAl+ 0.157 0.219 1.026 8.338
Ni+fAl+ 0.162 0.2802 1.2417 4.095
Na+fA1+ 0.0690 0.0626 1.014 14.389
Na+fAl+ 0.0857 0.0521 1.161 4.762
POzto~ 0.095 0.0979 0.2018 0.489
POzto~ 0.0709 0.0852 0.166 -POrto~ 0.096 0.0788 0.1736 0.592
P03/0~ 0.0775 0.0731 0.146 -
F~to~ 0.1814 0.0658 0.0406 0.035 p~to~ 0.0973 0.0568 0.0400 -c1~1o~ 0.0166 0.0172 0.0746 0.109 c1~1o~ 0.0172 0.0153 0.0546 -
The positive and negative-ion spectra of the alumina fracture surfaces of an as
prepared sample and a sample annealed at 450 oc are shown in fig. 7. The 58Ni+ 1 Al+, POz I o~ and P03 I o~ intensity ratios in Table 4 show a strongly increasing
coverage of Ni and P containing compounds with increasing annealing temper
ature. The nucleation elements Sn, Ag and Pd were also detected on the nickel
fracture surface but the signal intensity of these elements was too weak for signif-
138
icant changes in relative coverag~s to be observed. The relative coverage ofF de
creases with increasing temperature while increasing relative intensities are
observed for Cl and Na. This may be associated either with diffusion and segre
gation or with a different fracture path, a point which will be discussed in greater
detail in section 5.4.
Fig. 7:
t
Na+ 450 oc
~NiOH/ Pb+ ,-A-, ,......_
AI+
I Ni2H/
K+ Ag+/
Si/ ~~~o+ 10x u+
\ e+ l po+ 100 150 200
~~ r J l. .I
0 50 100 150 200 mass (amu)---..
Static-SIMS spectra of the alumina fracture surfaces of samples with
rough-type substrates and acetate-type Ni(P) before and after
annealing. Peeling was done in the vacuum of the analyser. A linear
intensity scale is used. A (top) shows the positive-ion spectrum of an
as-prepared sample and B (bottom) the positive-ion spectrum of a
sample annealed at 450 "C.
139
Fig. 7:
t
OW
~.;~:~ w /
p- I BO-I I 2 \ / po-\ /
40 60
0 20 40 60
+ SiOxHyx AIOxHy-
25 oc
'! ,x 10x
100 120
450 oc
/Po3-
N;o- Ni02W
\ I 10x
80 100 120 140
80 100 120 140 mass (amu) ___..,....
Continued from last page. C (top) shows the negative-ion spectrum of
an as-prepared sample and D (bottom) the negative-ion spectrum of a
sample annealed at 450 oc.
In order to obtain more insight into the thermal behaviour of the organic com
plexing agents present at the interface, samples were prepared with a solution
containing glycine instead of acetate. The thermal stability of glycine may differ
from that of acetate, but this compound is much more suitable for investigation
with static-SIMS because it is not present as a contaminant in air. Moreover,
better alternative techniques for such an analysis are not available. The charac
teristic fragment of glycine is the NH2CH2C02 fragment at m/z 74- . The intensity
of this fragment relative to oxygen decreased by a factor of lO to 20 by annealing
at 300 and 400 oc, compared to the intensity ratio of an as-prepared sample. The
fact that the signal at m/z 74- did not completely disappear upon annealing for 1
140
hour at 400 oc indicates that the. degradation had still not been completed. Frag
ments indicative of degradation products were not found in these static-SIMS
spectra.
5.4. Discussion
5.4.1. Mechanical behaviour
- Energy balance
During peeling, energy is consumed by fracture (Gc) and possibly by bulk plastic
deformation of the film (Gder), while energy is supplied externally by peeling (Gp)
and internally by relaxation of residual stresses present in the film (G.,). Therefore,
the following energy balance is valid for the peel test (6):
[1]
All energy terms are per unit area. The fracture energy term Gc is made up of an
intrinsic contribution, which represents the energy required for breaking interfacial
bonds, and a dissipation contribution which is due to crack-tip plasticity (9). The
influence of each of these terms upon the peel energy GP will be considered in the
discussion which follows.
- Residual strain energy
In a previous investigation (6) it has been shown that built-in elastic strain energy
due to the deposition process itself does not play a significant role in the energy
balance. As shown in the discussion below, the internal strain energy due to ther
mal effects is more important. The amount of energy per unit area G.~, stored in
the metal layer at elevated temperatures owing to the difference in thermal ex
pansion coefficients Aa between the layer and the substrate, can be estimated with
eq. 2 (6),
[2]
141
where D is the metal layer thickness, E is the Young's modulus of the layer and
AT is the temperature difference with the deposition temperature. Ge1 is about
11 Jfm2 for the current layer thickness of 7 J.tm in the peel test with AT is 550 oc, E is 200 GPa (10) and Atx is 6. lQ-6 oc-1 (10, 11). This means that the layer on the
smooth-type sample can spontaneously debond at 580 oc because the thermal
elastic strain energy exceeds the initial fracture energy of 8.5 J/m2 (fig. 1). The fact
that this spontaneous debonding was not observed, can be explained by increased
intrinsic interfacial bonding, exceeding the build-up of thermal elastic strain en
ergy. From eq. 2 it also follows that no residual strain energy is present in the
Ni(P) I Ni layers after annealing, owing to differences in thermal expansion be
tween the metal layer and the substrate, since the temperature at which the peel
measurements are carried out (room temperature) is only slightly lower than the
deposition temperature of the galvanic Ni layer (50 °C). Only irreversible changes
in the metal layers such as plastic deformation due to thermal stresses during
annealing, or crystallization shrinkage of Ni(P), can cause residual stresses. The
elastic strain energy Ge., associated with such stresses is released upon debonding
and lowers the peel energy GP measured; see eq. l. Accordingly, these processes
cannot cause an increase in peel energy with annealing temperature. If such irre
versible processes, such as the crystallization of Ni(P), have introduced additional
residual strain energy Geh then the actual fracture energy Gc has increased more
than the measured increase in peel energy GP.
- Changes in plasticity
As we see in eq. 1, various other phenomena may contribute to the increase in peel
energy with increasing annealing temperature. Firstly, the intrinsic adhesion due
to interfacial interactions may have become greater, thereby increasing Gc. Sec
ondly, more energy may be dissipated by plastic deformation at the crack tip. This
also increases Gc . Thirdly, more energy dissipation may take place in the bulk of
the film during peeling (Gder). The hardness measurements showed a decreased
hardness of the galvanic Ni film with increasing annealing temperature so that
more plastic deformation can be expected to take place during peeling in the bulk
of the film, Gder· On the other hand, the hardness of the underlying electrolessly
deposited Ni(P) layer, relative to the as-deposited value (10) increases by a factor
of two upon annealing at 450 °C. Therefore, near the crack tip, a decreasing
142
amount of energy can probably be dissipated with increasing temperature (Gc).
Since two counteracting effects may play a role in the contribution of plastic de
formation to the peel energy, it is difficult to estimate the overall change in
plasticity during peeling as a consequence of thermal treatments.
Because of changes in plasticity described above, it is difficult to draw definitive
conclusions as to changes in interfacial bonding on the basis of these peel tests
only. However, the samples without galvanic Ni, on which a relatively thick
electroless Ni(P) layer was deposited, provided additional evidence for enhanced
interfacial interactions. As described in section 5.3.1, an increase in peel energy
with annealing temperature was measured for these samples in the limited tem
perature range of 20 to 200 oc. In addition, the results of the DPO tests, for which
similar samples were used, showed the same trend as the peel test results. Conse
quently, the increased DPO strength cannot be ascribed to increased plasticity of
the metal layer, because the Ni(P) metal layer becomes more brittle during
annealing below 500 °C. Hence, it must be concluded from the whole set of ad
hesion measurements that the improved adhesion is brought about by stronger
interfacial interactions.
- Intrinsic adhesion
For the intrinsic adhesion, mechanical interlocking effects should be distinguished
from chemical interfacial effects. The large differences between the peel energies
on smooth- and rough-type substrates are probably caused by differences in me
chanical interlocking. This interlocking is likely to take place in the open pores
near the surface of the rough-type substrate shown in fig. 4, but not for the
smooth-type substrate which has little porosity. However, the change in adhesion
on each substrate type due to thermal treatments cannot be explained by differ
ences in mechanical interlocking. Ni(P) does not flow into pores upon heating and
increasing brittleness of Ni(P) would rather lead to a smaller mechanical inter
locking contribution. It is therefore more likely that chemical interfacial inter
actions are enhanced by annealing. The effect of water upon the peel energy
(fig. 2) is consistent with this interpretation. Water has a greater effect on the peel
energy of samples which have been annealed at the higher temperatures than that
of the samples annealed at the lower temperatures. Water may promote the
143
breaking of interfacial chemical bonds, including Van der Waals inte~actions, but
it is not expected to influence mechanical adhesion.
5.4.2. Interface chemistry
For the samples with the rough-type substrates, the XPS analyses did not provide
evidence for a change in the fracture path. The amount of Ni detected with XPS
after fracture on the substrate surface was the same for samples annealed at high
and at low temperature. The amounts of aluminum on the substrate surface were
also constant. The XPS measurements indicated rather a change in the chemical
composition of the interface. On the metal fracture surface of the sample annealed
at the higher temperature, less oxygen and oxidized Ni and P were measured, and
the nucleation material had largely disappeared, probably by diffusion into the
metal bulk. Nevertheless, the SEM micrographs show a significantly larger cov
erage of torn metal pieces on the alumina substrate surface for the sample annealed
at 450 oc, compared with the sample annealed at 150 oc. Since the Ni(P) cannot
flow into substrate surface pores, it must be assumed that the mechanical inter
locking remained constant during annealing, and therefore that the higher cover
age of nickel particles indicates a stronger intrinsic adhesion. This higher metal
particle coverage does not become apparent in a higher XPS Ni coverage of the
alumina surface of the sample annealed at 450 oc.
According to the XPS measurements, the Ni coverage of the alumina fracture
surfaces is similar to the AI coverage, although the SEM micrographs show a metal
particle coverage of only a few percent at most, as visually estimated; see fig. 5.
This confirms the assignment of the Ni and P XPS coverage to the interfacial layer
a few nanometers thick, observed with TEM. The crack proceeds through this
interfacial layer, leaving behind a Ni-containing surface layer all over the alumina
fracture surface, which layer is far too thin to be observed with SEM. Only the
small amount of P which is assigned to Ni(P) in the multiscan XPS measurements
of the alumina surface can be explained by the metal particles. Hence, the fracture
for the samples with rough-type alumina proceeds mainly through the interfacial
layer and passes through the metal only at interlocking sites. The crack does not
enter the ceramic. Because of a stronger intrinsic adhesion for the samples
144
annealed at the higher temperature, it is more difficult to pull out metal from
interlocking sites and this may explain the higher density of metal particles.
The changes in relative coverage of Ni and P containing species with annealing
temperature, measured by static-SIMS on the alumina fracture surfaces of samples
with smooth-type substrates, do not seem to agree very well with the XPS results
obtained from the alumina fracture surfaces of rough-type substrates discussed
above. With static-SIMS an increasing Ni and P coverage was found with in
creasing annealing temperature, whereas with XPS this coverage was constant.
To avoid possible uncertainties in the interpretation of the relative static-SIMS
intensities, the smooth-type alumina fracture surfaces were therefore also measured
with EDX and the increasing coverage was confirmed. For the samples on which
Ni was found with EDX, it proved to be present all over the smooth-type
substrates, not on interlocking sites because such sites could not be discovered on
the smooth-type substrates. A possible explanation might have been that more Ni
particles remained on the smooth-type alumina at higher annealing temperature.
Such pieces were not found with SEM up to the highest magnification of 40 000
times. It is therefore more probable that the remaining Ni and P on the smooth
type substrates originates from the interfacial layer discussed before. This means
that an increasing fraction of the interfacial layer remains on these substrate with
increasing temperature. Because of the high surface sensitivity of the static-SIMS
technique, a Ni and P surface layer with a mean thickness of 1 or 2 nm can dom
inate the spectrum.
The stress which caused the extensive microfracture in the Ni(P) layer which was
observed with TEM, may have arisen as a consequence of lateral shrinkage of the
Ni(P) material during crystallization. During annealing the adhesion of the Ni(P)
layer to both the substrate and to the galvanic Ni layer was apparently stronger
than the cohesion because cracks along the interface were not observed. Moreover,
the brittleness of the Ni(P) material increases during crystallization thereby pro
moting microcracking. Additional stress is probably introduced into the Ni(P)
layer during annealing owing to thermal-expansion differences between the
galvanic Ni layer and the substrate. Despite the increased brittleness of the Ni(P)
145
phase and despite the microcracks in this layer, the fracture is found to proceed
through the interfacial layer at the metal - ceramic interface.
5.5. Conclusions
An improvement by a factor of 2 to 3 in the adhesion of electroless Ni(P) to
alumina is observed with both peel tests and DPO tests after annealing at tem
peratures above 250 oc. Fracture surface analyses with SEM I EDX, XPS and
static-SIMS show that, irrespective of the annealing treatment, fracture occurs
through an interfacial layer of a few nanometer thickness, observed with cross
section TEM. It is therefore concluded that the adhesion improvement is due to
stronger cohesion within this interfacial layer. With TEM, indications were ob
tained for densification of the interfacial layer by annealing. With static-SI MS and
XPS changes were observed in the chemical composition of the fracture surfaces.
The contribution of mechanical interlocking for the rough-type substrate cannot
be changed with heat treatment of the metallized samples. Therefore, the adhesion
improvement is entirely ascribed to a larger contribution by chemical interactions.
The same holds for the smooth-type substrates, for which evidence for mechanical
interlocking was not obtained at alL The greater effect of water on the fracture
energy of samples annealed at the higher temperatures is consistent with this ex
planation.
References
l. H.F. Fischmeister, G. Elssner, B. Gibbesch and W. Mader, Materials Re
search Society International Meeting on Advanced Materials Vol. 8, (1989),
227.
2. T.S. Oh, R.M. Cannon and R.O. Ritchie, Mat. Res. Soc. Symp. Proc. 130,
(1989), 219.
3. H. Honma and S. Mizushima, J. Met. Finish. Soc. Jpn. 33, (1982), 380.
146
4. T. Osaka, E. Nakajima, Y. Tamiya and I. Koiwa, J. Met. Finish. Soc. Jpn.
40, (1989), 573.
5. H. Honma and K. Kanemitsu, Plating and Surface Finishing 74(9), (1987),
62.
6. J.W. Severin, R. Hokke, H. van der Wel and G. de With, submitted to 1.
Appl. Phys. (1992) and Chapter 4, this thesis.
7. J.W. Severin, R. Hokke, H. van der Wel and G. de With, accepted for pub
lication in J. Electrochem. Soc. (1993) and Chapter 3, this thesis.
8. K.S. Rajam, S.R. Rajagopalan, M.S. Hegde, B. Viswanathan, Mater. Chem.
Phys. 27, (1991), 141, and National Institute of Standards.
9. D. Broek, Elementary Engineering Fracture Mechanics, Kluwer Academic
Publishers, Dordrecht, 1986, p. 14.
IO. W. Riedel in "Funktionelle Chemische Vemicklung", E.G. Leuze Verlag,
Sau1gau, 1989, Ch. 10.
11. E. Dorre and H. Hiibner in "Alumina, Processing, Properties and Applica
tions", Springer Verlag, Berlin, 1984.
147
148
Chapter 6
The influence of substrate chemistry on the adhesion of
electroless Ni(P) on metal-oxide coated ceramics
Summary
The adhesion of electrolessly deposited Ni(P) was studied using alumina
ceramic substrates which were covered with Si02, Sn02, Ti02, Ah03,
Y20 3, Zr02 and (In, Sn)O, (ITO) coatings. The adhesion was measured
with the 90° peel test. Strong adhesion of Ni(P) was found on the
substrates with the Zr02 and AhOJ coatings, weak adhesion on the
substrates with the Si02, Ti02, Sn02, Y 203 and ITO coatings.
Tbe fracture surfaces were analysed with scanning electron microscopy,
energy-dispersive analysis of X-rays and with X-ray photoelectron
spectroscopy in order to obtain information on the fracture path and the
type of interfacial bonding. For the strongly adhering samples, fracture
took place through the metal layer and along the interface. The sample
with the Sn02 substrate coating showed fracture through the Sn02 at low
peel energy. For the other weakly adhering samples only interfacial failure
was observed between the Ni(P) layer and the metal-oxide coating. The
differences in peel energy values are tentatively ascribed to differences in
micromechanical interlocking due to microporosity of the metal-oxide
substrate coatings. The coverage with nucleation material as measured
with X-ray fluorescence was significantly different for the various metal
oxide substrate coatings, but did not show a correlation with the peel en
ergy.
149
6.1. Introduction
In this chapter the results of an investigation into the dependence of the adhesion
of electrolessly deposited Ni(P) on the chemical nature of the substrate are de
scribed. Up to now, we investigated the adhesion of electrolessly deposited Ni(P)
using polycrystalline alumina substrates (1 to 5). From these studies, evidence has
been obtained that the adhesion is influenced not only by mechanical interactions
but also by interface-chemical interactions ( 4, 5), in contrast to the common
opinion (2). This was reported also by Osaka et al. (16). This implies that it should
be possible to influence the adhesion not only by changing the surface roughness,
but also by changing the chemical composition of the substrate surface.
This is done by using a number of metal-oxide coatings on polycrystalline alumina
ceramic substrates. In order to keep the roughness constant as a factor which in
fluences the adhesion, the same polycrystalline alumina ceramic substrates are used
in this study as in the previous studies. Two substrate types, with different
roughnesses are used. On these alumina substrates, various metal-oxide films are
vapour-deposited. The thickness of these films is small compared with the
roughness of the ceramic substrates. This ensures a constant mechanical contrib
ution to the adhesion which allows conclusions to be drawn on interfacial chemical
effects. As reference measurements, also uncoated alumina substrates were
metallized and analysed.
For the adhesion measurements the 90° peel test is used, which can provide infor
mation on the intrinsic interfacial interactions (4). Before and after the nucleation
treatment (1), the surface composition is quantitatively analysed with X-ray
fluorescence (XRF) in order to obtain information on the interface formation. The
fracture surfaces are analysed with scanning electron microscopy (SEM), energy
dispersive analysis of X-rays (EDX) and X-ray photoelectron spectroscopy (XPS)
in order to obtain information on the fracture path and the type of interfacial
bonding.
ISO
6.2. Experimental procedures
An overview of the approach followed in this study is presented in Table l. Details
on the various steps in Table I are described in the subsequent sections.
Table 1: Overview of successive procedures and analyses described in this section.
Step Procedure I analysis Remark
1 Substrate cleaning Organic solvents, glow discharge
2 Metal-oxide coating Evaporation: Si02, Zr02, Ah01, Sn02 Y203, Ti02
Sputtering: (In, Sn)Ox (ITO)
)I Annealing (A) 1 hour, 300 °C, air
4 Surface analysis XRF
5 Cleaning Detergent solution
6 Nucleation2 Sn, Ag, and Pd solutions
7 Surface analysis XRF
j8 Ni(P) electroless deposition pH 4.7, 65 oc
9 Galvanic Ni deposition Sulphamate bath, 50 oc
10 Peel test 90", rate 1 mm/min, in air
11 Annealing (B) 1 hour, 150 oc, air
12 Peel test 90°, rate 1 mm/min, in air
13 Fracture surface analyses SEM/EDX and XPS
1: From step 3 onwards also uncoated rough-type and smooth-type substrates were metallized and analysed, as reference samples.
2: For the sample with the ITO coating, step 6 was carried out twice due to slow and inhomogeneous initiation after the standard single nucleation procedure.
151
6.2.1. Sample preparation
- Substrates and metal-oxide coatings
For the sample preparation a rough-type 96 % alumina from Maruwa (Seto,
Japan) and a smooth-type 99.5 % alumina from MRC/Coors (USA) were used
as the substrates. An impression from the surface roughnesses can be obtained
from the SEM micrographs in figs. l to 5. All metal-oxide substrate coatings were
deposited by e-beam evaporation, except the ITO coating, which was deposited
by sputtering. The coating thickness was always about 0.1 pm, for the exact
thicknesses, see Table 3. Prior to the deposition of these coatings, the substrates
were cleaned by rinsing with ethanol and hexane and by glow discharge in the
vacuum-deposition apparatus.
The evaporation process was carried out with a Balzers BA510 apparatus,
equipped with an e-gun. Oxidic starting materials were used. The process pressure
varied from I0-6 mbar, which is the background pressure, to J0-4 mbar. The de
position rate was 30 nm/min and the substrate temperature was about 300 °C.
ITO was magnetron-sputtered using a Perkin Elmer 2400 system starting from a
In/Sn alloy (85/15 at. %). During sputtering Ar and 02 as the reactive gas were
introduced at flow rates of 120 and 30 standard cubic centimetres per minute ,
respectively. The background pressure and the process pressure were I0-6 and
7. 10-3 mbar, respectively. The deposition rate was 10 nm/min and the substrate
temperature was 300 °C. Before further processing, all coated substrates were
annealed for 1 hour at 300 oc in air in order to obtain stable, completely oxidized
coatings. This step is denoted by "Annealing (A)".
- Metallization
The two substrate types, provided with the various metal-oxide coatings, were
metallized using the following procedure: The samples were cleaned with a deter
gent solution, and nucleated by successive immersion in solutions containing Sn,
Ag and Pd ions. By electroless metallization, Ni(P) layers of about 0.3 p.m thick
ness were deposited. On top of the electrolessly deposited Ni(P) layers, galvanic
Ni layers were deposited from a low-stress sulphamate bath. The metal layer
thickness was about 7 pm. The galvanic Ni layer was only applied in order to fa-
152
cilitate adhesion measurements. For the peel test, a metal layer with sufficient
strength and stiffness is required. Details and backgrounds of the conditions of
these wet-chemical processes are given in refs. (l) and (4).
6.2.2. Analyses
The adhesion measurements were carried out with 90° peel tests as described in
ref. (4). This test was carried out before and after treating the metallized samples
for 1 hour at 150 "C in air. This annealing treatment is denoted by "Annealing
(B)". SEM I EDX and XPS analyses were carried out as described in the same
paper (4). The equipment and measuring conditions for the XRF analyses are de
scribed in ref. (17).
6.3. Results
6.3.1. Adbesion measurements
In Table 2 the results of the peel measurements are listed.
Table 2: Peel energy (Gp) values of Ni(P) I Ni bilayers on rough- and smoothtype alumina substrates, provided with various thin metal-oxide coatings. Peel tests were performed before and after annealing the metallized samples for 1 hour at 150 oc.
1 Metal oxide Rough-type substrate Smooth-type substrate
• substrate As-deposited Annealed (B) As-deposited Annealed (B)
coating Gp (J/m2) GP (Jim2) GP (J/m2) GP (J/m2)
Uncoated 22.7 15.7 3.7 2.3
· Zr02 55.9 >97.3 17.9 15.7
i Ah03 39.5 39.0 8.3 13.9
Ti02 24.1 26.3 7.9 3.1
Si02 24.1 17.2 4.8 3.1
Sn02 30.8 15.5 5.7 1.2
Y203 25.4 19.1 3.3 2.2
ITO - - - -Below detection limit of 0.5 Jjm2
!53
The peel energy measurements show that:
The peel energy values on the rough-type ceramic are considerably higher
than on the smooth-type ceramic. This is in agreement with previous results
(4) that mechanical interlocking plays an important role in the adhesion on
the rough-type substrates.
For both the rough- and the smooth-type surfaces and both before and after
annealing, the highest peel energy values were measured with the Zr01
substrate coating.
The peel energy of the Ni(P) I Ni layers on samples with ITO substrate
coatings is below our detection limit of ea. 0.5 Jjm2, for both substrate types.
There is virtually no adhesion.
The peel energy of as-deposited Ni(P) I Ni on the Sn02 surface is higher than
on the uncoated alumina, for both substrate types. After annealing the peel
value strongly decreases and visual inspection shows that the (yellow col
oured) Sn02 layer is peeled off the alumina ceramic surface.
The peel energy values of the Ni(P) f Ni layers on the other metal-oxide
surfaces are higher than on the uncoated alumina ceramic, both before and
after annealing B, for 1 hour at 150 oc, except for the samples with the ITO
coating.
In most cases the peel energy after annealing step B, for 1 hour at !50 oc, was lower than before annealing.
6.3.2. Analyses of surface composition
The chemical composition of the substrate surfaces was measured with XRF be
fore and after the wet-chemical cleaning and nucleation treatments. The thickness
of the substrate coating and the coverage by nucleation material were determined.
The analysed area is about 20 mm2• The relative accuracy of these measurements
is estimated to be within 10 % for the lower coverages (nucleation material) and
within a few % for the higher coverages (oxidic layers). The results are presented
in Table 3.
!54
Table 3: XRF analyses of the chemical composition of the coated and uncoated substrate surfaces before and after the wet-chemical cleaning and nucleation procedures.
Substrate Thickness (nm) Coverage (1015 atomsfcm2)
coating Sn Ag Pd Cl
Si02r.b 73 - -
Si02r.a 75 0.74 14.7 0.55
SiOz•· b 99 - - - -Si0z'·" lOO 0.90 10.1 0.73 8.0
Ti02r,b 100 - - -Ti0z'·• 98 2.6 15.6 1.9 11.0
TiOzs,b 96 - - -TiOz•·• 96 4.1 19.5 2.6 14.9
SnOzr,b 100 - I - -
Sn0z'·• 109 _I 1.6 ~ Sn02s,b 96 - - - -SnOz•·• 95 - I 4.5 0.85 4.2
! A}zQ3r,a _2 0.23 0.43 0.52 1.3
AbQ3s,a _2 0.19 0.47 0.33 1.5
Zr02r,b 74 - - -Zr02r,a 74 2.4 12.4 3.2 14.1
ZrOz•·b 65 - - - -
ZrOz•·• 66 1.9 8.2 2.5 9.9
Yz03r,b 68 - -Y203r,a <l 2.0 2.7 0.93 3.0
Y,o,,~ 67 - - -
Yz03s.a <I 1.2 1.4 0.49 1.1
• Uncoa - 1.0 2.2 0.61 2.0
ITO •·• I - 1.8 2.2 0.55 1.9
r: Rough-type substrate. s: Smooth-type substrate. b: Before wet-chemical cleaning and nucleation treatments. a: After wet-chemical cleaning and nucleation treatments. 1: Sn from nucleation cannot be distinguished from coating material. 2: Coating material cannot be distinguished from substrate material.
155
The XRF results show that:
The oxide layer is affected by the nucleation treatment only for the samples
with the Y20 3 substrate coating.
On the samples with the Ti02 and Zr02 coated substrates, high coverages of
Sn, Ag, Pd and Cl from the nucleation treatment are found, compared with
the uncoated substrate after nucleation. Especially the coverages with Ag and
Cl are very high.
On the sample with the Si02 coated substrate, very high Ag and Cl coverages
are found. For the other elements from the nucleation treatment, the cover
ages are comparable with the uncoated alumina substrate after nucleation.
The Ag and Pd coverages on the sample with the Sn02 coated substrate, are
in the same range as those of the reference sample. This is an indication that
Sn from the oxide layer does not play an important role in the nucleation.
The same holds for sample with the ITO substrate coating, which has been
submitted to the nucleation treatment twice in order to achieve quick and
homogeneous initiation of Ni(P) deposition. The difference in the thicknesses
of the Sn02 coating before and after nucleation is. too large to be ascribed to
deposition of Sn by the nucleation treatment. It is therefore probably due to
inhomogeneiety in the as-deposited Sn02 thickness.
The Sn and Ag coverages on the sample with the Ab03 coating are a factor
of 2 or more lower than for all other samples. The Pd and Cl coverages on
these samples too, are relatively low.
The coverage of nucleation material on the uncoated alumina substrate dif
fers less than 30 % from the values reported earlier on the same type of ce
ramic (1).
6.3.3. Fracture surface analyses
- SEM / EDX
With SEM and EDX the following observations were made on the structure and
the chemical composition of the fracture surfaces:
!56
An irregular type of surface roughness was observed. The metal- and
substrate fracture surfaces were covered with particles. With EDX Sn
ITO:
was detected on both the metal and the ceramic side. Apparently,
fracture took place through the Sn02 layer.
A high coverage of the substrate surfaces with metal particles was ob
served, both at grain boundaries and on grain surfaces. The metal
sides showed corresponding images. Zr was not detected on the metal
side. The surface structure of the metal-oxide film was visible. It was
copied into the metal fracture surface. The fracture path was along
interface and .through the metal layer (figs. 1 and 5).
Similar observations were made as for the samples with the Zr02
substrate coating.
Complete interfacial failure was observed, both by SEM and EDX.
Fracture took place along interface only (fig. 2).
Similar to the fracture surfaces of samples with ITO substrate coatings.
More remaining metal particles were observed at grain boundaries on
the substrate surfaces than for the uncovered reference substrates, but
less than for samples with the Zr02 and Ah03 coatings. Fracture took
mainly place along the interface and for a small part through the metal
layer.
Y was not detected with EDX on either side. The structures of the
fracture surfaces are very similar to those of the samples without
substrate coatings.
Uncoated: Interfacial failure was observed. In gaps between surface grains small
amounts of remaining metal was observed for samples with rough-type
substrates (fig. 4) but not for samples with smooth-type substrates
(fig. 3).
For each substrate coating a similar fracture behaviour is observed on the two
substrate types, except for a difference in mechanical interlocking which is not
observed on the smooth-type substrates. The substrate surface structure is very
similar for the coated and the uncoated substrates. Only a microroughness be
comes apparent for some of the substrate coatings, while the uncoated alumina
grains generally expose smooth crystal faces.
157
Fig. 1: SEM micrographs of a smooth-type substrate fracture surface after
peeling-off the metal film for the sample with a Zr02 substrate coating.
- XPS fracture surface analysis
With XPS the composition and the chemical state of the Ni(P) and metal-oxide
fracture surfaces were analysed for a few samples with strong and with weak ad
hesion. The samples investigated were those with Zr02 , Si02 and Ab03 coatings
and the one without metal-oxide coating, all with smooth-type substrates. Details
on the experimental conditions of these measurements are given in section 6.2.2.
The fracture surfaces were analysed after peeling-off in a dry nitrogen atmosphere
and after sputtering several nanometers deep, see Table 4. The sputter rate
amounted to 0.6 nm/min.
158
Fig. 2:
Fig. 3:
SEM micrograph of a smooth-type substrate fracture surface after
peeling-off the metal film for a sample with an ITO substrate coating.
SEM micrograph of a smooth-type substrate fracture surface after
peeling-off the metal film for a sample without substrate coating.
!59
Fig. 4:
Fig. 5:
160
SEM micrograph of a rough-type substrate fracture surface after
peeling-off the metal film for a sample without substrate coating.
SEM micrograph of a rough-type substrate fracture surface after
peeling-off the metal film for a sample with a Zr02 substrate coating.
Table 4: Chemical composition m atom % of peeled metal-oxide and Ni(P) fracture surfaces measured with XPS before and after sputtering for timet.
Surface t (min) Surface composition (at. %)
C1s Ols Ni2p P2p S2p Zr3d Al2p Si2p Sn3d5 Ag3d Pd3d5 C12p
Sample with Zr02 coating
Zr02 0 29 46 9.5 4.8 - 10.5 - - 0.3 0.1 - -170 - 63 2.8 0.4 - 23 11 - - - - .
iNi(P) 0 33 23 30 10.5 - - - 0.3 2.4 0.5 0.3
30 0.7 90.2 8.3 0.4 ! - - - . . - - -
I Sample with AhOJ coating
1Ah03 0 22 47 8.2 3.8 . 1- 18 - 0.1 . - -
190 - 59 11 0.8 - - 29 - - - - -
Ni(P) 0 39 28 29 1.3 0.3 - - - 0.2 0.4 0.1 -
30 - - 92.6 7.4 - . I· - 1- - - -
Sample with Si02 coating
Si02 0 23 51 10.9 1.6 - - 0.9 13 0.07 om 1- -
190 - 67 l.8 - - - 13 18 - 0.1 - -
•Ni(P) 0 38 23 24 ll 1.0 - . 1.3 0.2 0.7 -
30 - . 93 7.4 - . - . - 0.6 . .
Sample with blanco AbOJ ceramic substrate
AhOJ 0 19 52 6 1.6 - . 21 - 0.05 1- - .
190 . 64 0.6 - . 35 . - - - -
Ni(P) 0 38 39 16 6.3 0.4 1- - . 0.6 0.3 - -
30 . 0.7 91 8.0 . - - - 0.2 - - -
161
Table 5: Assignment of exact XPS peak positions to chemical compounds on both fracture surfaces for the sample with the Zr02 substrate coating.
Element Position ( e V) Relative amount C%) Environment
Ni(P) I Substrate Cls 284.8 m 286.2 8 6
288.4 10 - -0-C 0 -
Ols 530.1 - 35 NiO I Zr02
531.2 78 NiO,
531.5 - 61 org. -C-H-0
532.4 22 - idem
533.6 - 4 idem
Ni2p3 852.7 lOO 48 Ni, NiP,
856.5 - 52 NiO,
P2p 129.4 29 23 Ni(P)
130.2 41 - Ni(P)
132.9 30 - P03, P04
133.4 77 P03,P04
S2p 161.8 68 - NiS
163.1 32 - org. S
Zr3d5 182.1 - 100 Zr02
Cl2p 198.3 100 -
The following remarks can be made on the XPS results:
On the metal fracture surfaces, the Ni was always in the metallic state. This
means that oxidation did not significantly take place during handling after
peeling in the glove box filled with nitrogen. Therefore, the oxidation states
measured for the other fracture surfaces represent the situation at the inter
face (Table 5 to 8).
!62
I
I
I
Table 6: Assignment of exact XPS peak positions to chemical compounds on both fracture surfaces for the sample with the Ab03 substrate coating.
Element Position ( e V) Relative amount (%) Environment
Ni(P) Substrate
Cls 284.8 86 lOO -C-H
286.4 7 - -C-O
288.7 7 - -0-C 0
Ols 531.0 61 56 Ni oxide, Ah03
532.0 39 44 -C-O
Ni2p3 852.7 lOO 8 Ni, Ni(P)
854.1 - 92 Ni oxide
P2p 129.0 - 20 Ni(P)
129.3 - Ni(P)
129.9 - 7 Ni(P)
130.2 38 - Ni(P)
130.5 - 6 Ni(P)
133.2 31 6 P03, P04
133.8 - 61 P03, P04
S2p 161.9 63 - NiS
163.3 37 - so3; so4 165.8 - 100 C-S org.
Al2p 74.5 - lOO Ab03
On all substrate fracture surfaces oxidized Ni and P were detected, indicating
that the weak boundary layer from remaining bath compounds is still present
for all sample types. Another indication for the presence of remaining bath
components is the XPS signal ascribed to an organic sulphur containing
compound, which can only originate from the electroless deposition bath
(Table 5 to 8).
On the samples with high peel energy also metallic Ni and P were detected
on the substrate side, in the case of the Zr02 substrate coating this was even
50 % of the total Ni coverage (Table 5).
163
Table 7: Assignment of exact XPS peak positions to chemical compounds on both fracture surfaces for the sample with the Si02 substrate coating.
Element Position ( e V) Relative amount(%) Environment
Ni(P) Substrate
C1s 284.8 86 95 -C-H
286.2 8 5 -C-O
288.1 6 - -0-C=O
Ols 531.2 77 44 Ni oxide, -C-0
532.8 23 56 Si02
Ni2p3 852.7 lOO Ni, Ni(P)
856.7 - 100 Ni oxide
P2p 129.3 31 - Ni(P)
130.1 36 - Ni(P)
132.9 33 P03, P04
133.2 lOO P03, P04
S2p 162.2 69 - NiS
163.9 31 - 1-S-C
Si2p 102.0 - 9 SP+ I
103.2 91 Si02
1: Substoichiometric silicon dioxide
164
For all samples Sn, Ag, Pd and Cl, originating from the nucleation treatment,
were detected. The highest coverages of these elements were found on the
metal fracture surfaces for all samples analyzed (Table 4). Generally, Sn and
Ag were found in the oxidized state and Pd in the metallic state (not listed
in the Tables).
The ratio of the Ni coverage versus the coverage of Zr, Si, or AI substrate
material varied between 1/1 and 1/3 (Table 4). It did not show a correlation
with the peel energy.
The Si02 substrate coating was not completely oxidized by the annealing
treatment at 300 oc in air (Table 7).
For the sample with the Si02 substrate coating, a small Si coverage was
found on the Ni(P) fracture surface (Table 4).
Table 8: Assignment of exact XPS peak positions to chemical compounds on both fracture surfaces for the sample with the uncoated alumina substrate.
Element Position ( e V) Relative amount(%) onment
Subst
97
286.2 14 1 -C-O
288.4 8 2 -0-C=O
Ols 531.4 70 80 Ni oxide, Ab03
532.8 30 20 0-C-
Ni2p3 852.4 lOO Ni, Ni(P)
856.2 100 Ni oxide
P2p 129.1 21
130.0 31
133.4 48
25 iS
75 S-C-, S03/ S04
lOO Ah03
6.4. Discussion
The influence of various processing parameters upon the adhesion of electrolessly
deposited Ni(P) on alumina ceramics is reviewed in ref. (2). The conditions of the
etching treatment were found to be of much more influence on the adhesion
strength than the nucleation and metallization conditions. Therefore, it is the
general opinion of most researchers in the field that the adhesion is controlled by
mechanical interactions between the Ni(P) layer and the rough ceramic surface due
to mechanical interlocking. By etching, glass phase is removed from between sur
face alumina grains, resulting in a surface roughness with a characteristic dimen
sion of several micrometers. Only Osaka et al. (16) found evidence that other,
non-mechanical interfacial interactions contribute to the adhesion.
In a previous study (4) with cross-section TEM we discovered that an amorphous
layer l to 2 nm thick is present at the interface between the electroless Ni(P) layer
165
and the alumina ceramic substrate. Fracture surface analyses with XPS,
static-SIMS and SEM I EDX showed that the fracture always took place through
this layer, except at interlocking places. The interfacial layer consisted of all com
pounds present in the metallization solution and of nucleation material. The
cohesion of the material within this layer was therefore of decisive influence on the
adhesion of Ni(P) on smooth-type substrates where evidence for the occurrence
of mechanical interactions was not obtained. On these smooth-type substrates a
peel energy of 8.5 J/m2 was measured. On the rough-type substrates a much higher
peel energy of 44 J/m2 was measured, which is largely explained by mechanical
interactions. By annealing these samples at temperatures above 250 °C, a two- to
threefold higher peel energy was measured (5). Since the fracture path remained
through the interfacial layer, the improved adhesion was ascribed to stronger
cohesion within the layer after annealing.
By using a vacuum-deposited Pd nucleation layer, with an underlying Ti adhesion
promotor layer, excellent adhesion was found of Ni(P) on alumina ceramics.
Fracture took place cohesively in these systems (3). The strong adhesion was ex
plained by the absence of the amorphous interfacial layer as shown by cross
section TEM. This allowed strong interfacial metal - metal bonds to be formed.
The results from the previous investigations described above, demonstrate the im
portance of the interfacial layer for the adhesion. We suppose that this layer is
formed by incomplete displacement of the metallization solution from the
hydrophylic substrate surface by the newly formed Ni(P) metal phase at the initial
stages of the metallization process. After evaporation of water, the dissolved bath
components remain at the interface and prevent intimate contact on atomic scale
and thus chemical interactions between the metal layer and the substrate. If the
interface is formed similarly on the substrates used in the present study, the adhe-'
sion improvement for some of the samples can only be explained in terms of me
chanical interactions.
The oxidized Ni and P species, the nucleation material and the sulphur com
pounds, detected with XPS on fracture surfaces of all sample types, provide indi
cations for the presence of such an interfacial layer for all sample types, in the
166
same way as found in the previous studies ( 4, 5). This may imply that the improved
adhesion, found for some sample types, cannot be explained by enhanced chemieal
interfacial interactions, for example due to differences in acid-base properties of
the metal oxide surfaces (8, 18, 19). Therefore, the only alternative explanation
for the improved adhesion of some of the sample types, is the occurrence of more
efficient mechanical interactions. However, all coated substrates have a very simi
lar roughness, on a scale observable with SEM. If there are any differences, the
coated substrates are smoother rather than rougher, compared with the uncoated
substrates. Alternatively, roughness on a smaller scale than observable with SEM
may play a role.
It is well known (6, 10, 14, 20, 21) that e-beam evaporated metal-oxide films have
a porosity of 10 to 30 %, in spite of a deposition temperature of about 300 oc. The material grows in columns and open gaps remain between these columns (12).
If these gaps are wider than a few nanometres, Ni(P) is likely to penetrate and
form microscopic anchoring sites. The Ni(P) films generally follow the surface to
pography of the substrate on this scale (1, 4, 5). If the adhesion depends on the
microstructure of the metal-oxide films, this also explains why on the rough- and
the smooth-type substrates similar influences of the metal-oxide coatings have been
observed. This type of roughness was termed microroughness by Venables (15),
who observed adhesion improvement of polymer layers on phosphoric acid
anodized aluminium due to roughness structures of 5 to 10 nm.
The porosity of the metal-oxide substrate coatings may also account for the high
coverages with nucleation material, as listed in Table 3. The coverages correspond
to up to 25 monolayers of solid material, and are much higher than observed for
nucleation of alumina ceramic surfaces, both in ref. (1) and in this work, see also
Table 3. Vapour deposited metal-oxide coatings can absorb water (11), and
therefore nucleation material can penetrate into the coating. The analysis depth
of the XRF measurement is such, that all nucleation material, present in the
substrate coating, is detected in contrast with XPS, which only measures the out
ermost 3 nm. This can explain the fact that the amount of nucleation material
measured with XPS is of the order of 0.1 monolayer for both fracture surfaces
together.
167
The high coverage of the nucleation elements for substrates with the Si02 , Ti02,
and Zr02 coatings can be explained by a higher effective surface area due to the
coating porosity. The deposition of the nucleation material can be described by
the same processes as reported in ref. (l) for the non-porous alumina ceramic
surface. The high coverages of Ag and Cl, relative to the other elements used in
the nucleation treatments can possibly be explained by precipitation of AgCl, as
also suggested in (1). The AgCl precipitation may be enhanced by inefficient rins
ing due to the porosity. The valencies of Ag and Pd, determined by the XPS
multiscan measurements on the fracture surfaces, are in agreement with the pro
posed processes during nucleation.
With EDX Y is not detected on the metal and substrate fracture surfaces of the
samples with the Y203 substrate coatings. The XRF measurements before and after
nucleation show that Y already disappears in the nucleation treatment. This may
explain the observation that these samples behave very similarly to the samples
with the uncovered substrates, concerning the coverage with nucleation material,
the peel energy values and the structure of the fracture surfaces.
The results obtained with the ITO substrate coatings are remarkable. For both
substrate types the peel energy was extremely low, see Table 2. This is an indi
cation that both mechanical interactions and chemical interfacial interactions are
very weak. The initiation on the ITO coated substrates was slowly and the Ni(P)
coverage was incomplete after a standard single activation procedure. Therefore,
for the samples used in this study, the nucleation procedure (step 6 in Table 1) was
carried out twice and the coverages listed for ITO in Table 3 refer to this repeated
nucleation procedure. With this repeated procedure quick initiation was observed,
but still the adhesion is very poor. More detailed investigations are required be
fore explanations or even reasonable speculations can be formulated.
It should be noted that even for the samples with high peel energy values, the
metal-oxide substrate coatings are not broken away from the alumina ceramic.
This means that the adhesion of the metal-oxide coatings on the alumina ceramic
is generally very strong. Only in the case of the SnOz substrate coating, failure was
observed in the coating. However, this was cohesive failure in the Sn02 layer and
168
it was only observed after annealing at 150 oc. A possible explanation is that
moisture, present in the SnOz substrate coating after the wet-chemical deposition
of Ni(P), has degraded the mechanical properties of the film upon annealing
(6, 20).
The peel energy values reported in this work for the samples with the uncoated
rough-type and smooth-type substrates of 22 and 3.5 Jjm2 respectively, are con
siderably lower than the values of 44 and 8.5 J/m2 measured using the same
substrates in previous studies (4, 5). This difference is caused by a higher internal
stress in the galvanic Ni top layer, as confirmed by a measurement with a deposit
stress analyser. This does not have consequences for the relative differences be
tween peel energy values measured for various sample types in this study as they
are all made with the same galvanic Ni solution, and have the same internal stress
in the galvanic Ni top layer. The differences in peel energy between the various
sample types are consistent for the rough-type and smooth-type samples. More
over, the differences in the fracture paths for samples with low peel energy and
samples with high peel energy strongly suggest that these differences are caused
by differences in intrinsic adhesion, not by differences in bulk properties of the
metal layers.
The reason for the change in peel energy upon annealing the metallized samples
at 150 oc is not clear. Similar changes have been observed in a previous study to
the influence of thermal treatments upon the adhesion of Ni(P) on uncoated
smooth-type and rough-type substrates (4). Only at temperatures above 250 ac a
rigid increase in peel energy and adhesion was measured. A possible explanation
for the decrease in peel energy can be the evaporation of water in the porous
substrate coatings or in the interfacial layer and builds up pressure under the Ni(P)
films. However, there are some arguments against this assumption: Firstly, blisters
have not been observed, secondly, the change is sometimes different for rough-type
and smooth-type substrates, thirdly, the adhesion on the non-porous uncoated
alumina substrates also decreases upon annealing and finally, debonded areas have
never been observed in TEM- and SEM cross-sections. Therefore, it must be
concluded that this question remains unclear.
!69
The real widths of the interfaces between Ni(P) and metal-oxide, and between
metal-oxide and the alumina substrate is probably much smaller than found with
the XPS depth profiling measurements. The apparent width of an interface is in
creased by several effects: Firstly, with an information depth of 3 nm, the under
lying material is detected already when the real interface is still 3 nm deeper.
Secondly, on a rough surface the angle between the sputter beam and the surface
strongly varies with the position on the substrate. Consequently, the sputter rate
is not constant over the substrate surface (22). Thirdly, the sputter beam generally
has a different angle with the surface from the analyser. This causes shadow effects
on a rough surface. Fourthly, redeposition of sputtered material may take place
in valleys and gaps on rough surfaces. Finally, when the chemical composition of
a surface is laterally inhomogeneous, local differences in sputter rate may arise as
a consequence of preferential sputtering.
6.5. Final remarks
The peel energy of electrolessly deposited Ni(P) is strongly increased by using ea.
0.1 pm thick e-beam evaporated Zr02 and Ah03 coatings on alumina ceramics,
relative to the adhesion on uncoated alumina substrates. Other oxidic substrate
coatings such as e-beam evaporated Si02 and Ti02 do not lead to a strongly in
creased adhesion. On a sputtered ITO coating the adhesion was very weak, with
a peel energy below our detection limit of 0.5 Jjm2• The interfacial analyses do not
allow definitive conclusions on the differences in the adhesion mechanisms on
samples with and without these various coatings. The Y20 3 coating was not stable
during wet-chemical processing and for the Sn02 coating cohesive fracture took
place in the oxidic coating at low peel energy. A correlation between the peel en
ergy and the coverage of nucleation material or the coverage of Pd was not found.
A possible explanation for the increased peel energy for some of the substrate
coatings is the following: Micro-mechanical interlocking on the oxidic coatings
may have taken place on nanometer scale due to porosity of the substrate coatings.
However, for obtaining conclusive evidence on the mechanism of adhesion in these
170
systems a cross-section TEM study on these interfaces is required. Preparations for
this study are in progress.
References
1. J.W. Severin, H. van der Wel, R. Hokke and G. de With, accepted for pub
lication in J. Electrochem. Soc. (1992) and Chapter 3, this thesis.
2. J.W. Severin and G. de With, accepted for publication in J. Adhesion Sci.
Technol. (1992) and Chapter 2, this thesis.
3. J.W. Severin, H. van der Wel, R. Hokke, M. Johnson and G. de With, ac
cepted for publication in J. Electrochem. Soc. (1993) and Chapter 7, this
thesis.
4. J.W. Severin, R. Hokke, H. van der Wel and G. de With, submitted to J.
Appl. Phys. (1993) and Chapter 4, this thesis.
5. J.W. Severin, R. Hokke, H. van der Wel and G. de With, submitted to J.
Electrochem. Soc. (1993) and Chapter 5, this thesis.
6. G.G. Long, D.R. Black, A. Feldman, E.N. Farabaugh, R.D. Spal, D.K.
Tanaka and Z. Zhang, Thin Solid Films 217, (1992), 113.
7. K.L. Chopra, S. Major and D.K. Pandya, Thin Solid Films 102, (1983), 1.
8. G.A. Parks, Chemical Reviews 52, (1965), 177.
9. H.A. Macleod, J. Vac. Sci. Technol. A 4, (1986), 418.
10. A.G. Dirks and H.J. Leamy, Thin Solid Films 47, (1977), 219.
11. H.A. Macleod and D. Richmond, Thin Solid Films 37, (1976), 163.
12. H.A. Macleod, Optical Thin Films, SPIE Proceedings 325, (1982), 21.
13. K.H. Guenther, Optical Thin Films, SPIE Proceedings 346, (1982), 9.
14. M. Lottiaux, C. Boulesteix, G. Nihoul, F. Varnier, F. Flory, R. Galindo and
E. Pelletier, Thin Solid Films 170, (1989), 107. ' 15. J.D. Venables, J. Mater. Sci. 19, (1984), 2431.
16. T. Osaka, Y. Tamiya, K. Naito and K. Sakaguchi, J. Surf. Finish. Soc. Jpn.
40, (1989), 835.
17. D.K.G. de Boer, J.J.M. Borstrok, A.J.G. Leenaers, H.A. van Sprang and
P.N. Brouwers, submitted to X-Ray Spectrometry (1992).
18. G.A. Parks and P.L. de Bruyn, J. Phys. Chem. 66, (1962), 967.
171
19. F.M. Fowkes, J. Adhesion Sci. Tech. 1, (1987), 7.
20. A. Feldman, X. Ying and E. Farabaugh, Applied Optics 28, (1989), 5229.
21. E. Farabaugh, A. Feldman, J. Sun and Y.N. Sun, J. Vac. Sci. Technol. A
5, (1987), 1671.
22. A. Benninghoven, F.G. Riidenauer and H.W. Werner in "Secondary Ion
Mass Spectrometry", John Wiley & Sons, New York, 1987, p. 204.
172
Chapter 7
The adhesion of electrolessly deposited Ni(P) on alumina ce
ramic using a vacuum-deposited Ti- Pd nucleation layer.
Summary
The adhesion of electrolessly deposited Ni(P) on Al20 3 ceramic substrates
using sputtered and evaporated Ti - Pd nucleation films has been studied.
The adhesion was measured using the direct pull-off test and the 90° peel
test. The morphology and the chemical composition of the fracture sur
faces of the samples with evaporated Ti - Pd nucleation films were studied
with SEM / EDX and static-SIMS. Failure did not occur along the metal
- ceramic interface, but mainly in the alumina and therefore the strength
of the system is determined primarily by the substrate material. Cross
sectional TEM and HR-TEM were used to study the interface structure
before failure. The oxidation state of Ti at the interface was measured
with XPS. This was carried out in the (sub)monolayer range by using a
Ti wedge deposited on alumina with a maximum thickness of 0.35 nm. It
is concluded that the strong adhesion at the metal - ceramic interface is
caused by chemical bonding of the first Ti monolayer with substrate oxy
gen atoms.
173
7 .1. Introduction
Generally, the adhesion between electrolessly deposited Ni layers and non
conducting substrates is weak. Often this can be improved by increasing substrate
surface roughness, thus making use of mechanical interlocking (1). In some cases,
however, this is not possible or not sufficient. In this chapter a procedure is de
scribed to improve the adhesion of electroless Ni(P) layers on smooth substrates.
Usually Pd, required as a catalyst for the initiation of the electroless deposition,
is deposited on the substrate surface as nuclei from aqueous solutions, e.g. by
immersion in SnCl2 and PdC12 solutions (2, 3). However, a vacuum-deposited Pd
layer can also act as a catalyst for the initiation of this process (3, 4). Unfortu
nately, adhesion properties deteriorate with increasing nobility of the metal. The
stronger adhesion of base metals like Ti, AI and Cr, compared to semi-noble and
noble metals like Ni, Cu, Ag and Pd is ascribed to the tendency of the former
group to form chemical bonds with oxygen atoms of the substrate surface (5). For
a number of metals a proportionality relationship has been established between
oxidation potential and adhesion strength (6). The adhesion of noble metals is
often improved by alloying with base metals or by applying a thin interlayer of
such a metal (5, 7). Furthermore, the adhesion of electroless Ni is stronger on
metal substrates than on non-metallic substrates (8). A combination of these ob
servations leads to the idea of using stacks of metal layers, for instance the stack
shown in fig. I. The relatively thick electrodeposited Ni layer was applied on top
of the electroless Ni(P) layer in order to increase the strength and the stiffness of
the metal layer which was necessary for the adhesion measurements. For the de
position of Ti and Pd films several techniques are in use. The two common tech
niques applied here are magnetron sputtering and evaporation.
In order to obtain information on the nature of the chemical interaction at the
metal - ceramic interface, the Ti layer has been investigated with X-ray
photoelectron spectroscopy (XPS) during the initial stages of deposition. Similar
investigations to interpret adhesion phenomena have been reported for Ti on silica
and sapphire (9), AI on silica (10) and Al on polymers (11).
174
4 Galvanic Ni (several f.Jm)
3 Electroless Ni(P) (0.3 f.Jm)
2 Pd (20 nm) 1 Ti (20 nm)
Al20 3 ceramic
Fig. 1: Stack of metal layers used in this investigation.
At first sight it may not seem useful to combine a relatively expensive vacuum
deposition technique for the nucleation layer with the simple wet-chemical depo
sition of the nickel layer. However, for a number of applications electroless Ni(P)
has to be used because of specific properties required, such as high strength and
hardness, good wear resistance, oxidation and corrosion stability and resistance to
various chemicals (12). In such cases, where high specifications have to be met, a
vacuum-deposition technique for the nucleation layer can be a suitable solution for
obtaining strong adhesion.
7 .2. Experimental procedures
Metal layers were deposited on two types of ceramic substrates; on relatively rough
96 % alumina (HCT, Hoechst Rubalit 708), with an R. value of 0.3 /liD and on
relatively smooth 99.5 % alumina (MRC 996), with an R, value of 0.06 J1m. The
substrates were cleaned with a fluorinated alkylsulphonate detergent solution and
175
etched for 4 minutes in a 2.5% HF solution. The changes in chemical composition
of the surface due to these treatments is reported elsewhere (13).
The magnetron-sputtered Ti and Pd films were deposited under the following
conditions: ea. 20 nm Ti was sputtered at a rate of 0.1 nm/s for 200 s, immediately
followed by the deposition of ea. 20 nm Pd at a rate of 0.4 nm/s for 50 s. The
substrates were not heated during deposition. A background pressure of 10-5 mbar
was maintained, the process pressure being 2.5 10-2 mbar Ar. Ti and Pd films 20
nm thick were also deposited by evaporation using a similar procedure, both at a
rate of 0.5 nm/s and at room temperature. The background pressure was
2.5 10-6 mbar.
Electroless Ni(P) was deposited using a commercial Enlyte 512 bath (OMI) at a
temperature of 60 to 65 oc. For the electrodeposition of a low stress Ni layer a
sulphamate bath was used (8, 12). For the adhesion strength measurements Ni
layers of 2 to 4 fLID thickness were used, except for the evaporated Ti-Pd samples
(14-30 fLID). For the fracture energy measurements a Ni layer of 40 }liD thickness
was used in order to avoid rupture of the layer during peeling.
The adhesion was measured by a direct pull-off (DPO) adhesion strength meas
urement (14) and by a 90° peel test which provides information concerning the
fracture energy (15, 16). For the DPO test, aluminum pull studs with an epoxy
adhesive were bonded to the Ni layer at 150 oc for 1 hour in air. This heat treat
ment may affect adhesion, but is difficult to avoid. The samples for the peel test
received the same heat treatment before attaching the peel strip to the load cell.
The fracture surfaces of the peeled sample were analysed with scanning electron
microscopy (SEM) and energy dispersive analysis of X-rays (EDX).
Cross-section TEM micrographs were made using a Philips EM 400 transmission
electron microscope at an electron energy of 120 keV. For the high-resolution
cross-section TEM micrographs a Philips CM 30 microscope was used at an
electron energy of 300 keV. Samples were prepared by grinding, polishing and ion
milling as described in ref. (17). The equipment and measuring conditions for the
static secondary ion mass spectrometry (static-SIMS) surface analyses are de-
176
scribed in ref. (18). A reflectron type Time-of-Flight Static-SIMS apparatus from
IonTOF GmbH was used for the surface analysis of the first monomolecular layers
of the fracture surfaces. The mass resolution of the spectra was such (3000 - 5000
in the mass range from 20 to 150 amu) that peaks from the metal ions could be
separated from those of hydrocarbon ions of the same nominal mass.
For the XPS analyses of the Ti layer at the initial stages of deposition, MRC
substrates were used which were cleaned by sputtering at elevated temperature.
Then Ti was evaporated in a VG Semicon V80M MBE chamber and the XPS
analysis was carried out in a VG Scientific ESCA!ab using Mg Ka: radiation. A
Ti wedge of about 40 mm length and a maximum average thickness of 0.32 nm
was deposited by evaporation from a resistively heated Ti filament ( > 99.9 %
pure) at a rate of the order of 0.1 monolayer/min. This wedge was prepared and
analysed in a similar manner to the Fe/Cr wedgejFe (100) sample in ref. (19). A
moving shutter was used for the wedge preparation. The background pressure was
equal to the process pressure, being 10-10 mbar. The Ti layer thickness was deter
mined by analysis of the integrated AI 2s and Ti 2p312 XPS peak intensities. A lin
ear increase of the Ti thickness along the wedge was confirmed, suggesting that
Ti initially grows as relatively flat patches. The absolute accuracy in determining
the layer thickness is about 10 %, however, by using the wedge geometry, the rel
ative thicknesses are extremely well defined.
7 .3. Results
7.3.1. Adhesion measurements
- Direct pull-off tests
In Table 1 the results of the DPO adhesion strength measurements are listed for
the samples with sputtered and evaporated interlayers, prepared as shown in
fig. 1 and as described in section 7.2.
177
Table 1: Mean fracture strength ur (MPa), sample standard deviation sn-l (MPa) and number of samples N for samples with sputtered and evaporated Ti-Pd layers. MRC and HCT refer to the smooth and rough types of alumina respectively.
Sample type Substrate nr. Ur Sn-1 N Failure*
Sputtered MRC lA 55 8.5 33 Substrate
lB 55 6.0 32 Substrate
Evaporated HCT 2 76 4.1 6 Stud
* See text for explanation
The strength values listed for the samples with sputtered Ti-Pd (nrs. lA and B) are
lower than those with the evaporated Ti-Pd layers (nr. 2). However, this does not
mean that the adhesion strength is lower. In the case of the sputtered interlayers
(nrs. lA and B) failure took place by fracture of the substrate. Therefore it is only
possible to conclude that apparently the adhesion strength is higher than the values
measured in these tests. In order to overcome this problem, the substrates were
strengthened by bonding a thick rigid body on the back of the substrate for the
subsequent adhesion strength measurements of the samples with evaporated Ti-Pd
interlayers (nr. 2). In this case the substrate did not break, but failure always took
place in the adhesive with which the studs are bonded on the samples. Again, the
adhesion is apparently stronger than the values measured in these tests. Since this
did not provide extra information on the metal - ceramic interfacial strength, only
a few samples of this type have been measured.
-Peel tests
Three peel measurements were carried out. For practical reasons (sample size) this
test was only done with the evaporated Ti'-Pd nucleation layer. From two meas
urements on one part of the sample a reproducible peel energy of 226 J jm2 was
measured. From a third measurement on another part of the sample a value of
306 Jjm2 is obtained. This last measurement was done near to the edge of the
substrate where the layer thickness was 30 ,urn instead of the average value of
40 ,urn. In fig. 2 the peel profile is depicted with a corresponding peel energy of
306 Jfm2•
178
-.§ 200 z -CD
~ 100 LL
CD CD 0..
0 2 4 6 8
Displacement (mm) ...
Fig. 2: Peel profile of Ni I Ni(P) layer with evaporated Ti-Pd nucleation layer.
For reasons described above, only from the peeled samples (with evaporated
Ti-Pd) could the fracture surfaces be analysed. A variety of structures are visible
on these surfaces as shown in the SEM micrographs in figs. 3A to D. On the metal
side a large fraction of the surface is covered by individual alumina grains or by
larger ceramic pieces (fig. 3A). On the alumina side, ceramic-ceramic fracture sur
faces are seen (fig. 3B). This means that fracture took place mainly throughout the
ceramic. On the relatively smooth surfaces of grains which remained on the
substrate, a micro-roughness becomes apparent (fig. 3C). This is, however, a rela
tively small fraction of the whole fractured area. EDX analysis shows the presence
of Ni on these alumina grain surfaces. Ni is also detected with a stronger EDX
signal at the grain boundaries than on the grain surfaces. On the metal side
(fig. 3D), metal metal fracture is observed on sites corresponding to substrate
grain boundaries. Apparently, the amounts of Ti and Pd on these fracture surfaces
are below the EDX detection limit. In order to obtain more detailed information,
additional analyses were carried out with static-SIMS.
179
Fig. 3A: SEM fractograph of the metal side: most of the metal fracture surface
is covered by alumina grains (right hand side). On the left hand side
fracture occurred close to the metal-ceramic interface.
Fig. 3B: SEM fractograph of the alumina side: ceramic-ceramic fracture surface
at the upper part of the figure. At the lower part the top grains re
mained on the ceramic.
180
Fig. 3C: SEM fractograph of the alumina side: By EDX it is shown that the
roughness on- the alumina grains at least partly consists of Ni. More
Ni is present at grain boundaries.
Fig. 3D: SEM fractograph of the metal side: Metal fracture at positions corre
sponding with substrate grain boundaries.
181
7 .3.2. Interface chemistry analysis by static-SIMS measurements
In fig. 4A the positive-ion static-SIMS spectrum of the Al20 3 side of the peeled
interface (with evaporated Ti-Pd nucleation layer) is shown. The spectrum is
dominated by the AI signal at mass I charge ratio (m/z) 27. Smaller peaks from
Ni, Ti and Pd are also observed. The relatively high intensity of the peaks from
alkaline and alkaline earth metal ions is caused by their high ionization probabil
ity. The corresponding spectrum recorded from the metal side of the interface
(fig. 4B) is very similar to the former one, also with a dominating AI signal. This
is in agreement with the observation with SEM/EDX that fracture took mainly
place through the ceramic. The peaks that are not assigned are mostly due to alkyl
fragments, generally observed in such measurements and probably mainly origi
nating from the laboratory atmosphere. These fragments are present in the spectra
at m/z 15, 29, 39, 41, 55 and 57. The signal at m/z 39 is due to the hydrocarbon
ions C3Hj and to K+ in a 1:1 ratio. The signal at mjz 27 is mainly due to AI+. The
peak of C2Hj at mfz 27 has an intensity which is about 10 times lower than the
intensity of AI+.
Once it had become apparent from the SEM/EDX and static-SIMS measurements
that fracture took place cohesively, it was also clear that further mechanical char
acterization could not give any additional information on the interface. Therefore,
it was decided that TEM and XPS measurements were more appropriate to in
vestigate the nature of the metal - ceramic interface itself.
7.3.3. Interface structure from cross-sectional TEM
Cross-sectional TEM micrographs of the samples with sputtered and with evapo
rated nucleation films on both types of substrates have been made. In fig. 5A the
micrograph of the sample with the lough type alumina is shown. The images ob
tained from samples with sputtered and evaporated layers are similar. In both
cases a stack of layers is seen as schematically given in fig. I. The ceramic
substrates are well covered and no interface voids are observed. Intimate contact
is also observed for the other metal-metal interfaces. At the interface between Ti
and alumina (fig. 5A) a contrast is observed which could be assigned to an
interfacial layer of less than I nm thickness. However, a clear interfacial layer
182
could not be distinguished in the lattice image observed with HRTEM (fig. 5B)
on the same sample.
>-...... "(i.i c (1) ...... c (1) > ~ (1)
a:
Fig. 4:
15x Mg+
15x
0 20 40 60 80 100 120
Mass (amu) ..,.
Static-SIMS positive-ion spectra of alumina (A, top) and nickel (B,
bottom) fracture surfaces. A linear intensity scale is used.
183
Fig. 5A: Cross-sectional TEM micrograph of sample with evaporated Ti-Pd
nucleation on rough type ceramic.
7 .3.4. Interface formation studied with XPS analyses
Fig. 6 shows the XPS survey spectra, recorded before (6A) and after (6B) deposi
tion of the Ti wedge. Subsequently, detail spectra of the main Ti peaks from the
layer and the main AI and 0 peaks from the substrate were recorded at five posi
tions on the wedge, each at a different layer thickness. In Table 2 the peak posi
tions (± 0.3 eV) are listed for the five thicknesses. A rigid shift in the position of
the Al and 0 peaks is observed along the wedge. The mean shift values of the Al
and 0 peaks relative to the reference values (20), were used to correct the position
of the Ti peaks for the minor electric charging of the insulating sample. The elec
tric charging decreased with increasing layer thickness, probably due to improved
conduction of the thicker Ti layer.
184
Fig. 5B: Cross-sectional HRTEM micrograph of sample with evaporated Ti-Pd
nucleation on rough type ceramic.
185
t 012ol c:: :::~ Al(2s) .£ Al(2p) ~ C{1s) ;:: Ar(2p) '(ii
~ 0(1s) E Q)
> ·~ Q5 a:
0 200 400 600 800 1000
Binding Energy (eV) ---
Fig. 6: XPS survey spectra recorded from the alumina substrate before (A, top)
and after (B, bottom) Ti deposition.
Table 2: XPS binding energy peak positions (eV) of Al, 0 and Ti peaks at various Ti layer thicknesses T (nm).
No. 1 2 3 4 5 Ref. (20)
T (nm) 0.02 0.055 0.095 0.165 0.32 -AI 76.1 76.1 75.7 75.1 74.6 74.7 (Al20 3)
0 532.6 532H 532.3 532.0 531.6 531.6 (Al20 3)
Ti 459.9 459. 458.5 455.7 453.8 458.5 (Ti02), 453.8 (Ti)
l·u 458.8 458 457.7 455.3 453.8
* After correction for the shift in substrate signals due to electric charging.
The gradual change in the spectra of the Ti 2p312 and Ti 2p112 peaks as a function
of Ti layer thickness can be seen in fig. 7. The binding energy scale of each curve
has been corrected for electric charging as shown in Table 2. At the lowest cover
age the spectrum is characteristic of Ti02, while the spectrum of Ti at the highest
coverage is characteristic of metallic Ti (see the reference binding energies in
186
Table 2). At intermediate coverages' the transition can be followed as a change in
relative contribution of both these spectra. For reasons of clarity of presentation
the intensity of the spectra has been normalized to the most intense peak of Ti at
all coverages. This explains the relatively large amount of noise in the spectra for
the lower coverages. Since the increase of the Ti 2p312 peak intensity in the
submonolayer regime (as determined with the AI 2s intensities) was exactly linear
with the position along the wedge as measured in the first 4 spectra (fig. 7, spectra
1, 2, 3, and 4), it is concluded that the Ti grows in a close to "layer by layer" mode.
Fig. 7:
450 455 460 465 470 475 480 Binding energy (eV) .,.
XPS detail spectra of the Ti 2p112 and 2p112 peaks recorded from 5 places
on the Ti wedge: 0.02 nm (1), 0.055 nm (2), 0.095 nm (3), 0.165 nm (4)
and 0.32 nm (5). Spectrum 5 'was recorded from place 5 after 20 hours
exposure to the vacuum environment.
187
It has been established that oxidation of the Ti layer in the vacuum did not influ
ence the measurements. Even after 20 hours storage in this environment, only a
small change in the shape of the Ti XPS peaks is observed, see curve 5 and 5' in
fig. 7. The peak position did not significantly shift. The other spectra shown in
figs. 6 and 7 were recorded within 1 hour after deposition of the metal. From this
observation it can also be concluded that other reactions, e.g. with the substrate,
do not play a role on this time scale.
7 .4. Discussion
7 .4.1. Adhesion
In Table 3 a comparison is made for the adhesion obtained with a wet-chemical
nucleation (21) and with the Ti-Pd nucleation. In this Table it is shown that with
the Ti-Pd nucleation fracture takes place cohesively and at higher strengths and
at higher peel energies than in the case of the wet-chemical nucleation, where ad
hesive fracture occurred.
Table 3: Comparison of mechanical data on adhesion with Ti-Pd nucleation layer and with wet-chemical nucleation layer.
Substrate Nucleation Strength (MPa) Peel energy (J/m2) Fracture path
HCT Wet-chemical (21) 12 41 Interfacial
MRC Wet-chemical (21) 5 8.5 Interfacial
HCT Ti-Pd > 76 226 Cohesive
MRC Ti-Pd > 55 - Cohesive
In fig. 8 the fracture path through the stack of fig. 1 is schematically shown as
observed after the peel test. The dips in the ceramic represent grain boundaries.
Fracture takes mainly place through the ceramic substrate but also on some places
through the Ni(P) layer. The fracture through the ceramic takes place at a more
or less constant depth, relative to the interface. At the places where fracture takes
place near to the interface, small particles are observed on the grain surfaces. These
particles not only consist of Ti or Ti plus Pd, but also Ni and P are detected at
these places with EDX. On the grain surfaces, which are relatively smooth, the
l88
possibility of mechanical adhesion is excluded. Therefore, these areas show that the
strong adhesion must be due to chemical bonding at the interface.
Fig. 8:
Galvanic Ni
Electroless Ni (P)
Pd Ti
Schematical representation of fracture path through stack. The dips in
the ceramic surface represent grain boundaries.
For the samples prepared with wet-chemical deposition of the Pd nuclei, it has
been found that the peel energy is equal to the fracture energy (21). For those
samples, the peel energy did not depend on the layer thickness. However, for the
samples with the vacuum-deposited nucleation layer, the situation is somewhat
more complicated. Since the adhesion is stronger, peeling only starts at higher
loads. In order to avoid plastic deformation of the Ni film, a layer thickness of
about 40 p.m is chosen, instead of about 10 p.m. Still, from the difference in shape
of the onset and the end part of the peel curve it is concluded that at least some
bulk plastic deformation of the metal film has occurred. This may also explain the
dependence of peel energy on location, probably due to variation in layer thick
ness. An additional phenomenon which may have contributed to the large peel
energy, is the formation of many small cracks (pulverization) in the brittle alumina
layer which is peeled from the substrate.
189
7.4.2. Chemical bonding
The results from the XPS analyses on the Ti layer on the MRC ceramic at the in
itial stages of deposition show a similar trend to that reported by Chaug et al. (9)
who used a polished single crystal sapphire as substrate. However, a notable dif
ference seems to be present in the binding energies. Chaug et al. found a shift in
Ti binding energy relative to metallic Ti of 2.6 eV, in between the values reported
for TiO (2.1 eV) and Th03 (3.7 eV). Therefore, they concluded that the oxidation
state of Ti at the interface is between 2 + and 3 +. In this work a shift of 4. 7 eV
is found, corresponding with a 4 + oxidation state. A possible explanation can be
found in the layer thicknesses. Chaug et al. used layer thicknesses from 0.1 to
0.7 nm Ti on sapphire, whereas in the present case this was 0.02 to 0.35 nm, which
is about 5 times less on the thinnest place. At 0.095 nm the shift (fig. 6,
spectrum 2) has already decreased to 3.9 eV, relative to metallic Ti, close to the
value corresponding with Te+ and approaching the value reported in ref. (9), ob-
. tained at a similar layer thickness. A further reduction in Ti thickness to the ex
treme submonolayer regime as reported here, is required to encounter the 4+
oxidation state.
On increasing the Ti layer thickness above about 0.02 nm, the absence of sufficient
surface 0 makes it increasingly difficult for the additional Ti atoms to achieve the
4 + oxidation state. Consequently, lower oxidation states are encountered at lower
binding energies. For the thickest layers, metallic Ti is encountered (fig. 7, spec
trum 4, 5). At the higher binding energy side of both these Ti 2p peaks, satellite
peaks are observed the absolute intensity of which remains fairly constant with
increasing Ti layer thickness. This suggests that at all Ti thicknesses considered, a
strongly chemically bonded interface with the Ah03 is formed in the first Ti
monolayer.
The results of the XPS measurements also agree with the observation with
HRTEM, that no separate, structurally different interfacial reaction layer is
formed. An atomically sharp interface is present, although small lattice defor
mations may be present within 0.5 nm from the interface. It is worth noting in
addition that these TEM images were made of samples which were several months
190
old and which had been heated for I hour at 150 "C before the DPO test. This
means that the sharp interface is very stable under these conditions.
7 .5. Conclusions
Strong adhesion is obtained between electroless nickel layers and alumina
substrates by using vacuum-deposited Ti - Pd nucleation layers. In the peel tests,
cohesive failure takes place mainly in the alumina ceramic. Due to the high
interfacial strengths, cohesive failure occurred and with the DPO test only lower
limits of the adhesion strength could be obtained. No differences are observed in
the adhesion with sputtered and evaporated Ti - Pd nucleation films. From the
fracture surface analyses it is concluded that the strong adhesion is brought about
by interface chemical interactions.
At small coverages, XPS indicates that Ti exists in an oxidized state on the
alumina surface, most probably in the 4 + state. With increasing coverage a de
crease in the relative amount of oxidized Ti is found. At a few monolayers thick
ness mainly metallic Ti is measured, but most probably, the original oxidized Ti
layer is still present. The Ti layer is bonded to the alumina substrate by an inter
action of the first monolayer of Ti atoms with oxygen in the top layer of the
alumina substrate. It is concluded that this interaction is responsible for the strong
adhesion at the metal - ceramic interface. The other interfaces are strong metal -
metal interfaces.
References
1. J.W. Severin and G. de With, accepted for publication in J. Adhesion Sci.
Technol. (1993).
2. C.H. de Minjer and P.F.J. v.d. Boom, J. Electrochem. Soc. 120, (1973), 1644.
3. T. Osaka, I. Koiwa and L.G. Svendsen, J. Electrochem. Soc. 132, (1985),
2081.
191
4. L.G. Svendsen, T. Osaka and H. Sawai, J. Electrochem. Soc. 130, (1983),
2252.
5. D.M. Mattox, Thin Solid Films 18, (1973), 187.
6. J.T. Klomp in "Fundamentals of diffusion bonding", Y. Ishida ed., Elseviers
Science Publishers, (1987), p. 3.
7. K.B. Guy and D.M. Jacobson in "Designing Interfaces for Technological
applications", Peteves ed., Elsevier science publishers, England, (1988), 33.
8. E.P. Saubestre in "Modern Electroplating", F.A. Lowenheim ed., Wiley
Interscience, (1974), Ch. 28.
9. Y.S. Chaug, N.J. Chou and Y.H. Kim, J. Vac. Sci. Technol. A 5, (1987),
1288.
10. M.H. Hecht, R.P. Vasquez, F.J. Grunthaner, N. Zamani and J. Maserjian,
J. Appl. Phys. 57, (1985), 5256.
11. P. Marcus, presentation at Ecasia '91 conference in Budapest.
12. W.H. Safranek in "The properties of electrodeposited metals and alloys",
Elsevier, New York, (1974), Ch. 22, 464.
13. J.W. Severin, R. Hokke, H. van der Wel and G. de With, accepted for pub-
lication in J. Electrochem. Soc. (1993).
14. K.L. Mittal, Electrocomponent Science and Technology 3, (1976), 21.
15. J.E.E. Baglin, Mat. Res. Soc. Symp. Proc. 47, (1985), 3.
16. K.S. Kim, Mat. Res. Soc. Symp. Proc. 119, (1988), 31.
17. L.C. Feldman and J.W. Mayer in "Fundamentals of surface and thin film
analysis", Elsevier Science Publishers, New York (1986).
18. H. van der Wel, P.N.T. van Velzen, U. Jiirgens and A. Benninghoven in
"Analysis of Microelectronic Materials and Devicestt, M. Grasserbauer and
H.W. Werner, eds., 1991, John Wiley & Sons Ltd., Ch. 2.10.
19. S.T. Purcell, W. Folkerts, N.W.E. McGee, K. Jager, J. aan de Stegge, W.B.
Zeper, W. Hoving and P. Griinberg, Phys. Rev. Lett. 67, (1991), 903.
20. Handbook of X-Ray Photoelectron Spectroscopy, Physical Electronics Divi
sion, Perkin-Elmer Corporation, Eden Prairie, Minnesota, 1979.
21. J.W. Severin, R. Hokke, H. van der Wel and G. de With, submitted to J.
Appl. Phys.
192
Chapter 8
Final discussion, conclusions and outlook
Summary
In this chapter an assessment is made of what is achieved in the work
described in the previous chapters. To this end, first the aim of the inves
tigation is recalled and a summary is given on the current status of
knowledge in the literature. Subsequently, for the experimental chapters
the essential new findings and interpretations are presented. This progress
is compared with the original aim. Remaining unclear aspects are dis
cussed and suggestions for further research are made.
193
8.1. Aim and status of current knowledge
The aim of this work was to obtain insight into the mechanism of adhesion be
tween electrolessly deposited Ni(P) and alumina, and to find procedures to im
prove the adhesion. Since for many applications etching or abrasion is undesired
for technological or economical reasons, it is the aim to attain strong adhesion
without making use of surface roughness.
As described in the literature overview in Chapter 2, the conditions of the etching
treatment were found to be of much more influence upon the adhesion strength
than the nucleation- and metallization conditions. Therefore, it is the general
opinion of most researchers in the field that the adhesion is controlled by me
chanical interactions between the Ni(P) layer and the rough ceramic surface, owing
to mechanical interlocking. By etching, glass phase is removed between surface
alumina grains, resulting in a surface roughness with a characteristic dimension
of several micrometres. Procedures to obtain strong adhesion on smooth non-me
tallic substrates, without the application of an etching treatment, are not known.
Moreover, in the literature no information is available on the structure and
chemical composition of the interface and of the fracture surfaces on molecular
scale, which is of paramount importance for the adhesion. In the literature, only
adhesion strengths are measured by direct pull-off tests. These tests do not allow
definitive conclusions on intrinsic interfacial interactions such as chemical bonding
or mechanical interlocking, because the adhesion strength is governed by the size
of interfacial flaws as well.
8.2. New insights
In the study to the interface formation, described in Chapter 3, it became clear that
the nucleation material is not homogeneously distributed over the substrate sur
face, but rather a structure of islands and chains is present after the sensitization
treatment which remains essentially unchanged by subsequent nucleation steps. It
was also shown that Ni(P) deposition starts on these islands and chains, covering
the largest part of the substrate surface by lateral growth. With cross-section TEM,
the resulting interface was analysed. It was observed that an amorphous interfacial
194
layer of 1 to 2 nm is formed. Apart from this interfacial layer, the Ni(P) deposit
fits very closely to the irregular substrate surface, within a few nanometers.
Fracture surface analyses with XPS, static-SIMS and SEM j EDX showed that the
fracture always takes place through this layer, except at interlocking places.
Therefore, this interfacial layer can be considered as a typical example of a weak
boundary layer. The weak boundary layer consisted of all compounds present in
the metallization solution and of nucleation materiaL The cohesion of the material
within this layer was therefore of decisive influence on the adhesion of Ni(P) on
smooth-type substrates where evidence for the occurrence of mechanical inter
actions was not obtained.
By using both adhesion strength- and fracture energy measurements, the influence
of critical flaw sizes upon the adhesion could be distinguished from the influence
of intrinsic interfacial interactions. On the smooth-type substrates a peel energy
of 8.5 J/m2 was measured. On the rough-type substrates a much higher peel energy
of 44 Jfm2 was measured, which is mainly explained by mechanical interactions.
It is shown that the adhesion strengths as measured with the DPO test are strongly
influenced by the size of interface critical flaws.
By heat-treating the above described samples, a two- to threefold higher peel en
ergy and direct pull-off strength was measured. Since the fracture path remained
through the interfacial layer, the improved adhesion was ascribed to stronger
cohesion within the layer after annealing. Unfortunately, relatively high temper
atures of more than 250 oc are required to obtain adhesion inprovement, which
may not be practical for a number of applications.
The results from the previous investigations described above, demonstrate the im
portance of the interfacial layer for the adhesion. We suppose that the formation
of this layer can be described by the following model: Due to incomplete dis
placement of the metallization solution from the hydrophylic substrate surface by
the newly formed Ni(P) metal phase at the initial stages of the metallization pro
cess, a layer of metallization solution is present at the interface during metalliza
tion. After evaporation of water, the dissolved bath components remain at the
195
interface and prevent intimate contact on atomic scale and thus chemical inter
actions between the metal layer and the substrate.
By using a vacuum-deposited Pd nucleation layer, with an underlying Ti adhesion
promotor layer, excellent adhesion was found of Ni(P) on alumina ceramics.
Fracture took place cohesively in these systems at a peel energy of 200 to
300 Jfm2• For this type of system, the amorphous interfacial layer was not present
at the interface with Ni(P), as shown by cross-section TEM. This allowed strong
interfacial metal - metal bonds to be formed. The absence of the weak boundary
layer can be explained by the higher affinity of Ni(P) for the Pd surface, compared
to the alumina surface. For this system the chemical bonding of the Ti base metal
with the ceramic substrate is studied in-situ in a UHV system. It is found that Ti
in the first monolayer forms oxidic bonds with the alumina substrate.
A strong increase of the intrinsic interfacial interactions was achieved also by the
application of Zr02 and Al20 3 substrate coatings. The interfacial analyses carried
out with these systems do not yet allow definitive conclusions on the type of in
terfacial interactions. However, XPS fracture surface analyses provided indications
that the weak boundary layer is still present for these systems. The most probable
explanation for the adhesion improvement is the occurrence of micro-mechanical
interlocking on nanometre scale due to porosity of these vapour-deposited subs
trate coatings.
8.3. Suggestions for further work
Due to the formation of the weak boundary layer, further efforts to improve the
adhesion on hydrophylic substrates by changing the conditions of the electroless
metal deposition process are not considered worthwhile. All further research ef
forts for adhesion improvement should be aimed at eliminating the role of the
weak boundary layer. This insight also explains the observations reported in the
literature, that conditions of the nucleation treatment and the metallization do not
significantly influence the adhesion strength. It is obvious that the weak boundary
layer can hardly affect the mechanical interlocking.
196
The role of micromechanical interlocking for the systems with the metal-oxide
substrate coatings should be verified. If the proposed model is valid, the influence
of wet-chemically deposited metal-oxide substrate coatings upon the adhesion
could be studied, as a simpler and cheaper alternative for the vapour-deposited
metal-oxide coatings.
Furthermore, future investigations to extend the applicability of the Ti-Pd proce
dure to polymeric substrates will be useful. The adhesion of electrolessly deposited
Ni(P) on polymer surfaces is an even more complicated problem than for oxidic
surfaces, with an even wider range of applicability.
More insight in the DPO test would be beneficial for future adhesion studies in
general. In the literature, most attention has been paid to the stress distribution
at the interface with this test. As shown in this work, the gradual growth of critical
defects is determining for the strength in a number of cases. More insight in this
growth is required for a better interpretation of DPO test results.
197
198
Summary
This thesis deals with a study of the adhesion of electrolessly deposited Ni(P) on
alumina ceramic substrates. The aim of this study was to obtain insight into the
adhesion mechanism and to improve the adhesion, preferably without making use
of surface roughness. In the introductory Chapter l some general backgrounds and
principles of adhesion, fracture mechanics and electroless metallization are de
scribed. Furthermore, the aim of this study and the approach followed are out
lined.
In Chapter 2 a literature overview is given on the adhesion of Ni(P) on alumina.
It is generally found that the conditions of the etching pretreatment are of much
more influence than the conditions of the subsequent nucleation and metallization
treatments. Therefore, it is the common opinion of most researchers that the ad
hesion is determined by mechanical interactions, rather than by interfacial chemi
cal interactions.
In Chapter 3 the formation of the interface between electrolessly deposited Ni(P)
and the alumina ceramic substrate is studied. The changes in the structure and
chemical composition of the substrate surface are analysed on monolayer scale or
nanometre scale after each successive process step, using plan-view TEM, XRF
and static-SIMS. With help of literature data, a model for the interface formation
is presented. In addition, with cross-section TEM the structure of the resulting
interface is analysed. An amorphous interfacial layer of 1 to 2 nm thick is ob
served.
In Chapter 4 the adhesion is measured with 90° peel tests and direct pull-off tests
using two types of substrates with different roughnesses. The peel test provides
information on the fracture energy and the direct pull-off test measures the adhe
sion strength. Quantitative aspects of the adhesion measurements are considered
and the results are interpreted using fracture mechanics with the Griffith-Irwin
approach. Weibull analysis of the adhesion strength data indicate a monomodal
distribution of critical flaw sizes. On glass model substrates it is shown that the
size of these flaws gradually increases during testing. Interface and fracture sur-
199
face analyses show that the fracture path is through the interfacial layer described
above, which is therefore a weak boundary layer. For the rough-type lsubstrates
the adhesion is strongly influenced by mechanical interactions, whereas the adhe
sion for the smooth-type substrates is probably most strongly influenced by inter
facial interactions via the weak boundary layer.
In Chapters 5 to 7 procedures are investigated in order to improve the adhesion
which is obtained by the standard procedures as described in Chapter 4. In
Chapter 5 the influence of thermal treatments upon the adhesion is studied. A two
to threefold increase in the peel energy and the direct pull-off strength is found by
annealing at temperatures above 250 oc, both for the rough-type and for the
smooth-type substrates. An analysis is made to distinguish between changes in the
plasticity of the metal layer and changes in the intrinsic adhesion. Cross-section
TEM reveals that the interfacial layer remains present upon annealing. SEM/EDX,
XPS and static-SIMS fracture surface analyses show that fracture still occurred
through this layer. Therefore, it is concluded that the improvement of the adhesion
of heat-treated samples is caused by a stronger cohesion of the material within the
weak boundary layer.
In Chapter 6 the influence on the adhesion of the chemical composition of the
substrate surface is studied. Alumina ceramic substrates are provided with various
metal-oxide coatings. The adhesion of Ni(P), deposited on these surfaces using the
standard procedures, is measured with the peel test. Strong adhesion of Ni(P) is
found for rough-type and smooth-type samples with Zr02 and Al20 3 substrate
coatings. Other coatings, such as Si02 and Ti02, do not lead to a significant ad
hesion improvement. Since XPS analyses provide evidence that the weak boundary
layer is still present, the adhesion improvement is tentatively explained by micro
mechanical interlocking in the small pores, present in metal-oxide coatings.
In Chapter 7 the effect of an alternative nucleation procedure is investigated. In
stead of the conventional wet-chemical nucleation procedure, vacuum-deposited
Pd is applied, with an underlying Ti base metal layer for providing strong adhesion
between the Pd nucleation layer and the alumina substrate. Both in peel- and direct
pull-off tests fracture takes place cohesively, which means that maximum adhesion
200
is obtained. For this system the chemical bonding of the Ti base metal with the
ceramic substrate is studied in-situ in a UHV system. It is found that Ti in the first
monolayer forms oxidic bonds with the alumina substrate. The other interfaces
in these systems are formed by strong metal-metal bonds. Cross-section TEM
showed that no interfacial layers were present like in the case of the wet-chemical
nucleation.
This thesis concludes with Chapter 8, in which an assessment is made of the pro
gress with respect to the aims described above. The most important new insight is
that a weak boundary layer is formed between the metal and the substrate during
the interface formation if the conventional wet-chemical nucleation procedure is
used. This explains the large influence of surface roughness and the small influence
of nucleation- and metallization conditions upon the adhesion. Three procedures
are found for adhesion improvement: Firstly, by annealing at temperatures above
250 oc a two- to threefold increase in peel energy and DPO strength is attained.
Secondly, with the standard wet-chemical nucleation procedure, strong adhesion
is obtained by using a Zr02 vapour-deposited substrate coating. Thirdly, strong
adhesion of electrolessly deposited Ni(P) is obtained by using a vacuum-deposited
Ti-Pd nucleation layer. As a suggestion for further research, the influence of wet
chemically applied metal-oxide substrate coatings could be studied. This can be
an interesting option for strong adhesion of Ni(P) by a simple and cheap process.
All further research efforts for adhesion improvement should be aimed at elimi
nating the role of the weak boundary layer.
201
202
Samenvatting
Dit proefschrift gaat over een onderzoek naar de hechting van stroomloos afgezet
Ni(P) op aluminiumoxyde keramiek substraten. Het doel van dit onderzoek was
het verkrijgen van inzicht in de hechting en het verbeteren van de hechting, bij
voorkeur zonder gebruik te maken van oppervlakteruwheid. In het inleidende
Hoofdstuk 1 worden een aantal algemene achtergronden en principes van
hechting, breukmechanica en stroomloos metalliseren behandeld. Daarnaast
worden hier het doel van dit onderzoek en de gevolgde aanpak beschreven.
In Hoofdstuk 2 wordt een overzicht gegeven van wat er bekend is in de literatuur
over de hechting van Ni(P) op aluminiumoxyde keramiek. Over het algemeen
wordt gevonden dat de condities van de etsprocedure veel belangrijker zijn voor
de hechting dan de condities van de bekiemings- en metallisatie procedures. De
meeste onderzoekers zijn dan ook van mening dat de hechting eerder bepaald
wordt door mechanische interacties, dan door grensvlak-chemische interacties.
In hoofdstuk 3 wordt een onderzoek naar de vorming van het grensvlak tussen
stroomloos afgezet Ni(P) en aluminiumoxyde keramiek beschreven. De veran
deringen in de structuur en de chemische samenstelling van het substraatoppervlak
zijn geanalyseerd op nanometer- of monolaag schaal na ieder van de opeen
volgende processtappen, daarbij gebruik makend van transmissie electronen
microscopic (TEM), Rontgen fluorescentie (XRF) en statische secundaire ionen
massa spectrometrie (statische SIMS). Met behulp van literatuurgegevens wordt
in dit hoofdstuk een model voor de grensvlakvorming gepresenteerd. Daarnaast
is met TEM een doorsnee van het gevonnde grensvlak bestudeerd. Een amorfe
laag van 1 a 2 nm dik is waargenomen, tussen de Ni(P) laag en het substraat.
In Hoofdstuk 4 worden hechtingsmetingen beschreven met 90° pelproeven en
trekproeven, gebruik makend van relatief ruwe en gladde substraten. De pelproef
geeft infonnatie over de scheurenergie en de trekproef over de hechtsterkte.
Kwantitatieve aspecten van beide hechtingsmetingen worden beschouwd en voor
een breukmechanische interpretatie van de resultaten wordt de Griffith-lrwin
theorie gebruikt. Weibull analyse van de hechtsterkteresultaten levert een
203
monomodale verdeling van critische defectgroottes op. Met glas modelsubstraten
is aangetoond dat defecten geleidelijk groeien tijdens het belasten. Grensvlak- en
breukvlakanalyses tonen aan dat de breuk plaatsvindt door de eerder beschreven
tussenlaag van 1 a 2 nm dik, die daarom als zwakke grenslaag beschouwd kan
worden. Voor het ruwe type substraat wordt de hechting sterk be:invloed door
mechanische interacties, terwijl voor het gladde type substraat de hechting waar
schijnlijk het sterkst wordt be'invloed door grensvlakinteracties via de zwakke
grenslaag.
In de hoofdstukken 5 tot 7 worden procedures beschreven om een betere hechting
te verkrijgen dan met de standaard procedure die gebruikt is voor het onderzoek
beschreven in Hoofdstuk 4. In Hoofdstuk 5 wordt de invloed van temperatuurbe
handelingen op de hechting beschreven. Een twee- tot drievoudige toename in de
pelenergie en de hechtsterkte is gevonden door monsters te verwarmen bij tempe
raturen boven 250 oc, voor zowel het ruwe als het gladde type substraat. Een
analyse is gemaakt om na te gaan in hoeverre de hechtingsverbetering veroorzaakt
wordt door veranderingen in bulkeigenschappen van de metaallaag of door ver
anderingen aan het grensvlak. TEM analyse van een doorsnede van het grensvlak
toont aan dat de grenslaag nog steeds aanwezig is na de temperatuurbehan
delingen. SEM/EDX, XPS en statische-SIMS breukvlakanalyses laten zien dat de
breuk ook nog steeds plaatsvindt door de grenslaag. Daarom wordt geconcludeerd
dat de hechtingsverbetering door de temperatuurbehandelingen tot stand komt
door een sterkere cohesie van het materiaal in de zwakke grenslaag.
In Hoofdstuk 6 wordt de invloed van de oppervlakchemie van het substraat op
de hechting van stroomloos afgezet Ni(P) behandeld. De aluminiumoxyde
substraten werden eerst voorzien van diverse metaaloxyde lagen. De hechting van I
Ni(P), afgezet op deze oppervlakken met de standaard procedure werd gemeten
met de pelproef. Hoge pelenergiewaarden werden gemeten voor monsters met
Zr02 en Al20 3 lagen. Andere metaaloxyde lagen, zoals Si02 en Ti02 leidden niet
tot significante hechtingsverbetering. Omdat XPS breukvlakanalyses aanwijzingen
opleverden dat de zwakke grenslaag nog steeds aanwezig is voor deze diverse
monsters, wordt de hechtingsverbetering die voor sommige van de monsters ge-
204
vonden voorlopig toegeschreven aan micro-mechanische verankering in kleine
porieen in de aangebrachte metaaloxyde lagen.
In Hoofdstuk 7 wordt het effect op de hechting beschreven van een heel andere
bekiemingsmethode dan de standaardmethode. In plaats van de conventionele
nat-chemische methode, wordt Pd in dit onderzoek door middel van vacuiimde
positie aangebracht, met een onderliggende Ti laag voor sterke hechting tussen de
Pd kiemlaag en het aluminiumoxyde substraat. Zowel bij de pelproeven als bij de
trekproeven trad de breuk op in het substraat, in plaats van aan het grensvlak.
Dit betekent dat maximale hechting verkregen was. Voor dit systeem is de che
mische binding van het Ti met het keramische substraat in-situ bestudeerd in een
utra-hoog vacui.im apparaat. Met XPS werd aangetoond dat het Ti in de eerste
monolaag oxydische bindingen vormt met het substraat. Aan de andere grens
vlakken in dit systeem kunnen sterke metaal-m0taal bindingen worden gevormd.
Analyse van een doorsnede van het grensvlak met TEM toonde aan dat in dit
systeem geen grenslaag gevormd wordt zoals voor de monsters gemaakt met nat
chemische bekieming.
Dit proefschrift besluit met Hoofdstuk 8, waarin wordt nagegaan wat er bereikt
is in verhouding tot de oorspronkelijke doelstelling. Het belangrijkste nieuwe in
zicht is dat er een zwakke grenslaag gevormd wordt tussen de metaallaag en het
substraat tijdens de vorming van het grensvlak, indien de gebruikelijke nat
chemische bekiemingsprocedure toegepast wordt. Dit verklaart de grote invloed
op de hechting van de oppervlakteruwheid en de kleine invloed van bekiemings
en metallisatiecondities. Drie procedures zijn gevonden voor hechtingsverbetering:
In de eerste plaats, door temperatuurbehandelingen boven 250 "C wordt een twee
tot drievoudige verhoging van de pelenergie en de treksterkte bereikt. Ten tweede
wordt met de standaard nat-chemische bekiemingsprocedures sterke hechting ver
kregen door gebruik te maken van een Zr02 Iaag op het substraat. In de derde
plaats wordt zeer sterke hechting van stroomloos afgezet Ni(P) verkregen, met
cohesieve breuk, door gebruik te maken van vacuiimdepositie om Pd af te zetten
en door de Pd laag te laten voorafgaan door een Ti hechtlaag. Als suggestie voor
verder onderzoek zou de invloed van nat-chemisch aangebrachte metaaloxyde
lagen op de hechting onderzocht kunnen worden. Dit kan een interessante optie
205
zijn om sterke hechting te verkrijgen van Ni(P) door een eenvoudige voorbehan
deling van het substraat. Algemeen gesproken, zouden alle verdere onderzoeksin
spanningen gericht moeten zijn op het verkleinen van de rol van de zwakke
grenslaag.
206
List of symbols
a defect size acr critical defect size ~ energy loss factor
order number in failure probability estimation (Weibull analysis) m Weibull constant sN_, sample standard deviation sx standard deviation in the mean t time
A interface area D layer thickness E Young's modulus E' effective elastic modulus FP peel force Gc fracture energy per unit area Gcter energy dissipated by bulk plastic deformation, per unit debonded area G01 elastic strain energy per unit area Gi intrinsic fracture energy per unit debonded area GP peel energy per unit debonded area GP, energy dissipated at the crack tip, per unit debonded area H height of plastically deformed metal zone K Griffith-Irwin geometrical factor L peel length N number of test samples Pr failure probability RP peel radius R' P film radius after peeling T temperature ("C) U energy V volume W peel strip width wa work of adhesion
rx linear thermal expansion coefficient e strain eT thermal strain v Poisson's ratio ri stress rir fracture stress rij internal stress ri0 Weibull normalization constant riu Weibull threshold stress riy yield stress
All units are S.l. units unless specified.
207
List of abbreviations
AES amu AU CVD DPO EDX ESCA IC ITO m/z OM RBS SEM Static-SIMS STM (HR)TEM TOF WBL XPS XRF XRD
208
Auger electron spectroscopy Atomic mass unit Arbitrary units Chemical vapour deposition Direct pull-off Energy-dispersive X-ray analysis Electron spectroscopy for chemical analysis Integrated circuit Indium tin oxide Mass to charge ratio Optical microscopy Rutherford backscattering spectrometry Scanning electron microscopy Static secondary ion mass spectrometry Scanning tunneling microscopy (High-resolution) transmission electron microscopy Time-of-flight Weak boundary layer X-ray photoelectron spectroscopy X-ray fluorescence spectrometry X-ray diffraction
Dankwoord
Graag wil ik iedereen bedanken die heeft bijgedragen aan het totstand komen van
dit proefschrift. Op de eerste plaats dank ik mijn promotor prof. dr. G. de With
voor zijn stimulerende belangstelling, boeiende discussies en voor het kritisch
doorlezen en becommentarieren van manuscripten. Ook dank ik prof. Brongersma
die zich enthousiast bereid verklaarde om als tweede promotor op te treden. De
directie van het Nat.Lab. ben ik erkentelijk voor de gelegenheid die zij mij heeft
geboden om dit proefschrift te schrijven. De prettige sfeer in de groep Grensvlak
chemie en de kundige leiding van deze groep door Ties van Maaren hebben ook
bijgedragen aan de voltooiing van dit werk.
Robin Hokke en Mark van Weert wil ik bedanken voor hun prettige samen
werking en voor het nauwgezet uitvoeren van vele experimenten en analyses. Dank
gaat ook uit naar andere collega's binnen en buiten de groep, waarmee ik heb
samengewerkt. In het bijzonder wil ik mijn kamergenoot Hans van der Wel be
danken voor de buitengewoon plezierige samenwerking en voor de vele
static-SIMS analyses. De collega's die mijn manuscripten in de interne rond
zending van opbouwende kritiek hebben voorzien wil ik hiervoor danken. De vele
mensen in de diverse diensten van het Nat.Lab. dank ik voor hun bijdragen. Ook
wil ik de mensen van het CFT, PMF en Passieve Componenten bedimken voor
hun prettige samenwerking. In het bijzonder wil ik Jos Janssen bedanken voor zijn
belangstelling.
Mijn ouders dank ik voor hun stimulerende belangstelling en mijn vader daarnaast
voor de nuttige suggesties en het kritisch doorlezen van een aantal manuscripten.
Marijke ben ik meer dan dankbaar voor haar voortdurende morele steun en voor
het feit dat zij als vanzelfsprekend gedurende enkele jaren vrijwel de volledige zorg
voor Dirk, Anne en mij op zich genomen heeft. Zonder haar steun had ik dit
proefschrift nooit kunnen schrijven.
209
210
Curriculum vitae
Jan Severin werd op 25 december 1963 geboren in Geldrop. Hij behaalde zijn
einddiploma VWO-B aan het Hertog Jan College te Valkenswaard in 1982. In
datzelfde jaar begon hij met de studie scheikunde aan de Rijksuniversiteit Utrecht.
Het doctoraalexamen werd behaald in februari 1987. Het hoofdvak vaste-stof
chemie bestond uit een onderzoek naar Rontgenluminescentie van zeldzame-aard
verbindingen. De bijvakken, op het gebied van chemische technologie, volgde hij
aan de TU Delft. In 1987 begon hij een verkorte opleiding tot ingenieur aan de
TU Eindhoven. Hiervoor verrichtte hij als hoofdvak vanaf februari 1987 aan het
Philips Natuurkundig Laboratorium een onderzoek naar de keramische processing
van een nieuwe klasse van supergeleidende oxides. In december 1987 behaalde hij
de graad van ingenieur aan de TU Eindhoven. In januari 1988 trad hij in dienst
van het Philips Natuurkundig Laboratorium, waar hij in de groep Grensvlakche
mie onder meer onderzoek verrichtte naar de thermostabiliteit van chemische op
pervlakmodificaties en de hechting van stroomloos afgezet Ni(P). Dit laatste
onderwerp leidde tot dit proefschrift.
211
Stellingen
behorende bij het proefschrift
Adhesion of Electrolessly Deposited Ni(P) on Alumina Ceramic
door
J.W. Severin
De hechting van stroomloos afgezet Ni(P) wordt in sterke mate belnvloed door
de aanwezigheid van een zwakke grenslaag. Aangezien stroomloze metaaldepositie
op niet-metalen alleen goed initieert op hydrofiele oppervlakken, is de vorming van
de zwakke grenslaag inherent aan de aard van het proces.
(Dit proefschrift)
11
Hechtkracht is een verkeerd begrip om inzicht in hechting te krijgen. In
rusttoestand is de netto hechtkracht altijd gelijk 0. Goede parameters om hechting
te begrijpen en te meten zijn de kracht en de energie per oppervlakte eenheid die
nodig zijn voor breuk.
III
Het is niet mogelijk om, zoals Fowkes stelt, op grond van inzicht in chemische
interacties aan het grensvlak voorspellingen over de hechtsterkte te doen. Andere
factoren, zoals bulk mechanische eigenschappen van de laag en het substraat en
de structuur van het grensvlak hebben ook grote invloed op de hechtsterkte.
(F.M. Fowkes, J. Adhesion Sci. Tech. 1, (1987), 7)
IV
Een goed inzicht in hechting kan pas verkregen worden wanneer, naast
mechanische karakterisering door middel van hechtingsmetingen, ook aandacht
besteed wordt aan karakterisering van de chemische samenstelling en de
microstructuur van het grensvlak, zowel op moleculaire schaal als op micrometer
schaal.
V
De aanwezigheid van chemische bindingen tussen atomen afk:omstig van het
substraat en atomen afkonistig van organosilaan oppervlakgroepen zoals
gedetecteerd in statische SIMS metingen, vormt geen bewijs voor het werkelijk
voorkomen van die chemische bindingen aan het oppervlak van het monster.
(M. Gettings and A.J. Kinloch, J. Mater. Sci. 12, (1977), 2049)
VI
Naast silanolgroepen en silaandiolgroepen kunnen ook stabiele silaantriolgroepen
op een silica oppervlak aangebracht worden. Dit is mogelijk door middel van een
reactie van SiC!4 met geisoleerde silanolgroepen, gevolgd door hydrolyse.
(J.W. Severin en J.M.J. Vankan, Philips J. Research)
VII
De thermische stabiliteit van trialkylsilylgroepen op silica oppervlakken wordt
behalve door oxidatie of ontleding van de alkylgroepen, zoals algemeen wordt
aangenomen in de literatuur, ook in hoge mate bepaald door hydrolyse van de
binding tussen de organosilaangroep en het oppervlak.
(J.W. Severin et al., Surf. Interface Anal. 19 (1992), 133)
VIII
Bij het introduceren van nieuwe consumentenelectronica produkten liggen de
grootste technische problemen over het algemeen bij de materiaalkunde en de
procesbeheersing bij de fabricage.
IX
In tegenstelling tot individuele mensen en dieren, leren organisaties maar weinig
van hun eigen fouten.
X
Het democratisch gehalte van een samenleving kan gemeten worden aan de
toename van het begrotingstekort van de overheid v66r de verkiezingen en de toename van de belastingdruk mi de verkiezingen.