Thin Solid Films, 146 (1987) 165-174 METALLURGICAL AND PROTECTIVECOATINGS
165
T E M P E R I N G E F F E C T S I N I O N - P L A T E D TiN F I L M S : T E X T U R E , R E S I D U A L STRESS, A D H E S I O N A N D C O L O U R A. J. PERRY*
lnstitut Straumann, CH-4437 Waldenburg (Switzerland)
(Received April 23, 1986; accepted August 19, 1986)
The contraction of the expanded crystal lattice, inherent in films made by plasma-enhanced vapour deposition, occurs on tempering with an activation energy of about 2.09 eV and is accompanied in the present films by an increase in residual stress and the development of a texture in the plane of the film. The intrinsic stress falls so the net stress increase is attributed to a concomitant substrate contraction. At the same time the colour of the films becomes more deeply yellow in accordance with recent discussion of the simple ionic model. The scratch test adhesion is not significantly affected by the tempering and the increase in the residual stress. Indeed, adhesion increases slightly with tempering probably owing to interdiffusion with the substrate.
l. INTRODUCTION It is well established 1-4 that the crystal lattice of TiN films, made by plasmaenhanced vapour deposition methods, is frequently expanded above the equilibrium value over a wide range of nitrogen contents. This expansion is reduced by tempering at elevated temperatures 5. The contraction is more or less complete by 900 °C although extended tempering to achieve the equilibrium value may not be advisable as nitrogen loss is likely to occur 6' ~ at such temperatures. Previous studies s - l ° have indicated that the amount of trapped argon and the type of lattice defects (vacancies or interstitials) depend upon the conditions under which the films were made. It seems reasonable to assume that all these defects follow the usual diffusion relations so that a lower tempering temperature can be used if the time is correspondingly increased. The objective of the present study is to establish the effect of tempering on ionplated TiN taken as being typical of plasma-enhanced vapour-deposited films. The primary interest is in reducing and stabilizing the stress level through diffusioninduced relaxation from the high values 11 found in the as-received condition. It was anticipated that the colour of the films would also change, as found previously 9' ~o. A * Present address: GTE Valeron Corporation, 1711 Thunderbird, Troy, M148084, U.S.A. 0040-6090/87/$3.50
© ElsevierSequoia/Printed in The Netherlands
166
A . J . PERRY
secondary interest is to establish whether this can also be stabilized by tempering which could thus be used to combat the aging effects reported before 5. Tempering at temperatures of the order of 900 °C 5.9 may not be an optimum treatment either for cemented carbide because of possible decarburization and consequent embrittlement, or for stainless steel because of possible stress relief 11 and distortion effects in heavily formed objects such as watch cases when coated for decorative purposes. Hence a lower tempering temperature is desirable. 2. EXPERIMENTAL PROCEDURE AND RESULTS
The samples chosen for study were designed to be of practical significance. A series of industrial cemented carbide P50 quality S N U N 190416 inserts were coated with about 3 txm TiN in a single process, to ensure uniformity of properties, by ion plating following standard industrial procedures. The films were chosen at r a n d o m from the series for the tempering treatments to improve the statistical significance. The lattice parameters, residual stress, colour and scratch adhesion of the films were then studied. Tempering was carried out in a programmable tempering furnace operating at a vacuum of better than 1 m P a in the presence of a titanium getter. The m a x i m u m available values of 99 h and 930 °C were used as upper limits in this study. X-ray diffraction studies indicated the films to be very strongly (111) textured as before 9. Lattice parameter measurements were made with a counter diffractometer the calibration of which was confirmed with the Si(111) diffraction peak. The lattice parameters, peak width at half-maximum ( P W H M ) and peak intensity of the (111) diffraction peak, determined with Cu K0t radiation, are shown in Fig. 1. The diffraction peaks become narrower and more intense, moving to higher 20 values, i.e. to lower lattice parameters, as tempering proceeds. The residual stress in the plane of the films was determined using the sin2~k method as given in detail elsewhere 12. In spite of the strong texture, reliable measurements were possible using the (420) diffraction peak from Co K0t radiation. The data were evaluated using a straight line fit assuming E to be 616 G P a 13 and v = 0.20 14 and the compressive stresses are given in Fig. 2(a). For completeness, the values of the lattice parameters of the (420) plane for ~, = 0 are included in Fig. 2(b). It is seen that they follow the same trend as the (111) plane data in Fig. l(a). Whilst a greater film thickness would have made the experimental data more precise, it had been found previously11 that the stress in thicker TiN films on cemented carbide was lower and can be accompanied by a coarsening of the grain structure. The sin2~k plots showed neither curvature nor ff splitting due to stress gradients or shear stresses respectively 11. However, films annealed at temperatures of 800°C and above showed a distinct serpent's tail behaviour (Fig. 2(c)), which is due to the development of a texture in the plane of the film 15'16. An attempt was made to confirm this by making pole figures from selected samples. Unfortunately, the (111) texture is so strong that no changes could be established within the experimental accuracy of the methods. All the samples were cleaned in an industrial degreasing and Freon drying sequence before the colour and adhesion measurements were made. The CIE 1976
TEMPERING EFFECTS IN ION-PLATED TiN FILMS
167 36.4
•
0.6260
B
36.6 ~ (D o4
0.42~0 36.8
-
0.4220
I
200
(a)
I
I
I
I
400 600 BOO 1000 tempering temperature (C)
0.60
-r 3~
O.t~O
0.20
I
I
200
1-00
I
600
I
I
800
1000
tempering temperature ( C}
(b) 100 c
,4
b ~'-
50
o
~
o~
.c_ I
2OO
(c)
I
I
I
I
800 1000 tempering temperature {'C) 400
600
Fig. 1. X-ray diffraction data from the (111) peak measured with Cu Kct radiation, showing (a) lattice parameters, (b) P W H M and (c) intensity, as a function of tempering temperature and time: O, as received; O , tempered for 1 h;/% tempered for 10 h; [], tempered for 99 h.
L*,a*,b* colour coordinates were determined under diffuse D65 illumination with d/0 and d/8 measurement geometries. Minolta C R 1 0 0 and Macbeth 1500 calorimeters respectively were used. As found previously 5, the latter returned rather higher values of all the parameters but the trend of both sets of curves is the same. The rough surface of the coated cemented carbides is not ideal and affects z7 the results of colour measurements so that they should not be compared quantitatively with the earlier data 5'9. Such comparison would require the films to be diamond polished or the substrate to have been polished before coating. The latter could have affected the state of residual stress in the films, the central interest of the study. The colour data are given in Fig. 3 as a function of the reciprocal lattice volume (RLV); this is more convenient for a discussion in terms of charge carrier concentration. The colour data of an unpolished film made by chemical vapour deposition (CVD), assumed to be in crystallographic equilibrium, are included for reference.
168
A.J. PERRY 3O00
's-
20O0 &
1000
tJ
o
0
I
I
i
200
t+00
600
(a)
I
I
' 800
1000
tempering temperofure ( [ )
E
0.~55
0
0.t,250 I
I
I
200
600
600
I
1000
tempering temperofure (C)
(h) --
I
8OO
0.&255 0.&250 0.~245
0.~2~0
I
I
0.1
0.2
I
I
l
~
!
0.3 0.4 0.5 0.6 (c) sin 2 dr, Fig. 2. X-raysin2q~stress measurements from the (420) peak with Co Kct radiation, showing (a) residual compressive stress in the plane of the film and (b) lattice parameter at ~b= 0, as a function of tempering temperature and time (symbolsas in Fig. 1),(c) a sinZ~bplot showing serpent's tail behaviour taken from a film tempered for 1 h at 900 °C.
Adhesion was studied using a Revetest unit is with a steadily increasing load. The critical loads for the onset of film loss by flaking 19 are recorded in Fig. 4; each film was tested twice. The loads were d e t e r m i n e d microscopically a n d confirmed from the accelerometer traces which were quite typical 2°. T w o samples were also tested which had exceptionally become slightly tarnished d u r i n g the tempering treatment. The critical loads were some 5 - 1 0 N lower which is p r e s u m e d to be due to increased friction 21 between the d i a m o n d stylus a n d the film. Finally, the m i c r o s t r u c t u r e of selected films was studied in fracture-section with a s c a n n i n g electron microscope. I n a d d i t i o n to an as-received sample, the samples chosen had all been a n n e a l e d for 1 h a n d their residual stresses had been determined. N o m i c r o s t r u c t u r a l differences could be detected; typical fractographs are given in Fig. 5.
TEMPERING EFFECTS IN ION-PLATED T i N FILMS
169
L~ 0~ b~ IO0 - 2 0 . 1 0 0
80
8O
15 60-
~ I
~,
~
8O
60
_
10 • t+O.
20 -
•
0 12.8
o
•
~ /
I
I
I
I
I
I
i
I
13.0
13.1
13.2
I/o3
..Qo_
/
"6 '-'
12.9
R LV
~ ./
•
20
O-
60
z
b*
/.0
5
O.
L~
I
0
(nrn-3)
I
I
I
I
200
400
6OO
800
tempering
temperafure
I
1000
(C)
Fig. 3. CIE 1976L*, a*, b* colour coordinates measured with D65 illumination under d/0 (O, A, IS])and d/8 (O, A, l ) geometry as a function of RLV; symbols for tempered filmsas in Fig. 1. Fig. 4. Scratch adhesion test data showing the critical load as a function of tempering conditions; symbols as in Fig. 1.
(a)
(b)
Fig. 5. Typical scanning electron fractographs from the present films(a) as receivedand (b) tempered for 1 h at 800 °C. 3. DISCUSSION
3.1. Lattice parameters The lattice p a r a m e t e r d a t a in Fig. l(a) allow a n estimate to be m a d e of the a c t i v a t i o n energy of the diffusion process. It is usual in diffusion studies of alloys to analyse the diffusion profile formed between the two c o m p o n e n t s as a function of time a n d t e m p e r a t u r e by such m e t h o d s as radioactive labelling or metallographic slicing. I n dealing with the kinetics of vacancy decay, however, such a n a p p r o a c h is n o t practicable because the defects are decaying t h r o u g h o u t the whole sample. A s o l u t i o n to this exercise in a n n e a l i n g theory has been given by D a m a s k a n d
170
A . J . PERRY
Dienes 22 where samples containing vacancies are heated at a constant rate and the defects anneal out over a narrow range of temperatures. The analysis of the experimental data is simplified if two samples are tested starting from the same initial conditions but with different heating rates ~1 and ~2. If T, and T2 are the corresponding temperatures at which an identical fraction of vacancies has been annealed out then the activation energy E can be derived from the equation \~1T2, / = ~Assuming that the isochronal curves in Fig. l(a) approximate to constant heating rate data, the temperatures at which a quarter, the midpoint and three-quarters of the lattice parameter change (taken as 0.4257nm, 0.4250nm and 0.4243 nm respectively} are achieved for the isochronal steps of 1 h and 99 h can be used to derive an activation energy of 2.09 eV. Very few data are available for comparison of the self-diffusion of nitrogen in the group IVb nitrides. A value of 2.26 eV was derived from titanium nitriding data by Levinskii et al. 23 This is much lower than values for the diffusion of nitrogen or carbon in Z r N given as 3.39eV 24 and 3.99 eV 25 respectively. In accordance with the inference drawn from the observation of annealing phenomena at such low temperatures (the melting point of TiN is about 2950 °C), the present result may well indicate that we are dealing with the diffusion of a non-equilibrium species such as self-interstitials. In connection with the interest in finding a moderate annealing temperature, it is not known whether a complete contraction of the equilibrium lattice parameter is necessary for stability or whether an intermediate value would suffice. However, an estimate of the order of magnitude of the tempering times can be made. The achieving of the midpoint value by tempering at about 550 °C would lead to surface oxidation at atmospheric pressure. The annealing times at temperatures of 450 °C and 350 °C can be evaluated. These are 6 weeks and 3½ months respectively, periods which are probably not viable for industrial application. 3.2. R e s i d u a l s t r e s s
Possibly the most surprising result in this study is the increase in residual stress found on tempering (Fig. 2(a)) coupled with the development of texture effects in the plane of the films. This increase in stress is in contrast with the findings of Suzuki et al. 26 who reported that the residual stress in TiC, TiN and Ti(CN) films could be reduced to negligible proportions after tempering for less than 2 h at 900 °C. The present results are not unique, however; stress increases on tempering have been found in Cr7C 3 and TiC films on steel by Janowski et al. 27 and in the oxide formed during the oxidation of nickel by Aubrey 2s. It is not at all likely that differential thermal expansion can be having any effect here. Cooling can only lead to a tensile stress in the film as the expansion coefficient of TiN is greater than that of cemented carbide 29. The increase in the compressive stress on tempering appears to be in contradiction to the falling lattice parameter (Figs. l(a) and 2(a)) and, more particularly, the falling diffraction peak width (Fig. 2(b)) because the latter effect is often associated with stress relief or grain growth. The scanning electron micro-
TEMPERING EFFECTS IN I O N - P L A T E D
TiN
FILMS
171
graphs (Fig. 5) show no evidence of the latter and can be contrasted with the very marked growth reported by Wendler 3° in metastable T i 2 N - T i N films tempered at 900 °C and above. It is possible that an additional effect is to be sought in a third contribution to the diffraction peak width. The diffusion of self-interstitials onto lattice sites promotes lattice perfection and this would cause a reduction in the peak width. The observations can be understood in the following way. If it is accepted that the reduction in the values of at111), a~=o and the P W H M indicates a reduction in the intrinsic stress arising during the deposition process, then the stress increase found on tempering can be attributed to a concomitant contraction in the cemented carbide substrate causing a contractile stress in the film. This can be estimated as follows. Rickerby 3~ showed recently that the intrinsic stress and the value of a,= o are linearly related. In the present work the latter does not reach (Fig. 2(a)) the equilibrium value (Fig. l(a)) during tempering. Assuming its linear relationship with a~= 0, the intrinsic stress falls from some 1700 M P a in the as-received condition to about 1200 M P a after a tempering treatment at 900 °C where the experimentally measured value is about 2400 MPa. The difference is attributed to a contractile strain imposed by substrate contraction. Taking Exis to be 616 G P a 13 as above then this strain is about 0.2~o. TiN is elastically anisotropic as is evident from its low Poisson's ratio value ~4. Indeed, the differences in the lattice parameters measured in stressed films in a previous study 9 indicate that the ratio is different for different crystallographic directions. It might be expected that such a variation is also to be found in the Young's modulus. It is therefore plausible that the change in the texture reflects a stress-induced (and lattice-vacancy-aided) re-ordering of the crystallites to favour a direction of lower Young's modulus in order to offset the stress increase accompanying the substrate recovery. Studies of TiC and HfC indicate 32 that the (110) plane is favoured for slip at low temperatures so possibly this texture is developed by tempering of TiN. It is worth adding that some stress relief is expected to accompany the interdiffusion thought to be responsible for the increased adhesion found on tempering discussed below. The measurements also support the contention that the state of residual stress in the substrate needs to be determined concurrently to understand changes in the stress in the films. This can also be done using the X-ray sin2~k method 33,34. 3.3. Colour
It is convenient in the following discussion to consider (Fig. 3) the colour coordinates as a function of the RLV which is proportional to the charge carrier concentration. If the CVD film is considered to be in equilibrium then the RLV fall, in moving from it towards the as-received film, is accompanied by a decrease in both a* and b* (the red cast and yellowness respectively). The original simple ionic model 9 predicted that a fall in RLV should promote yellowness under otherwise constant conditions (composition, texture etc.) because the increase in carrier concentration shifts the reflectance curve en bloc to lower energies. This is in direct contrast with the present data.
172
A . J . PERRY
Departures from the simple model have been recorded 1°. In a series of films of constant composition, those with a small RLV change had a less intense colour than a C V D film. The knee of the reflection curves was found to have been displaced to lower photon energies, i.e. the slope of the reflectance curve had been reduced. As the RLV was increased further, the yellow chroma began to increase in conformity with the simple ionic model. These data implied that an initial small decrease in RLV (not included in the range of films then available) would have led to a steadily decreasing yellow chroma. The present experimental data represent a confirmation of this. There is an additional contribution. The present interpretation of the X-ray diffraction peak widths indicates that the films with expanded lattices are crystallographically less perfect. The value of the reflectance at the knee of the reflectance curve is then lower so that the colour is less intense. These two effects are additive. Let us return to the objective of stabilizing both stress and colour. Although tempering has increased the total stress it has also reduced the intrinsic stress and increased the crystallographic perfection. The lattice parameters a~l ~1) are close to equilibrium so that it seems unlikely that further aging is to be expected: the remaining lattice defects are most probably vacant lattice sites which require bulk diffusion to remove. Of practical significance is the observation that their presence is favoured by high deposition rates whereas the proportion ofinterstitials is reduced ~°. On the question of general stability, a further point is worth noting. A limited amount of tempering at temperatures of the order of 400°C has not had any measureable effect on the X-ray diffraction data (Fig. 1) but has reduced the scatter in both the residual stress (Fig. 2(a)) and the scratch adhesion (Fig. 4). Hence, storage even at ambient temperatures should be beneficial as has indeed been demonstrated so clearly in his studies of carbide film adhesion by Sprou135. 3.4. Adhesion
The increase in the scratch test critical load with increasing tempering is assumed to be caused by interdiffusion between the film and the substrate coupled with the increased lattice perfection of the films. It is surprising that the increase in residual stress appears to have no effect or, at least, that its effect is not significant in well-adhering films. 4. CONCLUSIONS Tempering of ion-plated TiN films shows that the lattice contraction occurs by a diffusion process with an activation energy of 2.09 eV. It is attributed to the diffusion of self-interstitials into excess lattice vacancies both of which were formed during the coating process. Both stress and colour can be stabilized but tempering does not appear to be an industrially viable process. The residual lattice stress is compressive in the plane of the film and increased during tempering concomitant with the lattice contraction; this has been observed in other systems 27'2a and is caused by a contraction of the substrate in the present samples. The change in the colour of the films is in direct contrast with the predictions of
TEMPERING EFFECTS IN ION-PLATED TiN FILMS
173
the simple ionic model. It agrees with more recent observations and relates to the variation in the position of the knee and slope of the reflectance curve 1° where an initial shift to lower photon energies for small volume expansion occurs before the shift of the reflectance curve as a whole dominates the colour change at larger volume increases, in accordance with the ionic model. Tempering causes a slight technologically insignificant increase in the adhesion scratch test critical load. ACKNOWLEDGMENTS
The author wishes to express his thanks to Dr. R. Schmid for ion plating the samples, Mr. R. Christ for the X-ray diffraction and scanning electron microscope studies, Dr. L. Chollet for the stress measurements and Dr. W. D. Miinz and Mr. H. Pedersein for making colour (Macbeth 1500) and Revetest scratch adhesion tests. A Minolta CR 100 colorimeter unit was kindly made available by Mr. A. Ullrich.
REFERENCES 1 B. Karlsson, J. E. Sundgren and B. O. Johansson, Internal Rep. UPTEC 8198 R., October 1981 (Institute of Technology, University of Uppsala). 2 J.E. Sundgren, B. O. Johansson, S. E. Karlsson and H. T. G. Hentzell, Thin Solid Films, !05 (1983) 367. 3 S. Schiller, B. Beister and W. Sieber, Thin Solid Films, 111 (1984) 259. 4 I.N. Martev, G. I. Grigorov, I. G. Petrov and E. Dynowska, Thin Solid Films, 131 (1985) 303. 5 A.J. Perry, J. Vac. Sci. TechnoL, in the press. 6 R. Kieffer, P. Ettmayer and M. Freudhofmeier, Metall (Berlin), 25 (1971) 1335. 7 M. Fukuhara and H. Mitani, Jpn. Inst. Met. Trans., 21 (1980) 211. 8 A. PanandJ. E. Greene, ThinSolidFilms, 78(1981)25. 9 A.J. Perry and J. Schoenes, Vacuum, 36 (1986) 149. 10 A.J. Perry, M. Georgson and C. G. Ribbing, J. Vac. Sci. Technol., in the press. 11 L. Chollet and A. J. Perry, Thin Solid Films, 123 (1985) 223. 12 L. Chollet, H. Boving and H. E. Hintermann, J. Mater. Energy Systems, 6 (1985) 293. 13 K.J. Portnoi, A. A. Mukaseev, V. M. Gribkov and Yu. V. Levinskii, Soy. Powder Metall. Met. Ceram., 65 (5) (1968) 406. 14 E.A. Almond, Powder Metall., 25 (1982) 146. 15 H. D611e, J. Appl. Crystallogr., 12 (1974) 489. 16 H. D611e and J. B. Cohen, Metall. Trans., A, 11 (1980) 831. 17 W.D. Miinz and D. Hoffmann, Metalloberfliiche, 37 (1983) 279. 18 H.E. Hintermann, P. Laeng and P. A. Steinmann, Proc. Int. Ion Engineering Conf., IPA T83, Kyoto, 1983, Institute of Electrical Engineers of Japan, Tokyo, 1983, p. 1115. 19 A.J. Perry, Thin Solid Films, 107 (1983) 167. 20 P.A. Steinmann and H. E. Hintermann, J. Vac. Sci. Technol. A, 3 (1985) 2394. 21 J. Valli, J. Vac. Sci. Technol., in the press. 22 A.C. Damask and G. J. Dienes, Point Defects in Metals, Gordon and Breach, New York, 1963, p. 148 et seq. 23 Yu. V. Levinskii, Yu. D. Stroganov, S. E. Salibekov, M. Kh. Levinskaya and S. A. Prokofiev, Izv. Akad, Nauk SSSR, Neorg. Mater., 4 (12) (1968) 2068. 24 I.I. Spivak, Izv. Akad. Nauk SSSR, Neorg. Mater., 5 (6) (1969) 1938. 25 Y.F. Khromov, V. P. Yanchur and V. S. Eremeev, Fiz. Met. Metalloved., 33 (1972) 642. 26 H. Suzuki, H. Matsubara, A. Matsuo and K. Shibuki, J. Jpn. Inst. Met., 49 (1985) 773. 27 S. Janowski, W. Szyrle and J. Tacikowski, Metalozn. Obrobka Cieplna, 67 (1984) 14.
174
28 A. Aubry, Thesis, Tech. Univ. Compiegne, 1985. 29 J.J. Oakes, Thin SolidFilrns, 107(1983) 159. 30 B. Wendler, Thin Solid Films, 141 (1986) 223. 31 D. Rickerby, J. Vac. Sci. Technol., in the press. 32 D.J. Rowcliffe and G. E. Hollox, J. Mater. Sci., 6 (1971) 1270. 33 T. Yamamoto and K. Kamachi, J. Jpn. Inst. Met., 49 (1985) 120. 34 A.J. Perry and L. Chollet, J. Vac. Sci. Technol., in the press. 35 W.D. Sproul, J. Vac. Sci. Technol., in the press.
A . J . PERRY