8
Elementary surface processes during reactive magnetron sputtering of chromium Sascha Monje, Carles Corbella, and Achim von Keudell Citation: Journal of Applied Physics 118, 133301 (2015); doi: 10.1063/1.4932150 View online: http://dx.doi.org/10.1063/1.4932150 View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/118/13?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Nanopatterning of PMMA on insulating surfaces with various anticharging schemes using 30 keV electron beam lithography J. Vac. Sci. Technol. B 29, 06F304 (2011); 10.1116/1.3636367 Structure and optical properties of pulsed sputter deposited Cr x O y ∕ Cr ∕ Cr 2 O 3 solar selective coatings J. Appl. Phys. 103, 023507 (2008); 10.1063/1.2831364 Evolution of film temperature during magnetron sputtering J. Vac. Sci. Technol. A 24, 1083 (2006); 10.1116/1.2210947 Surface flattening processes of metal layer and their effect on transport properties of magnetic tunnel junctions with Al–N barrier J. Appl. Phys. 97, 10C920 (2005); 10.1063/1.1854452 Metallic tin reactive sputtering in a mixture Ar–O 2 : Comparison between an amplified and a classical magnetron discharge J. Vac. Sci. Technol. A 22, 1540 (2004); 10.1116/1.1759349 Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 134.147.160.151 On: Tue, 12 Apr 2016 09:54:19

Elementary surface processes during reactive magnetron

  • Upload
    others

  • View
    3

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Elementary surface processes during reactive magnetron

Elementary surface processes during reactive magnetron sputtering of chromiumSascha Monje, Carles Corbella, and Achim von Keudell Citation: Journal of Applied Physics 118, 133301 (2015); doi: 10.1063/1.4932150 View online: http://dx.doi.org/10.1063/1.4932150 View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/118/13?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Nanopatterning of PMMA on insulating surfaces with various anticharging schemes using 30 keV electron beamlithography J. Vac. Sci. Technol. B 29, 06F304 (2011); 10.1116/1.3636367 Structure and optical properties of pulsed sputter deposited Cr x O y ∕ Cr ∕ Cr 2 O 3 solar selective coatings J. Appl. Phys. 103, 023507 (2008); 10.1063/1.2831364 Evolution of film temperature during magnetron sputtering J. Vac. Sci. Technol. A 24, 1083 (2006); 10.1116/1.2210947 Surface flattening processes of metal layer and their effect on transport properties of magnetic tunnel junctionswith Al–N barrier J. Appl. Phys. 97, 10C920 (2005); 10.1063/1.1854452 Metallic tin reactive sputtering in a mixture Ar–O 2 : Comparison between an amplified and a classical magnetrondischarge J. Vac. Sci. Technol. A 22, 1540 (2004); 10.1116/1.1759349

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 134.147.160.151 On: Tue, 12 Apr

2016 09:54:19

Page 2: Elementary surface processes during reactive magnetron

Elementary surface processes during reactive magnetron sputteringof chromium

Sascha Monje, Carles Corbella,a) and Achim von KeudellResearch Group Reactive Plasmas, Ruhr-University Bochum, Universitystr. 150, 44801 Bochum, Germany

(Received 18 February 2015; accepted 20 September 2015; published online 2 October 2015)

The elementary surface processes occurring on chromium targets exposed to reactive plasmas have

been mimicked in beam experiments by using quantified fluxes of Ar ions (400–800 eV) and

oxygen atoms and molecules. For this, quartz crystal microbalances were previously coated with

Cr thin films by means of high-power pulsed magnetron sputtering. The measured growth and etch-

ing rates were fitted by flux balance equations, which provided sputter yields of around 0.05 for the

compound phase and a sticking coefficient of O2 of 0.38 on the bare Cr surface. Further fitted pa-

rameters were the oxygen implantation efficiency and the density of oxidation sites at the surface.

The increase in site density with a factor 4 at early phases of reactive sputtering is identified as a

relevant mechanism of Cr oxidation. This ion-enhanced oxygen uptake can be attributed to Cr sur-

face roughening and knock-on implantation of oxygen atoms deeper into the target. This work,

besides providing fundamental data to control oxidation state of Cr targets, shows that the extended

Berg’s model constitutes a robust set of rate equations suitable to describe reactive magnetron sput-

tering of metals. VC 2015 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4932150]

I. INTRODUCTION

Magnetron sputtering is a widespread technology for the

deposition of optical and ceramic thin films. Tailoring of ma-

terial properties, film homogeneity, and scalability is some of

the advantages of this plasma deposition technique.1,2 For

this, accurate control of surface state of the sputtering target

(composition and morphology) is mandatory. Important short-

comings observed in the deposition of metal oxides by reac-

tive magnetron sputtering are hysteresis and reduction in

growth rate due to target poisoning. In order to address the

poisoning issue, Berg et al. have modelled the basic sputtering

mechanisms with reactive gases in terms of sputter yields and

sticking coefficients.3,4 The flux balance equations include

processes, such as ion implantation, chemisorption, and sec-

ondary electron emission, which must be experimentally veri-

fied for each sputtering atmosphere and target material.

Previous studies on reactive sputtering of aluminium by

means of particle beam experiments showed that ion-

induced oxidation is a relevant process of aluminium poison-

ing.5,6 Quartz crystal microbalance (QCM) and Fourier

transform infrared spectroscopy (FTIR) diagnostics moni-

tored in-situ mass variation rates and chemical states on Al

target surface, respectively. An extended version of the

Berg’s model for the surface coverage was implemented,

showing that in this case, oxidation was enhanced preferen-

tially by knock-on implantation of oxygen into Al upon bom-

bardment with Ar ions. Further ion-induced oxidation

mechanisms considered in this model were surface activation

and electric field-driven oxidation. The suitability of this

model to other sputtered metals has not been tested so far

and it constitutes the objective of this article.

The present work investigates the basic surface reactions

during reactive sputtering of chromium, which is extensively

used as sputtering target in industrial applications. Effective

rates were measured in real time by sending quantified

beams of Ar ions and O/O2 species to a Cr-coated QCM in a

vacuum beam reactor. The experimental setup permitted to

explore the atomistic surface processes underwent by Cr tar-

gets at different oxygen partial pressures and ion energies

ranging between 400 and 800 eV.

The main contributions reported in this article are (1)

the description of surface processes during reactive sputter-

ing of Cr, especially the knock-on implantation and oxide

site density increase as mechanisms that enhance oxidation

and, therefore, induce target poisoning; and (2) the confirma-

tion that the surface coverage model applied in Al sputtering

also holds for Cr sputtering. Therefore, this work demon-

strates the versatility of this approach in elucidating funda-

mental surface processes controlling surface state during

reactive sputtering.

II. EXPERIMENTAL DETAILS

Cr thin films were deposited on QCM by high-power

pulsed magnetron sputtering (HPPMS). This deposition tech-

nique provides coatings with superior surface and mechani-

cal properties.7 Second, the Cr-coated QCM was bombarded

by argon ions and oxygen atoms and molecules in an ultra-

high-vacuum (UHV) particle beam reactor which is thor-

oughly described elsewhere.8

A. Cr film deposition

HPPMS depositions were carried out in a reactor

pumped down to a base pressure of 3� 10�3 Pa and with a

distance magnetron-substrate of 10 cm. Cr films of 500 nm in

thickness were deposited with a pressure of 0.48 Pa using a

a)Author to whom correspondence should be addressed. Electronic mail:

[email protected].

0021-8979/2015/118(13)/133301/7/$30.00 VC 2015 AIP Publishing LLC118, 133301-1

JOURNAL OF APPLIED PHYSICS 118, 133301 (2015)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 134.147.160.151 On: Tue, 12 Apr

2016 09:54:19

Page 3: Elementary surface processes during reactive magnetron

pulsed signal of 560 V and 16 A with a frequency of 60 Hz

and pulse time of 200 ls. The topography of the films was

characterized by means of a LEO (Zeiss) 1530 Gemini field

emission scanning electron microscope (FESEM) and a

Keyence VK-9700 laser scanning microscope (Fig. 1). Film

composition was measured with an OXFORD AZtecEnergy

electron dispersive X-ray microanalysis (EDX) system. The

deposited Cr films contained around 5 at.% O as character-

ized by EDX.

B. Beam treatments

The UHV reactor comprised a load-lock chamber and a

beam chamber, where the particle sources were focused to a

target sample. An ECR plasma source (Gen2 from TecTra

GmbH) with a double-grid ion optics provided Arþ ion

beams orthogonal to the sample at energies from 400 to

800 eV and fluxes comprised between 5� 1013 and

1.4� 1015 cm�2s�1. The aperture of this ion gun is located

92 mm away from the sample. With a base pressure better

than 5� 10�6 Pa, the working pressure was around

2� 10�2 Pa. Therefore, the mean free path of the beam spe-

cies in the reactor was longer than the source-sample distan-

ces. The O/O2 species were produced with a hot capillary

oxygen beam source (OBS from Dr. Eberl MBE-

Komponenten GmbH) at a distance of 80 mm and oriented

45� with respect to the sample. Its operation at temperatures

up to 1800 �C provided oxygen beams with a dissociation

degree of approximately 15%. Then, the flux reaching the

target surface is basically formed by oxygen due to direct

beam and diffuse background, which is formed by the colli-

sion flux of particles with thermal velocities profile.5 Thus,

the total incident flux of oxygen atoms and molecules onto

the sample surface, jO2, is the sum of the direct fluxes,

jO2,dirþ jO,dir (tabulated by the fabricant of OBS) and the flux

from the gas background, jO2,back, which is estimated from

gas kinetics theory

jO2;back ¼pO2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2pmO2jBT

p ; (1)

where pO2 is the oxygen partial pressure, mO2 is the molecu-

lar mass of O2, jB is the Boltzmann’s constant, and T is the

gas temperature. It is worth noting that the diffuse wall flux

constituted by background particles is one order of magni-

tude higher than the direct fluxes, as reported by Kuschel

and von Keudell.5 Therefore, the dominant contribution to

oxygen flux comes from the diffuse background.

In-situ measurements of mass variation rates were car-

ried out with water-cooled QCM (Maxtek BDS-250) consist-

ing of circular slabs of AT-cut quartz resonators with 0.5 mm

thickness and 14 mm diameter. The piezoelectric crystal,

which is coated with Al electrodes, showed a resonant fre-

quency of 6 MHz. The active site of such crystal was previ-

ously coated with Cr as indicated in Sec. II A, and it

constituted the target sample in the beam experiments. The

control unit consisted of an SQC-310 Series Deposition

Controller from Inficon. For kinetic measurements of

the effective sticking coefficient of oxygen on Cr surface, the

adsorption rate of oxygen is obtained by measuring the

growth rate of oxide layer on a QCM at a constant oxygen

flux. Before each experiment, the target surface was pre-

sputtered with Ar ions (800 eV) during several minutes to

remove oxide and contamination layers in order to provide a

clean metallic surface.

III. RESULTS AND DISCUSSION

A. Modeling

The experimental data were fitted using a surface cover-

age model based on two coupled rate equations. There, cov-

erage fraction H is defined as the ratio of oxidized surface

sites over total site density, n0. Hence, an oxide-free metal

surface corresponds to H¼ 0, whereas a fully oxidized sur-

face corresponds to H¼ 1. On one side, the effective rate of

Cr during reactive sputtering, C, which is directly measured

with a QCM, is defined as the net flux of particles being de-

posited on or etched from the target. To obtain this rate, first,

growth/etching rate was measured with QCM. Such mass

rates are converted to particle rate or incident flux. The effec-

tive incidence rate is determined from the balance of O2

adsorption on Cr (positive term) and etching of Cr and chro-

mium oxide (negative terms)5

C ¼ 2jO2sO2;0ð1�HÞ2 � jArYMð1�HÞ � jArYOH; (2)

FIG. 1. SEM micrographs showing (a) uncoated (clamp shadow) and coated

parts of the QCM after Cr deposition and (b) the surface topography of Cr

film deposited onto QCM by HPPMS. RMS surface roughness is �1 lm.

The profile was measured with the Keyence laser scanner microscope.

133301-2 Monje, Corbella, and von Keudell J. Appl. Phys. 118, 133301 (2015)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 134.147.160.151 On: Tue, 12 Apr

2016 09:54:19

Page 4: Elementary surface processes during reactive magnetron

where H is the coverage, jO2 is the oxygen molecule flux,

sO2,0 is the sticking coefficient in the zero coverage limit, jAr is

the flux of incident Ar ions, YM is the sputter yield of the metal

(Cr), and YO is the sputter yield of oxygen. Both sticking coef-

ficients and sputtering yields of oxygen atoms were calculated

with error bars of 65� 10�3. Sputter yield of Cr at different

ion energies was taken from TRIM (Transport of Ions in

Matter) simulations of Cr bombardment with argon ions.

Best fittings were performed with term (1 -H)2 instead

of using the 1 -H approach typical in the flux balance equa-

tions in reactive sputtering theory. The (1 -H)2 term accounts

for the dissociative chemisorption of O2 on Cr, an oxidation

mechanism that has been also validated in the case of Al

sputtering.5

The coverage H is necessary to fit C from Eq. (2) and is

obtained from the second rate equation, where the coverage

rate dH/dt is proportional to the balance between oxygen

sticking (positive term), oxide etching, and ion-induced oxi-

dation (negative terms)

n0

dHdt¼ 2jO2

sO2;0 1�Hð Þ2 � jArYOH� jArkH: (3)

Here, n0 is the surface site density, which is of the order of

�1015 cm�2 in crystal surfaces. Analogous to the Al sputter-

ing model of Kuschel et al., the coverage rate as function of

ion flux includes an additional term �jArkH, which accounts

for oxygen incorporation by creation of new chemisorption

sites due to ion bombardment. This term comprises all the

ion-induced effects, which according to Kuschel et al.,include (1) knock-on implantation, (2) electric-field-driven

diffusion, and (3) ion-induced surface activation as the most

relevant effects, leading to target oxidation.5

The implantation of adsorbed oxygen atoms within the

target by impinging Ar ions induces the transport of the oxy-

gen atoms from the very surface to the subsurface region of

the target. Thus, for each buried oxygen atom, one metallic

site is again available for receiving another oxygen atom,

i.e., the density of oxidized sites decreases. Then, assuming a

decrease in oxide site density proportional to the flux of

incoming Ar ions, jAr, one obtains the proportionality factor

k, which quantifies the oxidation efficiency by all the ion-

induced effects listed above.

In steady-state coverage, a constant oxidized top layer is

created on the target surface and, therefore, the only net

changes in mass are due to Cr removal. It means that oxygen

implantation, oxygen adsorption, and sputtering of oxygen

and chromium atoms are parallel processes without involv-

ing changes in the surface coverage state. In these condi-

tions, the efficiencies or probabilities of sputtering, sticking,

and implantation can be fitted to the QCM measurements

since the coverage fraction is constant.

B. Kinetics of oxygen adsorption duringAr1 bombardment

Fig. 2 shows the temporal evolution of the effective

sticking coefficient, which is calculated from the ratio

between adsorption rate and incident flux of oxygen mole-

cules as function of time. Series of data I and II show the

effective sticking coefficients at the conditions without and

with simultaneous Arþ etching, respectively.

In both series I and II, the values of sO2 at t¼ 0 extrapo-

lated from the fitted curves constitute the sticking coefficient

of a clean metallic surface, i.e., zero coverage sticking coef-

ficient, sO2,0. The fitted sticking coefficients of sO2,0 between

0.35 and 0.40 are in good agreement with literature.9 In con-

trast, a much lower sticking was measured for Al (0.015).5

These different behaviours can be explained in terms of a

quantum mechanical effect.10 The ground state of the spins

of O2 adsorbates is a triplet, whereas the bonding state shows

a singlet configuration. The transition is in principle forbid-

den, which accounts for the relatively poor reactivity of Al

with oxygen. This selection rule is weakened for heavier ele-

ments, such as Cr, resulting in a higher sticking coefficient

of O2.

As time increases, the sticking coefficient decays since

surface sites are being progressively occupied by incoming

oxygen atoms and molecules. Saturation is achieved when

(1) there are no more available sites for oxidation on the Cr

surface or (2) the etching and adsorption of oxygen are bal-

anced. In our case, the decay time associated with series I is

significantly smaller than in series II. The kinetics of O2

adsorption on Cr and sputtered Cr has been fitted with a cov-

erage model and discussed below.

1. Study of O2 adsorption (series I)

The effective sticking coefficient displayed by data se-

ries I in Fig. 2 as function of time, sO2(t), was fitted with the

following expression:

sO2ðtÞ ¼ sO2;0ð1�HÞ2; (4)

where the surface coverage H is solution of balance equation

(3) without the Ar etching terms and with initial condition

H(0)¼ 0. By fitting Eq. (4) to the data series I, one obtains

FIG. 2. Temporal evolutions of the measured (symbols) and fitted (lines)

effective sticking coefficient of oxygen molecules on Cr. Cases without (I,

open circles) and with (II, solid squares) simultaneous Arþ bombardment

(400 eV) are displayed. The oxygen molecule flux is jO2¼ 7.5� 1015 cm�2s�1

and, in the etching experiment, argon ion flux is jAr¼ 5� 1014 cm�2s�1. The

low ion-to-neutral ratio was chosen to sputter in fully poisoned mode. Fitted

parameters are obtained from Eqs. (2) and (3). k¼ 0.1 is chosen as initial value

and is varied in further fittings.

133301-3 Monje, Corbella, and von Keudell J. Appl. Phys. 118, 133301 (2015)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 134.147.160.151 On: Tue, 12 Apr

2016 09:54:19

Page 5: Elementary surface processes during reactive magnetron

values of s0,O2¼ 0.37 and n0¼ 1.5� 1016 cm�2. The last

value is surprisingly high compared to the expected site den-

sity for a crystal surface (n0� 1015 cm�2). Such discrepancy

in n0 by one order of magnitude can be introduced by the

surface roughness of the Cr target. In fact, QCM does not

have atomically flat surfaces, instead they are very rough due

to the orientation of the crystal cut. Fig. 1 shows microscopic

top view images of an untreated sample. The Cr film repro-

duces well the surface morphology of the substrate, as

expected from the conformal deposition by HPPMS.7 An

RMS roughness of approximately 1 lm was measured with

the Keyence laser scanning microscope. Moreover, the spe-

cific surface area was estimated to be at least a factor 5

higher than the apparent area. Thus, the large roughness

yields an elevated density of oxidation sites through the

increase in reactive surface area.

2. Study of O2 adsorption 1 Ar etching (series II)

The kinetics of oxygen sticking on Cr was also fitted dur-

ing simultaneous Arþ bombardment (series II in Fig. 2) with

the same flux of molecular oxygen. The model fitted to series

II is based on Eq. (4) with the surface coverage H obtained

from solving balance equation (3), including the Ar etching

terms and with the initial condition H(0)¼ 0. sO2 shows a

slower decay compared to the sticking experiment with only

oxygen. This effect was also observed in experiments of Al

oxidation by Kuschel and von Keudell.5 The best fitted stick-

ing coefficients at zero-coverage and compound sputter yield

are sO2,0¼ 0.39 and YO¼ 0.05, respectively.

The value of n0 rises up to 6.0� 1016 cm�2 when oxy-

gen adsorption takes place simultaneously to Arþ bombard-

ment compared to the case of simple adsorption of oxygen.

Hence, the already high surface site density due to surface

roughness undergoes a subsequent increase by a factor 4.

Such increase is responsible for the slower decay in the curve

of sO2 for series II compared to the data in series I (Fig. 2).

In fact, n0 determines only the decay time of the effective

sticking, while YO and k are also controlling the offset of the

asymptotic value of sO2. Thus, the influence of both sets of

parameters on sO2 can be decoupled. An accepted explana-

tion of the increase in n0 consists of the oxygen implantation

into the target subsurface, which was proposed by Depla and

De Gryse11 in their model of target poisoning and was exper-

imentally confirmed by G€uttler et al.12 Such increase in n0

can be interpreted likewise as the evolution of an ion-

enhanced surface roughness during reactive sputtering. Both

possibilities are discussed in Sec. III D.

An important parameter characterizing target oxidation

during reactive sputtering is the knock-on implantation effi-

ciency of chemisorbed oxygen atoms, k. In contrast with re-

active sputtering of Al, the best fittings are obtained for very

low knock-on parameters k in Eq. (3). In order to understand

the relevance of k in the sticking process, Fig. 3 plots again

the experimental data series II together with fitting curves

with the values of parameter k equal to 0.0, 0.1, and 1.0. The

main effect of increasing either k or YO is the offset of sO2

towards higher values at very long time. The fitting around

k¼ 0.1 for 400 eV is compatible with our measurements and

is further considered in Sec. III C for higher ion energies.

C. Reactive sputtering of Cr in stationary conditions

This section tests the fundamental parameters obtained

above in reactive sputtering of Cr at different ion energies.

Fig. 4 shows the QCM-measurements of the effective rates

of Cr targets during reactive sputtering in steady-state cover-

age conditions with different oxygen molecule fluxes at Ar

ion energies of 500 eV, 600 eV, and 800 eV. The curves are

located at the poisoned regime of the target. Concretely, oxy-

gen surface coverage calculated with balance equation (3) in

steady state (i.e., dH/dt¼ 0) is higher than H¼ 0.5 in all the

sputtering experiments. The metal regime of the target,

which is characterized by a linear evolution of effective rate

with ion flux,5 was not reached at the given conditions due to

the combination of high sticking coefficient of oxygen on

chromium and low sputtering yield of chromium oxide.

Thus, higher ion fluxes and/or lower oxygen fluxes are

required to operate in the metal regime.

Effective rates measured with QCM were fitted using

the surface coverage model described in Sec. III A. The fit-

ting curves were obtained by setting sO2,0¼ 0.38, as fitted in

Sec. III B, and by setting the metallic sputter yields YM with

the values obtained from TRIM simulations at the corre-

sponding ion energies. The fitted parameters were the oxy-

gen sputter yield, YO, and the knock-on parameter, k.

The characteristic sputter yield of the compound phase,

YO, follows approximately the dependence YO� 0.05xYM.

However, the empirical relation for metal oxides reported in

the literature is YO� 0.1xYM,13 which accounts for the stron-

ger dissociation energies of the Me–O bonds compared to

the Me–Me bonds. The discrepancy between these two

empirical relations is solved by considering that sputter

yields for compounds are material dependent and that these

yields can vary a few orders of magnitude depending on the

oxide phase. Kubart et al. studied this effect for different

forms of titanium oxide.14 Chromium admits a number of

FIG. 3. Temporal evolution of the measured (symbols) and fitted (lines)

effective sticking coefficient of oxygen molecules on Cr with simultaneous

Arþ bombardment (400 eV) assuming different values of k (0.0, 0.1, and

1.0). The oxygen molecule flux is jO2¼ 7.5� 1015 cm�2s�1 and argon ion

flux is jAr¼ 5� 1014 cm�2s�1.

133301-4 Monje, Corbella, and von Keudell J. Appl. Phys. 118, 133301 (2015)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 134.147.160.151 On: Tue, 12 Apr

2016 09:54:19

Page 6: Elementary surface processes during reactive magnetron

combinations with oxygen as well due to the many different

valence states of the metal. The bond dissociation energy of

Cr–O is approximately 423 kJ/mol, whereas values as high as

531 kJ/mol and 477 kJ/mol correspond to O–CrO and O–CrO2

bonds, respectively.15 This relative increase in 25% in the

bond energy of highly oxidized states could explain the very

low values fitted for YO, which are determined by the domi-

nant oxidation state of the Cr target. Another factor contribut-

ing to modify substantially the sputter yield is the roughness

of the target surface. Boydens et al. studied the variation in

sputter yield depending on the surface topography for differ-

ent metals.16 In their study, it was shown that surface rough-

ness could end up in lower or higher sputter yields depending

on how the angular distribution of ejected particles interacts

with the microscopic configuration of surface topography.

Hence, the resulting effect is finally determined by the compe-

tition between yield increase due to oblique ion incidence and

recapturing of sputtered particles by the surface asperities.

In order to study the influence of ion-induced effects in

oxidation of Cr, which are implemented through knock-on

parameter k, Fig. 5 plots the same effective rates from Fig.

4(b) together with fitting curves with the values of parameter

k equal to 0, 0.1, and 1. The values of k are effective on

balance equation (2) through its dependence with H, which

is determined from Eq. (3). The knock-on parameter k¼ 0.1

fits to the experimental data and is consistent with TRIM

simulations. The ion-induced transport of oxygen atoms

across the first atomic monolayer has been simulated as the

generation of recoils in a Cr2O3 target using default TRIM

input parameters. Only those oxygen recoils generated near

the surface, i.e., within a depth interval of 0 and 2 A, are sup-

posed to be originated from the process of knock-on

FIG. 4. Effective rates during reactive sputtering of Cr with varying argon

fluxes at energies of 500, 600, and 800 eV. The rates were measured in sta-

tionary conditions. The oxygen molecule fluxes are (a) 2.7� 1014 cm�2s�1

(flow rate of 0.05 sccm) and (b) 6.0� 1015 cm�2s�1 (flow rate of 0.20

sccm). The solid, dashed, and dotted lines refer to the modelling according

to Eq. (2). k¼ 0.1 is chosen as initial value and is varied in further fittings.

FIG. 5. Effective rates during reactive sputtering of Cr with Arþ (same data

as Fig. 4(b), jO2¼ 6.0� 1015 cm�2s�1) at ion energies (a) 500 eV, (b)

600 eV, and (c) 800 eV. Balance equation (2) was fitted to these data plots

using different values of parameter k (0.0, 0.1, and 1.0). The other fitting pa-

rameters remain the same.

133301-5 Monje, Corbella, and von Keudell J. Appl. Phys. 118, 133301 (2015)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 134.147.160.151 On: Tue, 12 Apr

2016 09:54:19

Page 7: Elementary surface processes during reactive magnetron

implantation and, therefore, are selected to calculate the

coefficient k. The simulation for 500 eV Ar ions impinging

onto a chromium oxide target yields k� 0.1. Therefore, the

effects of additional ion-induced surface activation mecha-

nisms, which may increase the value of k in several units as

observed in reactive sputtering of Al, are less important in

the case of oxidation of sputtered Cr. Concerning Al oxida-

tion, surface activation by Arþ bombardment was evidenced

through the positive values of mass variation rate at very low

ion-to-neutral flux ratios.5

Besides the significant difference in sticking coefficients

of O2 on Al and Cr, another important factor that influences

the different surface oxidation of both metals is oxygen incor-

poration by knock-on implantation. As stated above, oxygen

implantation events in Al (k� 1�3)5,6 are more frequent

compared to oxygen implantation in Cr (k� 0.1). The lower

probability of oxygen implantation in Cr may be related to a

higher oxygen affinity than Al. This is in apparent contradic-

tion with the higher enthalpy of formation of Al2O3

(�1670 kJ/mol) compared to that of Cr2O3 (�1130 kJ/mol)

from the metallic states. However, this argument can be con-

tested in terms of the kinetics of metal oxidation. In fact, the

formation of Al2O3 requires overcoming an activation energy

of approximately 150 kJ/mol,17 whereas the oxidation of

chromium shows energy barriers of only 35 kJ/mol and even

lower due to the availability of different oxidation states.18

Such effects, which point to a stronger oxygen affinity for

chromium, could result in weak ion-induced implantation of

oxygen into Cr due to preferential oxidation at the surface.

D. Ion-induced increase in surface site density

Here, we discuss the physical origin of enhanced oxygen

sticking observed in Cr targets submitted to simultaneous

fluxes of O2 and Arþ. Such sticking is correlated with an

increase in oxide sites at the target surface. A similar behav-

iour was found in the oxidation of Al targets as well, where

the saturation of oxidation sites during sputtering was

delayed with respect to the process of oxygen adsorption

without sputtering.5 In principle, two mechanisms are

equally responsible of ion-enhanced target oxidation: oxygen

implantation and surface roughening.

• Oxygen implantation: Its role in target poisoning, as indi-

cated above, has been already addressed in past works.11,12

However, the conditions of our QCM experiments re-

stricted the presence of reactive species to only oxygen

neutrals. Therefore, oxygen implantation in chromium

was exclusively caused by recoil implantation of adsorbed

oxygen atoms instead of reactive ion implantation (Fig.

6(a)). In principle, this implanted layer should be thicker

than a native oxide layer formed without Arþ bombard-

ment. Indeed, Kuschel and von Keudell found out by XPS

profiling that the oxidized layer of Al with Ar bombard-

ment was thicker than without bombardment.5 Also, insitu FTIR measurements by Kreiter et al. provided an esti-

mate of the total oxygen concentration which combines

chemisorbed oxygen on the top layer and implanted oxy-

gen in the subsurface, corroborating the increase in oxide

layer thickness.6

• Increase in surface roughness: It is an alternative factor

contributing to higher site density besides oxygen implanta-

tion. The initial etching phase in reactive sputtering with

Arþ in an oxygen atmosphere shows a very small H, i.e.,

H� 1. Indeed, the exposed area of metal existing between

oxide islands must be large enough in order to permit dif-

ferential erosion that promotes surface roughening. We

assume that the surface of this weakly oxidized layer leads

to laterally inhomogeneous sputtering because a local

charging of oxide surface islands will bend the incident

ions to neighbouring oxygen free surface areas. Since the

sputtering yield of chemisorbed oxygen is much lower than

for pure metal, the metal valleys will develop faster than

the oxide islands. In this sense, the top layer of chemisorbed

oxygen atoms plays the role of etching mask resembling a

nanolithography process of interstitial ion beam etch-

ing.19,20 As soon as these islands overgrow the metal val-

leys, a homogeneous charging may occur and the sputtering

should be laterally homogeneous (Fig. 6(b)). In the initial

state of oxidation, this unstable surface sputtering leads to

an ion-induced enhancement of the surface area and thereby

to an average increase in the surface site density. However,

when applying this scenario to surfaces in contact with re-

active plasmas, the interaction between target surface and

plasma sheath must be taken into account. The analysis of

this complex scenario is out of the scope of this article.

Our surface coverage model is restricted to the very sur-

face of Cr in contact with reactive plasma and does not con-

sider subsurface oxygen concentration due to implantation

events. This condition requires the coverage fraction, H, to

quantify exclusively the oxidation degree on the top layer.

Thus, due to the assumptions of the model, it is not possible

to consider separately the effects of roughness increase and

FIG. 6. Magnetron sputtering of Cr with oxygen as reactive gas may pro-

mote oxygen uptake by means of (a) oxygen knock-on implantation and (b)

ion-induced surface roughening.

133301-6 Monje, Corbella, and von Keudell J. Appl. Phys. 118, 133301 (2015)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 134.147.160.151 On: Tue, 12 Apr

2016 09:54:19

Page 8: Elementary surface processes during reactive magnetron

broadening of the oxide layer submitted to Ar bombardment.

The separation of both effects would require data treatment

with a more sophisticated model.

IV. CONCLUSION

Surface processes during reactive sputtering of Cr have

been mimicked by means of vacuum beam experiments.

Basic parameters in plasma-surface interactions are provided

by a very simple model based on rate equations of the effec-

tive deposition/erosion flux and surface coverage fraction.

The input parameters for the model, which are consistent

with literature values, have been evaluated from QCM meas-

urements of mass variation rates in dynamic and stationary

coverage conditions. Ion-enhanced surface activation is the

main mechanism promoting Cr oxidation. More precisely,

the most relevant effect of target bombarding with Arþ si-

multaneous to O2 adsorption is the increase in oxidation site

density n0 by a factor 4. Such effect is not restricted to Cr

and is also observed in other sputtered metals. The efficiency

of oxygen knock-on implantation by Ar ions agrees with

TRIM simulations, discarding any other ion-induced activa-

tion mechanism. Finally, implantation of oxygen adsorbates

and surface roughening have been proposed as possible

mechanisms that lead to surface site density increase during

reactive sputtering. In summary, this study shows that n0 is

the central parameter determining surface oxidation of sput-

tered Cr through the modulation of sticking coefficient of O2

on Cr surface.

This work proves that the system of balance equations

based on the extended Berg’s model is very robust and is

adequate to describe elementary surface processes on metal

targets exposed to reactive plasmas. The input parameters of

this model are very helpful for the coatings industry in order

to design and scale up deposition recipes of thin films by

means of reactive magnetron sputtering.

ACKNOWLEDGMENTS

This work is supported by the DFG (German Science

Foundation) within the framework of the Coordinated

Research Centre SFB-TR 87 and the Research Department

“Plasmas with Complex Interactions” at Ruhr-University

Bochum. The authors thank N. Grabkowski for his technical

assistance. Thanks are also to Dr. A. Hecimovic for the

deposition of Cr films and Dr. R. Neuser for the SEM and

EDX characterizations.

1M. Ohring, Materials Science of Thin Films: Deposition and Structure(Academic Press, San Diego, 2002).

2G. Br€auer, B. Szyszka, M. Verg€ohl, and R. Bandorf, Vacuum 84, 1354

(2010).3S. Berg, H. O. Blom, T. Larsson, and C. Nender, J. Vac. Sci. Technol. A 5,

202 (1987).4S. Berg and T. Nyberg, Thin Solid Films 476, 215 (2005).5T. Kuschel and A. von Keudell, J. Appl. Phys. 107, 103302 (2010).6O. Kreiter, S. Grosse-Kreul, C. Corbella, and A. von Keudell, J. Appl.

Phys. 113, 143303 (2013).7K. Sarakinos, J. Alami, and S. Konstantinidis, Surf. Coat. Technol. 204,

1661 (2010).8C. Corbella, S. Grosse-Kreul, O. Kreiter, T. de los Arcos, J. Benedikt, and

A. von Keudell, Rev. Sci. Instrum. 84, 103303 (2013).9J. C. Peruchetti, G. Gewinner, and A. Jaegle, Surf. Sci. 88, 479 (1979).

10C. Carbogno, J. Behler, A. Groß, and K. Reuter, Phys. Rev. Lett. 101,

096104 (2008).11D. Depla and R. De Gryse, Surf. Coat. Technol. 183, 184 (2004).12D. G€uttler, B. Abendroth, R. Gr€otzschel, W. M€oller, and D. Depla, Appl.

Phys. Lett. 85, 6134 (2004).13D. Depla and S. Mahieu, Reactive Sputter Deposition (Springer, Berlin,

2008), p. 153.14T. Kubart, D. Depla, D. M. Martin, T. Nyberg, and S. Berg, Appl. Phys.

Lett. 92, 221501 (2008).15B. deB. Darwent, Bond Dissociation Energies in Simple Molecules,

National Standard Reference Data Series Vol. 31 (National Bureau of

Standards, Washington, 1970), p. 26.16F. Boydens, W. P. Leroy, R. Persoons, and D. Depla, Thin Solid Films

531, 32 (2013).17C. Rossi, Al-Based Energetic Nanomaterials. Volume 2: Design,

Manufacturing, Properties and Applications (Wiley, London, 2015),

p. 20.18G. C. Fryburg, F. J. Kohl, and C. A. Stearns, “Enhancement of oxidative

vaporization of chromium (III) oxide and chromium by oxygen atoms,”

(National Aeronautics and Space Administration, Washington, DC, 1974),

Report No. NASA TN D-7629.19H. W. Deckman and J. H. Dunsmuir, Appl Phys. Lett. 41, 377 (1982).20S. Portal, M. Rubio-Roy, C. Corbella, M. A. Vallv�e, J. Ign�es-Mullol, A.

Canillas, and E. Bertran, Thin Solid Films 518, 1543 (2009).

133301-7 Monje, Corbella, and von Keudell J. Appl. Phys. 118, 133301 (2015)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 134.147.160.151 On: Tue, 12 Apr

2016 09:54:19