Home Search Collections Journals About Contact us My IOPscience

Influence of particle and energy flux on stress and texture development in magnetron

sputtered TiN films

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

2013 J. Phys. D: Appl. Phys. 46 055301

(http://iopscience.iop.org/0022-3727/46/5/055301)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 194.167.47.1
The article was downloaded on 29/12/2012 at 11:15

Please note that terms and conditions apply.

IOP PUBLISHING JOURNAL OF PHYSICS D: APPLIED PHYSICS
J. Phys. D: Appl. Phys. 46 (2013) 055301 (9pp) doi:10.1088/0022-3727/46/5/055301

Influence of particle and energy flux on

stress and texture development in
magnetron sputtered TiN films
G Abadias1 , W P Leroy2 , S Mahieu2 and D Depla2
1
Département Physique et Mécanique des Matériaux, Institut P’, UPR 3346, CNRS-Université de
Poitiers-ENSMA, SP2MI, Téléport 2, F86962 Futuroscope-Chasseneuil, France
2
Research Group DRAFT, Department of Solid State Sciences, Ghent University, Krijgslaan 281 (S1),
9000 Ghent, Belgium
E-mail: gregory.abadias@univ-poitiers.fr

Received 8 August 2012, in final form 26 October 2012

Published 28 December 2012
Online at stacks.iop.org/JPhysD/46/055301

Abstract
The real-time stress evolution during reactive dc magnetron sputter deposition of TiN films in
Ar+N2 plasma discharge was measured in situ using a multiple-beam optical stress sensor,
while the film texture was determined ex situ using x-ray diffraction. The influence of atomic
N/Ti flux and energy flux, previously quantified by combining plasma characterization and
Monte Carlo simulations (2009 J. Phys. D: Appl. Phys. 42 053002), was investigated by
varying either the N2 partial pressure at fixed total pressure, the total working pressure or the
bias voltage applied to the substrate. The contribution of thermal stress was carefully taken
into account from thermal probe measurements to evaluate the intrinsic (growth) stress from
the measured film force data. A clear correlation between stress, film texture and energy flux is
evidenced: while underdense (1 1 1)-textured TiN films with ‘V’-shaped columnar growth
(zone T) are under tensile stress (up to +0.6 GPa), dense TiN films with zone II microstructure
develop a (0 0 2) texture and large compressive stress (up to 3 GPa) when the energy flux is
higher than ∼150 eV per incoming particle. However, it is shown that a positive or negative
bias voltage, though increasing the energy flux, did not promote a (0 0 2) texture. It is
concluded that compressive stress development and (0 0 2) preferential growth are both
kinetically driven processes in magnetron sputtered TiN layers, but exhibit distinct dependence
with the substrate fluxes.
(Some figures may appear in colour only in the online journal)

1. Introduction metals growing in the Volmer–Weber mode, for which

surface stress, attractive forces at the grain boundaries and
Understanding the physical origin of intrinsic stress during incorporation of excess atoms have been identified as the main
thin film growth is an important issue in materials science driving forces of stress build-up [3–5]. Recently, in situ stress
for designing functional coatings with reduced stress levels measurements were also applied during reactive magnetron
and enhanced mechanical integrity. Titanium nitride (TiN) sputter deposition of transition metal nitrides (TiN, ZrN) [6–8].
films deposited by physical vapour deposition techniques on a The development of stress was found to be dependent on
rigid substrate are usually in a stressed state [1, 2], which may the film thickness: the stress was largely compressive (up
significantly alter their performance, as excessive stresses can to −4.5 GPa) in thinner (<10 nm) films, then decreased for
lead to coating failure by cracking or buckling. increasing film thickness, and became eventually tensile in the
Real-time stress measurements using the wafer curvature 100–200 nm range, depending on the working pressure. It was
technique implemented in situ during growth have provided demonstrated that this behaviour was related to the existence
new insights into the atomistic mechanisms of stress generation of intrinsic stress gradients over the film thickness and not
and relaxation. Studies were mainly focused on high-mobility due to a stress relaxation process. These gradients originate

0022-3727/13/055301+09$33.00 1 © 2013 IOP Publishing Ltd Printed in the UK & the USA
J. Phys. D: Appl. Phys. 46 (2013) 055301 G Abadias et al

from two kinetically competing stress generation mechanisms: a dc power supply. Another samples series was synthesized at
incorporation of defects associated with compressive stress and higher Ptot , from 0.40 to 0.80 Pa, by adjusting the total gas flow.
attractive forces at column boundaries which induce tensile The quantitative determination of energy flux and particle
stress [6]. Experiments carried out under higher energetic flux (both ion and metallic flux) towards the substrate
particle bombardment, e.g. using a lower working pressure was carried out previously by Mahieu and Depla [11] by
and/or applying a negative bias voltage to the substrate, combining Monte Carlo simulations using the SIMTRA code
were found to critically affect the overall film stress, by [22–24], energy resolved mass spectrometer and retarding field
enhancing the compressive stress contribution due to atomic energy analyser measurements [21]. This method enabled
peening [7, 8]. determining the atomic N/Ti flux,
N/Ti = (
N +
N+ )/
Ti ,
The deposition conditions also have an influence on where
N ,
N+ and
Ti are the atomic N neutral, N+ ions
the microstructure and texture evolution [9, 10], especially and metallic Ti flux, respectively. Under the condition used,
for transition metal nitride films which exhibit a distinctive
N+ was always one order of magnitude lower than
N . In
columnar microstructure, evolving from ‘zone T’ to ‘zone II’, this work, the deposition conditions were identical to those
with increasing adparticle mobility [9, 11]. A texture change used in [11] for plasma measurements, except the inclination
from (1 1 1) to (0 0 2) is concomitantly observed [11]. Several angle ϕ = 25◦ between the Ti target and the substrate normals.
studies have been reported in the literature over the last three However, this geometrical change does not significantly affect
decades to study the influence of several deposition parameters, the value of
Ti , as SIMTRA calculations performed at
such as the N2 partial pressure, substrate temperature, ion- ϕ = 25◦ showed a loss of only ∼7% in the metallic Ti
to-atom ratio or bombarding particle energy on the preferred particles reaching the substrate compared with ϕ = 0◦ (normal
orientation in TiN films [12–18]. It is only recently that incidence).
Mahieu et al have categorized the microstructure and texture The in situ stress evolution during growth was measured
evolution of magnetron sputtered TiN films in terms of the in real-time using a multi-beam optical stress sensor (MOSS)
extended structure zone model [19] and quantitatively linked designed by kSA and implemented on the top flange of the
these evolutions to the particle fluxes reaching the substrate deposition system. The measured quantity is the relative
[11], in agreement with the atomistic model of Gall et al [20]. change in the wafer curvature, κ, which is related to the
They also showed that some of the film’s properties, such as average stress × film thickness product, σav × h, according
the hardness and elastic modulus, were directly related to one to Stoney’s equation: σav × h = 16 Ys h2s κ, where Ys is the
of these substrate fluxes [21]. However, it is not yet elucidated biaxial modulus of Si (Ys = 180.5 GPa) and hs is the substrate
how the intrinsic stress during growth relates to these fluxes thickness (hs = 170 ± 2 µm). Note that σav × h is equivalent
and whether stress and texture development are driven by the to the film force per unit width, F /w, quantity [25]. In a
graph of F /w versus h, the incremental stress is given by
same physical process.
the slope and represents the stress variation due to the newly
In an effort to address this issue, we carried out real-time
deposited layer or relaxation in the buried layers. The Si
stress measurements during growth of magnetron sputtered
wafers (1 × 1 cm2 ) were placed horizontally on the sample
TiN films under the same deposition conditions that were used
holder, without any external fixation, allowing unconstrained
by Mahieu and Depla to quantify the particle fluxes reaching
bending. The geometrical parameters of the system were
the substrate [11]. In particular, the dependence of the intrinsic
determined using reference mirrors with known radius of
stress on the atomic N/Ti ratio and energy per arriving particle
curvature. Details on the MOSS device and detection set-up
EPA was critically investigated by varying either the N2 flow,
can be found elsewhere [25, 26].
the total working pressure or the bias voltage applied to the
Although no intentional substrate heating was used, the
substrate.
plasma discharge caused an unavoidable temperature rise in
the film, and consequently a variation in thermal stress due
2. Experimental procedures to the difference in thermal expansion coefficients between
the TiN film and Si substrate. To get the intrinsic (growth)
TiN thin films with thickness up to 300 nm were deposited on stress contribution only, the thermal stress contribution must be
(0 0 1) Si wafers covered with native oxide using unbalanced subtracted [27]. For this purpose, a passive thermal probe was
magnetron sputtering from 2 inch diameter planar Ti (99.97% used as a heat flux monitor [28] to determine the temperature
purity) targets under mixed Ar +N2 atmosphere. The target-to- change due to plasma heating. This set-up also enabled
substrate distance was fixed at 13 cm and the angle of incidence measuring the total energy flux (power per area) reaching the
of the Ti flux was ∼25◦ with respect to the normal of the substrate, Etot . This latter quantity is related to EPA (in eV/Ti)
substrate surface. Sputtering occurred at constant current through the relation Etot = EPA ×
Ti .
mode (0.85 A), with typical target powers of 0.3–0.45 kW. For The crystallographic orientation of the TiN films was
the first film series, the total working pressure, Ptot , was fixed investigated from ex situ x-ray diffraction (XRD) using a
at 0.30 Pa, corresponding to a total gas flow of 32 sccm at a D8 Discover Bruker diffractometer operating in the Bragg–
pumping speed of 183 l s−1 . The N2 flow was varied between Brentano configuration with CuKα wavelength and linear
0 and 18 sccm, resulting in a variation of the target discharge detector (LynxEye Silicon Strip Detector). Pole figures were
voltage from 425 to 530 V and of the deposition rate from 0.43 also performed on the selected samples using a Seifert XRD
to 0.16 nm s−1 , due to the target poisoning effect. Substrates 3000 four-circle diffractometer in point-focus geometry. Plan-
were either grounded or positively or negatively biased using view scanning electron miscroscopy (SEM) images were

2
J. Phys. D: Appl. Phys. 46 (2013) 055301 G Abadias et al

(a) 200 can be also observed that the subsequent magnitude of the
N flow = 12 sccm incremental stress is the same as prior to the first growth
2
stress×thickness (GPa.Å)

0 interrupt. A similar stress evolution is observed after a second

growth interrupt/resumption, although the magnitude of the
-200
– 3.2 GPa
effect is larger.
This stress evolution upon growth interrupt seems
-400 reminiscent of what has been observed for high-mobility
metals (e.g. Ag, Al, Sn) during thermal evaporation
-600 [3, 29–31] or sputter deposition [5, 31], where the reversible
stress changes were attributed to diffusing adatoms inside
-800 (respectively outside) grain boundaries during growth
0 100 200 300 400 500
(respectively upon interruption) [29, 30], changes in adatom
t (s) population [32, 33] or recrystallization processes [34].
(b) 200 However, for TiN films grown at room temperature, the
on homologous temperature, defined as T̃ = Tdep /Tm , is ∼0.08,
where Tdep and Tm are the deposition and melting temperature,
stress×thickness (GPa.Å)

0
respectively. This value reaches T̃ ∼ 0.12 at Tdep = 150 ◦ C
on
-200 (see figure 2(a) and related discussion hereafter), which
remains much lower than the threshold value of T̃ ∼ 0.2
-400 reported for high-mobility materials. Therefore, the above
mentioned mechanisms hardly take place here.
-600
The time-dependent temperature variation monitored
off under the same deposition condition (Ptot = 0.30 Pa, N2
off
-800
h=262 nm flow of 12 sccm and grounded substrate) using a passive
h=136 nm
thermal probe located at the same target-to-substrate distance
0 400 800 1200 1600 2000 2400 of 13 cm is shown in figure 2(a). The temperature variation
t (s) T when plasma is turned ‘on’ and ‘off’ exhibits the same
trend as the film force evolution measured during growth
Figure 1. Time-dependent evolution of the film force per unit width
(stress*thickness product) during magnetron sputter deposition of interrupt/resumption (see figure 1). At a N2 flow of 12 sccm,
TiN films at Ptot = 0.30 Pa and N2 flow of 12 sccm on the grounded the temperature rises rapidly up to 120 ◦ C during the first
Si substrate: (a) early growth stages (0–500 s) and (b) sequence of hundred seconds of deposition, then saturates to ∼150 ◦ C.
interruption (plasma ‘off’) and growth resumption (plasma ‘on’). When plasma is turned off, the temperature falls down with
The corresponding film thickness at the growth interrupt is also a similar time constant. Taking into account the thermal
indicated.
expansion coefficients of Si (αSi = 3.1 × 10−6 K−1 ) and
acquired on a FEI Quanta 2000F FEG-SEM to observe the TiN (αf = 9.35 × 10−6 K−1 ) [35], the corresponding change
film surface morphology, which gives an indication on the type in thermal stress, σth , can be calculated (see figure 2(b)),
of columnar growth regimes, as usually classified into zone assuming in a first step that mechanical properties of bulk TiN
I, zone T and zone II, according to structure zone diagrams (Ebulk = 456 GPa and ν = 0.24) also applies for the TiN thin
[10, 19]. film, where Ebulk and ν are Young’s modulus and Poisson’s
ratio of TiN [2]. A tensile (respectively compressive) thermal
3. Thermal stress stress evolution is obtained during interrupt (respectively
growth), in qualitative agreement with the experimental data.
Figure 1 shows a representative stress evolution, plotted as However, calculations yield to larger σth values and one has to
σav × h quantity versus time t, during and after growth of use a correction factor of ∼0.5 to fit the experimental data.
a TiN film sputter deposited at Ptot = 0.30 Pa and N2 flow The temperature rise during deposition, T , at various
of 12 sccm on a grounded substrate. In the early growth N2 flows and fixed working pressure, Ptot = 0.30 Pa, is
stages (figure 1(a)), a non-monotonic stress evolution is shown in figure 3(a), together with the corresponding energy
noticed. The incremental stress, initially compressive (3.2 GPa flux Etot , as measured from the thermal probe. One can
for t < 100 s), reverses back and becomes slightly tensile observe an increase in T from 120 to 165 ◦ C, as well as an
afterwards (t > 200 s). increase in Etot from 350 to 650 mW cm−2 , when the N2 flow
Figure 1(b) is plotted on a larger time scale (up to 2400 s), increases from 3 to 18 sccm. This evolution is consistent with
showing also a sequence of two interruptions and growth previous measurements, for which the different contributions
resumptions. At the first interrupt (t = 640 s, h ∼ 136 nm), to the energy flux were carefully investigated [11]. Even
a relatively abrupt tensile stress is observed, which saturates on a grounded substrate, it was found that the predominant
with time after ∼250 s. When deposition is resumed, a rapid contribution to EPA was the ion energy flux, as measured
compressive variation occurs up to approximately the same using a retarding field energy analyser [21]. This is primarily
film force level reached before interruption, ruling out any due to the relatively high ion-to-atom ratio under the present
possible stress relaxation mechanism in the buried layer. It deposition conditions. This ratio was shown to increase from

3
J. Phys. D: Appl. Phys. 46 (2013) 055301 G Abadias et al

(a) off (a)

160 N flow
600

E
2

tot
12 sccm
120
∆T (˚C)
160

(mW.cm )
∆T (˚C)
500
80

40 120
400

-2
0
0 100 200 300 400 500 600 700 800 80 300
on 0 5 10 15 20
t (s)
N flow (sccm)
2
(b)
calc. (E ) N flow (b) 0.9
0.6
bulk 2
Calc. (E , α)
fit 12 sccm bulk f

data Calc. (E , α *ρ )
meas f r
0.7 data
(GPa)

0.4

σ (GPa)
on 0.5
th

th
σ

0.2
0.3

0
0.1
600 700 800 900 1000 0 5 10 15 20
off t (s)
N flow (sccm)
2

(c)
300 N flow 9 sccm
2
Figure 3. (a) Influence of the N2 flow on the temperature change
(T ) and energy flux (Etot ), as measured using passive thermal
bias + 20 V probe at Ptot = 0.30 Pa; (b) Influence of the N2 flow on the thermal
stress σth change during growth interrupt: measured from in situ
200
∆T (˚C)

stress evolution (squares), calculated using bulk Ebulk and αf data

(solid line) and calculated using Emeas and modified thermal
grounded expansion coefficient (αf × ρr ) data (dashed line).
100

Figure 3(b) shows the thermal stress evolution, calculated

bias – 30 V from the T variation of figure 3(a), as a function of the
0
0 100 200 300 400 500
N2 flow at fixed Ptot = 0.30 Pa and the grounded substrate.
t (s)
The solid line corresponds to calculations performed assuming
bulk thermal and elastic properties of TiN. A discrepancy is
Figure 2. (a) Variation of the sample temperature during (red line) found between these calculated values and the stress changes
and after (blue line) plasma discharge at Ptot = 0.30 Pa and N2 flow measured in situ during growth interrupt. However, one
of 12 sccm, as measured using a grounded thermal probe; (b) has to take into account that TiN films deposited under
evolution of the thermal stress σth : measured from in situ stress
evolution (symbols), calculated using Ebulk = 456 GPa and these conditions are not fully dense, as attested to by the
αf = 9.35 × 10−6 K−1 (solid lines) and best-fit σth values to relatively low elastic modulus values, Emeas , measured by
experimental data (dashed lines), (c) variation of the sample nanoindentation [21]. For a series of TiN films deposited
temperature at different bias voltages applied to the thermal probe at at 0.55 Pa, reported values ranged from 150 to 210 GPa with
fixed deposition conditions (Ptot = 0.30 Pa, N2 flow of 9 sccm). increasing N2 flow. At 0.30 Pa and N2 flow of 9 sccm, the
measured elastic modulus was 270 GPa, which is 65% lower
∼6 to ∼13 with increasing N2 flow from 6 to 18 sccm at than the value E1 1 1 = 416 GPa of bulk (1 1 1) TiN [2]. In
0.55 Pa, mainly caused by a decrease in
Ti due to target addition, relative mass density values ρr = ρfilm /ρbulk of ∼0.5–
poisoning [21]. Therefore, the increase in T with increasing 0.8 were deduced from x-ray reflectivity measurements [36],
N2 flow can be explained by the larger ion-to-atom flux ratio. confirming that TiN films were underdense. Therefore, these
At fixed N2 flow, an increase in Ptot resulted in a decrease films should contain a significant fraction of voids, which
in both T and Etot (not shown here). This can be explained would lessen their thermal and elastic properties. Using Emeas
by the presence of less energetic sputtered atoms and ions, as and αf × ρr as input values for the elastic modulus and thermal
incoming species experience more collisions in the gas phase expansion coefficient, respectively, the calculated thermal
before reaching the substrate. Also, a lower target discharge stress is found to closely match the experimental data (see
voltage is obtained at higher pressure, causing a decrease in dashed line in figure 3). This suggests that the magnitude of
the magnetron power (the discharge current was kept constant), the reversible stress changes noticed during growth interrupts
which contributes to a decrease in Etot . Conversely, when a is most likely to be accounted for by thermal stress. In what
positive bias is applied to the substrate, a larger amount of follows, unless otherwise stated, the film force graphs will
electrons arrived at the film surface, inducing a substantial display only the intrinsic (growth) component of F /w, which is
increase in Etot and T . One illustration is given in figure 2(c) obtained by subtracting the thermal stress component (σth × h)
for Ptot = 0.30 Pa and N2 flow of 9 sccm. to the total film force per unit width.

4
J. Phys. D: Appl. Phys. 46 (2013) 055301 G Abadias et al

which the net stress is almost zero. It has been shown

previously that two competing stress sources act during TiN
film growth [6, 7]: a tensile stress component originating from
attractive forces at grain boundary and/or void formation, and
a compressive stress source due to the incorporation of point
defects in the film induced by the bombardment of energetic
species [37], a mechanism known as ‘atomic peening’. The
regions of low (<9 sccm) and high (>12 sccm) N2 flows
correspond to deposition conditions for which one of these
sources dominate, while at intermediate N2 flow, a balance
between these acting stress sources is reached.
A close correlation is observed between the stress
behaviours of figure 4(a) and the preferred orientation of
TiN films evidenced from the XRD patterns of figure 4(b)
and confirmed by pole figure measurements. These latter
revealed a clear fibre-texture with the absence of any in-plane
biaxial alignment. It is demonstrated that TiN films deposited
at low N2 flow exhibit a pronounced (1 1 1) texture, while
those deposited at high N2 flow have a pure (0 0 2) texture.
In the latter case, a substantial peak broadening is observed,
which can be related to smaller grain sizes as well as larger
lattice distortions (increase of microstrain contribution) due to
a higher concentration of point defects. The XRD patterns of
TiN films deposited at intermediate N2 flow are characterized
by a mixed (1 1 1) + (0 0 2) preferred orientation and relatively
much lower diffracted intensities.
This change in the film texture is accompanied by
a drastic change in film surface morphology due to two
distinctive microstructural film growth regimes [11]: a
‘zone T’ columnar microstructure with ‘V-shaped’ columns
along the growth direction and faceted column tops (giving
rise to a rough top surface, see figure 5(d)), which is
characteristic of an evolutionary growth regime, and a ‘zone II’
columnar microstructure with approximately straight columns
throughout the whole film thickness and featureless, smooth
top surfaces (as in figure 5(c)).
Under the present deposition conditions, it has been shown
Figure 4. (a) Evolution of the intrinsic film force per unit width that this change from a ‘zone T’/(1 1 1) texture to a ‘zone
(corrected from σth contribution) during magnetron sputter II’/(0 0 2) texture was driven by the increase in diffusion length
deposition of TiN films at different N2 flows and fixed Ptot = 0.30 Pa
on a grounded Si substrate; (b) corresponding evolution of XRD
L of adparticles (see figure 19 of [11]). It was also correlated
patterns of TiN films (h > 250 nm) with increasing N2 flow. with the increase in N/Ti flux. At fixed Ptot , the increase
in N2 flow results into both Etot and N/Ti flux increase. A
closer and quantitative inspection of the relation between these
4. Growth stress and texture
fluxes, texture and stress state of TiN films will be examined
4.1. Effect of the N2 flow rate at fixed working pressure in section 5.
(Ptot = 0.30 Pa)
4.2. Effect of working pressure and bias voltage
Figure 4(a) shows the intrinsic stress evolutions during sputter
deposition of TiN films as a function of the N2 flow and film Figure 5(a) shows the intrinsic stress evolution with increasing
thickness. Three distinctive behaviours can be rationalized, working pressure of TiN films deposited on the grounded Si
according to the final stress state being reached. At low substrate at constant N2 flow (18 sccm). The corresponding
N2 flows (3 and 6 sccm), a tensile stress develops since the evolution of the XRD patterns is shown in figure 5(b). It is
beginning of the growth and a steady-state value of +0.6 GPa observed that when Ptot increases from 0.30 to 0.55 Pa, the
is observed. At high N2 flows (15 and 18 sccm), TiN films intrinsic stress changes from compressive to tensile. However,
are under compression; however, no steady-state regime is at this high N2 flow, the incremental stress at the beginning of
reached, as the incremental stress decreases continuously with growth remains compressive but reverses back to tensile upon
increasing film thickness. The stress evolution of TiN films increasing film thickness and working pressure. This suggests
deposited at 9 and 12 sccm is an intermediate situation for that attractive forces between column/grain boundaries rapidly

5
J. Phys. D: Appl. Phys. 46 (2013) 055301 G Abadias et al

(a)
0
0.40 GPa

stress×thickness (GPa.Å)
-2000

P = 0.30 Pa
tot
-4000 18 sccm, + 30 V bias
15 sccm, –60 V bias
15 sccm, grounded
-6000 3 sccm, –60 V bias –7.3 GPa
3 sccm, grounded

0 200 400 600 800 1000

h (Å)
35
(b) (111)
grounded P = 0.30 Pa
tot

–60 V bias 3 sccm N

2

Intensity (cps)
25

15

35.5 36 36.5 37 37.5 38

2 θ (˚)

(c) grounded P = 0.30 Pa

tot (002)
– 60 V bias
15 18 sccm N
+ 30 V bias 2
Intensity (cps)

(111)

36 38 40 42 44
2 θ (˚)

Figure 6. (a) Influence of the substrate bias on the film force per
unit width during magnetron sputter deposition of TiN films at
various N2 flows; (b) comparison of XRD patterns of TiN films
(h > 250 nm) deposited at grounded and −60 V bias conditions at
Ptot = 0.30 Pa and N2 flow of 3 sccm; (c) comparison of XRD
patterns of TiN films (h > 250 nm) deposited at +30 V, grounded
and −60 V bias conditions at Ptot = 0.30 Pa and N2 flow of 18 sccm
Figure 5. (a) Influence of the working pressure Ptot on the intrinsic
film force per unit width (corrected from σth contribution) during
magnetron sputter deposition of TiN films at fixed N2 flow of Bragg reflection is found to increase with Ptot (see figure 5(b)).
18 sccm on a grounded Si substrate; (b) corresponding evolution of Another striking feature is the asymmetry of the (1 1 1) XRD
XRD patterns of TiN films (h > 250 nm) with Ptot , (c) top view line, which exhibits a tail on the left angular side: this is
SEM image of the TiN film deposited at 0.30 Pa and N2 flow of particularly visible for the film deposited at 0.80 Pa, and attests
18 sccm, (d) top view SEM image of the TiN film deposited at
0.55 Pa and N2 flow of 18 sccm.
of the presence of vacancy-type point defects [38] and voids
in the film. This is consistent with the lower mass density of
these films. The surface morphologies of the films deposited
become the dominant stress source and govern the growth at 0.30 (figure 5(c)) and 0.55 Pa (figure 5(d)) are in agreement
stress evolution at Ptot
0.40 Pa. This is related to the with the XRD observations. Note also that TiN films deposited
decreasing contribution of the atomic peening process with at Ptot
0.40 Pa are systematically (1 1 1)-textured and under
increasing Ptot , as energetic sputtered atoms and reflected tensile stress, independently of the N2 flow used.
neutrals experience more collision in the gas phase, thereby In order to increase the contribution of atomic peening,
decreasing their kinetic energy [24]. This stress change from a negative substrate bias was applied to the substrate at the
compressive to tensile is correlated with a texture change lowest Ptot investigated (0.30 Pa). The consequence on the
from (0 0 2) to (1 1 1). The diffracted intensity of the (1 1 1) stress evolution is reported in figure 6(a) for a substrate bias of

6
J. Phys. D: Appl. Phys. 46 (2013) 055301 G Abadias et al

−60 V and for two distinct conditions of N2 flow corresponding bulk position. We also observe that the application of a
to low (3 sccm) and high (15 or 18 sccm) atomic N/Ti flux. As positive substrate bias has the same effect on the (0 0 2) texture
expected, a larger compressive stress is obtained for negatively degradation (see figure 6(c)).
biased substrates compared with grounded ones. However,
the texture appears distinctively altered with the application 5. Relation with particle and energy flux
of a negative bias: at low N/Ti flux, the texture remains
(1 1 1) (see figure 6(b)), while at high N/Ti, a loss of (0 0 2) As mentioned above, Mahieu and Depla [11] have categorized
diffracted intensity is found (see figure 6(c)). Note that, for the microstructure and texture development in magnetron
both conditions, no (2 2 0) peak could be detected around sputtered TiN films in terms of the extended structure zone
2θ ∼ 62◦ . model. The basis of this model is the adparticle mobility, which
Two important conclusions may be drawn from these is controlled by EPA . They clearly showed that the transition
experimental results: (1) the presence of compressive stress is from zone T to zone II was correlated with surface diffusion
not necessarily associated with a (0 0 2) preferred orientation processes (ripening and island diffusion) rather than a bulk
and (2) the application of a negative substrate bias does process (grain boundary migration). In this
case, the relevant
not favour a (0 0 2) or (2 2 0) texture, at least in the bias quantity is the diffusion length L ≈ Exp(−1/EPA )/
Ti
voltage range [−60 V, 0] investigated. The first conclusion rather than EPA itself. Also, under high atomic N/Ti flux,
clearly advocates against a strain energy minimization model typically for
N/Ti > 5, a (0 0 2) texture was favoured [11].
to explain the development of a (1 1 1) preferred orientation In order to evaluate the relation between the substrate
in TiN films, as claimed earlier by several authors [39–41]. fluxes, texture development and growth stress build-up, we
The second point suggests that, under high N/Ti flux, EPA reproduce in figure 7 the experimental results obtained in
is large enough to promote a (0 0 2) texture and that the this study (intrinsic stress state and preferred orientation) as
contribution of an extra energetic ion flux (due to accelerated a function of
N/Ti (figure 7(a)) and EPA (figure 7(b)). Since
Ar+ and N+ ions in the plasma sheath) acts in a deleterious these fluxes were previously quantified for grounded condi-
way: ion bombardment is likely to induce excessive defects in tions [11], only the data obtained on grounded Si substrates
the surface layer, thereby reducing the diffusion length of Ti will be reported for these plots. Note that direct measurements
adatoms, and promoting renucleation. This leads to a decrease of Etot from the thermal probe gave higher EPA values than
in the average grain size [16] and could explain the observed those reported in [11]; however, the trends in figure 7 remain
(1 1 1) + (0 0 2) mixed texture (see figure 6(c)). The present unchanged. One can see that the development of compres-
data are also in good agreement with the results of Petrov sive stress requires sufficient adatom mobility, i.e. typically for
et al on magnetron sputtered TiAlN films [42], for which the
N/Ti > 5. However, a more direct relation is obtained when
increase in ion energy Ei at a fixed ion-to-atom ratio had little considering EPA , i.e. the energy delivered to the growing layer
effect on the film texture but significantly altered the lattice per incoming particle. In this representation (figure 7(b)), a
parameter due to lattice damage and increased compressive clear transition between intrinsic tensile and compressive stress
stress. They also found that for Ei = 100 eV, the XRD is obtained at a critical threshold EPA ∼ 150 eV/Ti, also corre-
peaks decreased in intensity. Finally, the absence of a (2 2 0) lated with a transition from (1 1 1) to (2 0 0) preferred orienta-
preferred orientation at −60 V is not surprising as this texture tion. TiN films deposited under this critical energy flux exhibit
is usually observed at very high negative substrate bias (above a mixed (1 1 1) + (2 0 0) texture, which corresponded to the
−400 V) and under strongly ionized plasma conditions, such intermediate N2 flow conditions reported in figure 4(a). Com-
as cathodic arc [40, 43] or plasma immersion ion implantation parison of figures 7(a) and (b) suggests that high enough ion
[41], as a result of ion channelling along the preferential flux (this flux is the dominant contribution to EPA ) is required
crystallographic directions. to induce compressive stress generation.
The influence of a positive substrate bias on the stress Figure 8 shows the evolution of the average stress
and XRD pattern evolutions of TiN films deposited at high (σav ) of TiN films, calculated at the same film thickness of
N2 flow has also been investigated. Results are shown in 120 nm, as a function of EPA for the different deposition
figures 6(a) and (c) for a bias voltage of +30 V. Compared conditions investigated in this study. Markers also indicate the
with grounded conditions, a larger compressive stress is again preferred crystallographic orientation. For grounded substrate
observed. Since in this case ion bombardment is no longer conditions, a direct dependence between the stress magnitude
effective, the enhancement of compressive stress cannot be and EPA is here demonstrated. It may be noticed, in particular,
due to the atomic peening process. However, a large flux that all tensile TiN films exhibit a (1 1 1) texture, corresponding
of electrons can reach the substrate, significantly increasing to low deposited energy conditions. Above a certain energy
the substrate temperature (see figure 2(c)). Therefore, the threshold (EPA ∼ 150 eV/Ti), compressive stress builds up.
total energy flux is again increased compared with grounded In this stress regime, the magnitude of σav scales with EPA ,
conditions, and a larger compressive stress is expected due to reaching ∼−3 GPa at EPA ∼ 325 eV/Ti.
larger thermal stress. It is remarkable that the stress level at +30 However, the data obtained under biased substrate
and −60 V bias is approximately the same: at a thickness of conditions (see figure 6) suggest that stress and texture
120 nm, the average stress in the TiN layer reaches −6.5 GPa. development exhibit distinct dependences with EPA . Indeed,
This is further qualitatively confirmed by the same angular while larger compressive stresses (up to −7 GPa) are reached
shift of the XRD lines towards lower angles compared with when a negative or positive bias voltage is applied to the

7
J. Phys. D: Appl. Phys. 46 (2013) 055301 G Abadias et al

(a) [111]
1.0
TiN films (h=120 nm)
[111] +[200]
[200] 0.30 Pa 0.0
20

-1.0 – 60V

σ (GPa)
15
0.40 Pa -2.0
N/Ti

av
compressive

10 -3.0
grounded
[111] bias
-4.0 [111]+[200]
5 [200]

tensile -5.0
0.55 Pa 50 100 150 200 250 300 350 400

0 E (eV/Ti)
PA
3 6 9 12 15 18

N flow (sccm) Figure 8. Evolution of the average residual stress of 120 nm thick
2 TiN films, σav , with EPA . Circles correspond to grounded substrate
conditions and squares to biased ones. The preferential orientation
(b) is also shown, according to the symbol type (empty = [1 1 1],
[111]
[111] +[200] grey = [1 1 1] + [2 0 0] and black = [2 0 0]).
300 [200]
compressive

agreement with the experimental results. At high N2 flow,

it is assumed that ion (respectively electron) bombardment
250
leads to excessive levels of EPA under negative (respectively
(eV/Ti)

0.30 Pa positive) substrate bias. This causes crystal damage, grain

200 size refinement and a weakening of the (0 0 2) texture degree.
Therefore, although stress and texture development appear
PA

150 0.40 Pa both as kinetically mediated processes, distinct dependences

tensile
E

on atomic N/Ti and energy fluxes are evidenced.

100
6. Conclusions
50
0.55 Pa The present investigation of the dependence between stress,
texture and particle flux reaching the substrate during reactive
0
3 6 9 12 15 18 magnetron sputter deposition of TiN films has contributed to
N flow (sccm) clarify some of the fundamental questions often debated in
2 the literature regarding growth of transition metal nitride thin
films.
Figure 7. Relation between (a) the atomic N/Ti flux and (b) the
energy per arriving particle, EPA , with the preferential orientation – It is confirmed that stress evolution during TiN growth
and intrinsic stress state (compressive or tensile) of 120 nm thick is governed by two competing stress sources, attractive
TiN films, as a function of N2 flow. forces between column boundary and the atomic peening
process. The thermal stress due to plasma heating plays
substrate, TiN films exhibit either a (1 1 1) or a mixed (1 1 1) + a minor contribution compared with the intrinsic stress,
(2 0 0) texture. At low N2 flow, this observation can be unless a positive bias is intentionally applied to the
understood by considering both changes in
N/Ti and EPA substrate.
when a substrate bias is used. For example, at 0.30 Pa and – A clear transition from tensile to compressive stress is
N2 flow of 3 sccm, when Vs changes from 0 to −60 V, EPA obtained when the energy flux EPA exceeds a critical
increases from ∼60 to ∼280 eV/Ti, exceeding the critical value threshold of ∼150 eV/Ti, while the development of
of 150 eV/Ti. This increase in EPA is due to the ion energy a (2 0 0) preferred orientation requires a high enough
flux contribution Eion = Ei
ion /
Ti , where Ei is the average atomic N/Ti flux (>5) and moderate EPA values (up to
energy of positive ions, given by the difference between the ∼300–400 eV/Ti).
plasma potential and substrate bias, |Vp −Vs |. In the same time, – The application of a negative substrate bias, under high

N/Ti is estimated to reach ∼1–2, but remains much lower enough adparticle mobility conditions, leads to larger
than the critical value of 5 required for the (1 1 1) to (0 0 2) compressive stress but contributes to degrade the (0 0 2)
texture transition (figure 7(a)). As a consequence, a (1 1 1) preferred orientation, due to the generation of excessive
preferred orientation and compressive stress are expected, in defects in the growing layer when EPA  400 eV/Ti.

8
J. Phys. D: Appl. Phys. 46 (2013) 055301 G Abadias et al

Therefore, although stress and texture development appear [18] Abadias G, Tse Y Y, Guerin Ph and Pelosin V 2006 J. Appl.
both as kinetically mediated processes, distinct dependences Phys. 99 113519
[19] Mahieu S, Ghekiere P, Depla D and De Gryse R 2006 Thin
on atomic N/Ti and energy fluxes are evidenced.
Solid Films 515 1229
[20] Gall D, Kodambaka S, Wall M A, Petrov I and Greene J E
Acknowledgments 2003 J. Appl. Phys. 93 9086
[21] Mahieu S, Van Aeken K and Depla D 2008 J. Appl. Phys.
104 113301
GA thanks Ghent University and the WOG: ‘Structural and
[22] Van Aeken K, Mahieu S and Depla D 2008 J. Phys. D: Appl.
chemical characterization of materials at the micro- and Phys. 41 205307
nanoscale’ for financial support during his stay in the DRAFT [23] SIMTRA, Simulation of transport of sputtered particles,
group in which this work has been performed. available at http://www.draft.ugent.be
[24] Depla D and Leroy W P 2012 Thin Solid Films 520 6337
[25] Floro J A, Chason E, Lee S R, Twesten R W, Hwang R Q and
References Freund L B 1997 J. Electr. Mater. 26 969
[26] Fillon A, Abadias G, Michel A and Jaouen C 2010 Thin Solid
[1] Perry A J 1990 J. Vac. Sci. Technol. A 8 1351 Films 519 1655
[2] Abadias G 2008 Surf. Coat. Technol. 202 2223 [27] Shull A L and Spaepen F 1996 J. Appl. Phys. 80 6243
[3] Floro J A, Chason E, Cammarata R C and Srolovitz D J 2002 [28] Stahl M, Trottenberg T and Kersten H 2010 Rev. Sci. Instrum.
MRS Bull. 27 19 81 023504
[4] Janssen G C A M 2007 Thin Solid Films 515 6654 [29] Chason E, Sheldon B W, Freund L B, Floro J A and Hearne S J
[5] Pletea M, Koch R, Wendrock H, Kaltofen R and Schmidt O G 2002 Phys. Rev. Lett. 88 156103
2009 J. Phys.: Condens. Matter. 21 225008 [30] Shin J W and Chason E 2009 Phys. Rev. Lett. 103 056102
[6] Abadias G and Guerin Ph 2008 Appl. Phys. Lett. 93 111908 [31] Koch R 2010 Surf. Coat. Technol. 204 1973
[7] Abadias G, Koutsokeras L E, Guerin Ph and Patsalas P 2009 [32] Friesen C and Thompson C V 2002 Phys. Rev. Lett. 89 126103
Thin Solid Films 518 1532 [33] Spaepen F 2000 Acta Mater. 48 31
[8] Machunze R, Tichelaar F D and Janssen G C A M 2010 Surf. [34] Koch R, Hu D and Das A K 2005 Phys. Rev. Lett. 94 146101
Coat. Technol. 205 1313 [35] Hultman L, Barnett S A, Sundgren J-E and Greene J E 1988
[9] Petrov I, Barna P B, Hultman L and Greene J E 2003 J. Vac. Cryst. Growth 92 639
Sci. Technol. 21 S117 [36] Mahieu S, Leroy W P, Van Aeken K, Wolter M, Colaux J,
[10] Anders A 2010 Thin Solid Films 518 4087 Lucas S, Abadias G, Matthys P and Depla D 2011 Sol.
[11] Mahieu S and Depla D 2009 J. Phys. D: Appl. Phys. Energy 85 538
42 053002 [37] Sarakinos K, Alami J, Karimi P M, Severin D and Wuttig M
[12] Greene J E, Sundgren J-E, Hultman L, Petrov I and 2007 J. Phys. D: Appl. Phys 40 778
Bergstrom D B 1995 Appl. Phys. Lett. 67 2928 [38] Ehrhart P 1994 J. Nucl. Mater. 216 170
[13] Hultman L, Sundgren J-E, Greene J E, Bergstrom D B and [39] Pelleg J, Zevin L Z, Lungo S and Croitoru N 1991 Thin Solid
Petrov I 1995 J. Appl. Phys. 78 5395 Films 197 117
[14] Banerjee R, Chandra R and Ayyub P 2002 Thin Solid Films [40] Zhao J P, Wang X, Chen Z Y, Yang S Q, Shi T S and Liu X H
405 64 1997 J. Phys. D: Appl. Phys. 30 5
[15] Huang J-H, Lau K-W and Yu G-P 2005 Surf. Coat. Technol. [41] Lim S H N, Mc Culloch D G, Bilek M M M and
191 17 McKenzie D R 2003 J. Appl. Phys. 93 4283
[16] Petrov I, Hultman L, Sundgren J-E and Greene J E 1992 [42] Petrov I, Adibi F, Greene J E, Hultman L and Sundgren J-E
J. Vac. Sci. Technol. A 10 265 1993 Appl. Phys. Lett. 63 36
[17] Li T Q, Noda S, Okada F and Komiyama H 2003 J. Vac. Sci. [43] Bendavid A, Martin P J, Wang X, Wittling M and Kinder T J
Technol. B 21 2512 1995 J. Vac. Sci. Technol. A 13 1658