Some titanium germanium and silicon compounds

Some titanium germanium and silicon compounds: Reaction and properties O. Thomas,a) F. M. d'Heurle, and S. Delageb) IBM T. J. Watson Research Center, P.O. 218, Yorktown Heights, New York 10598 (Received 27 December 1989; accepted 20 March 1990) Titanium reacts with pure Ge in two different ways: At low temperatures one observes the formation of Ti6Ge5 with some characteristics typical of diffusion-controlled reaction. Upon completion of this first stage Ti6Ge5 reacts with remaining Ge to form TiGe2, isomorphous with C54 TiSi2, in a process which is clearly controlled by nucleation. The same observations apply to reactions with a Ge alloy containing 25 at.% Si. With an alloy containing 50 at.% Si the two stages become merged, so that while remaining identifiable, they are much less distinct than with the previous conditions. The reaction behavior observed with a Ge alloy containing 80 at.% Si resembles that generally obtained with pure Si: there are no easily identifiable steps between the initial Si-Ti sample and the final one, Si-TiSi2. With both the 50-50 and 80-20 Si-Ge alloys the formation of the C54 structure is preceded by that of the C49 structure (ZrSi2 type), as with pure Si. The gradual merging of the diffusion-controlled reaction and that controlled by nucleation as the concentration of Si in the substrate increases implies that nucleation plays a significant role in the formation of TiSi2, even if that role cannot be easily isolated. Effects due to gas impurities on the path of the metal-substrate reaction have been analyzed. The resistivities of several pure and alloyed phases have been measured. Alloy scattering in the system TiSi2-TiGe2 is briefly discussed.

I. INTRODUCTION

The subject of this paper has already been the object of two short publications1'2 that focused on the behavior of Ti with layers of pure Ge and 50-50 Si-Ge alloys on (100) Si substrates. Thus for the sake of being brief, both this introduction and the usual description of the experimental details will be kept to a minimum. In this ternary system the phases with the highest metalloid content, TiGe2 and TiSi2, have the same structure, orthorhombic (C54) with unit cell dimensions that are very close: a = 8.594 A, b = 5.030 A, c = 8.864 A, and a = 8.252 A, b = 4.783 A, c = 8.540 A, respectively.3 (One notes here some ambiguity in the literature about the respective definitions of the a and c axes so that it is not exactly clear whether the a lattice parameters are smaller than c parameters for both compounds. It will be assumed here that the axes are properly defined as presently given.) Thus with lattice parameter differences of the order of 6%, well within the limits of 15% given by the Hume-Rothery rule for

"'Permanent address: Laboratoire des Materiaux et de Genie Physique, ENSPG, BP 46, 38402 Saint Martin d'Heres, France. b) Permanent address: Thomson-CSF, Domaine de Corbeville, 91401 Orsay, France. J. Mater. Res., Vol. 5, No. 7, Jul 1990

extended solubilities between elements4 (and presumably between chemically equivalent compounds also), the two compounds should be mutually soluble, as verified2 for samples containing 50 at.% Si (and 50 at.% Ge, ignoring Ti). On the other hand, in the middle of the diagram there exists3'5'6 no exact equivalent in the Si-Ti system to the phase35'7 Ti6Ge5. Most of the investigations were carried out with layers of Ge and Si-Ge alloys deposited on (100) Si substrates via molecular beam (MBE) deposition techniques. Although these were not specifically intended to be epitaxial, they were in fact found to be so. In the results reported thus far one notes the two-stage reaction of Ti with the epitaxial layers of pure Ge: at low temperatures the formation of Ti6Ge5 with features that strongly imply a diffusion-controlled mode of growth, and quite distinctly at higher temperatures the nucleation-controlled8 formation of TiGe2. With the 50-50 Si-Ge alloy, two stages can still be recognized but the overall picture is much less clear than with pure Ge. Here attention will be focused on the behaviors of the 25-75 and 80-20 alloys, and on the electrical properties of the final disilicide-germanide solid solutions. Some relevant observations made on the reaction of Ti with single crystal Ge and on bilayers of Ti and Ge on Si will also be reported. © 1990 Materials Research Society

1453

0. Thomas, F. M. d'Heurle, and S. Delage: Some titanium germanium and silicon compounds

II. REACTION KINETICS A. Pure Ge layer and single crystal Ge

The Ti reaction with epitaxial layers of pure Ge has been sufficiently described in Refs. 1 and 2, and the results outlined in the introduction above. It will not be discussed any further here except to recall that for heat treatments of 1 h each, Ti6Ge5 forms between 450 °C and 500 °C, while the nucleation and growth of TiGe2 is observed between 550 °C and 600 °C. Figure 1 shows the partial backscattering spectra for a film of Ti deposited over single crystal Ge. The spectra presented are for samples as-deposited and after heat treatments from 480 °C, 500 °C, 650 °C, and 700 °C, respectively, for 1 h each. Because Ge is heavier than Ti, the Ge part of the spectra shows at higher energy than Ti, which for the as-deposited sample appears in the spectrum as a "hat" between about 1.60 MeV and 1.65 MeV, on top of the Ge. Precisely as with the MBE Ge layer (see Fig. 1 of Ref. 2) the formation of Ti6Ge5 is proceeding at 480 °C and is complete after the heat treatment at 500 °C. Spectra obtained with samples annealed at 550 °C and 600 °C (not shown) reveal that no further reaction beyond this Ti6Ge5 formation occurs at these temperatures. After a heat treatment at 650 °C there is considerable formation of TiGe2. Examination of the surface of the sample, through an optical microscope equipped with a Nomarski objective, Fig. 2, reveals islands of TiGe 2 surrounded by a smooth background of Ti6Ge5, in a configuration which is characteristic of a new phase growing through a nucleationcontrolled8 reaction. Nucleation is difficult so that it requires relatively high temperatures, and even then it occurs only at selective sites, quite distant from one another. The temperature is sufficiently high that once

FIG. 2. Surface view of a sample of Ti over single crystal Ge, annealed at 650 °C for 1 h, as seen in the optical microscope with a Nomarski objective. The small islands of TiGe2 are visible because of their ruffled appearance against the smooth background of Ti6Ge5.

nucleated the new phase grows rapidly through the thickness of the layer where it becomes visible. It then spreads laterally, forming islands of increasing diameter, until final consumption of the available material. B. Si-Ge (25-75) alloy

o s> z o

1.6 1.8 BACKSCATTERING ENERGY (MeV)

FIG. 1. Partial backscattering spectra for samples of Ti over single crystal Ge, as-deposited, and after heat treatments at 480 °C, 500 °C, 650 °C, and 700 °C. 1454

The reaction in this case started at about 500 °C, some 50° higher than with plain Ge, with the formation also of Ti6(Si, Ge)5. This initial reaction appears to be diffusion-controlled with a smooth increase of the thickness of the new layer, of finite composition, as a function of temperature. The reaction is completed at 560 CC (1 h). Figure 3 shows that no further reaction occurs at 590 °C, while an increase of the annealing temperature by only 10° to 600 °C is sufficient to cause the almost complete transformation to Ti(Si, Ge)2. This sudden change in behavior is one of the characteristics of nucleation-controlled reactions. Pictures similar to Fig. 2 could be obtained from samples not totally reacted. The reaction of the 50-50 alloy (see Refs. 1 and 2) is not too different from that with the 25-75 alloy. In backscattering one may observe a separation between a

J. Mater. Res., Vol. 5, No. 7 Jul 1990

O. Thomas, F.M. d'Heurle, and S. Delage: Some titanium germanium and silicon compounds

1.2 1.4 1.6 1.8 BACKSCATTERING ENERGY (MeV) FIG. 3. Partial backscattering spectra for samples of Ti over a SiGe (25-75) alloy, after respective heat treatments at 560 °C, 590 °C, and 600 °C, showing the sudden nucleation of Ti(Si, Ge)2 at 600 °C.

germanium-rich surface layer and a silicon-rich bottom layer, which disappears upon completion of the reaction to Ti(Si, Ge)2. This has been tentatively ascribed to phase separation resulting from the already mentioned absence of a Si phase isomorphous with Ti6Ge5. Another significant difference, in comparison with the samples prepared from pure Ge or from the 25-75 alloy, is that the nucleation and growth of a metalloid-rich phase, penetrating through the whole thickness of the samples, occur when some of the initial Ti film still unreacted remains at the surface of the samples. C. Si-Ge, 80-20 alloy Figure 4 shows the progression of the reaction as observed by backscattering spectrometry. As often

1.0

found in the reaction of Ti with pure Si (with remarkably few exceptions, e.g., Ref. 9), it is almost impossible to describe in simple terms any intermediate conditions between the original unreacted state and the final state totally transformed into Ti(Si, Ge)2. One salient feature is the fact that the region next to the free surface becomes rich in Si-Ge faster than the region adjacent to the Si-Ge substrate itself. Although such a situation has not been too often encountered with Ti, it has been described with respect to the formation of silicides with other refractory metals, e.g., with Mo and W,1011 WTa,1213 and W-Ti.14 Presumably, in reactions where Si (or Ge) is the mobile element the initial reaction, transforming the metal into a compound with an accompanying increase in volume and formation of compressive stresses,15 puts the remaining metal in tension, increases the grain boundary diffusion coefficient, and favors the diffusion of Si (Ge) to the free surface. The reaction would proceed preferentially in this location because of the relatively easy accommodation of volume change and favorable conditions for stress relaxation.1013 Assuming the validity of this model, it remains to answer the question as to when does this mode of reaction predominate over the anticipated growth from the bottom up. III. ELECTRICAL (AND X-RAY) MEASUREMENTS A. Kinetic aspects and phase identification

The resistance of samples prepared from MBE layers of various Si-Ge compositions but approximately equal thickness of Ti is shown in Fig. 5 as a function of annealing temperatures (1 h each) in the range from 500 °C to 700 °C. The behaviors appear to be almost

1.2 1.4 1.6 1.8 BACKSCATTERING ENERGY (MeV)

FIG. 4. Partial backscattering spectra for samples of Ti over an (80-20) Si-Ge alloy, showing the increased rate of reaction at the free surface relative to the metal-substrate interface.

7

5

6

7

5

6

7

5

6

7

ANNEALING TEMPERATURE (IOO°C)

FIG. 5. The resistance of samples made from MBE layers of Si, (80-20), (25-75) Si-Ge, and Ge, as a function of annealing temperatures ranging from 500 °C to 700 °C.


1455

O. Thomas, F. M. d'Heurle, and S. Delage: Some titanium germanium and silicon compounds

the same, a first decrease in resistance followed by a second one corresponding to the final phase TiSi2, Ti(Si, Ge)2, and TiGe2, all with the C54 structure characteristic of TiSi2 and of TiGe2 as well. Samples with the 50-50 composition (not shown, but see Ref. 2) display a behavior fitting between those of the 80-20 and 25-75 alloys. Upon closer scrutiny, however, the behaviors can be distinguished into two groups. Those familiar with the reaction of Ti with Si will easily recognize on the left-hand side of Fig. 5 that the step at about 2.5 H between 550 °C and 600 °C, with the final drop to a value about four to five times smaller, corresponds to the transition16"19 from the C49 structure (characteristic of ZrSi2) to the low resistivity form C54. For the 80-20 alloy one sees about the same comportment, with increased resistance values due to alloy scattering. Proof of this change of structure is provided in the partial diffraction patterns (Cu Ka) shown in Fig. 6. The peak at about 19° for the sample annealed at 560 °C could be the (001) peak of some remaining Ti (see Ref. 20), or belong to some Ti-rich compound. There is, however, no possible ambiguity (see Ref. 15) for the peaks at about 20°, 20.5°, and 25.6°; these are the (060), (131), and (200) peaks of the C49 structure, with the following unit cell dimensions21 for TiSi2 (without any Ge ad-

12

Si-Ge (80-20) 5 6 0 °C

;

A, " 0

—i

1

1

dition): a = 3.562 A, b = 13.531 A, and c = 3.550 A. According to the RBS spectra in Fig. 4, such a sample should indeed be almost entirely transformed into Ti(Si,Ge)2. The pattern corresponding to the sample annealed at 650 °C is characteristic of the C54 structure, while in between, for a sample annealed at 600 °C, one sees the superposition of both diffraction patterns. No trace of the C49 structure could be found in the diffraction patterns for the samples prepared from pure Ge or from the 25-75 Si-Ge alloy. The intermediate resistance plateaus seen in Fig. 5 for these samples at about 8 Cl and 6 Cl correspond to the formation of the Ti6Ge5 phase and the subsequent transition to TiGe2 (C54). With the 50-50 alloy the transitions from Ti6(Si, Ge)5 to Ti(Si, Ge)2 and from C49 to C54 are less distinct than in the cases discussed here, but proof of this latter transition could be obtained from a close examination of the resistance and diffraction results.2 In all cases the final state of the samples was one phase with the C54 structure. There was no indication that the anticipated total solubility of the disilicide and digermanide should be violated. B. Resistivities

In Fig. 7 the room temperature resistivity of the final C54 structure is shown as a function of composition. The resistivities were obtained from samples annealed to their resistance minimum, either 650 °C or 700 °C, since these samples, as those made with pure Si, display a tendency to increased resistance upon annealing at high temperature. This is due22'23 to surface tension causing the films to agglomerate into separate islands (above 800 °C, for example, with the 25-75 alloy). The data in Fig. 7 follow a parabolic-like

C49 -

I

I

i

Ti(Si,Ge)2 "E

° 30

-

a

jt

:SISI

I

20

-

-r

. /

LU

C54 20 25 BRAGG ANGLE (0)

i

FIG. 6. Partial diffraction patterns showing the characteristic transition from the C49 to the C54 structure for samples prepared with an (80-20) Si-Ge alloy. 1456

o: 10

0 Si

25

i

50 75 COMPOSITION (at.%)

100 Ge

FIG. 7. The resistivities of the solid solutions TiSi2-TiGe2 as a function of their composition.

J. Mater. Res., Vol. 5, No. 7, Jul 1990


curve typical of a Nordheim24 plot anticipated for alloy scattering. Data for Si alloy with 5 and 10 at.% Ge (not plotted) indicate that within experimental errors the resistivity of TiSi2 increases linearly with increasing Ge content, almost, indeed, up to 50 at.%. According to Linde25 this type of behavior is expected of most alloys, at least within limited concentrations. Taking into account the straight line from pure TiSi2 to pure TiGe 2 , the residual resistivity amounts to about 0.23 nil cm/mol % for TiGe2 in TiSi2, that due to TiSi2 in TiGe2 to about 0.33 /JL£1 cm. (The residual resistivity per atom would be half as big as the value per mole.) For a comparison with a simpler metallic system one may recall that the residual resistivity26 of Ag in Cu is 0.077 IA£1 cm/at.% Ag, but that of Cu in Ag is about twice as high, 0.14 fiil cm. The room temperature resistivity of TiGe2 is estimated to be about 23 /id cm, that of Ti6Ge5 at 150 /xH cm. This latter value should be considered an upper limit since further annealing could cause a decrease in resistivity; however, this could not be done in the present investigation because of the formation of TiGe2. One would wish to know more about the germanides; unfortunately, the published equilibrium diagram27 extends only from pure Ti to Ti5Ge3, which is isomorphous with Ti5Si3, but even the melting temperature of Ti 5 Ge 3 is not shown. For a comparison of known data about silicides and germanides one may turn to the Appendix in Ref. 28. IV. DISCUSSION A. Nucleation

The evidence for the nucleation behavior of TiGe2 with single crystal Ge, MBE layers of Ge, and Si-Ge alloys with respective compositions of (25-75 at.%) and (50-50 at. %) has been well documented in this work. It is quite certain that this behavior is mandated by the small free energy change AG between Ti6Ge5 and TiGe2- Unfortunately this, too, is not known from published data, but it is likely that comparison with the transition from TiSi to TiSi2 with an enthalpy change of only - 1 kcal/mol (from the difference -31 kcal for TiSi2 and -30 kcal/mol for TiSi)6 provides sufficient information to assert that AG is small for the germanide transition as well. Somewhat paradoxically since published values of thermodynamic data are often accompanied by relatively large error limits, for small AGs the kinetic behavior observed here may provide more direct information about the magnitude of AG than actual thermodynamic data. The density of observed nuclei has not been measured accurately, but casual observation reveals that it is not a strong function of temperature, indicating that nucleation is heterogeneous, namely that it occurs at preferential sites at the

interface. What these sites are is not known, but some information about this may be found in the observation that under similar circumstances nucleation of TiGe2 occurs at a temperature some 50° higher on single crystal Ge than on the MBE layer. One may want to attribute this difference to strain in the MBE layer of Ge, which would increase AG; this may play only a minor role, if at all, since the measurements of the lattice parameter of the MBE layer showed that it was quite relaxed. It is likely that the somewhat easier nucleation on the MBE layer was the result of more "effective" nucleation sites. Curiously, in this case also one did not observe large differences in the density of nuclei, so that one may not simply assume that accommodation dislocations provide the nucleation centers in the MBE layers. The picture (one would like to say Gestalt) of the well identified nucleation of TiGe2 over pure Ge and over the (25-75) Si-Ge alloy, of the merging of nucleation and growth phenomena over the (50-50) alloy, of the absence of a clearly identifiable kinetic process in the case of the (80-20) alloy as well as in the formation of pure TiSi2, and of the observed roughness of reacted layers of TiSi2 strongly suggests that this roughness results from nucleation processes hidden during the course of the reaction between Ti and Si. This has been proven29 to be the case for CoSi2, which also forms according to somewhat ambiguous kinetic laws, and where nucleation could be isolated. One likes to think in terms of well isolated and separated phase formations, e.g., the diffusion controlled growth of Ni2Si followed by that of NiSi, and finally the nucleation controlled formation of NiSi2. Yet nature does not always have to provide effects that are both discrete and simple. In the case of TiSi2, the formation of some Tirich compound, Ti5Si3 and/or TiSi (both of which have been observed individually under specific circumstances, alloying with As or Sc, but others are possible also: Ti5Si4, not counting metastable forms, perhaps Ti6Si5), and the nucleation of TiSi2 could well occur simultaneously, making analysis extremely complex. One is aware also of experiments involving the formation3031 of metastable amorphous phases. Because the formation of the final phase would then involve a small AG, it would necessarily require a somewhat difficult nucleation step, as has been well demonstrated32"34 for amorphous CoSi2 and Cr3Si. Yet if this involves only small composition changes and consequently small volume modifications (as in melting), it would not explain the final roughness of the reacted layers; this requires discontinuous phase nucleation and growth accompanied by considerable composition and volume changes. The role of nucleation in the reaction of Ti with single crystal Si first mentioned in Ref. 35 is discussed in greater detail in Ref. 36.


1457


B. Surface growth, diffusion barriers, and impurity effects

1

I.9

FIG. 8. Backscattering spectra showing the reaction of Ti with the (80-20) Si-Ge alloy. The samples are the same as in Fig. 4, but aged about six months in the as-deposited condition in a N2 atmosphere. 1458

1

'

1

'

1

"AA

The reason why surface phase formation has been observed in the case of the (80-20) alloy is not understood. An attempt was made to investigate this question through further annealing of as-deposited samples which had been kept in an atmosphere of dry N2 for approximately six months. These samples were otherwise identical to those used for Fig. 4. Not only had they been deposited simultaneously with them, but they were taken from the same wafer. The behavior shown in Fig. 8 was, however, totally different from what was seen previously (compare with Fig. 4). The marked surface reaction exhibited by these samples has disappeared. It is believed that the difference is due to the adsorption of nitrogen and oxygen (the cabinet in which the samples were kept is flushed with N2, but is opened to air quite often). Evidence for the presence of such gases is presented in Fig. 9, where small backscattering peaks for surface nitrogen and oxygen can be seen in the spectrum for a sample annealed at 590 °C. Figure 8 shows that in such a sample the reaction has proceeded almost to completion. The surface oxygen and nitrogen could not be seen in samples annealed at lower temperatures. Quite certainly the gases that had diffused into the Ti film during aging and found themselves in solution were snow plowed into the metal ahead of the reacted layers during the various heat treatments. Such a behavior has been well identified37 with respect to oxygen during the formation of TiSi2. As in this latter reference it is seen (Fig. 9 after heat treatment at 600 CC) that annealing at increased temperatures leads to expulsion of the gases out of the samples and completion of the reaction. This latter effect can be observed in the

.3 I.6 BACKSCATTERING ENERGY ( MeV i

'

590°C

3 O

\

SURFACE -NITROGEN

U O

,600°C

-

SURFACE

o

Si-Ge (80 -20) 0

1

i

0.5

1

0.7

i

i

0.9

-

M ,

I.I

t

1.3

BACKSCATTERING ENERGY (MeV) FIG. 9. Partial backscattering spectra showing the buildup of surface nitrogen and oxygen in the samples from Fig. 8, after an anneal for 1 h at 590 °C. After an anneal at 600 °C the signals for these surface gases have disappeared and the reaction has proceeded to completion.

straightening of the signal for surface Si at 1.3 MeV. The surface appears to be slightly silicon-rich in this case, but this is only a step in the final homogenization which eradicates the surface disproportionation, germanium-rich and silicon-poor, which occurs during heat treatments at 570 °C or 590 °C (see Fig. 8 at 1.3 MeV and 1.85 MeV). Since one expects reactions to occur smoothly from the interface up, there is an innate tendency to ascribe irregularities such as rapid surface reactions to some flaws in sample preparation. The evidence presented here points in the opposite direction, namely, gas contamination of the metal layer induces "regular" compound formation in samples which do not behave so regularly in the absence of such a contamination. The model offered for surface growth involves grain boundary diffusion through the metal layer; such grain boundary diffusion is greatly reduced or eliminated in samples that contain impurities. There is ample published material bearing witness to this effect. Copper alloying reduces grain boundary diffusion38 in Al and thereby reduces also the rate of electromigration failure. Moreover, such Cu alloying is known to enhance a smooth reaction between Al and transition metals39 (in pure Al the reaction proceeds mostly along grain boundaries). Impurity effects on grain boundary diffusion are known40 also for boron and carbon in iron. Some systematic studies can be quoted, e.g., Refs. 42 and 43. The importance of impurities in diffusion barriers has been the object of many reports.44"46 Two ideas need to be retained here, (a) Gas impurities may favor a seemingly normal compound formation (from the substrate upward), (b) These gas impurities in metals such



as Ti can be incorporated at room temperature so that, especially with very thin layers, one should be cautious about the storing of samples lest the reaction behavior be fundamentally altered. For the sake of completeness one should mention here another anomaly in reaction behavior, in Fig. 1 for compound formation over Ge single crystal. That one, however, remains to be satisfactorily explained. The reaction appears to be complete after annealing at 700 °C, yet the little peak between 1.60 and 1.65 MeV corresponds to a surface region which has remained rich in Ti. That this belongs to some titanium-rich compound may be evidenced by the presence of an extraneous diffraction peak (not belonging to TiGe2) at about 19°. It disappears after annealing at 900 °C, but the titaniumrich surface peak in the backscattering spectrum is already no longer present after heat treatment at 850 CC. Thus there is no one-to-one relationship between the backscattering and diffraction evidence. No gas presence could be detected in the backscattering spectra (which are notoriously insensitive to this type of impurities). Evidence for incomplete reaction at 700 °C and for the high resistivity of the remaining titanium-rich layer is provided by the resistivity of the sample, some 25% higher than the value obtained with the sample grown over the MBE layer of pure Ge. The matter has not been explored any further. C. Most mobile species

An experiment was conducted with a bilayer of Ge and Ti deposited in this order over Si. There was excess Ti so that the final compound should correspond roughly to Ti(Si05, Geo.5)2. Annealing at 480 °C caused the Ti-Ge to transform into Ti6Ge5, as anticipated. The Si remains unaffected, as seen in Fig. 10. Reaction with Si results immediately into a solid solution of TiSi2 and TiGe2 at a temperature approaching 580 °C. In Fig. 11 the backscattering spectrum for such a sample is shown together with the spectrum for a fully reacted sample, after annealing at 650 CC. On the same figure modelized spectra for TiSi2 and TiGe2 have also been superimposed. It is seen that as soon as the reaction shifts to the disilicide-germanide, while there still remains some titanium-rich region at the surface, the Si and Ge are nearly homogeneously distributed in the Ti(Si, Ge)2 layer, providing evidence for the high mobility of the metalloids in this compound and confirming earlier results6'47-48 showing that TiSi2 (as well, evidently, as TiGe2) forms by metalloid atom motion. Since these compounds are built from the same structural elements as MoSi2 and WSi2, their diffusion behaviors should be similar to that described for these two latter compounds10 where the Si atoms were shown to be more mobile than the metal atoms by several orders of magnitude.

1.2

1.5

1.8

BACKSCATTERING ENERGY (MeV) FIG. 10. Partial backscattering spectra of a bilayer sample with Ge and Ti over a Si substrate, respectively, as-deposited and after heat treatment at 480 °C for 1 h.

D. Resistivity

Measurements down to liquid He temperature of the resistivity of the TiGe2 samples prepared either from MBE or single crystal Ge reveal that the high room temperature resistivity of TiGe2 as compared with TiSi2 is due to extrinsic effects, such as impurities or stacking faults. The evidence is provided in Fig. 12 where the low temperature evolution of the resistivity of TiGe 2 (annealed at 700 °C) and that of

BACKSCATTERING ENERGY (MeV)

FIG. 11. Partial backscattering spectra for bilayer samples of Ti and Ge over a Si substrate, respectively, after heat treatments for 1 h each at 580 °C and 650 °C. Model spectra for TiGe2 (continuous line) and for TiSi2 (dashed line) are also included. The position of surface Si is at 1.3 MeV, that of surface Ge at 1.85 MeV, and that of surface Ti at about 1.65 MeV, as indicated by the upper (high energy) edges of the two model spectra.


1459


100

200

300

TEMPERATURE (K)

FIG. 12. The evolution of the resistivity of TiGe 2 and of Ti(Sio.5, Geo.5)2 from room temperature down to liquid helium temperature.

Ti(Sio.5, Ge o . 5 ) 2 are displayed simultaneously. The intrinsic value of the room temperature resistivity of TiGe 2 (temperature effect) is about the same as that of TiSi2, in the vicinity of 12 jiil cm. An attempt at studying the structure of TiGe 2 by transmission electron microscopy failed because the sample was transformed into some titanium-rich compound during thinning by means of ion milling. This observation is consistent with the small difference in free energy between Ti 6 Ge 5 and TiGe 2 , which is revealed by the nucleation behavior. (NiSi2 is known to revert to NiSi under moderate ion bombardment.49) Since there are no reasons to attribute the high resistivity to impurities, one wonders about the possible presence of twins and stacking faults. Stacking faults have not been reported for the C54 structure of TiSi2, but they have been observed14'50 in the similar structures of WSi2, MoSi2, and alloys thereof. They could exist as well in TiGe 2 since this corresponds to an ABCDA stacking of identical planes, while other stacking orders would appear to be possible. In metals twins and stacking faults are frequently observed in materials such as Ag with stacking order ABCA or Zn with stacking order ABA. In the silicides, TiSi2 is intriguing since the absence of twins in the four-layer repeat sequence implies long-range atomic interactions, which would appear to be less prominent in similar silicides with repeat sequences of only two or three layers. Perhaps TiGe 2 differs from TiSi2 with respect to the rigidity of its stacking order? One also notes that TiSi2 in its C54 form is endowed with very large grains, because of the extremely difficult nucleation from C49 to C54 and of the resulting low density of nucleation sites. Other things being equal, one would anticipate TiGe 2 to have smaller grains than TiSi2 and a higher residual resistivity.

1460

The alloy scattering visible in Fig. 7 is in marked contrast with what is observed in another system with complete solubility, CoSi 2 -NiSi 2 , where almost no alloy scattering is observed.51'52 Considering the relatively small difference in lattice parameters between TiGe 2 and TiSi2, the alloy scattering appears to be remarkably large as compared to what happens in Cu-Ag 26 where lattice effects that are more than twice as large produce residual resistivities that are more than twice as small. One may venture some speculation about the asymmetry in the residual resistivities of Si in TiGe 2 and of Ge in TiSi2, the latter being smaller than the former. In Cu-Ag this has been attributed to an asymmetry in the lattice effects, the contraction due to Cu in Ag being smaller than the expansion due to Ag in Cu, with resulting asymmetry in the localized charge density. Such effects may play a role in TiSi 2 -TiGe 2 , but confirmation of this possibility would require the determination of unit cell dimensions with greater accuracy than is possible on stressed thin film samples.

V. CONCLUSIONS (1) A study of the reaction of Ti with Ge and SiGe alloys provides strong evidence for the role that nucleation is likely to play in the reaction of Ti with Si. (2) With Ge and (25-75) Si-Ge alloys Ti first forms Ti 6 Ge 5 in a reaction which appears to be diffusion controlled. The transition from Ti 6 Ge 5 for both pure Ge and the Si-Ge alloy is clearly nucleation-controlled. This latter effect is attributed to the small free energy change that accompanies this reaction. (3) Evidence has been provided that gas impurities in solution in the Ti layer can modify significantly the reaction path. Such impurities can accumulate during storage of the metal film at room temperature. The effects observed have been explained in terms of what is known about the role of impurities in slowing grain boundary diffusion. (4) The room temperature resistivity of TiGe 2 has been measured as close to 23 /iCl cm, that of Ti 6 Ge 5 as 150 /JLCI cm. This latter value should be considered as an upper limit since further annealing could lower it. The intrinsic value of the resistivity of TiGe 2 is close to that of TiSi2, about 12 nil cm. (5) TiSi2 and TiGe 2 seem to form a continuous series of solid solutions as anticipated from the small difference in unit cell dimensions of the two compounds. (6) No evidence was found for the formation of the C49 (ZrSi2 type) during the reaction with either pure Ge or the 25-75 Si-Ge alloy. However, as with pure Si, the C49 structure precedes the formation of the final C54 structure in the reactions with the 50-50 and 80-20 alloys.



Sigrmiller, in Thin Film Interfaces and Interactions, edited by J. E. Baglin and J. Poate (The Electrochemical Society, Pennington, NJ, 1980), p. 213. 16 F. M. d'Heurle, P. Gas, I. Engstrom, S. Nygren, M. Ostling, and C.S. Petersson, IBM Research Report RC 1151, Yorktown Heights, NY, 1985. 17 R. Beyers and R. Sinclair, J. Appl. Phys. 57, 5240 (1985). 18 I. J. M. Raaijmakers, A. H. Reader, and H. J.W. Van Houtum, J. Appl. Phys. 57, 5240 (1985). 19 C. L. Jia, J. Jiang, and X. F. Zong, Philos. Mag. A59, 999 (1989). 20 Standard Diffraction Powder Pattern no. 5-0582. 21 Standard Diffraction Powder Pattern no. 23-964 (1973). 22 L.R. Zheng, L.S. Hung, S.Q. Feng, P. Revesz, J.W. Mayer, and G. Miles, Appl. Phys. Lett. 48, 767 (1986). ^T. Brat, C. Osburn, T. Finstad, J. Liu, and B. Ellington, J. Electrochem. Soc. 133, 1451 (1986). 24 L. Nordheim; Ann. Phys. 9, 607 (1931). 25 J. O. Linde, Ann. Phys. 15, 219 (1932). 26 F. J. Blatt, Phys. Rev. 108, 285 (1957). 21 Binary Alloy Phase Diagrams, edited by T. Massalski, L. H. Bennett, and H. Baker (American Society for Metals, Metals Park, OH, 1986). 28 P. Gas, G. Scilla, A. Michel, F. K. LeGoues, O. Thomas, and F. M. d'Heurle, J. Appl, Phys. 63, 5335 (1988). 29 ACKNOWLEDGMENTS F.M. d'Heurle and C.S. Petersson, Thin Solid Films 128, 283 (1985). The authors wish to express their appreciation to 30 K. Holloway and R. Sinclair, J. Appl. Phys. 61, 1359 (1987). D. Moy and T. Gallo for the deposition of the Ti films, 31 K. Holloway and R. Sinclair, J. Less-Common Met. 140, 139 to G. Coleman for performing the backscattering stud(1988). 32 ies, and to C. S. Petersson for reading the manuscript. K. N. Tu, D.A. Smith, and B.Z. Weiss, Phys. Rev. B 36, 8948 (1987). We are indebted to W. Gallagher for access to the low 33 D. A. Smith, K. N. Tu, and B. Z. Weiss, Ultramicroscopy 23, 405 temperature resistivity equipment and for assistance (1987). with its use. 34 B.Z. Weiss, K. N. Tu, and D.A. Smith, Acta Metall. 34, 1491 (1986). 35 F.M. d'Heurle, IBM Research Report No. RC 10422, Yorktown REFERENCES Heights, NY 10598, 1983. 36 'O. Thomas, S. Delage, F. M. d'Heurle, and G. Scilla, Appl. Phys. I. J. M. M. Raaijmakers, Fundamental Aspects of Reactions in TitaLett. 54, 228 (1989). nium-Silicon Thin Films For Integrated Circuits, Ph. D. Thesis, 2 O. Thomas, S. Delage, F. M. d'Heurle, and G. Scilla, Appl. Surf. Technical University of Eindhoven, Eindhoven, The NetherSci. (1989). lands, 1988. 3 P. Villars and L. D. Calvert, Pearson's Handbook of Crystallo- 37M. Berti, A.V. Drigo, C. Cohen, J. Siejka, G.G. Bentini, R. graphic Data for Intermetalllc Phases (American Society for Nipoti, and S. Guerri, J. Appl. Phys. 55, 3558 (1984). 38 Metals, Metals Park, OH, 1986), pp. 2435, 3194. F. M. d'Heurle and A. Gangulee, in The Nature and Behavior of 4 L.S. Darken and R.W. Gurry, Physical Chemistry of Metals Grain Boundaries, edited by H. Hu (Plenum Press, New York, (McGraw-Hill, New York, 1953), p. 79. 1972), p. 339. ^Binary Alloy Phase Diagrams, edited by T. Massalski, L. H. Ben- 39F. K. LeGoues, M. Wittmer, T. Kwok, H-C. Huang, and P. S. Ho, nett, and H. Baker (American Society for Metals, Metals J. Electrochem. Soc. 134, 944 (1987). Park, OH, 1986), Vol. 1, p. 1252. •"V.T. Borisov, V. M. Golikov, and G.V. Scherbedinsky, Fiz. Met. 6 M-A. Nicolet and S. S. Lau, in VLSI Science and Technology: Mi- Metalloved 17, 881 (1964). 41 crostructure Science, edited by N. G. Einspruch and G. B. T. Rosso, M. Aucouturier, and P. Lacombe, Scripta Metall. 2, 393 Larrabee (Academic Press, New York, 1983), p. 453. (1968). 42 'Standard Powder Diffraction Pattern no. 21-357 (1971). P. Gas, S. Poize, and J. Bernardini, Acta Metall. 34, 395 (1986). 8 43 F. M. d'Heurle, J. Mater. Res. 3 (1), 167 (1988). P. Gas, S. Poize, J. Bernardini, and J. Cabane, Acta Metall, 37, 17 9 C. A. Pico and M. Lagally, J. Appl. Phys. 64, 4957 (1988). (1989). 10 44 J. E. Baglin, J. Dempsey, W. Hammer, F.M. d'Heurle, C.S. H. Kattelus and M-A. Nicolet, in Diffusion Phenomena in Thin Petersson, and C. Serrano, J. Electron. Mater. 8, 641 (1979). Films and Microelectronic Materials, edited by D. Gupta and P. S. U J. Torres, A. Pelissier, A. Perio, J. C. Oberlin, and G. Bomchil, Ho (Noyes Publications, Park Ridge, NJ, 1988), p. 432. 45 Le Vide-Les Couches Minces 42-236, 91 (1987). O. Thomas, A. Charai, F.M. d'Heurle, T.G. Finstad, and R.V. 12 A. Appelbaum, M. Eizenberg, and R. Brener, Vacuum 33, 227 Joshi, Thin Solid Films 171, 343 (1989). 46 (1983). J. M. E. Harper, S. E. HSrnstrom, O. Thomas, and L. Krusin13 P. Gas, J. Tardy, and F. M. d'Heurle, J. Appl. Phys. 60, 2211 Elbaum, J. Vac. Sci. Technol. A7, 875 (1989). 47 (1987). W. K. Chu, H. Krauttle, J.W. Mayer, H. Miiller, and K. N. Tu, 14 P. Gas, J. Tardy, F. K. LeGoues, and F.M. d'Heurle, J. Appl. Appl. Phys. Lett. 25, 454 (1974). Phys. 61, 2203 (1987). ""A. P. Botha and R. Pretorius (Proc. Mater. Res. Soc. Symp.) (Ma15 J. Angilello, J. E. Baglin, F. M. d'Heurle, C. S. Petersson, and A. terials Research Society, Pittsburgh, PA, 1982), Vol. 10, p. 129.

(7) Evidence provided from the reaction of Ti with a thin film of Ge over Si shows that Si and Ge mix extremely rapidly in the disilicide-digermanide at the temperature of formation, thus confirming that these metalloid atoms constitute the dominant moving species during the formation of the compound layers. (8) Considering the small difference in lattice parameter and the similar electronic configurations, the system TiSi2-TiGe2 exhibits a relatively large alloy scattering effect. Ge additions in TiSi2 have a larger scattering cross section than Si additions in TiGe2; with respect to atomic size this asymmetry is the same as encountered in a simple metallic system such as Cu-Ag. (9) The relatively high residual resistance of TiGe2 leads one to speculate about the possible presence of twins and stacking faults in this compound, and about the surprising absence of these planar defects in TiSi2 (C54).


1461


49

E. C. Cahoon, C. M. Comrie, and R. Pretorius (Proc. Mater. Res. Soc. Symp.) (Materials Research Society, Pittsburgh, PA, 1984), Vol. 25, p. 57. 50 'F. F. M. d'Heurle, F. K. LeGoues, R. Joshi, and I. Suni, Appl. Phys. Lett. 48, 332 (1986).

1462

51

F. M. d'Heurle, J. Tersoff, T. J. Finstad, and A. Cros, Appl. Phys. Lett. 59, 177 (1986). 52 A. White (private communication, 1988).

J- Mater. Res., Vol. 5, No. 7, Jul 1990