www.elsevier.com/locate/tsf

Thin Solid Films 473

Improved adhesion of Au thin films to SiOx/Si substrates

by dendrimer mediation

Xiao Lia, Feng Huanga, M. Currya,b, S.C. Streeta,b, M.L. Weavera,c,*

aCenter for Materials for Information Technology, The University of Alabama, 205 Bevill Building, Tuscaloosa, AL 35487-0209, USAbDepartment of Chemistry, The University of Alabama, 133 Lloyd Hall, Tuscaloosa, AL 35487-0336, USA

cDepartment of Metallurgical and Materials Engineering, The University of Alabama, A129 Bevill Building, Tuscaloosa, AL 35487-0202, USA

Received 2 January 2004; received in revised form 13 June 2004; accepted 29 July 2004

Available online 15 September 2004

Abstract

Quantitative evidence of significantly improved interfacial adhesion between Au films and SiOx/Si substrates induced by an organic

dendrimer monolayer was presented. For dendrimer-mediated Au films, nanoscratch tests revealed a critical load that was two times higher

than that for films without dendrimer mediation. Atomic force microscopy (AFM) examination of nanoindents revealed much constrained

lateral flow of metals in the dendrimer-mediated Au films during nanoindentation, indicating enhanced adhesion due to the presence of the

dendrimer layer.

D 2004 Elsevier B.V. All rights reserved.

Keywords: Adhesion; Dendrimer; Au film; Nanoscratch

1. Introduction

Poor adhesion between gold films and oxides is well

known [1,2]. Although Au adhesion to oxides can be

improved by introducing metallic interlayers, such as

tantalum or chromium [1], this method suffers from some

undesirable effects, in particular alloying and interdiffusion.

Recently, Baker et al. [3] showed that interfacial adhesion

between Au films and glass or Si substrates could be

improved by substituting a thin organic layer, such as

amine-terminated polyamidoamine (PAMAM) dendrimer

for the metallic underlayer [3]. Results from peel tests

revealed qualitatively improved adhesion by the PAMAM

dendrimer, with the most marked effect exhibited by the

Generation 8 (G8) dendrimer [3]. Introduction of the G8

dendrimer also influenced the nanoindentation response of

Au films. Street et al. [4] found that dendrimer-mediated Au

0040-6090/$ - see front matter D 2004 Elsevier B.V. All rights reserved.

doi:10.1016/j.tsf.2004.07.080

* Corresponding author. Department of Metallurgical and Materials

Engineering, The University of Alabama, Box 870202, Tuscaloosa, AL

35487-0202, USA. Tel.: +1 205 3487073; fax: +1 205 3482164.

E-mail address: mweaver@coe.eng.ua.edu (M.L. Weaver).

films on silicon wafers covered by native oxides (hereafter

indicated as SiOx/Si) were more resistant to indentation

penetration than dendrimer-free films. However, detailed

attempts to quantify the adhesion improvement via den-

drimer-mediation and related deformation characteristics are

still lacking.

This paper reports our recent investigations of dendrimer-

mediated adhesion of Au films to SiOx/Si substrates.

Experimental results obtained from nanoscratch tests,

scanning electronic microscopy (SEM), nanoindentation,

and atomic force microscopy (AFM) are presented.

2. Experimental details

Two inch (50.4 mm), (100)-oriented silicon wafers,

Si(100), with a layer of native oxides (SiOx) were used in

this study. The thickness of the SiOx layer was ~2 nm, as

previously determined by X-ray reflectivity (XRR [4]). To

investigate dendrimer-mediated adhesion, some of the SiOx/

Si wafers were covered by a self-assembled monolayer

(SAM) of G8 amine-terminated PAMAM dendrimer prior to

(2005) 164–168

Fig. 1. Ramped load (4 mN) nanoscratch surface displacement and friction

coefficient profiles for Au/D/SiOx/Si (a) and Au/SiOx/Si (b).

X. Li et al. / Thin Solid Films 473 (2005) 164–168 165

Au film deposition. To form the G8 SAM, wafers were first

cleaned by placing them in a freshly prepared piranha

solution (3:1 H2SO4/30% H2O2) for 1 h to remove organic

impurities. The clean wafers were then placed in a ground-

glass-sealed weigh bottle along with a 1 AM ethanolic

solution of the PAMAM dendrimer. The substrates were

allowed to remain in the solution for 12 h, where a self-

assembled monolayer of dendrimer molecules was absorbed

stably onto the surface. The substrates were then removed

and rinsed in pure ethanol followed by drying in a stream of

dry N2.

Au films were deposited by thermal evaporation using

one of two deposition systems, both of which utilize

resistive heating of W boats to achieve metal evaporation.

The first system, which can hold up to six wafers, is

cryopumped with a base pressure of 1�10�8 Torr. The

second system, which can accommodate one whole wafer,

has a base pressure of ~5�10�8 Torr. The film thickness

12.5 monitored in situ using a quartz crystal microbalance

and confirmed ex situ by XRR analysis [4]. Matched pairs

of Au on dendrimer-mediated and dendrimer-free substrates

were obtained from the same deposition system. The film/

substrate systems will be designated as Au/D/SiOx/Si and

Au/SiOx/Si, respectively.

Ramped load scratch tests were performed using a Nano

IndenterR II instrument equipped with a nanoscratch

attachment (MTS Systems, Oak Ridge, TN). A Berkovich

diamond tip oriented in the face-forward mode was applied

at the rate of 1 Am/s [5–7]. Each scratch test, 100 Am in

length, was performed by linearly increasing the normal

load from 0.02 mN to the desired level. Both the surface

profile and the friction coefficient were recorded in situ.

Surface profiles of the scratch before and after testing were

also obtained by scanning the tip at an extremely low load

(0.02 mN). The morphology of scratch tracks was also

examined ex situ by SEM.

Nanoindentation tests were performed on a Hysitron

TriboIndenterR system (Minneapolis, MN) using a Berko-

vich diamond tip. In addition to depth-sensing capability,

the TriboIndenterR system is also equipped with an AFM

attachment, which allows for in situ imaging of the

indentation area, using the same tip as a surface probe.

Generally, the AFM imaging is finished within 5 min after

the unloading.

3. Results and discussion

Fig. 1 shows two selected surface profiles collected

before, during, and after scratching Au/D/SiOx/Si (Fig. 1(a))

and Au/SiOx/Si films (Fig. 1(b)), respectively. The surface

displacement and the coefficient of friction l are plotted as a

function of scratch travel distance (and the applied normal

load). The displacements corresponding to the bduring-scratchQ profiles are greater than those corresponding to the

bafter-scratchQ profiles in that the former includes all elastic/

plastic contributions from the entire film/substrate system

[5], while the latter mainly reflects the plastic deformation

of the film. With an increase of the applied normal load, the

Berkovich tip gradually scratched through the sample. In

both cases, the profile drops suddenly at a well-defined

critical position/load, which suggests film failure [6]. These

critical loads (Lc) are marked by the vertical arrows in Figs.

1(a) and (b). The sudden drops at these critical points were

also reflected in the l profiles. The corresponding wear

tracks of the two scratches were further examined ex situ by

SEM, as shown in Fig. 2. Each track shows a clear critical

point, beyond which the track quickly expanded. These

critical points indicate the onset of film failure and match

well with those shown in the corresponding surface profiles

(Figs. 1(a) and (b)). With dendrimer mediation, film failure

was delayed (Fig. 2(a)). Accordingly, for the same scratch,

both the surface profiles (Fig. 1) and microscopic observa-

tions (Fig. 2) lead to comparable critical loads, which are

~3.5 and b1 mN in the systems with and without dendrimer

mediation, respectively.

Clearly, dendrimer mediation of the substrate has resulted

in a significant improvement in Au film adhesion (i.e., a

Fig. 3. Representative SEM track images for 10 mN ramped load scratches

for Au/D/SiOx/Si (a) and Au/SiOx/Si (b).

Fig. 2. SEM track images with normal load and travel distance scales for

scratches in Fig. 1: Au/D/SiOx/Si (a) and Au/SiOx/Si (b).

X. Li et al. / Thin Solid Films 473 (2005) 164–168166

greater than twofold increase in Lc). To the authors’

knowledge, no standard value of Lc has been reported for

Au films on SiOx. However, it is our contention that the

abovementioned ~3.5 mN Lc value for Au/D/SiOx/Si (Fig.

1(a)) should be taken as a lower limit. In 4 out of 10 scratch

tests run up to the 4 mN peak load, the samples did not fail.

This is an indication that the critical load was very near to

or slightly greater than 4 mN in some of the films. To

confirm this hypothesis, we therefore raised the peak load in

the subsequent scratch tests up to 10 mN. Testing under this

condition revealed average Lc values of 4.4F0.4 mN for

Au/D/SiOx/Si and 1.5F0.3 mN for Au/SiOx/Si, respec-

tively. The higher Lc values for 10 mN ramped load

scratches compared to 4 mN scratches can be attributed in

part to the larger scratch loading rate [8]. Fig. 3 presents the

corresponding SEM observation of selected sliding wear

tracks for films with and without dendrimer mediation.

These tracks exhibited excellent reproducibility in delaying

film failure via the dendrimer layer. Furthermore, we

believe the higher Lc value in dendrimer-mediated Au films

is really due to enhanced adhesion in that examination of

the wear tracks shown in Fig. 3 by ex situ AFM (not shown

here) suggested an interfacial adhesive failure mode.

Therefore, in terms of the Lc values, the dendrimer-

mediation-induced improvement in the adhesion of Au

films is very impressive.

The dendrimer-mediated adhesion also manifested itself

in nanoindentation tests. Besides the fact that dendrimer-

mediated Au films were more resistant to penetration [4],

the morphology of the indents also revealed the influence

of the dendrimer layer on the Au flow during nano-

indentation. Two representative plan-view AFM images of

indents, made at a 900 AN peak load, are shown in Figs.

4(a) and (b). This relatively high load (as compared with

that previously reported in Ref. [4]) was selected for the

convenience of AFM imaging. The maximum displace-

ment at the peak load is ~50 nm. The light-colored regions

immediately around the indents (the dark triangular region)

are the pileup of the material [9]. In the presence of the

dendrimer, the pileup is more uniform around the indent,

while the less-uniform deformation exhibited by the

dendrimer-free Au film might reflect some localized poor

adhesion or decohesion at the Au/SiOx interface [10]. The

black lines in the plan-view images, chosen to bisect the

triangular projections, indicate where typical cross-sec-

tional topographies (Figs. 4(c) and (d)) were traced. Some

of the experimental quantities involved in the following

analysis [11] were also defined in Figs. 4(c) and (d).

Although significant pileup around the indent can be seen

in both, a comparison revealed marked differences. The

pileup in Au/D/SiOx/Si is more pronounced (the maximum

pileup h is ~20% larger), while measurements of the

impression size and the radius of the plastic zone, a and c,

are ~25% smaller than in Au/SiOx/Si. Please note that

these trends were observed in all similar topographies.

These differences can only be rationalized by assuming

that during nanoindentation, the lateral flow of metals in

Au/D/SiOx/Si was more restrained than in Au/SiOx/Si; that

Fig. 4. Typical plan-view AFM images of indents (1.4�1.4 Am; 50 nm

vertical scale) and selected cross-sectional profiles for Au/D/SiOx/Si

(a and c) and Au/SiOx/Si (b and d).

X. Li et al. / Thin Solid Films 473 (2005) 164–168 167

is, this behavior is evidence of the improved adhesion in

Au/D/SiOx/Si.

Regarding the mechanism(s) underlying the adhesion

improvement, Baker et al. [3] and Tokuhisa et al. [12]

indicated that adhesion between dendrimers, Au and SiOx,

comes from the high level of multidentate interactions,

specifically the physical/chemical interactions between the

–NH2 functional terminal groups and the respective

layers. Each G8 PAMAM dendrimer molecule has a high

surface area and a dense exterior that contains 1024

terminal amine groups, which results in strong van der

Waals interaction with the Au and SiOx surfaces.

Furthermore, a large number of those end groups

chemisorb to the metal film and substrate surfaces which

stabilizes the amine/Au interaction and increases adhesion

between them and the dendrimer monolayer. As noted by

Baker et al. [3], these bonds were found to be stronger

than the adhesion of Au/SiOx.

In addition to physical and chemical adsorption,

mechanical interlocking is also expected to occur due to

geometrical matching of the deposited Au film with the

surface profile of the self-assembled monolayer, which

consists of an orderly arrangement of oblate-shaped

molecules [13]. Provided that the growing Au films

conform to the surface profile of dendrimer monolayer,

the geometric shape of the Au/D interface will inhibit Au

grains from sliding [14–16] or twisting [17] under applied

loads, which will increase their resistance to penetration

and/or failure.

4. Conclusions

Improved adhesion of Au films to SiOx/Si substrates due

to dendrimer mediation by a PAMAM monolayer was

characterized by a combination of techniques, including

nanoscratch, SEM, nanoindentation, and AFM. The critical

load determined by nanoscratch studies in the dendrimer-

mediated samples was approximately two times higher than

that dendrimer-free ones. This improved adhesion was

further confirmed by microscopic examination of the

characteristics of plastic flow during nanoindentation. It is

postulated that the improved adhesion is caused by strong

physical and chemical adsorption coupled with mechanical

interlocking between the Au film and the dendrimer

monolayer.

Acknowledgement

We gratefully acknowledge the financial support through

the NSF under award No. CMS-0324601 and shared

facilities from the Materials Research Science and Engineer-

ing Center (MRSEC) Program at UA through NSF-DMR

0213985.

References

[1] K.E. Haq, K.H. Behrndt, I. Kobin, J. Vac. Sci. Technol. A. 6 (1969)

148.

[2] M.A. George, Q.C. Bao, I.W. Sorensen, W.S. Glaunsinger, J. Vac. Sci.

Technol. A. Vacuum, Surfaces, and Films 8 (1990) 1491.

[3] L.A. Baker, F.P. Zamborini, L. Sun, R.M. Crooks, Anal. Chem. 71

(1999) 4403.

[4] S.C. Street, A. Rar, J.N. Zhou, W.J. Liu, J.A. Barnard, Chem. Mater.

13 (2001) 3669.

[5] T.W. Scharf, J.A. Barnard, Thin Solid Films 308–309 (1997)

340.

[6] H. Deng, T.W. Scharf, J.A. Barnard, J. Appl. Phys. 81 (1997)

5396.

[7] J.N. Zhou, A. Rar, D. Otte, J.A. Barnard, J. Appl. Phys. 88 (2000)

1880.

[8] P.A. Steinmann, H.E. Hintermann, J. Vac. Sci. Technol. A. 7 (1989)

2267.

[9] A. Gouldstone, H.-J. Koh, K.-Y. Zeng, A.E. Giannakopoulos, S.

Suresh, Acta Mater. 48 (2000) 2277.

[10] P. Tangyunyong, R.C. Thomas, J.E. Houston, T.A. Michalske, R.M.

Crooks, A.J. Howard, Phys. Rev. Lett. 71 (1993) 3319.

X. Li et al. / Thin Solid Films 473 (2005) 164–168168

[11] S. Harvey, H. Huang, S. Venkataraman, W.W. Gerberich, J. Mater.

Res. 8 (1993) 1291.

[12] H. Tokuhisa, M. Zhao, L.A. Baker, V.T. Phan, D.L. Dermody, M.E.

Garcia, R.F. Peez, R.M. Crooks, T.M. Mayer, J. Am. Chem. Soc. 120

(1998) 4492.

[13] V.V. Tsukruk, Adv. Mater. 10 (1998) 253.

[14] H.V. Swygenhoven, Science 296 (2002) 66.

[15] J. Schibtz, F.D. Di Tolla, K.W. Jacobsen, Nature 391 (1998) 561.

[16] J. Schibtz, K.W. Jacobsen, Science 301 (2003) 1357.

[17] I.A. Ovid’ko, Science 295 (2002) 2386.