Structure and properties of titanium nitride thin films deposited at low temperatures

using direct current magnetron sputtering

F. Elstner, A. Ehrlich, H. Giegengack, H. Kupfer, and F. Richter

Citation: Journal of Vacuum Science & Technology A 12, 476 (1994); doi: 10.1116/1.579155
View online: http://dx.doi.org/10.1116/1.579155
View Table of Contents: http://scitation.aip.org/content/avs/journal/jvsta/12/2?ver=pdfcov
Published by the AVS: Science & Technology of Materials, Interfaces, and Processing

Articles you may be interested in

Numerical investigation via three-dimensional Monte Carlo modeling of sputtering and deposition processes in a
direct current unbalanced magnetron discharge
J. Vac. Sci. Technol. A 26, 103 (2008); 10.1116/1.2816940

Low temperature, fast deposition of metallic titanium nitride films using plasma activated reactive evaporation
J. Vac. Sci. Technol. A 23, 394 (2005); 10.1116/1.1874152

Structure and mechanical properties of Ti–Si–N films deposited by combined DC/RF reactive unbalanced
magnetron sputtering
J. Vac. Sci. Technol. A 22, 2351 (2004); 10.1116/1.1798771

Magnetic properties of Ti/Fe double layers grown on MgO(100) by direct current magnetron sputtering
J. Appl. Phys. 85, 4943 (1999); 10.1063/1.370053

Deposition and properties of yttriastabilized zirconia thin films using reactive direct current magnetron sputtering
J. Vac. Sci. Technol. A 9, 3054 (1991); 10.1116/1.577172

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 169.233.7.23 On: Fri, 21 Nov 2014 18:20:35
 Structure and properties of titanium nitride thin films deposited at low
temperatures using direct current magnetron sputtering
F. Eistner, A. Ehrlich, H. Giegengack, H. Kupfer, and F. Richter
Technical University of Chemnitz, Department of Physics, Oberfrohnaer Strasse 33,
D-OYIl7 Chemnitz, Germany
(Received 22 June 1993; accepted 6 November 1993)
Titanium nitride layers on silicon substrates have been deposited by magnetron sputtering at a
constant temperature of 200 °C using nitrogen partial pressures (PN2) from 0.001 to 0.36 Pa. Distinct
dependences between PN2 in the reactive gas and the mechanical and structural properties of the
layers could be found by x-ray diffraction and other methods. The general absence of the Ti2N phase
in layers deposited at low temperatures can be explained by the high energy impact of energetic
particles during the deposition process as well as by the limited mobility of adatoms due to the low
substrate temperature. Two different cases of lattice distortions were detected for low and high PN2'
ab
respectively. In the higher pressure range the lattice parameters l1 are greater than the 0
a6°
. For
this the variable density of interstitial sites in different planes together with the macroscopic film
a6
stress was found to be responsible. At lower PN2 the 11 was smaller than the 0
a6°
• This effect
could be explained by the presence of a high number of vacancies in the lattice influencing the
chemical bonding and thus leading to an anisotropy of the stability of the interplanar spacings. From
that macroscopic stresses results in an anisotropic deformation of the film crystallites. Both
explanations could be supported by annealing experiments.

I. INTRODUCTION mechanical, and corrosion properties of low temperature

magnetron sputtered protective coatings.
The number of publications in the fIeld of deposition of In this study, we investigated TiN layers deposited on
nitrides of group IVB elements by physical vapor deposition silicon substrates at a constant substrate temperature of about
(PVD) techniques did not decrease within recent years. That 200 DC at different nitrogen partial pressures. The chemical
indicates, that in spite of significant progress in this field a composition, structure, morphology, and microhardness of
strong interest remains in the understanding of the layer for- the films have been analyzed. Continuing earlier investiga-
mation processes and the influence of process parameters on tions of our group,16 the influence of an ex situ annealing
stmcture and mechanical properties of the films. Titanium treatment at comparatively low temperatures (100-400°C)
nitride is commonly accepted as a representative example of was examined, too. From the experimental data, conclusions
this class of compounds. As Kadlec et al. pointed out, 1 the are drawn about the nature of lattice distortions and the rea-
final goal of the abovementioned investigations should be a sons for the general absence of Ti2N in our layers.
unified model of the dependence of film stmcture on the
deposition parameters. Such a model should include all data
available. Since it is nearly impossible to find a single depo- II. EXPERIMENTAL DETAilS
sition parameter whose influence on the stmcture and me- The layer deposition by reactive dc magnetron sputtering
chanical properties of magnetron sputtered TiN layers has was performed using a high vacuum assembly equipped with
not been subject of detailed studies, recent publications dealt an oil diffusion pump system yielding a base pressure of
often with very special problems, as for example the influ- about] Xl 0. 3 Pa. We used a circular magnetron with a di-
ence of x-ray elastic constants (XEC); of different planes on ameter of 95 mm. The target plate was made of vacuum-
the macroscopic stress values,25 the origin of lattice melted high-purity titanium. For the sputtering process, a
distortion 6- 9 or the behavior of vacancies especially at high mixture of nitrogen (purity 99.99%) and argon (purity
temperatures.lO,11 99.999%) was fed into the vacuum chamber. The total pres-
The influence of substrate temperature on stmcture and sure of the gas mixture during the sputter deposition was
properties has been investigated, too. l •12 14 However, it 0.36 Pa in all cases. The gas composition was varied between
should be emphasized, that only few investigations dealt pure nitrogen and pure argon. The discharge power was 520
with TiN layers deposited by magnetron sputtering at tem- W, connected with a discharge current of about 1.3 A In all
peratures below 300°C. 8,12.15 Petrov 14 pointed out that some the experiments, we chose a deposition time of 20 min. Prior
substrate materials as glass and integrated circuits where TiN to deposition, a 10 min ion-etching process was applied to
is used for instance as a diffusion barrier allow only deposi- clean the substrates, using a pure argon atmosphere (I' = 1.0
tion temperatures below 500°C. Covering plastics with TiN Pa) and a discharge current of 50 rnA
and other hard coatings to protect the surface against abra- The polished single-crystal silicon substrate [(111)-
sive wear will get greater importance in the future as well. In oriented, resistivity p= 10 fl cm] were placed on a substrate
such cases, the deposition temperature should not exceed holder 30 mm from the magnetron cathode. During sputter-
200 dc. Therefore it is necessary to know the exact structure, ing, the substrate temperature was controlled to a value of

476 J. Vac. Sci. Techno!. A 12(2}, Mar/Apr 1994 0734-2101194/12(2)/476181$1.00 @1994 American Vacuum Society 476

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 169.233.7.23 On: Fri, 21 Nov 2014 18:20:35
 477 Eistner et sl.: Structure and properties of TiN thin films 477

TABLE I. Parameters and conditions of the deposition process

Discharge CUlTent 1.3 A

Discharge power 520W
Working pressure (PAr +PNz) 0.36 Pa
Residual gas pressure 10. 3 Pa
Target-to-substrate distance 30 mm
Deposition time 20 min
Bias voltage -lOOV
Ion current hitting the substrate holder O.2A
Substrate Silicon
Substrate temperature during deposition process 200°C

about 200 °C using a resistance heating and a thermocouple. a.oOl 0.005 0.01 0.05 0.1 0.5
The substrate holder was electrically isolated and exposed to Nitrogen Partial Pressure / Pa
a bias voltage of - 100 V referring to anode potentiaL In
earlier investigations of our groupl6 we measured the electri- FIG.!. The deposition rate dependence on PN,' This value decreases will]
cal conductivity of titanium nitride layers in dependence on the onset of the reactive deposition modc due tela TiN covering of the target.
the applied bias voltage. For a bias voltage of -100 V, the
resistivity was found to be 37 ;..dl cm at room temperature.
To improve the layer adhesion,17 we always started the B. Layer composition and hardness
deposition using a pure argon atmosphere yielding a titanium
layer of about 50 nm thickness. After that, the desired Layer composition was investigated using EPMA tech-
nitrogen-to-argon ratio was adjusted and the TiN x layer was nique. It should be noted that a direct determination of the
deposited without interruption. Because of an extremely nitrogen content was impossible because the nitrogen- K a-
strong (002) texture of the intermediate titanium lavers, in line is overlapped by a titanium Ene in the x-ray spectrum.
our x-ray diffraction (XRD)-pattems (002) peaks of titanium Only titanium and oxygen were found in detectable quanti-
appear in all cases (see Sec. III D). ties. Thus, the nitrogen concentrations were evaluated under
An optical microscope EPIVAL equipped with a special the assumption, that the layers consist mainly of nitrogen,
microhardness tester mhp 160 (Carl Zeiss Jena, Gennany) titanium, and oxygen, i.e., cN+co+cTi=100 at. % (ex is the
was used to measure the hardness of the layers according to elemental concentration in atomic percent).
the Knoop method. We found out a nearly fitoichiometric composition of the
Special attention was paid to XRD investigations includ- TiN at the pressure range above 0.02 Pa (Fig. 2). With in-
ing phase analysis (by comparing the observed data with creasing nitrogen partial pressure, we found a rapid rise in
standard bulk values), the determination of peak intensity the nitrogen concentration of the layer up to an amount of
ratios as well as the evaluation of lattice constants, grain size 40-45 at %0 This is accompanied by a change of the film
and fluctuations of the d hk1 • On selected samples macro-
scopic stress was determined using the sin2 1{1 method
(SSPM).IR
The chemical layer composition was investigated by elec- xx Ti
tron probe microanalysis (EPMA). The measurements were 00 N
carried out on an area with a diameter of 40 ,urn. The error of
the dctected concentrations was about ±5%. In addition to
EPMA the homogeneity of the TiN coating's composition
x
across the mm thickness was proved by Auger electron spec-
troscopy (AES) depth profiling. The deposition data are sum-
marized in Table L o 9IY 0
o

20
III. RESULTS
A. Growth rate
0.001 Q.005 0.01 0.05 0.1 0.5
Figure 1 shows deposition rate dependence of the nitrogen Nitrogen Parti al Pressure I Pa
partial pressure. The first three samples contain only the hex-
agonal titanium phase (see below). Beginning with the onset
of the reactive deposition mode characterized by the forma- FIG. 2. Elemental concentrations of titanium and nitrogen in the layers in
dependence on PN2 measured by EPMA. The method's error is about :'::5%.
tion of titanium nitride the growth rate decreases by about For the first three concentration values only the hexagonal a-Ti lattice could
three-quarters due to the well known target coverage with be detected by XRD due to a solid solubility of nitrogen in the titanium
titanium nitride. 19 lattice up to 25 at. %.

JVST A - Vacuum, Surfaces, and Films

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 169.233.7.23 On: Fri, 21 Nov 2014 18:20:35
 478 Elstner et al.: Structure and properties of TiN thin films 478

35,------------------------------------------,
load = 20 p
30

:2 20 :

j~l
~~1
o~i--~--,-~,iT'T'~'n'i~--.-~'~'~i-r'r,n"'i----r--r~ e) PN2- 0.Q1 Pa f) PN2- 0.03 Pa g) PN; 0.05 Pa h) PH - 0.36 Pa
0.001 0.005 0.01 0.05 0.1 0.5
(pure nitrogen)
Nitrogen Partial Pressure I Pa
FIG. 4. Layer morphology of selected samples for different PNz investigated
PIG. 3. Microhardness of the layers measured by the Knoop method with a by TEM of erpt replica (magnification 2000X). In thc region of substoichio-
load of 20 P in dependence on PN?' The first three values are those of metric TiN (cf., Fig. 2) slightest changes in PN, cause severe changes in the
titanium. Beginning with the reactive mode the microhardness reaches a film morphology (PN2 =0.0] Pa and PN2 =0.01 i Pal. The layer deposited at
maximum value; it decreases when the lattice reaches the stoichiometric PN2=O.36 Pa shows a different change in the character of the morphology.
region (c.f., Fig. 2).

structure (cf., Fig. 10) which correlates to a different film

morphology (see the electron micrograph Fig. 4). the film during growth. 2o Igasaki et at. 21 reported a nitrogen
At nitrogen partial pressures of about 0.01-0.02 Pa, a solubility in the hcp titanium lattice up to 25 at. % and a
maximum of the microhardness was detected (Fig. 3, M H K small distortion of this lattice, too. At PN? =0.01 Pa (accord-
=29.0±3.5 GPa). In this region the film stress has a maxi- ing to about 3% N2 in Ar), the diffraction pattern of TiN is
mum, too (see Sec. III H). fully developed. The Ti peaks of the diffraction pattern of
The hardness decreases to about 22 GPa for stoichio- this layer result from the abovementioned intermediate tita-
metric TiN with a distinct golden color (PN2=O.03 Pa). The nium layer. Measurements on layers deposited under the
further decrease of the hardness above the pressure and its same conditions but without such a sublayer showed the
nearly constant value for PN2?-:O.05 Pa indicate a structure usual TiN-phase diffraction pattern only. The E-Ti2N phase
modification in the films (see XRD investigations below), could not be detected.
which also change their color to bronze for PN2 above 0.1 Pa.
Despite this, for pressures higher than 0.02 Pa the chemical
composition is within an error of about ±5 at. % constant.

C. Surface morphology
The surface morphology shown in Fig. 4 indicates a fun-
damental change of the film structure in the nitrogen pressure
range around 0.01 Pa characterized by a decreasing grain PN2 ~ 0.01 Pa

size. Figures 4(d) and 4(e) show surfaces of samples depos-

ited at practically the same pressure (about 0.01 Pa) with
distinctly different film morphology. Both microhardness and
film structure point to the difficulty to deposit TiN films in Pm ~ 0,005 P.
the transition range reproducibly.
ATHD02)

D. Crystallographic phases
For XRD measurements a Bragg-Brentano diffractometer ~\ Pm ~ 0.001 Pa

with monochromized Cu-K a ray was used. Some XRD pat- 30 35 45 50

terns in dependence of the nitrogen partial pressure are given
in Fig. 5. Layers deposited in a pure argon atmosphere ex-
hibited the usual diffraction pattern of titanium. With in- FrG.5. XRD patterns of samples in the PN2 region of the transition from Ti
creasing nitrogen content of the working gas, a distinct dis- to TiN. The two patterns below show only peaks of the hep a-Ti phase, but
tortion of the hexagonal (hcp) titanium lattice, especially a the nitrogen in the titanium lattice leads to a strong lattice distortion indi-
cated by a shift of the Ti-(002) peak to a lower angle. At PN2 =0.01 Pa the
widening in the (001) direction occurs. This corresponds
cubic 8-TiN phase is fully developed, The Ti-(002) peak visible even in this
with the high solubility of nitrogen atoms in the hcp Ti lattice spectrum is due to the titanium underlayer, the NaCl-peaks result from rock
up to about 21-23 at. %, depending on the temperature of salt powder applied as a standard.

J. Vac. ScI. Technol. A, Vol. 12, No.2, Mar/Apr 1994

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 169.233.7.23 On: Fri, 21 Nov 2014 18:20:35
 479 Elstner sf af.: Structure and properties of TiN thin films 479

l~O.------------------------------------------'
""" 3.5
....
..-l

1000 ;;:3
<:>
N

ex:( 800 :is 2.5

--
Q)
N ~ 2
U5 600 (l)

C C

.yo
- 1.5
co 4-

(5 400 o

I'l~o
o
200

a 'I ' \ I
0.01 0.02 0.04 0.06 0.08 0.1 0.01 0,02 0<01. 0.06 0.08 0<1
Nitrogen Partial Pressure / Pa Nitrogen Partial Pressure I Pa

FIG. 8. The relation between the intensities of the (200) and of the (1 J 1)
FIG. 6. Grain size of the TiN samples in dependence of PN2 evaluated by the
peaks (texture). The dotted line gives the value of the ideal polycrystalline
Kochendorfer equations using the integral breadths of different peaks to
TiN.
distinguish grain size and microstrain (ef., Fig. 7). The maximum of the
grain sizes corresponds to that PNo region, where the films just become
stoichiometric (cf., Fig. 2). "
erage grain size decreases, while the d hkl fluctuations in-
crease. This dependence seems to be expectable, if we take
E. Grain size into account, that the smaller the grains are the more grain
Considering the Cu K a1- K a2 splitting, the peaks were boundaries exist and thus more disturbances in the crystal
fitted with square Cauchy functions. Fine-grained NaCl pow- structure appear.
der applied to the sample surface was used a..<; a standard for The grain sizes of the TiN phase are significantly smaller
a precise determination of the d hkl . The grain size and the than that of the Ti phase in the films deposited at nitrogen
average fluctuations of the d hkl were computed using the partial pressures up to 0.005 Pa. The grain size in the latter
Kochendolfer equations 22 considering the (Ill), (200), and ones decreases with increasing nitrogen partial pressure (see
(220), if detectable, the (311) and the (222) peaks of TiN. Fig. 5), while the d hkl fluctuations increase. That means, ni-
The instrumental broadening of the peaks was determined by trogen atoms on interstitial places of the hexagonal Ti lattice
the TiN-standard powder and subtracted from the integral affect the crystallite growth, what has been discussed in de-
peak width using a con-ective function for Cauchy peaks. tail in Ref. 21.
The results of this investigation are presented in Figs. 6
and 7. With increasing the nitrogen partial pressure, the av-
F. Film texture
Regarding the 111111200 ratio of the TiN layers, i.e., the
ratio of the intensities of the TiN (111) peak and the TiN
<>« 0.014.,---....:....----------------- (200) peak (see Fig. 8), we find a distinct correspondence
---cf.::L

0.012
between the nitrogen partial pressure and the II 11/l200 ratio.
In this connection, one has to take into account that this ratio
4- depends also on the film thickness,23 which was not held
o
<I) 0.01
constant in our series (cf., Table I and Fig. However,n
C
o additional XRD investigations of films with one and the
:;::; same thickness proved that the observed dependence is not
(fj 0<008

--
:::l
U
:::l
c;::: 0.006
due to the film thickness.

Q)
G. Lattice distortion
~O.004
L 0 0 ,
Lattice distortion in magnetron sputtered TiN layers has
~ 0.002 -'n---,---,--......-r--r--.--,-r-,--,C'"""T-,--~~~~~~I
i I : iii I I I i I r I if i • I
often been found. 4 - 9,12 The lattice is not exactly cubic, but
can be described in certain cases in terms of a rhombohedral
0.01 0.02 0.04 0.06 0<08 0.1
Nitrogen Partial Pressure I Pa cell with an angle a between 89.4 0 and 89.9°,24 Le., a certain
(yet quite small in this case) "elongation" of the lattice cell
in one of the (111) directions. Therefore, the lattice constants
FIG. 7. Average fluctuations of d hkl computed with the Kochendorfer equa- calculated by means of the formula for cubic lattices
tions (cf., Fig. 6), which are a measure for the microstrain of the layers. The
lowest values correspond with a microstrain of about 0.08%, the largest with ao= d hkl (h 2 + k 2 + [2) JI2,
one of about 0.3%.

JVST A - Vacuum, Surfaces, and Films

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 169.233.7.23 On: Fri, 21 Nov 2014 18:20:35
 480 Elstner et sl.: Structure and properties of TiN thin films 480

4.31 ~------.----------

00- (111l-plane

4.3
x x - (200l-plane
3 0\TH002)
~ ""~"o"' """r' I
J LJ l_ N~'_,.
-- 4.29
-!:l
C
2UB

~r
U) °OM
C
o
U 4.27 ATd002)-----------'
ill
u ~ I N.CH220l

o
x
x
11 ,,,,U,, I
x
x 0
o Pm ~ 0.1 P.

0.01 0.02 0.04 D.06 O.OB 0.1 30 35 45 50

Nitrogen Partial Pressure / Pa

FIG. 9. Lattice constants computed from d 11 j and d 200 , respectively, Llsing FIG. 10. XRD patterns for the two layers deposited at the highest P"2 under
the formula for the cuhic lattice. The splitting indicates a strong latlice investigation. The Ti-(002) peak is due to the strongly (002)-textured tita-
distortion in the regions of low as well as of high PN2' In the medium region nium ;nderlayer. The most significant difference betvicen both patterns is
the latti<:e is within the measuring error of about +-0.002 A cubic. the disappearance of the (200) peak at PNz·=0.36 Pa. The grains of this layer
have a distinct <11 t> orientation. It should be mentioned that the difference
between XRD patterns is accomp~l1lied by a distinct change in the surface
morphology (see Fig. 4).
either from the position of the (200) peak or the (111) peak
of the diffraction spectrum are, in general, different. These
lattice constants (I) II and a 0
a6°
can be used to describe the and 0'2 are the principal stresses parallel to the surface. The
lattice distortion. stress 0'3 normal to the surface is assumed to be zero. Then,
In Fig. 9, ab 11 and a6 00 in dependence of the nitrogen the lattice strain E¢4 for a known direction if> and ifi is given
partial pressure are given. As it can be seen, a splitting be- by
a6
tween ab II and 00 occurs in the lower and the upper pres- c<,t>1f=[(l 2
+ IJj)(O'l COS if>+0'2 sin2cP) sin 2 1/1
sure regions, while in the medium range no significant split-
ting can be detected. In the lower PN2 reglOn, . t he aoJ11·IS - VA(YI + 0'2)]IEj ,
smaller than a~oo and in the upper significantly greater.
where E,pv,=(d",Ij,-do)ld o , E j equal& the Young's modulus,
Stoichiometric layers deposited at pressures between vf equals the Poisson's ratio of the film, and
0.cn4 and 0.030 Pa show no significant splitting between the . ~ . 2
(T",=(TI cos"' if>+(T2 S111 cP the measured ~tress component
lattice constants computed from d III and d 200 . Layers de- proportional to the slope of Eq,q, versus sin" ifi.
posited at p =3.6x Hr 1 Pa in pure nitrogen exhibited an In practice, the angle 2fl should be as large as possible to
extraordinary strong (111) texture. Probably for this reason it ensure high precision. On the other hand, the intensities of
was impossible to detect the (200) peak. First, because of the high order peaks are too small to detect lattice spacing over a
absence of the (200) peak in connection with the presence of wide iff range. In this work, the compromise was found by
the (111) peak and the (220) peak, we supposed the existence using the (200) planes generally. The sin 2 1/1 plots given in
of a TiN2 phase,25 as it had been described by Manory and Fig. 11 have been approximated by straight lines. The
Kimme1. 26 But recording diffraction patterns with a ifi angle stresses were calculated from the slopes using E=640 GPa
of 35.3° [under this angle (200) planes of (lll)-oriented and v=O.295 (Ref. 3) and are given in Table U. Because we
crystallites are directed perpendicular to the surface] showed, used only the (200) planes for our SSPM, we had not taken
that the (200) peak exists, and that it absence in the former
XRD pattern was only a texture and no structural etlect This
behavior is shown in Fig. 10. TABLE II. Stress measurements by SSPM using (200) peaks showing an
isotonic increase of CJ'if! with 0
a60
(PN2-nitrogen partial pressure,
O'<b-biaxial film stress, a~ao_lattice parameter calculated from (/200 mea-
H. Film stress sured at ifJ=OO). Curvature (cf.. Fig. 11) indicates a stress component pt'r-
pendieular to the surface and I/i-splitting the existence of a shear stress
X-ray stress measurements were carried out on several component.
TiN coatings. In addition to the square Cauchy iltting of the
peaks, their centers of gravity were detected. That has been P N/Pa( X 1(
2
) aiioo/A O'q/MPa Comments
important for stress measurements, where the peaks became 5 4.266 ·10100 Curvature
often rather asymmetric with if! angles above 10°. 1 4.2M 8250 Curvature,
Table II contains the main feature of the samples used in ifJ-splitting
2 4.261 -5650 Curvature,
this study. The internal stress in the films was measured hy
</i-splitting
the well established SSPM. It is assumed for a first approxi- 1.7 4.250 -1750 Curvature
mation that a biaxial state of stress exists in the film. Here (FI

J. Vac. Sci. Techno!. A, Vol. 12, No.2, Mar/Apr 1994

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 169.233.7.23 On: Fri, 21 Nov 2014 18:20:35
 481 Elstner €It al.: Structure and properties of TIN thin films 481

of -100 V will be too high to form this phase during the

deposition processo On the other hand, regarding the quite
-[-{lOl
:>-
• [ow value of the ratio TJTm=O.07 [substrate temperature
OJ T,=200 DC; melting temperature T m =2949 DC (Ref. 30)],
!T . <t02
oj the mobility of adatorns is expected to be very IOW. 13 •31 Thus
~ ..0.03 o it is unlikely, that the initially formed cubic TiN phase is
<l fJ. o
partially transformed into the €- Ti2N phase in deeper film

Fl
C

areas.
)( This assumption is assisted by an experiment described by
u~J
& & PN2 "0.01 Pa
00 PN2 '" 0.02 Pa Valvoda et a1. 15 In this experiment a TiN layer (containing
.... PN2 = O.017Pa 34 at. % N) has been deposited onto chromium steel sub-
3~) x)( PN2 '" 0.05 Pa
strates at a temperature of 150°C. XRD showed only a pure
~L __ ~_----:"O~':...~
--
TiN phase in the diffraction pattern. After annealing the
sample for 3 h at 700 "C, the E- Ti2N phase appeared dearly
in the diffraction pattern in addition to the TiN peaks. The
Ho. 11. Lattice distortions fla calculated from d zo () in dependence of sin? 1/1. annealing process appeared to be sufficient to form the
The corresponding stress values are given in Table II. E- TizN - 8- TiN composition, which is stable at 700°C in the
concentration range.
The transition region between metallic and reactive depo-
sition mode (PN, =0.005-0003 Pal is characterized by a
into account the anisotropy factors of the XEC. 3 The stresses
maximum of grain size as revealed by XRD (cf., Fig. 6).
are compressive and of a high level. Their values vary mark-
Different film structures can occur. In this region film
edly for different nitrogen partial pressures but increase iso-
structure and properties like hardness are extremely sensitive
tonically with 0
a60
(cf., Table II). This good correlation be-
against thc slightest nitrogen partial pressure fluctuations.
tween the biaxial stress in the surface plane and the lattice
While XRD detects a significant dependence between the
spacing perpendicular to it is expected from the classical
nitrogen pa11iai pressure and the film structure, EPMA shows
theory of elasticity.
that from Px, =0002 Pa up to the highest nitrogen pressure
Deviation from the linear behavior was observed for all
under investigation a stoichiometric relation between Ti and
samples (cf., Fig. 11). Geometrical consideration shows that
N exists. This relation seems to be nearly independent of the
the path of the radiation through the layer and thus the ab-
nitrogen partial pressure. A similar observation has been
sorption of x-rays increase with !/J. That means, that less
made by Stanislav et al. 19
information from the deepcr layer regions will contribute to
the diffracted signal. Thus, nonzero perpendicular stress
leads to a deviation from the normally linear dependcnce Erp", B. Lattice distortion
versus sin 2 ifl.27 Lattice distortions with 116 11 significantly greater than
In addition to that, we observed if! splitting in some cases, a3° o had been detected in earlier investigations of our group
i.e., that the slope has been different for negative and posi- in numerous TiN samples deposited by dc magnctron sput-
tive !/J. Such a behavior can be explained by the presence of tering and hollow cathode arc deposition onto steel and sili-
shear stress components in the layers. 28 con substrates. The observed distortion has been always
greater in layers on steel substrates in comparison to those on
IV. DISCUSSION silicon. 24 There exist models of Goldfarb et at., Knotek
A. Crystallographic phases et at., and Valvoda et at}·'! which give important contribu-
tions for the interpretation of this behavior. These models
In the region of nitrogen partial pressures between 0.001 taken into account the anisotropic influence of nitrogen or
and 0.005 Pa only the hexagonal titanium phase could be inert gas atoms on octahedral (Knotek et al.) and/or tetrahe-
detected by XRD, whereas in all samples deposited at higher dral (Goldfarb et al.) interstitial places or of dislocations
pressures exclusively the cubic TiN phase was found. (Valvoda et al.) on the picostructurc of the TiN lattice. Rick-
It is usually accepted that in magnetron sputtered TiN erbv et af. 7 assumed this distortion to be only a Poisson ef-
coatings the €- Ti2N phase is seldom present. 19 However, fect caused by macroscopic stresso
some authors report about its presence in such layers. 15,29 We In this work, we measured the lattice distortion in depen-
could not find any peak of the €- Ti2N phase in our layers. dence on the nitrogen partial pressure (cf., Fig. 9). The
Regarding the Ti-N phase diagram,20 the formation of yielded results are partially similar like those described in
E- Ti2N occurs only at temperatures below 1050 °C in the Ref. 12; we found in the range of lower nitrogen partial
range of 3 at % N to 41 at. % nitrogen. Because sputtering is prcssure a lattice widening in the (100) direction in compari-
a nonequilibrium process, we cannot use exactly this tem- son to the (Ill) direction, and at higher pressures a (Ill)
perature value to explain the absence of E- Ti2N in our layers. directed widening. First we want to discuss the latter case.
But taking into account the phase diagram conditions, we
suppose that there are two mechanisms which suppress the 1. a~l1 >#000
formation of this phase. On one hand, the particle energies The observed effect could result from two different but
due to the sputter parameters as for example the bias voltage interacting mechanisms. First, there are point defects in form

JVST A - Vacuum, Surfaces, and Films

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 169.233.7.23 On: Fri, 21 Nov 2014 18:20:35
 482 Elstner et sl.: Structure and properties of TiN thin films 482

of vacancies in both sublattices as well as argon or nitrogen of a lattice relaxation process i.n TiN below 400°C. So we
atoms on interstitial places. Due to the higher density of assume that the low temperature sufficient in our experi-
possible interstitial places in the (111) planes than in the ments to obtain significant effects is due to the low deposi-
other ones, these point defects will influence the stability of tion temperature.
the elementary cells and thus of the crystallites in an aniso-
tropical manner. Second, there exist macroscopic stresses in 2. agoo>8Jll
the layer, which cause forces on the grains. Because of the A sample deposited atpNo=O.Ol Pa showing a~OO>abll
anisotropy of the stiffness of the crystal lattice the grains are after deposition was annealed in vacuum at 350°C. After 1 h
defonned in dependence on their orientation. <lll)-oriented of annealing, the macroscopic stress had been decreased
crystallites will be deformed in such a way, that a widening from initially 8250 MPa (see Table If) to 5900 MPa. Due to
of the (111 )-pJane spacing perpendicular to the sample sur- annealing, the lattice parameter at room temperature, calcu-
face will occur. The (lOO)-oriented crystallites will be less lated from d III did not change within the measuring error of
deformed than the (111) directed. ±O.002 A (a6 11 =4.257 A before, a6
11
=4.255 A after),
The result of such a behavior is shown in the higher pres- whereas that calculated from d 200 decreased from 4.270 A
sure range of Fig. 9. From geometrical considerations fol- (before) to 4.258 A (after). The lattice loses practically its
lows, that, if d 111 becomes greater, the spacing of the (Ill) distortion but stays expanded. This experiment in addition to
place decreases and vice versa. In earlier investigation,24 we the other results gives cause to suppose that other mecha-
could show that the peak positions of TiN samples can be nisms affect these layers than those discussed for the high
described by a rhombohedral cell with rhombohedral angles PN2 case. Especially the thermal stability of the interplanar
a of about 89.5". But we could not detect the peaks of the spacing of the (11 1) plane is higher than that of the (200)
(111) or of other planes with partially negative hkl values planes.
[(311), (220)]. Applying the above described hypothesis on Considering the chemical film composition (Fig. 2), it can
the <11] )-directed grains, a rhombohedral distortion of the be seen that the region of the low PN2 type of splitting of the
lattice should be possible, but with XRD at Bragg-Brentano lattice parameter is characterized by substoichiometry refer-
geometry only the peaks caused by the (111) planes will ring to nitrogen, Le., a high percentage of vacancies in the
appear in the diffraction pattern, whereas those of the (111) nitrogen sublattice.
and (11 1) planes are not detectable. Necke1 32 investigated theoretically t,l}c influence of vacan-
This hypothesis is supported by the fact that the cies on the chemical bonding in a hypothetical ordered
a611
- a5 00
splitting occurs only in samples produced at TiNo.75 stmcture. The elementary cell of this structure can be
rather low and rather high nitrogen partial pressures, i.e., described as Ti(6)Ti~4)N30 (Ti(6)-titanium atom with 6 nitro-
conditions of sample preparation where a high concentration gen neighbors as In stoichiometric TiN; Ti(4)-titanium atom
of point defects is likely. In the layers grown at low nitrogen with 4 nitrogen neighbors; D-vacancy). The vacancy is
partial pressures exist many vacancies (this will be discussed placed in the center of the elementary cell. This highly hy-
below), whereas in the layers deposited at high pressures a bridized Ti 3d - N 2p bonding is not changed significantly,
surplus of nitrogen atoms is to be expected. but the number of this kind of bondings is decreased due to
Another indication supporting this explanation is the fact the absence of 25 at. % of the nitrogen atoms. from that the
described in detail in Sec. III H (see Table II and Fig. 10), Ti 3d - Ti 3d bondings become more important. Three types
that the biaxial stress determined by the SSPM corresponds of them can be distinguished: At first, the Ti(4)_ Ti(6) bonding
with the splitting shown in Fig. 9. is weaker than the Ti-Ti bonding in the stoichiometric com-
The presumption of weakly bound N or Ar atoms situated pound. Second, a weak Ti-D-Ti bonding appears between
on interstitial sites has been supported by earlier investiga- titanium atoms which have a distance of ao from each other.
tion of our group:16 TiN layers comparable to those used in Third, the bondings between the Ti(4) atoms along the edges
this investigation and initially exhibiting a similar splitting as of the octahedron surrounding a vacancy become stronger,
that presented in Fig. 9 were annealed in air at different because in stoichiometric TiN an electronic charge transfer
temperatures (100, 200, 300, and 400°C). Annealing time happens from titanium to nitrogen. This change is now avail-
was 6 h in each case. Monitoring the layer structure during able for the titanium atoms and makes the ddf7 bondings
annealing in situ by special x-ray diffractometer we found better overlapping.
that the splitting started to decrease at a temperature as low Necke1 32 did computer simulations of the density of states
as 100 °C and disappeared completely during the 400°C an- (DOS) and found a vacancy peak about 2 eV below the
nealing. After the heat treatment the lattice had practically Fermi level caused in part by the three types of Ti-Ti bond-
lost its distortion and the lattice parameters a(\ II (at room ings described above. Other authors had found practically the
temperature) had been altered from initially 4.263 and 4.268 same DOS experimentally in TiN containing randomly dis-
A to 4.241 and 4.247 A, respectively, while a600=4.240 A tributed vacancies on nitrogen sites?3.34 From that he con-
stayed constant. cluded, that the chemical bonding in ordered TiNo.75 and in
Perryll reported the decrease of the TiN lattice parameter stmctures with randomly ordered vacancies is similar.
by a high temperature treatment at temperature values be- In the ordered TiNo.75 , the arrangement of the Ti-Ti bond-
tween about 500 and 900°C, but did not mention any lattice ings is different in different directions. In the (100) direction
distortion. At temperatures below 500 °C no effect could be one of four titanium atoms does not contribute to the bond-
detected by this author. In fact, we could not find any report ing between adjacent planes because the sequence -Ti-O-

J. Vac. Sci. Techno!. A, Vol. 12, No.2, Mar/Apr 1994

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 169.233.7.23 On: Fri, 21 Nov 2014 18:20:35
 483 Elstner et al.: Structure and properties of TIN thin films 483

Ti-O- continues in straight lines in this direction. In the (llI) neutrals during the deposition process and the limited mobil-
direction each Ti(4) atom is connected with the next nitrogen ity of adatoms due to the low substrate temperature.
plane by two instead of three (in comparison to stoichio- Two different cases of lattice distortions were detected for
metric TiN) Ti 3d-N 2p bondings. That means that each low and high PN2' respectively. In the higher pressure range,
titanium atom contributes to the strong Ti-N bonding in the the ail II is greater than the a~oo. For this the variable density
(111) direction between neighboring planes. In addition to of interstitial sites in different planes together with the mac-
this, the vacancies are not ordered straightly in this direction. roscopic film stress were found to be responsible. At lower
Considering our results (cf., Fig. 9), we suppose that this p N2' the a 6I I was smaller than the a ~OO. This effect could be
difference in the distribution of bondings between the (111) explained by the presence of a high number of vacancies in
and the (100) planes, respectively, makes the mechanical sta- the lattice. These vacancies weaken the chemical bonding in
bility depending on the direction. a different way for different planes thus leading to an anisot-
Figure 11 shows that in the low pressure region the mac- ropy of the stability of the interplanar spacings. Therefore
roscopic stress is compressive and of a remarkably high macroscopic stresses result in an anisotropic deformation of
value. This gives cause to suppose that the aA II_a~OO split- the film crystallites. Both explanations could be supported by
ting arises from an interaction between structural and chemi- annealing experiments.
cal properties on the one hand and the high compressive
stress on the other hand. If the different arrangement of Ti
3d-N 2p bondings leads to a higher mechanical stability in IS. Kadlec, J. Musil, and I. Vyskocil, Surf. Coat. Techno!. 54/55, 287
the (Ill) planes in comparison to that of the (200) planes, the ([992).
compressive stress would be able to widen the (200) more 2A. J. Perry, Thin Solid Films 170, 63 (1989).
3A. J. Perry, Thin Solid Pilms 193/194, 463 (1990).
than the (111) planes due to the Poisson effect. This would
4A. J. Perry, V. Valvoda, and D. Rafaja, Thin Solid Films 214. 169 (1992).
be in agreement with our observations (Fig. 9). 5J. A Sue, Surf. Coat. Techno!. 54/55, 154 (1992).
We believe that these results can also contribute to under- "B. O. Johansson. M. K. J. Hibhs. and J. E. Sundgren. J. App!. Phys. 58,
stand the experimentally found plane dependence of the 3104 (1985),
7D. S. Rickerby, B. /I.. Bellamy, and A M. Jones, Surf. Eng. 3, 138 (1987).
XEC 2- 5 on an atomic scale.
~I. Goldfarb, 1. Pelleg, L. Zevin, and N. Croitoru, Thin Solid Films 200,
117 (1991).
9v. Valvoda, A. J. Perry, L. Hultman, J. Musil, and S. Kadlec, Surf. Coat.
C, Film stress
Techno!. 49, 181 (1991).
The stress in our films is the result of at least two mecha- 10J. Brunner and A J. Perry, Thin Solid Films 163, 49 (1988).
11 A 1. Perry, 1. Vac. Sci. Tec~nol. A 6, 2140 (1988).
nisms being effective in connection with the layer deposi- 12v. Valvoda, R. Kuzel Jr., R. Cerny, L. Dobiasova, 1. Musil, and V. Poulek,
tion. One is the thermal stress due to different thennal ex- Cryst. Res. Techno!. 23. 1483 (1988).
pansion coefficients of substrate and film. The other is the 13S. J. Bull, A. M. Jones, and A. R. McCabe, Surf. Sci. Techno!. 54/55, 173
intrinsic stress caused by the growth process itself. Possible (1992).
14L Petrov, L. Hultman, J.-E. Sundgren, and J. E. Greene. 1. Vac. Sci.
mechanisms are the formation of point defects (vacancies or Techno!. A 10, 265 (1992).
incorporated impurities) or the impact of energetic ions on 15v. Valvoda, R. Cerny, R. Ku~C! Jr., L. Dobia.~ova, 1. Musil, V. Poulek, and
the growing film surface. For a simple estimation, the ther- J. Vyskocil. Thin Solid Films 170, 201 (1989).
16E Richter, H. Kupfer. H. Giegengack, G. Schaarschmidt, F. Scholze, F.
mal stress due to cooling down from coating temperature to
Elstner. and G. Hecht, Surf. Coat. TechnoL 54/55, 338 (1992).
room temperature can be calculated from: 17Y._L Chen and .T.-G. Duh. Surf. Coat. Technol. 46. 371 (1991).
IgE. Macherauch and S. Miiller, Z. Angew. Phys. 13,305 (1961).
O"th= (af- fXs)(Ts- Tn,)EI(I - vf)' .91. Stanislav, J. Sikac, and M. Cermak. Thin Solid Films 191,255 (1990).
lOT. B. Massalski, Binary Alloy Phase Diagrams (American Society for
Here, 0:f and as are the thermal expansion coefficients for
Metals. Metals Park, OH, 1986). Vol. 2, p. 1652.
film and substrate; T\. is the deposition temperature, and Tm 21y' Igasaki. H. Mitsuhashi. K. Azuma, and T. Muto, lpn. J. App!. Phys. 17,
is the room temperature. Using bulk material values for the 85 (1978).
expansion coefficients from the literature it can be shown 22R. Glocker, Matcriaipriijung mit Rontgenstrahlung (Springer, Berlin,

that the thermal stress in TiN layers on silicon substrates is 1985), p. 414.
23v. Valvoda, R. Cerny, R. Kuzel Jr., M. Musil and V. Boulek. Thin Solid
tensile in nature and about one order of magnitude smaller Films 158, 225 (1988).
than the total stress measured by the x-ray method. That 24H. Giegengack and F. Elstncr, in Verh. 10 Tagung Grenzjliichell/Diinne

means that the intrinsic stress is predominantly in our case. Schichten. edited by W. Schwenke (Dresden, 1990), Vol. 10, p. 152.
25H. Kupfer, F. Elstncr, G. Hecht, and F. Richter, poster at the Twelfth
International Vacuum CongressiEighth International Conference on Solid
V, CONCLUSIONS surfaces. Den Haag, October 1992 (unpublished).
26R. R. Manory and G. Kimmel, Thin Solid Films 150,277 (1987).
Titanium nitride layers on silicon substrates have been 27H, Dolle ami J. B. Cohen, Metall. Trans. A 11, 159 (1980).

deposited by magnetron sputtering at different nitrogen par- 28L. Chollet and A. J. Perry, Thin Solid Films 123, 223 (1985).
29J. E. Sundgren. Thin Solid Films 149,21 (1985).
tial pressures and at a constant temperature of 200°C. We
.lOB. Matzke, in The Physics and Chemistry of Carbides, Nitrides, and
found distinct dependences between PN2 in the reactive gas Borides, edited by R. Freer (Kluwer. Dordrecht, 1990), p. 359.
and the mechanical (hardness, film stress) and structural 31 A. Niedhardt. U. Reinhold. E. Schroeter, and W. Wuttke, Thin Solid Films

properties (chemical composition, grain size, texture, lattice 192. 263 (l990).
32A. Neckel. in Ref. 3D. pp. 485--511.
parameter, and lattice distortion) of the layers.
33It Hochst, R. D. Bringans, P. Steiner, and Th. Wolf, Phys. Rev. B 25,
A general absence of the Ti2N phase in our layers can be 7183 (1982).
explained by a too high energy impact by energetic ions and 34L. Porte, L. Roux, and J. Hanus, Phys. Rev. B 28, 3214 (1983).

JVST A - Vacuum, Surfaces, and Films

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 169.233.7.23 On: Fri, 21 Nov 2014 18:20:35