Origins of the residual stress in CVD diamond films

(1)

ELSEVI ER Thin Solid Films 290-291 ([996) 254-259

Origins of the residual stress in CVD diamond films

Cheng Tzu Kuo *, Chii Rucy Lin, How Min Lien

lnsl~lute of Matedals Science and EnglMeering, National Chino Timg Univer,ffty, I001 Ta-lfsueh Road, Hst~cfiu 30050, Taiw~n

Ahstrant

Diamond films were deposited on (100) Si wafer, WC (5%C0) and quartz substrata materials by a microwave plasma chemical vapor deposition (CVD) system. The effects of deposition and substrata conditions on residual stress of the films were systematically investigated. The films were characterized by scanning electron microscopy, X-ray diffraction, Raman and indentation adhesion testing, The film structure including its non.diamond carbon content, crystal size. texture co~fficiant, film thickness and surface roughness were examined. The results show that the residual stress of the films is a tunction of the surface pretreatment, in addition to the substrata material and deposition conditions. The origins of the residual stress are mainly the thermal stress and the intrinsic stress. The intrinsic stress is mainly from the effect of the non- diamond carbon content in the diamond crystals, not at the nr3,stal boundaries. A greater non-diamond carbon content in diamond crystals results in a greater residual stress. The texture of the films has no significant effect on the residual stress. A low compressive restdnal stress on Si wafer is beneficial to the adhesion of the film.

geywords: Diamond film; Residual s ~ s g Interface structure

1, Introda¢flon

It is well known that diamond is a unique material in both fundamental science and engineering applications. There- fore, the growth of chemical vapor deposition (CVD) diamond films on various substrates has been the subject of expanding interest, One of the major issues of concern for CVD diamond films is to understand the origins of the residual stress of the films, so that we can manage and control these stresses. In general, the possible origins of the residual stress include thermal stress, phase transformation stress, epi- t~xial stress, and intrinsic stress, etc, Thermal stress is formed during the cooling down stage after the deposition, and is due to the difference in thermal expansion coefficient between the film and the substrata. Stress can also be induced due to phase transformations of the films or substrateduring cooling from the deposition temperature to the lower temperatures. For a substrata and film with similar crystal structure but different lattice constants, the interface structure between the film and the substrate can be semiooherent or coherent, The stress so developed is called "epitaxial stress". The stress can also formed due to the presence of defects, such as grain boun&~'ies, dislocations, voids and impurities, etc., and this stress is called "intrinsic stress". There are many techniques used to analyze residual stresses in the films. Roman spee-

* Cortes~ndtag author.

0040-6090/96/$15.00 O 1996 Elsevier Science S,A, All righ[s reserved

Jell SO 040-6090 (96) 09016-5

troscopy is one of the methods used by many authors to assess the residual stress in the diamond films [1-8], and is also one of the most important tools for characterizing CVD diamond. In the present report, Raman spectroscopy was used to analyze the stresses in the diamond films. The origins of the residual stress in diamond films are discussed.

2, Experimental details

Diamond films were deposited on (100) Si wafer, quartz and cemented 5% Co--WC substrates by a microwave plasma CVD system with CH4 and Hz as the source gases. The surface morphologies of the films and substrates were examined by scanning electron microscopy (SEM). The substrates were subjected to a few different surface protreatments to obtain different surface roughness values from I ~ 50 am, including polishing with 0.1, 1.0 and 6 ~m diamond paste, 30 ~,m diamond-coated sheet, and 1 ~m polishing cloth. The deposition conditions were: 0.4/100~1/100, el-h/H2 ratio; 973 ~ 1173 K; 200,., 480 W, microwave power; 15 ~ 35 Ton, total pressure; 0.2 ~ 17 h.

X-ray diffraction (XRD) analysis was performed on the films. The texture coefficient,

T(hkl),

for ca~h

(hkl)

reflec- tion was calculated from

(2)

C.7",/t'uo eraL / Thin So|/d Films 290-291 (1~6) 254-259 255 where i~stj) is the relative intensity of the (hkl) plane of the

diamond film, and 1 o chhJ~ is the relative intensity of the (hkl)

plane of a randomly oriented polycrystailine film, and is assumed to be the same as the JCPDS powder diffraction files, Tchtl) represents a measure of the growth randomness or preferred orientation. For a randomly oriented poly- crystalline film, the theoretical value is 1.0.

The adhesion of the films was determined by the indentation adhesion testing method, which was described in detail elsewhere [9]. The diameter of the cracked ~ea, X, versus the indentation load, P, was recorded, The slope of the curve, OX/dP, is used as a measure of the quality of adherence between the film and the substrata. A greater value indicates a poorer adherence.

The film quality, non-diamond carbon content andresidual stresses of the films were assessed by the Raman spectroscopy. Renan spectroscopy was developed by LeGrice et at. [4 ] to analyze the stress in diamond films. Although the beam area of Re.man spectroscope is about 10 Itm in diameter, which is much greater than the grain size of the films in the present conditions, it is realized that the Renan spectroscopy measures the relatively lo~alized stresses by comparing with the X-ray and substrata curvature methods, it is found that both the stress and the microcO'stalline domain size can con- tribute to a frequency shift of Renan peak, The peak will shift to lower frequency as the crystalline domain size decreases, but also the full width at half-maximum (FWHM) increases. In general, a material which is under tensile strain will exhibit a Renan peak which is shifted to lower frequency, while the Renan peak of a material undergoing a compressive strain is shifted to higher frequency. Therefore, the total observed shift of the peak, noob, can be expressed

by

Acaob = Acod+ Ace, (2)

where A O) d is the shift due to domain size effect, and A m, is the shift due to stress. According to Van Acker et el. [5], the Renan peak shift of the film under stress, A ~o,, is given by

A~,= - e ~ (3)

where o- is the in-plane balanced biaxial stress, and P is a function of Grueneisen parameter, Poisson ratio and bulk modulus of the film. The value of f for diamond films was evaluated by many authors to be 1.,70~ 3.05 on-~ GPa" 1 [4-7]. In the present report, P = 2.23 era- ' GPa- ' was used to estimate the internal stress of diamond films. By consid- ering the sensitivity of Renan signal for non-diamond carbon phase is about 75 times of that for diamond, Renan spectroscopy was also used to estimate the non-diamond carbon content, Cma, in the film by the relation [8]

¢~

= II [ [ +75ff,/I,d)] (4)

where l,j is Renan peak intensity for diamond crystals, and 1,~ is the Raman peak intensity for a non-diamond carbon phase.

3. Results and dlseu~on

3.1. Filmmorphologyondquality

The 3 E M morphologies of diamond films revealed that they depend upon the deposition conditions, substrata protreatments and subsU'ato materials. Fig. I shows some typical film morphologies under varions conditions for different substrata materials (WC (5%Co), quartz and (i00) Si wafer). The films are smooth, small grained, and polycrystafiine. Raman spectroscopy revealed that the films contain diamond and non-diamond carbon phases, as evidenced by a narrow peak around 1332.5 cm -1 and a broad peak around 1560 cm-i. Fig. 2 presents the typical Raman spectra for three different substrata materials. The spectra under various con. ditions all showed the characteristic diamond peak, The non. diamond carbon content of diamond films was determined by Eq. (4). Fig. 3 shows effects of deposition time, substrata pretreaunents and materials on non-diamond carbon content of the films.

3.2. Effect of aon.diomond carbon content on residual

Stress

The magnitude of the Renan shift from the peak position of 1332.5 c m - ' for natural diamond, ~ , and the FWHM of the diamond film peak were measured. Based on the relation between A ~ and FWI-IM, as determined by LeGrice et ai. [4], the magnitude of the shift due to domain size effect can be estimated by measuring the ~/I-IM. The shift due to s ~ s s alone can then be determined from Eq. (2). For Si wafer substrates, the residual s ~ s s of diamond crystals, which has not yet formed a film, is from -2.55 to -0.42 GPa. The deposition conditions for these two boundary oases are 1173 K, I h, 35 Torr, 480 W, 16 nm roughness, 1% methane, and 1173 K, 0.2 h, 35 Tort, 480 W, 8 nm roughness, respectively. The negative sign indicates a compressive stress. Similarly, the residual stress of diamond crystals is from - 2.53 to - 0.84 GPa, and from - 1.01 to 0.23 GPa for WC end quartz substrata materials, respectively. It implies that the r~idual stress is not related to the non-diamond carbon phase at the grain boundaries. In terms of the grain boundary relaxation model, it was proposed by Windisch- mann et at. [ I0] that the tensile intrinsic stress is related to a voided microstrucmre at the grain boundaries. The comparable strong compressive residual stresses for both diamond crystals and diamond films reveal that the contribution of grain boundary relaxation is small in the present renditions. For Si wafer subslrate materials, the residual stress of diamond films is from -2.42 to -O.0g GPa. Fig. 4 shows the effect of deposition time on the internal stress of diamond films on three different subsWates. Curve b in Fig. 4 is for the films on Si wafers. The deposition temperature in this case is 1173 K. Therefore, the thermal stress is about - I..68 GPa ( - 1200× ( 2 . 6 - I) X 1O-eX (!173-298)), and the epitaxial stress is about 17 GPa (assuming coherent boundary to match three carbon atoms with two Si atoms) [ I I ]. There is no phase transformation between the deposition tempera.

(3)

~6 ¢,T. K ~ ¢t ~f,/ Tldn sotld Pib.s 290..291 ( I ~ ) ~ ¢ - 2 D

Fig. 1. 'i~pical diamond film tno~hologies ca different sub~ti~es and at different deposl41on conditions: (a) on ~, SI wafer, I0 h, 1173 K; (b) o. a Si w~fer, 5 h, 973 K; (0) end (d) on WC (Sg~Co), 1173 K, 1 am and 2() am roughness; (e) end (f) on qati~,, 1073 K ~d 973 K,

tare sad room temperature for both the substrata and film, In other words, the epitaxial stress and phase transformation stress should not be the main contributions, and the thermal stress and inainsic stress may be the main factors. It is note- worthy that curve b in Fig, 4 is comparable with curve d in Fig, 3 for non.diamond carbon content in the films. By assuming Young's modulus of graphite is 10 GPa and the volume expansion from amo~bous carbon (density, 3 g cm -3) to graphite (density, 2.4 g cm-3), the corresponding compressive stress is about - 0 . 8 GPa ( = 10× [(1/3) '~ s _ ( 1/2.4) t/s] / [ ( 1/3) i/~1 ). Therefore, the ina'iosie stress due to transformation of non-diamond carbon to the more stable graphite in thefilmsmay be one of the important factors to the residual stress of diamond films. A greater non-diamond carbon content in the films gives rise to a greater compressive stress. A similar relation is aiso seen by comparing both curve tin Fig. 4 sadFig, 3 for the filmson WC substrata. Curves b and c in Fig, 5 reveal that a greater methane con-

cenlzation in the source gases results in a D'eater compressive stress in the film due to a higher non-diamond carbon conteut in the film. This point is in agreement with the conclusions of Windischmann etal. [ 10].

For the WC substrata material, curve c in Fig, 4 shows the effect of deposition time on the stress of the films, The thermal stress in this case is about - 3,99 GPa [ 12], sad the intrinsic stress due to pres,.nce of non.diaraond carbon is also compressive in nature, Therefore, the residual stresses of the films on both WC and Si wafer are all compressive in nature,

For quartz substrata material, the thenaal slress for a deposition temperature of 1173 K is about 0.47 GPa [13], sad it is a tensile stress. Therefore, as shown in curve a of Figs. 4 and 5, the residual stress cart be either compressive or tensile, depending upon the sum of tensile thermal stress and compressive intrinsic stress, which is due to the presence of non. diamond carbon in the films,

(4)

C,T, K~o ¢ ~ ~t, f Tldn $#/td Films 2PO-2Pl ( IPP~) 2$4.-2YP ~'/

(,)

(b)

Pig. 2. The typical Rnman speara of dlamond fllrm on tlwee different sub- ~ale m~eriMs: (n) Sisubsmite, 117~ K,$ nmroughaess; (b) WCsubstn~, 1 IT3 K, Inm rouB~ess~ tc) quartz substtate, 973 K, 8 Bm zcuglme~s,

2,4 | i 1,4 J 0,I) 0,4 Time (ho~r)

Flg, 3, ~ffcct~ of depoqidon tire, subslr~e pl~,Mmenqs m~d materiels on th~ non.dlamond czzbon fontent of llt~ films: curve a, qemlz, 973 K, 8 nm rouzhness; carve b, WC, 1173 K, 20 nm muBhne~; curve e, WC, [ 173 K, Inm ~Bhnemz~ curve d, $i, J 172. K, $ nm rouslmen; curve e, $I, [ I?3 K, 16 rim mujlhn~s,

0" 2 I ~

. n I •

*3

2 4 0 M |0 lg |4 |6 Pill, 4, Effect of deposition time on the i n s t a l su~a of diamond films on throe different sul~trates: cuwe a, films on qmlz, 973 K~ Crave b, films on $t wafers, 11"/3 K' curve c, films on WC (S~o), 1173 K,

0,~ 0 -I

l

-IJ 4 -1,5 J,$ h i i 2 13 17 n cz-z~mmo O0 ~ )

Pt s, 5. B/re~t or CJ.~/H~ ralo on the internal ~ of dizm1~d ltlraz o~ thz'c~ different Hbzln~,s: Cur/e *, films on quan~ 973 K, 2,5 h't, ~ b, 61m~ on $t warm, 1173 K, 2. I~ curve c, lilmz on WC ($q6Co), 1172. K, Sh,

The effect of s u r f a ~ roughness of the W C s u b s ~ t c on ~sidual s ~ s s of the films is shown in FiB, 6, A zrougher surface, e@ curve b in FiB. 5, gives a gz~.ater compressive stress in the film due to a ~re~ter non.dinmondcm'bon content in the films. The effect of surface roushz~.~s on ndlzesion of the film was i n v e s t i s a t ~ by our lpoup [ 14]. It was concluded mat there exis~ an appropriate ran~ of the subs~te rough. hess values for an optimum adhesion strensth. This is related toagreaternucleation densityofdiamondcrysudsnndalower non-diamond carbon con~ent in the film in certain r o ~ h n c s s range,

(5)

2 5 8 C,T, K.o et al. / Thin $~lid Ftlsu ?~291 ff 996) 2 5 4 - - 2 5 9 -I .! ,$ .Z 4+$ .4,S ..4 e ~ IO Is Fig. 6. E r e c t s o f d e ~ s i l i o n t i m e and, s u r f a c e r o u g h n e s s o f t h e s u h e t r a t e o n t h e i n t e m M s t r e s s o f d l ~ n o n d f i l m s o n W C ( 5 % C o ) s u b s t m t e s : c u r v e a,

roughness = 1 nm; curve b, roughness- 20 .m,

?,3, Effect o/the fire thickness and texture coe/~cient on resid~l stress

Fig. 7 shows effect of the film Sickness on residual stress of the films on three different subsu'ate materials. For athick- ness greater than 8 p,m, it revealed that a greater thickness results in a less compressive stress in the film, This may related to a lower non.diamond carbon content near su~ace of the thicker film, and to the fact that the penetration depth of the laser beam for generating the Raman signals is so limited, Therefore, fox n thicker film the R a m ~ shift may nor respond the stress state of the e n t ~ film,

Fig. 8 shows effect of texture coefficient on residual stress of the films on WC and Si wafer substrat~s [t ;.ndieates that the texture of the film is not a significant factor on the residual

s t r e s s .

3.4. Effecto/restdgatstressonadhesinnofrhe~lm

It is interesting to know how the residual stress affects the adhesion of the films• In many engineering applications, a

-O,S

-2.a •

+2+! -.1,0

Thickness ( p m )

Fig, 7. Effect of the film thiGknees on residual stress of the films on WC, quonz and SI su~Urate matMi~ls,

• ,

" ° , s . . . '

,.~ "2,0

,4,0 •

- U

Texture Coefficient, T I1|

Fig, 8, Effect of the texture coefficient on residual stress of the films ors WC and Si substrate m~terlals,

. , . , . . - , . . . .

' A~ ' .~.o J A ' A .,.s .1.e Stress (GPa)

Fig, 9, Effect of the resid~ stress on erdhesien of t ~ dimaond films on WC and SI subsuale msmrials,

compressive stress is often intentionally applied to the surface of a component to increase its cracking resistance, Fig+ 9 shows residual stress versus slope, where the slope is a measure of the crack propagation ability of the film, i.e. the slope of the cracking diameter versus indentation load curve. A greater slope indicates a smaller cracking resistance or adhesion, At lower compressive stresses, a greater compressive stress seems to result in a favorable cracking resistance, At higher compressive stresses, there is not enough data to make a definiti'~ statement.

4. Conclusions

The residual stress of CVD diamond films on three different substrata materials were determined by micro.Raamn spectroscopy. The X-ray diffraction was used to determine the texture coefficient of the films, and the indentation testing method was used to evaluate the cracking resistance of the film on the substrata. It is found that the ~sidua] stress of the films on (1{30) Si wafer and WC materials is compressive in nature, and is both compressive and tensile in nature for quartz snbstrntes. The origins of the residual stress are mainly the sum of thermal stress and intrinsic stress, The thermal stresses for WC and Si w.afer suhstrates are compressive, whereas those for quartz substrates are tensile. The intrinsic stress arises mainly from the effect of non.diamond carbon

(6)

C,Z Kuo ¢~ a~ ~Thin $otld Fltms 2~.-29! 1'1996) 254-259 ~ 9 conte.t in the diamond crystals, not at the grain boundm'ies,

A greater non-diamond conU:nt in the diamond crystals will result ia a larger compressive residual suess. It is also found that the texture of the films has no significant effect on the

residual stress, For a high cracking resistance of the films on

Si substxatas, a smaller compressive stress is necessary.

Admowledgements

This work was supported by the National Science Council of Taiwan under contract No, NSC-84-2221-I/009-037, We would also like to thank Prof. K,H, Ch©n and Prof, L.C. Clzen of Academica Sinica at Talpei, and Prof. Y.Z. Hsieh of NCTU at Hslnohu to use their Raman Spectroscopes,

References

[11 J,A, Ba81io~ B.C, FlZrlnSWoz'~, S, HItI~.I~, G, Hamilt and D, O'Netl,

"A'kfn SolidFitm, 212 (1992) 180.

[2] Wang Wanlu, Llao Kojun, Oao JinyJn 8 and Llu Almia, Thin Solid F/Ires, 215 (1992) 174.

[3] P.g. Bachmnn. H,D. Bausen. H. [agl¢, D, Leers, D,U, Wieel~,M. N. Heats, R. Kold and P. Koidl, Di,~aoad Rdat*d Mulet,. 3 (|994) 1308,

[4l Y,M, LcOrtce. R J. I',lenmnloh, J.l". Glass. Y.ii. Lee~ R A. R~tdder tad RJ, MtrkzmM. Male~. Re*. Sou, Syrup. Proo.. 162 (1990) 219, 15] K, V u Ar.ker, H, Mohtm~hra, B. Blanpaln, P, Van Hou~ and J,P,

Cells, Matd:, R*:. $oe. Syrup. P~I, ~ (1~3) 677,

[6] C. Johnr.oa, A. Ormsley, P,R. Chalger end |,M, Buekley-Oolder,

Diamond Rdated Mater., I (1992)450.

[7] 1t. Bqspert, 1. van Streelea end ],E $11vent. Pkys. Ray. B, 32 (1985) 1423.

[8] M. SIIveffa, M, Beeu~i, E, Castellucci, F.P, MtllJot~ V, Blabi/'os~ R, Temciello and F, Galluzzl, Dlamoad Rdatcd Mater., 2 (1993) 1257.

[9] CX. KUO, T.Y. Yen and T,H, Huan8, J, Maser, Re~,, 5 (1990) 2515,

lie] H. Windlschmann, G,P. Bpps, Y, Cong and R,W, Collins, J, App,, Pky~., 69 (1991) 2231,

[ 11 ] M, Aim, D~, Peebles and J,A,Oh[!mt~rt, J, Arllzeztmz ~i, Tedeio|,,

7(1993) 1309,

[ 12] M, Nesladek, DiamondReiated Miner., 2 (1993) 9B,

[ 13] K, Fabisitk, J,M, lqdlippoz and H. vim tier Berg, Dimmed Relafed Mater., i (1992) 77.