Thin Solid Films, 128 (1985) 321-332 PREPARATION
BEAM III AND D. D. L. CHUNG
Deportment n/Metallurgical PA 15213 (U.S.A.) (Received
andMaterials Science, Carnegie-Mellon University, Pittsburgh.
X-ray diffraction was used in situ to study the phase transitions which occurred in 1500 A Au/GaAs( 100) on heating and cooling. The reaction between gold and GaAs took the form Au +Ga + cr-(Au-Ga). On heating, cz-(Au-Ga) completely dissolved in liquid Au-Ga. On subsequent cooling, P-(Au-Ga) (or Au,Ga,) formed. In nitrogen at 1 atm, phase transitions were observed reversibly at 525 + 25 “C (as a result of the complete dissolution of cc-(AuGa) on heating) and at 415 f 5 “C (as a result of the peritectic transformation of P-(Au-Ga) to cl-(Au-Ga) and liquid Au-Ga on heating). In a vacuum of 56.7 Pa, similar phase transitions were observed at 425 k 25 “C and 387 f 13 “C respectively. The contact morphology remained quite smooth provided that the dissolution of a-(Au-Ga) was avoided.
The thermal reliability of gold-based contacts to GaAs has been a problem in solar cell and integrated circuit technologies. This reliability problem affects the electrical behavior, contact uniformity, adhesion and bondability. To solve this problem, numerous workers have investigated the interdiffusion and reactions which occur between gold and GaAs. However, because of the ex situ nature of most of these investigations, the phase transitions which occur in the contacts on heating have received relatively little attention, even though these phase transitions are of direct bearing to the thermal reliability of the contacts. Nakanisi’s’ internal friction results on gold thin films on GaAs suggested the occurrence of three melting transitions on heating in a vacuum of 1 x lop4 Torr (1.3 x 10m2 Pa). These transitions were labeled T,, T, and T, (and the corresponding temperatures T,, T2 and T3). Transitions T, and T, were observed on first heating at 440°C and 500°C respectively; transition T3 was observed on second and subsequent heating at 340°C r. Moreover, the internal friction results showed that transition T2 was absent for gold film thicknesses less than 3000 A, so that only transition T, was observed on first heating r. As the gold film thickness increased, transition T, became increasingly significant until T, was buried in the shoulder of T, in the plot of internal friction versus temperature’. In second and subsequent 0040.6090/85/$3.30
((> Elsevier Sequoia/Printed
in The Netherlands
E. BEAM III. D. D.
heating, transition T, appeared strongly while transitions T, and T, were not observed ‘. By using in situ X-ray diffraction during heating and cooling at a pressure of 1 atm, Zeng and Chung’ observed in 1000 8, Au/GaAs a transition at 478k22 ‘C during first heating and a transition at 413 k 5 “C during second and subsequent heating. They showed that the transition at 413 f 5 “C was due to the melting of an Au-Ga phase, namely p-(AuGa) (or Au,Ga,)’ (see the AuGa phase diagram in Fig. 1 for the AuGa phases). 1080°
100 1080 -1000
491 3’ il:
“’ 1 400
OoO Fig. 1. Au-Ga
In this work, we used the same X-ray diffraction technique as Zeng and Chung to study Au/GaAs with a gold film thickness of 1500 A. As is also reported in the companion paper4, we found that, for 1500 8, Au/GaAs, the reaction took the form Au+Ga
where cr-(Au-Ga) is the terminal gold-rich solid solution. Using in situ X-ray diffraction, we observed that cc-(Au-Ga) completely dissolves in liquid Au-Ga (L)
r+L-+L This phase transition, although reversible, degrades the contact uniformity, induces phase formation and changes the crystallographic orientation of the c( phase4. It is therefore of practical importance to investigate the phase transitions in gold contacts to GaAs. 2. EXPERIMENTAL
Au/GaAs(lOO) wafers were kindly provided by P. Lindquist of HewlettPackard Corporation. The GaAs substrates were tellurium doped ((2-3) x 1017 cm 3), and gold was boat evaporated onto the substrate at 3 8, s- ’ at a temperature of less than 100 “C. The gold thickness was 1500 A. X-ray diffraction was performed using a Rigaku D/MAX II powder X-ray diffractometer system and its 1500°C high temperature attachment. The sample chamber was either purged with nitrogen gas or under a dynamic vacuum of 56.7 f 3.3 Pa (purged with argon prior to evacuation) during heating. The accuracy of the temperature measurement was k 5 “C. A fine-focus copper X-ray tube was used. Detection was achieved with a scintillation counter. Because a much thicker aluminum window was used in vacuum than at 1 atm (1.01 x 10’ Pa), the diffraction intensities were much lower in vacuum than at 1 atm. Scanning electron microscopy (SEM) was performed at room temperature using a Cambridge scanning electron microscope. 3. EXPERIMENTAL
Because of experimental limitations, we did not distinguish between gold and a-(AuGa) diffraction peaks in this work. However, this distinction was made in our ex situ X-ray diffraction work, which showed that a formed at 350 “C or below and had a lattice constant about 0.997 of that of gold4. Since B and Au,Ga, have almost the same lattice constants, distinction between these two phases was not made in this work. 3.1. At I atm (1.01 x IO5 Pa) Figure 2 shows X-ray diffraction patterns obtained at room temperature before heating and obtained in situ at 450 “C after being held at 450 “C for 2 h. The two patterns are essentially identical. On further heating to 550 “C, the gold (or CL)peaks were observed to vanish completely, as shown in Fig. 3. Although Fig. 2 shows the Au (or ~1)111 peak and not the Au (or CL)200 peak, the latter was observed outside the 20 range of Fig. 2. On heating to 550 “C, both gold (or CL)peaks disappeared, while the GaAs peaks remained. This is attributed to the complete dissolution of CLat 525 + 25 “C. Figure 3 shows a series of X-ray diffraction patterns obtained in situ at different temperatures on cooling after the first heating. The specimen was cooled stepwise from 550 “C to 500,450, 400 and 350 “C, with the time at each temperature being
9NllH3 '-I.a .a ‘III MIV38 '3
On second or subsequent heating, the peritectic transformation of p to c1+ L and the dissolution of CLwere again observed at the same respective temperatures. The reversibility of the peritectic transformation is further shown in Figs. 4 and 5. On heating (Fig. 4), the p (or Au,Ga,) peaks vanished while the gold (or CL)peaks grew; on cooling (Fig. S), the p (or Au,Ga,) peaks appeared while the gold (or a) peaks diminished in intensity. The fact that the peritectic transformation cl+ L + p did not cause the gold (or ~1)peaks to vanish totally was due to the remaining gold (or x), which did not participate in the peritectic transformation. Figure 4 shows that the peritectic temperature is between 410 and 420°C on heating. On cooling, the transformation was observed between 425 and 400 “C, as shown in Fig. 5. B(300)
,I 40 so6
Fig. 4. X-ray diffraction the peritectic
in situ on heating
of p (or Au,Ga,)
to (a) 410 ‘C and (b) 420 “C at
I atm to show
to c( (or gold) and liquid AuPGa.
aAs (200) hAs
Fig. 5. X-ray diffraction patterns obtained in situ at 1 atm on cooling 375 “C to show the reverse of the transformation shown in Fig. 4.
to (a) 425 “C, (b) 400°C
p~q qxq~ aIdcues aql30 amwadrual u100.1 112apmu SCM uo~~l?u!Luexa ~2s ‘&OS 1 le Quanbasqns pauyqo ulawd uogmgp aql s~oqs 6 a.uI%d %1goo3 JaqwIJ uo syead d aql 01 IDadsal ~J!M Ma.I8 yead ?zgznv aqL .syaad d aql ~JJM pals!xaoD qxq~ ‘yead EON.~o/pue 1 [s vDznv aqJ30 ammadde aql pasnm so SUE o~Bu~~oo3 laqjlnd .syead d aql30 qwo.18 aqj pasnm qD!qM ‘SC,OSE 01 paloo3 Jaq$.m3 SBM aldmzs aql ‘(8 %!A) 3, S,IE ]e $J30 amweaddeal aql _‘a]3v aql snqL syead d aql30 amemaddeal aq$ pasnm 3, SLE 01 Bug003 tuanbasqns ~paqs!ueA day4 3, oop $I?3nq ‘luasald [[gs alaM syead $j aq$ 3, SLE IV ‘8 +?!A ur u~oqs paleaq SBM a[dums aq$ ‘(L %a) ~,OSE 1~ d 30 amamadde aq$ .tayV sv ‘[email protected]
aq] L#M ~se.1~~03 u! u!I&? ‘x, JO uogedymd aql jnoqly 7 u10.13su1~03 d leqj s~oqs L 3,~ ‘a~otmaq~~n~ .ux~r?1 le uo~~~sue.~~]s!q430 Al!yqrslaAal aq3 qI!M ~sm~uo~ u! sr ;o 30 uopnloss!p aql3o iCuyq!slaAag aqL .pa.maddv syead 00~ pm E I 1 d aql mq ‘$uasqt paupma1 syead (E IO) ~108 aql‘&os~ 01 Buqooc~ laqlm3 uo .]uasqt? II!JS aJaM #yvad (xl JO) ~108 aq) ‘3,oop 04 8uqoo~ luanbasqns uo ‘so SZTSZP 1~ x 30 uognloss!p [email protected]
aql 01 pavq!.w SFs!qL ‘3, asp w paqs!uvA mq 3,001~ 1~ pwyad amwadural WOOI )e paalasqo syead (D JO) ~108 aqL %I![oo~ asydals luanbasqns uo & 0s~ pur! oop $r! put? (dais amlwadural q%a II? U!UI gz-0~) 8uyzaq as!Mdals uo & ()gp pm 009 le ni!s uz pauyqo sulawd UO!W?JJJ!~ h-~ srnoqs L a.m8!d “d L’yg,[o urnmnn v lq ‘Z’&
30 uo!ln~oss!p aql auo%apun lou p’?q a[dum JawoJ aql ltzqli%3 aq$ qly lua]ysuoD s! s!qL .sl!d qms alow hmu smoqs layit?1 aql sr?alaqhz ‘slrd [email protected]
?u%?i~a~ pau%yz awes SMoqs alduws Jawo3 aqJ. ‘(S-E ‘s%;i u! uaA!8 ale sulailed uo!1cw~~!p i(m-x asoqM a[dms aq$ .a.!) u0y?ux.103sue.11 yDal!lad aql 30 iCJq!q!s.taAaJ aql Moqs 01 [email protected]
uaq) pue [email protected]
01 dn pavaq uaaq peq qD!qM aldum aqi (q) put? (z 2%~ u! uaA!S aIk? swawd uo!lm~g~p Icm-x asoqm aldms aql .a.!) q z 103 3, asp 1~ paleaq uaaq pvq qD!qM aldum aqi (I?) 30 sqd&?oloqd ~3s amiwadwai u1o0.1 sMoqs 9 a.mZ!d
‘7 ‘U ‘Cl ‘111 b’lV38 ‘3
Fig. 8. X-ray diffraction to 375 “C.
to(c) 400 “C and (d) 350 “C.
Fig. 7. X-ray diffraction
I 40 28
I I 30
to show the reversible
of p (or Au,Ga,):
of b on cooling
(a) 375 ‘C;(b) on heating to 400 “C;(c) on cooling
of a on heating to (a) ~$00‘C and (b) 450 ‘C and the formation
to show the dissolution
in situ in vacuum
in situ in vacuum
E. BEAM III, D. D. L. (‘HUNG
Fig. 9. X-ray diffraction
the sample used for Fig. 8 to 150°C.
been heated to 450 “C and then cycled to show the reversibility of the melting of p (i.e. the sample whose X-ray diffraction patterns are given in Figs. 7-9). The microstructure was similar to that of Fig. 6(b), except that the aligned rectangular pits were several times smaller and denser than those of Fig. 6(b). This effect of reduced pressure during heating is as previously reported’. The similarity with Fig. 6(b) gives further support for the notion that the dissolution of (Ycaused substantial degradation of the surface uniformity. 3.3. Summary of phase transition temperatures Table I summarizes the phase transition (1.01 x 10’ Pa) and 56.7 Pa for 1500 A Au/GaAs.
at 1 atm
4. DISCUSSION As indicated
in the Au-Ga
(Fig. l), the dissolution
Hydrostutic pressure (Pa)
Temperuture of complete dissolution of
Tempera&w of appearance qfp (“C)
ci (“C) 1.01 x 105 56.7 ’ Reversible. b Observed during heating. ’ Observed during cooling.
500~550 a 400~450 b
Temper&we oJ appearance of Au,Ga
410-420 a 375-400 a
over a range of temperatures such that the dissolution is complete at 510 “C for an overall composition with 21 at.% Ga. The fact that p (Au-21at.“/,Ga) was formed on subsequent cooling suggests that CLand L together have a composition with 21 at.% Ga. Thus, if an overall composition with 21 at.% Ga is assumed, at 500 “C CLand L coexist, whereas at 550 “C only L is present. This is just what we observed in 1500 A Au/GaAs at 1 atm, during heating as well as cooling. For 1500 8, of gold in 1 atm, we observed the peritectic transformation of p to c1+ L. The transition temperature is between 410 and 420 “C. The Au-Ga phase diagram (Fig. 1) shows two peritectic transformations: 0 -+ a’+ L
(at 409.8 “C)
We did not observe ~1’.However, in Au/Ni/Au-Ge/GaP( 11 l), (x’was observed as the reaction product and the peritectic transformation of cz’to c1+ L was observed6. On reducing the hydrostatic pressure from 1 atm (1.01 x lo5 Pa) to 56.7 Pa, the temperature for the complete dissolution of CLdecreased by about 100 “C, while that for the melting of p decreased by about 28 “C. A low pressure enhances the arsenic evolution that accompanies the Au-Ga phase formation, thereby affecting the amount of arsenic dissolved in the “Au-Ga” phases. This might in turn affect the equilibrium transition temperatures. The sharp variation in the Au-Ga liquidus temperature with the gallium concentration (O-22 at.% Ga) suggests a sharp variation with the arsenic concentration as well. This argument is consistent with the large pressure dependence of the temperature for complete dissolution of cc Because the transition temperatures reported here were determined by stepwise heating, with 20 min or more at each temperature step, the transition temperatures obtained are equilibrium transition temperatures. This means that the dependence of the transition temperature on the pressure is due to thermodynamics rather than to kinetics. Other than the effect on the transition temperature, pressure had three additional effects. Firstly, a low pressure caused the dissolution of CI to be irreversible, probably as a result of the evacuation of the arsenic vapor. Secondly, a low pressure caused the melting of p to occur in the form p + L, whereas it occurred in the form fl + a+ L at 1 atm (1.01 x lo5 Pa). Thirdly, Au,Ga coexisted with p below 325 “C in vacuum but was not observed at 1 atm (1.01 x 10’ Pa). The origins of these effects are still to be understood. Further decrease in the pressure below 56.7 Pa is expected to decrease the phase transition temperatures further. This is suggested by the observation’ that the goldto-silver color transition temperature and the arsenic evolution temperature decreased with decreasing pressure below 56.7 Pa. For example, in situ mass spectrometry showed’ that 2100 8, Au/GaAs(lOO) underwent arsenic evolution at 358 “C at a pressure of about 5 x 10e6 Torr (6.7 x lop4 Pa). Our phase transition temperatures obtained for 1500 A gold films at 56.7 Pa are in agreement with those (Tr and T3) obtained’ by internal friction on gold films less than 3000 A thick at 1 x 10m4 Torr (1.3 x 1O-2 Pa). TI was determined by internal
BEAM III, D. D. L. CHUNG
friction to be 440 C, while our X-ray diffraction results showed it to be 425 + 25 “C (corresponding to the complete dissolution of a). T3 was determined by internal friction to be 340°C while our X-ray diffraction results showed it to be either 387 F 13 “C (corresponding to the melting of B) or 337 k 13 “C (corresponding to the appearance of Au,Ga on cooling). Because we did not heat Au,Ga after its formation, we have not yet observed the melting of Au,Ga. However, it is likely that the melting of Au,Ga occurs reversibly at a temperature close to 337+ 13 “C; the Au-Ga phase diagram suggests that this happens at 348.9”C; the dissolution of Au,Ga had been observed reversibly at 320 “C in 600 A Au/GaAs(lOO) by in situ transmission electron microscopy’ (TEM). On the basis of internal friction results, Nakanisi’ interpreted Tl as due to the melting of a portion of gold and T3 as the Au-Ga-As ternary eutectic. These interpretations are in contrast with those we arrived at using X-ray diffraction. For 1000 8, Au/GaAs, Zeng and Chung2 observed the vanishing of the B (or Au,Ga,) peaks on heating at 413 + 5 “C and 1 atm, in agreement with the temperature found in this work for 1500 A Au/GaAs at 1 atm. However, for 1000 8, Au/GaAs, the vanishing of the l3 (or Au,Ga,) peaks was not accompanied by any significant growth of the gold (or cx)peaks and the melting of B appears to take the form
rather than that of a peritectic transformation2. The behavior (this work) for 1500 A of gold at 1 atm agrees with the Au-Ga phase diagram better than that’ for 1000 b; of gold at 1 atm. This is reasonable since the phase diagram is for the bulk material system. Ex situ X-ray diffraction of 1500 8, Au/GaAs (the same sample as used in this work) showed that the reaction between gold and GaAs took the form Au + Ga -+ cl-(Au-Ga) and that this reaction occurred at 350 “C or below4. In this work, in situ observation of the dissolution of u gives additional evidence for the formation of cc The difference between the results (this work) for 1500 8, of gold at 1 atm and those2 for 1000 8, of gold at 1 atm may be due to the difference in sample preparation conditions, as the 1500 8, gold sample (this work) was prepared by boat evaporation whereas the 1000 A gold sample2 was prepared by electron beam evaporation. However, it may also originate from the fact that the interfacial reaction takes the form Au+Ga
for 1500 A Au/GaAs
4, whereas the interfacial
Au + Ga -+ Au-Ga
takes the form
for 1000 A Au/GaAs 2. It should be mentioned that 1500 A of gold was not due to insufficient time (tentatively AuGa) to take place. After 2 h at 450 “C, diffraction pattern was observed for the 1500 A gold
the behavior observed here for for the formation of Au-Ga no additional peak in the X-ray sample, whereas the reaction to
form AuGa had gone to completion for the 1000 A gold sample. Furthermore, the absence of AuGa phase formation for 1500 8, Au/GaAs after 2 h at 450 “C did not arise because the thick gold film hindered arsenic evolution, since arsenic evolution has been observed’ from 2100 A Au/GaAs at about 5 x 10m6 Torr (6.7 x 10m4 Pa). Therefore, we believe that the observed effect is not due to kinetic reasons. Internal friction’ indicated that the gold film thickness affects the phase transitions such that a change from “T,-T, behavior” to “T,-T, behavior” occurs at a gold film thickness of 3000 A. The details of this effect of the gold film thickness remain to be investigated. It is quite probable that even other different reactions occur for gold films considerably thinner than 1OOOA. For a gold film of thickness 600& ex situ TEM revealed’ the formation of Au,Ga precipitates at 250°C. For a gold film of thickness about 100 A, in situ TEM revealed the formation of rectangular features on heating to about 380°C the vanishing of the gold diffraction rings on further heating to 470°C and the formation of p (or Au,Ga,) on subsequent cooling to room temperature’. The microstructure of heated Au/GaAs also varies with the gold film thickness. Ex situ TEM showed that, for a gold thickness of 4000 A, p (or Au,Ga,) and gold precipitates formed after heating at 550°C whereas, for a gold thickness of about 400 A, the gold precipitates only appeared within near-surface zones after heating at 550°C lo. Ex situ SEM showed that, for a gold thickness of 2000 A, 8% of the projected surface area was encompassed by pits whereas, for a gold thickness of 1000 A, 18% of the projected surface area was encompassed by pits”. Ex situ SEM also showed” that deeper penetration of gold occurred for a gold thickness of 2000 A than for a gold thickness of 400 A. This work has shown that the dissolution of c1 is accompanied by a sharp increase in the density of the aligned rectangular pits, whether the dissolution is completed at 525k25 “C (observed at 1 atm or 1.01 x lo5 Pa) or at 425k25 “C (observed at 56.7 Pa). This implies that a relatively smooth contact morphology can be maintained by avoiding the dissolution of ~1.This can be done by keeping the temperatures below the dissolution temperature and/or keeping the pressure sufficiently high. 5. CONCLUSION In situ X-ray diffraction was used to observe the phase transitions in 1500 A Au/GaAs. In nitrogen at 1 atm (1.01 x lo5 Pa), we observed transitions at 525 + 25 “C and 415 + 5 “C; at 56.7 Pa, we observed similar transitions at 425 f 25 “C and 387 f 13 “C respectively. Thus the decrease in pressure decreased the transition temperatures, as also observed by Zeng and Chung2 for the transition which they observed at 478 + 22 “C at 1 atm for 1000 A Au/GaAs. For 1500 A Au/GaAs, we found that the higher temperature transition (525 + 25 “C at 1 atm or 1.01 x 10’ Pa, and 425 + 25 “C at 56.7 Pa) is due to the completion of the dissolution of cL,i.e. a + L -+ L, and that the lower temperature transition (415 f 5 “C at 1 atm or 1.01 x lo5 Pa, and 387+ 13 “C at 56.7 Pa) is due to the melting of p (or Au,Ga,). At 1 atm or 1.01 x lo5 Pa, the
BEAM III, D. D. L. CHUNG
melting takes the form of a reversible peritectic transformation, i.e. p --t CY+ L; at 56.7 Pa, the melting is also reversible but is simply of the form p -+ L. The contact morphology remained quite smooth provided that the dissolution of cI was avoided. The phase transition temperatures observed in this work for 1500 8, Au/GaAs heated in a dynamic vacuum (56.7 Pa) are quite close to those (Tr and 7”) observed’ by internal friction on less than 3000 8, Au/GaAs heated in vacuum (1 x 10m4 Torr or 1.3 x lo- 2 Pa). Applying the X-ray diffraction results for 1500 A Au/GaAs at 56.7 Pa to the internal friction results for less than 3000 8, Au/GaAs at 1 x 10 -4 Torr (1.3 x 10 ~’ Pa), it can be seen that transition T, is due to the complete dissolution of c1,i.e.
This transition can also be viewed as the melting of cI. In contrast, on the basis of the internal friction results, Nakanisi’ suggested that transition T, was due to the melting of a portion of gold. The transition temperatures which we observed at 1 atm (1.01 x lo5 Pa) are in good agreement with the Au-Ga phase diagram (Fig. l), whereas those observed at 56.7 Pa are lower than those indicated by the phase diagram. This effect of pressure is attributed to the fact that arsenic evolution accompanies the reaction and is enhanced by a lower hydrostatic pressure. Hence, pressure affects the amount of arsenic dissolved in the “Au-Ga” phases. ACKNOWLEDGMENTS
The X-ray diffraction equipment grant from the Division of Materials Research of the National Science Foundation under Grant DMR-8005380 was essential for this work. Equipment support from the Materials Research Laboratory Section, Division of Materials Research, National Science Foundation, under Grant DMR 76-8 1561 A01 is also acknowledged. The technical assistance of T. Kim of CarnegieMellon University is greatly appreciated. REFERENCES
1 2 3 4 5 6 7 8 9 10 11 12
T. Nakanisi. Jpn. J. Appl. Phys., 12(1973) 1818. X.-F. Zeng and D. D. L. Chung, Solid-State Electron., 27 (1984) 339. C. J. Cooke and W. Hume-Rothery, J. Less-Common Met., 10 (1966) 42. D. D. L. Chung and E. Beam III, Thin SolidFilms, 128 (1985) 299. X.-F. Zeng and D. D. L. Chung, Thin Solid Films, 93 (1982) 207. R. A. Ginley, D. D. L. Chung and D. S. Ginley, Solid-State E/ectron., 27 (I 984) 137. S. Leung, L. K. Wong, D. D. L. Chung and A. G. Milnes, J. Electrochem. Sot.. 130 (1983) 462. T. Yoshiie, C. L. Bauer and A. G. Milnes, Thin Solid Films, I I I (1984) 149. K. Kumar, Jpn. J. Appl. Phys., 18(1979)713. T. J. Magee and J. Peng, Phys. Status Solidi A, 32 (1975) 695. C. J. Todd, G. W. B. Ashwell, J. D. Speight and R. Heckingbottom, Met&Semiconductor Contacts. Inst. Phys. Cor$ Ser. 22 (1974) 1Il. J. Gyulai, J. W. Mayer, V. Rodriguez, A. Y. C. Yu and H. J. Gopen, J. Appl. Phys., 42 (1971) 3578.