Applied Surface Science38 (1989) 72-79 North-Holland, Amsterdam
FOPdViAT~ON O F CoSiz O N A M O R P H O U S SILHCON BY RTA Gyi~zi~ D R O Z D Y *, H a n n u RONIQ~INEN and Ilkka S U N I Semiconductor Laboratory, Technical Research Center of Finland, Otakaari 7 B, SF-02150 Espeo, Finland
Received 19 March 1989; accepted for publication 23 March 1989
The formation of cobalt silicide by RTA from a cobalt layer on top of amorphous silicon was compared to the usual case of cobalt on single crystalline silicon. The reaction between cobalt and amorphous silicon is faster and provides better surface morphology, it also results in an improved edge definition. The final resistivitiesof the two cobalt disilicide layers are the same.
L ltn~r~uc~on Cobalt silicide is one of the most potential alternatives for self-aligned gate metaUizations in integrated circuits. A simple salicide process is possible b y combining a rapid thermal annealing (RTA) step of vacuum deposited cobalt with selective etch of unreacted Co. CoSi2 with minimal resistivity a n d n o bridging over or encroachment under the oxide sidewall can b e obtained by proper selection of process parameters . We have investigated cobalt disilicide (Cogi.2) based salidde structures for possible implementation in our CMOS process "2]. Normally the silieide would be formed by interaction between a metal layer (Co) a n d polycrystalline (gate) or single crystal (source and drain) silicon. To improve the gate pattern delineation we have considered the possibility of substituting L P C V D grown amorphous silicon (a-3i) For the standard polycrystalline silicon gate. This spurred our interest in comparing the reaction rates of evaporated Co films with amorphous and crystalline silicon under RTA. It has been established that the amorphous state favors the nucleation of C o S i : at relatively low annealing temperatures. Thermodynamic calculations yield an estimate of 200 K for the nucleation temperature of CoSi 2 over a-Si while on e-Si the transformation of CoSi to CoSi z does not occur until the temperature is raised to about 5 0 0 ° C . This entails an attractive possibility of lowering the salicide formation temperature in a practical C M O S process. * Permanent address: Central Research Institute for Physics, P.O. Box 49, H-1525 Budapest, Hungary 0169-4332/89/$03.50 © Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)
G. Drozdy et al. / Formation of CoSi 2 on amorphous silicon by RTA
2. Experiments We have used 100 mm p-type (100) silicon substrates with a resistivity of 10-50 fl cm. Three sets, three wafers each, were processed. Each set contained one blank wafer, one oxidized silicon wafer with 440 nm a-Si grown by LPCYD at 575 ° C and one oxidized wafer without a-Si as a control sample. 100 nm Co was deposited by e-gun evaporation on all wafers. The wafers were subjected to repeated isothermal annealing steps in a Heatpulse 210-T rapid thermal annealing system in argon atmosphere. Wafers from the first set were annealed at 5 3 0 ° C up to 1500 s. Wafers from the second and the third set underwent heat treatments at 575°C up to 1500 s and at 7 0 0 ° C up to 180 s respectively. Wafers from one set were always annealed in parallel to get an identical heat treatment for each wafer.
3. Results and discussion
The sheet resistances were measured after each annealing step. The results are shown in figs. 1-3 for each set respectively. To estimate the time scale of the sil/cide formation sequence, theoretical res/stance curves were derived from a simple model. This model was based on the assumption that the three consecutive silicide phases appear one at the time in the sequence: Co2Si, Cogi [oN~q lO 9 o R
go I s T
2 1 o
Fig. I. Calculated and measured sheet resistances of Co/Si structures and the C o / S i O 2 control sample as a funcdon of the total R T A time at 530°C: (m) single crystal, (Q) amorphous, ( ~ )
G. D r o z d y et al. / Formation o j C o S i 2 on amorphous silicon by R T A
~o 9 6 R E s I S 7; A
7 6 5 4
3 2 1 0 o
Fig. 2. S a m e as fig. 1 at 5 7 5 ° C : (N) single crystal: ([:]) a m o r p h o u s , ( ~ ) control, ( O ) theoretic.
and CoSi 2. This assumption seems to Le well justified ha the light of r e c e n t reports dealing with the formation of Co2Si on e-Si  and the formation of CoSi 2 on c-Si [31. The growth of the silicJdes was assumed to be diffusion lJmhed. This should also be correct with a possible exception of CoSi 2 on c-Si [oH~
9 8 R E S
T 5 A N 4 C E 3 2 1 o 0
Fig. 3. Same as fig. 1 a n d 2 at 7 0 0 ° C :
(@) single crystal, (ra) a m o r p h o u s , ( @ ) control, ( O ) theoretic.
G. Drozdy et al. / Formation of CoSiz on amorphous silicon by RTA
Table 1 Parameters used to model the time dependence of the sheet resistance in Co/Si structures under RTA
Growth rate (nmZ/s) at 530 o C 575o C Co Co 2Si CoSi CoSi 2
187.9 58.3 33.3
700 o C -
1080 242 242
37600 6670 33300
Resistivity (/~ cm)
1 1.47 1.35 1.76
6.25 110 147 18
which at 5 3 0 ° C could still form by a nudeat/onAimited process . The parabolic growth rates for the model were taken from the literature [4-6]. The resisfivities and the atomic volumes of the siIicides have been adopted from previously published data [7,8]. The actual values have been listed in table 1. From fig. 1 it is clear that the silicides are not fully formed at 530°C. This follows from the experimental procedure where the samples are exposed to the room atmosphere after the first annealing step (5 s) without completing the Co2Si phase. The RBS spectra shown in fig. 4 reveal that only 25 nm of the original Co film has been transformed to a silicide. That corresponds to 32 nm of CozSi, exactly the thickness that should grow during 5 s of annealing. Further annealing even at 750 ° C did not increase the silJcide thickness. It is therefore suggested that a thin SiO2 layer is formed at the interface during the
• ". :,~ ....
Fig. 4. RBS spectra of Co/c-Si sample.~after the full series of RTA steps at 530°C (dotted line) and 575o C (continuous line).
G. Drozdy et aL / Formation of CoSi 2 on amorphous silicon by R T A
exposure to the room atmosphere. The amount of oxygen in the cobalt film is below the detection limit of RBS. The temporal behavior of the sheet resistances in fig. 1 also suggests that immediately after the supply of Co has been exhausted, first CoSi and then CoSi 2 grow. There was no significant difference between Co on a-Si and Co on c-Si. At 575°C (fig. 2) the formation of Co/Si follows the model with a reasonable accuracy. After the first phase has been completed the model suggests saturation of the sheet resistance. This follows from the fact that the resistivity of CoSi is slightly higher than that of Co2Si, just enough to offset the effect of volume expansion. In reality, the samples with c-Si display significant overshoot. This is not unexpected because the silicide layer transforms from the orthorhombic structure of CozSi to the cubic structure of CoSi with concomitant transitional disorder. In fact, similar anomaly in the sheet resistance has been recorded during an oxidation experiment of CoSi 2 on SiO2 where silicon is gradually depleted from the silicide. After approximately two thirds of the CoSi has been formed a rapid drop in the sheet resistance takes place. This must be attributed to the onset of CoSi2 formation. CoSi2 is the only silicide phase in the sequence to have sufficiently low resistivity to produce such a drop. For the initial reduction of the sheet resistance from approximately 9.6 to 6.5 ~2 only 10 nm of CoSi 2 is required. The parabolic growth rates of CoSi and CoSi 2 are roughly the same at this temperature (see table 1) but since the thickness per formula unit is 1.76 times larger for CoSi z the CoSi grows into completion well ahead of it. This is marked by a small kink in the sheet resistance curve corresponding to a final thickness of 190 nm for CoSi and 14 nm for CoSi 2. It is quite evident that the simple picture for the sequential growth of silicides does not hold in these experimental circumstances although it may still be valid for different film thicknesses and annealing temperatures. The early onset of the CoSi2 formation is even more pronounced in the case of a-Si, also shown in fig. 2. In that case no overshoot of the sheet resistance can be observed and the two silicide phases CoSi and CoSi 2 seem to grow in parallel from the start. This is evidenced by the rapid decrease of the sheet resistance immediately after completing the Co2Si phase. The shape of the annealing curve suggests that also in this case a high resistivity layer is formed at the CozSi/CoSi interface. When Co2Si is fully depleted the thickness ratio between the two remaining silicides is t(CoSi2) / t(CoSi) = 1/6. Hence, in spite of the nominally identical parabolic rate constants at 575°C CoSi seems to grow much faster than CoSi2. The growth rates have different activation energies, E a = 1.9 eV for CoSi  and Ea = 2.6-2.8 eV for CoSi 2 [3,6]. The thickness ratio is therefore expected to be sensitive to small variations in the annealing temperature. At 700 ° C the same general behavior is observed. The measured annealing times for fully reacted samples are much longer than predicted by the model. This can safely be attributed to the sequential annealing procedure where the
G. Drozdy et al. / Formation of CoSi 2 on amorphous silicon by RTA
Fig. 5. SEM micrographof a CoSi2 layer formed at 575°C on singlecrystallineSi. samples are repeatedly cooled down and exposed to the room atmosphere. It is also most probable that the rate constant for CoSi2 extrapolated from data in ref.  overestimates the growth rate at this temperature. The optimistic values would predict 10 s minimum annealing time for full reaction. This is clearly too short to agree with previously reported annealLng times for thinner Co films [1,9]. The difference between Co films annealed on c-Si and a-Si is also
Fig. 6. S E M m i c r o ~ p h of a CoSi 2 layer formed at 575 ° C on a-Si.
G. Drozdy et aL / Formation of CoSi, on amorphous" sificon by RTA
in tbJs case quite distinguishable. The sheet resistance maximum is much l-6gher and the overall annealing time to achieve minimum resistance is longer on c-Si than on ~-Si. The effect of the substrate material on the final resistivity of CoSi 2 was further investigated by amorphizing part of a single crystal substrate by Si ion-implantation. Rapid thermal annealing of a 60 nm thick Co film for 20 s at 7 0 0 ° C resulted in a sheet resistance of 1.16 ~ on the amorphized surface and 1.07 ~ on the single crystalline silicon respectively. The difference is within the limits of film thickness uniformity. Scanning electron micrographs of the two different CoSi z layers formed at 575 ° C on c-Si and a-Si are shown in figs. 5 and 6 respectively. Notwithstanding the relatively low annealing temperature the CoSi2 layer formed on c-Si has developed an orange peel like appearance whereas the same on a-Si displays a significantly smoother surface morphology. Ignoring the small contribution to the sheet conductance of the silicide layer by reduced surface scattering the smooth surface certainly helps patterning very small dimensional device structures.
4. Conclusion The experimental results show that amorphous silicon allows shorter annealing times or lower annealing temperatures than crystalline silicon for CoSi 2 formation by thin films reactions. The enhancement is mainly attributable to the easier onset of transformation from CoSi to CoSi2. The low temperature option and the observed smooth surface morphology can significantly relax the constraints of CMOS processes at the submicron level. The inherently low resistivity of CoSi 2 and the improved edge definition on a-Si make this combination very attractive for very small dimensional salicide structures.
Acknowledgement The authors would like to thank E. Zsoldos, J. Saarilahti, J. Salrni and L. Gr~3nberg for assistance in Co evaporations, RBS measurements and X-ray analyses.
References  A.E. Morgan, E.K. Broadbent, M. Delf'mo,B. Coulman and D.K. Sadana, J. FAectrochem. So¢. 134 (1987) 925.
G. Drozdy el al. / Formation of CoSi 2 on amorphous silicon by RTA
 G. Drozdy, H. Ronkainen, M. Orpana and I. Suni, in: Proc. 13th Nordic Semiconductor Meeting, 5-8 June 1988, Saltsj~baden" Sweden, pp. 49-52.  F.M. d'Heurle and C.S. Petersson, Thin Solid Films 128 (1985) 283.  B.S. Lira, E. Ma, M.-A. Nicolet and M. Natan, .L Appl. Phys. 61 (1987) 5027. [51 S.S. Lan" J.W. Mayer and K.N. Tn" L Appl. Phys. 49 (1978) 4005.  A. Appelbanm, R.V. Knoell and S.P. Murark,a, J. AppL Phys. 57 (1985) 1880.  C.-D. Lien, M. Finetti, M.-A. Nicolet and S.S. Lan, 3. Electron. Mater. 13 (1984) 95.  M.-A. Nicolet and $.S. Lan, in: VLSI Electronics: Microstructure Science, Materials and Process Characterization, Vol. 6, Ed. (3. Larrabee (Series Ed. N. Einsprech) (Academic Press, New York, 1983) pp. 329-464.  M. Tabasky, E.S. Bulat, B.M. Ditehek, M.A. Sullivan and S. Shatas in Rapid Thermal Processing, Materials Research Society, Pittsburgh, PA, 1986, VoL 52, Eds. T.O. Sedgwick, T.E. Seidel, B.-Y. Tsaur, pp. 272-277.