Observation of gold-silicon alloy formation in thin films by high resolution electron microscopy

Acta metall, mater. Vol. 39, No. 2, pp. 179-186, 1991

0956-7151/91 $3.00 + 0.00 Copyright © 1991 Pergamon Press plc

Printed in Great Britain. All rights reserved

OBSERVATION OF GOLD-SILICON ALLOY FORMATION IN THIN FILMS BY HIGH RESOLUTION ELECTRON MICROSCOPY W. ROBISON,t R. SHARMA~ and L. EYRING Department of Chemistry, Arizona State University, Tempe, AZ 85287-1604, U.S.A. (Received 18 December 1989; in revised form 22 June 1990)

Abstract--Thin films (<30nm) containing mixtures of gold and silicon were co-deposited on (100) surfaces of NaC1 from the pure elements co-evaporated at equivalent molar rates. The resulting samples, believed to be homogeneous distributions of the elements, were used to observe alloy formation during electron beam heating in an electron microscope. The formation of the alloy was shown to accompany the displacement and subsequent crystallization of excess silicon. The crystallographic data derived from the microscopic study were found to agree within experimental error with those reported from studies of bulk samples prepared by the rapid quenching of gold/silicon melts and assigned the composition Au2Si. Samples heated under intense electron beam irradiation resulted in the decomposition of the alloy and the formation of crystalline gold and silicon. Thicker samples (~ 100 nm) were prepared by evaporation on (100) NaCI and subsequently heated at temperatures up to 250°C. These specimens examined by the "thin-film" X-ray diffraction technique confirmed the formation of Au2Si as well as its subsequent decomposition at temperatures above 100°C. R~sumf----Des rims minces ( < 30 nm) contenant des mflanges d'or et de silicium sont dfposfs simultanfment sur des surfaces (100) de NaC1 5. des vitesses molaires 6quivalentes. Les 6chantillons rfsultants, considfrfs comme des distributions homog/~nes des 616ments, sont utilisfs pour observer la formation d'un alliage pendant le chauffage par le faisceau d'61ectrons dans un microscope 61ectronique. On montre que la formation de l'alliage s'accompagne du dfplacement et de la cristallisation ultfrieure de l'excfs de silicium. Les donnfes cristallographiques dfrivfes de l'ftude microscopique sont en accord, dans les limites des erreurs expfrimentales, avec celles qu'on a rapportfes pour des 6tudes sur des 6chantillons massifs prfparfs par trempe rapide d'or/silicium fondu et qui correspondent ~, la composition Au2Si. Dans les 6chantillons chaufffs par un faisceau intense d'61ectrons, on observe une dfcomposition de l'alliage et la formation d'or et de silicium cristallisfs. Des 6chantillons plus 6pais (~-100nm) sont prfparfs par 6vaporation sur une face (100) de NaC1, puis chaufffs zi des tempfratures allant jusqu'fi 250°C. Ces 6chantillons, examinfs par une technique de rayons x en film mince, confirment la formation de Au2Si ainsi que sa dfcomposition ultfrieure aux tempfratures supfrieures ~. 100°C. Zusammenfassung--Dfinne Filme (< 30 nm) einer Mischung aus Gold und Silizium wurden auf NaCI(100)-Oberfl~ichen durch Verdampfen mit gleichen Molrate aus Elementquellen abgeschieden. An den Schichten, von denen homogene Verteilung der Elemente angenommen wurde, wurde di Legierungsbildung w[ihrend der Elektronenbestrahlung in einem Elektronenmikroskop beobachtet. Die Legierungsbildung wird begleitet yon der Bewegung und der nachfolgenden Kristallisation des OberschuBsiliziums. Die aus den Beobechtungen erhaltenen Informationen fiber die Kristallografle stimmen innerhalb der experimentellen Ungenauigkeit mit denen fiberein, die an Volumenkristallen, hergestellt durch rasches Abschrecken einer Gold-Silizium-Schmelze, erhalten und mit Au2Si beschrieben wurden. Legierungsproben, die intensiv mit dem Elektronenstrahl aufgeheizt wurden, zerfielen zu kristallinem Gold und Silizium. Dickere Proben (~ 100 nm) wurden durch verdampfen auf (100)-NaCI hergestellt und anschlieBend bei bis zu 250°C ausgeheilt. Diese Proben wurden mittels der Rfntgenbeugungsmethode ffir dfinne Filme untersucht; die Ergebnisse best/itigen die Bildung yon Au2Si und dessen nachfolgende Zersetzung bei Temperaturen oberhalb von 100°C.

INTRODUCTION

term performance of advanced electronic devices. The chemical interaction between gold and silicon substrates in electronic devices is of great interest. The gold-silicon system has been under investigation since the late 19th century [1, 2], although studies of it were not widespread until the 1960's. The advent of semiconductor technology provided a new impetus as the need arose for very small conductors with high current carrying capacities. The initial bulk studies [1-3] of the gold-silicon system revealed a simple eutectic behavior at a

The introduction of thin film technology to the semiconductor industry has fostered an increased interest in the interactions of metals, non-metals and metalloids. The chemical as well as electronic properties are important in the design, fabrication and long tCollege of Engineering and Technology, Northern Arizona University, Flagstaff, AZ 86011-15600, U.S.A. :~Center for Solid State Science, Arizona State University, Tempe, AZ 85287, U.S.A. 179

180

ROBISON et al.: HIGH RESOLUTION STUDY OF Au-Si ALLOYS

composition of ,-~17% Si at a temperature of ,-, 370°C. See Fig. 1 [4]. Other workers [5-8], however, using X-ray diffraction, detected compounds which formed at room temperature or at temperatures well below 370°C. Anderson et al. [5], reported the formation of an orthorhombic compound (a = 7.82/~, b = 5.55/~, and c = 11.16/~) in rapidly quenched bulk samples of eutectic composition (17.9% Au). Anderson et aL reported that at room temperature, the surface layer of these samples dissociated over a period of a few hours yet the remainder of these specimens was unaffected. At elevated temperatures up to 370°C, however, the bulk samples also dissociated. Suryanarayana and Anatharama [6], performing X-ray diffraction on bulk samples, reported the formation of three structures in samples that were liquisol quenched (splat-cooled) from melts. In samples containing 16-25% Si, a cubic structure was reported to be of the E-manganese type with a unit cell containing 20 atoms and a = 6.75/L In samples containing 5-25% Si a cubic structure with a = 7.84/~ was found and in samples of 25-40% Si another cubic structure with a = 19.503/~ is reported [7]. Pedecki et al. [8] have determined that a ),-brass structure (a = 9.60/~) with the approximate composition of Au31Sis also exists in quenched bulk samples. Gold-silicon samples prepared as thin gold films on silicon (ll 1) surfaces have been widely studied [9-13], however, the goal of most of these studies has been to understand the relationship between the interactions of gold-silicon layers and their resulting electronic properties. Green and Bauer [9], investigated thin gold film deposition on Si [111] surfaces utilizing Auger electron spectroscopy (AES) and Low Energy Electron Diffraction (LEED). They determined that several structures are formed at temperatures well below the eutectic temperature of the bulk

system but could not relate these structures to any reported in the literature. Similar work was also performed by Oura and Hanawa [11], on the [100] silicon surface and with similar results. Hiraki et al. [10], studied the reconstruction of silicon layers on gold films deposited on silicon substrates. Using Rutherford Backscattering methods, a "coulombic screening" (sic) of the silicon surface atoms by gold was proposed to account for both the rapid diffusion of silicon in gold films and the formation of gold silicides at reduced temperatures. Using AES and Electron Energy Loss Spectroscopy (EELS) Cros et al. [12], confirmed the diffusion described above and reported the diffusion constant for silicon in gold. Previously, no studies have been made that demonstrate the formation of gold silicides directly, or conclusively isolate the different crystals that have formed in samples of widely varying compositions. The present work reports observations of the formation of a gold silicide in co-deposited thin films of gold and silicon. Analysis of the resulting thin films was accomplished by high resolution electron microscopy and electron and optical diffraction. The HREM observations were confirmed using separate samples heated under an inert atmosphere and examined by X-ray diffraction. The films themselves represent materials that are highly unstable with respect to either compound formation or the unmixing and phase separation of gold and silicon. The analysis by HREM allows the observation of the details of the chemical reactions at the atomic level. EXPERIMENTAL

Thin films containing gold and silicon ( ~ 200--300/~ thick) were co-deposited onto the [100]

Weight 0

16001400

Percent

Silicon

~

10

16

20

30

40 ~060 60100

I

=

=

~

i

~ =14=14/oc

iF

............•................

P 1200-11~4.43~c t. 1000-I

S + ~-

t~ t.

Q.

600 -

E 1--

600 400 20O

370°C 0

110

210

3=0

410

5=0

6=0

7i0

810

9i0

100

Atomic Percent Silicon Au

Si

Fig. 1. Phase diagram of the gold-silicon system [13].

ROBISON et al.:

HIGH RESOLUTION STUDY OF Au-Si ALLOYS

faces of rock salt (NaCI) by co-evaporating the pure elements in a 1/1 atomic ratio. A base pressure for this system was maintained at 4.0 x 10 -7 torr during the deposition. The rate of film deposition and its thickness were monitored using a quartz crystal oscillator. Individual evaporation rates were maintained at 0.4 A/s and were established prior to deposition. The rock salt substrate was maintained at room temperature (23°C) with a water cooled sample platform. By evaporating in this manner, a homogeneous distribution of gold and silicon atoms was probably attained within the final film. Film samples were then mounted on 400-mesh gold specimen grids using standard procedures. In order to observe alloy formation it was necessary to have freshly prepared specimens. If the codeposited films were left in air no reaction was observed in the microscope. It was assumed that oxidation of silicon forms barrier layers preventing interdiffusion. Strong beam heating at lower voltages (100 kV) resulted in the crystallization of gold and no alloy formation was observed, whereas at higher voltage (400 kV) no alloy was formed supposedly due to insufficient heating or other radiation effects. Transmission electron micrographs were obtained using two different microscopes: a JEOL 200-CX and a JEOL 4000-EX. The 200-CX has a "point to point" resolution of 2.4 ,~ and was used to follow the events of electron beam treatment of the samples. The controlled reaction in sample films was induced by the 200 kV electron beam in the JEOL 200-CX electron microscope using spot size No. 2 and removing-the condenser aperture. High resolution electron micrographs of films heated in the 200-CX electron microscope were obtained with the 4000-EX microscope which can resolve details as small as 1.7 ~. The high resolution images obtained were then used to determine the identity of crystalline features by optical diffraction. Optical diffraction patterns were taken of each crystalline feature and the corresponding crystallographic d-values and interplanar angles were used for comparison with the ASTM powder diffraction files on gold-silicon alloys [14]. Matches between observed and calculated data were determined using the "CRYSTALS" program at the Center for Solid State Science [15]. Five additional gold-silicon samples were prepared for heat treatment and X-ray analysis. Approximately 1000-~ films were prepared as described above. Each sample was then heated for 5 min in a vertical "tube" furnace under a nitrogen atmosphere. Temperatures chosen for these treatments were 100, 150, 200, 250°C and room temperature (23°C) (control sample). During X-ray determinations, the diffractometer was operated in the "thin-film" mode to insure sufficient intensity for diffracted X-rays. In this method the incident beam was fixed at 4 ° 2-0 and the diffracted beams detected by sweeping from 4 ° to 60 ° 2-0. This insured sufficient interaction of the

181

beam and the thin film sample. These samples were then stored in a vacuum desiccator for one month and reexamined by X-ray diffraction.

RESULTS AND DISCUSSION

The amorphous Au-Si films were initially examined in the JEOL 200-CX. To attain high resolution images, this microscope was operated with beam spot size No. 3 and condenser aperture No. 2. This procedure established a coherent beam but of rather low intensity. Under this low intensity beam, the films were quite stable and no changes were noted in the film structure. Figure 2(a) is a selected area electron diffraction pattern of the film after only 1.0 min exposure to the electron beam. Figure 2(b) is a high resolution micrograph of the corresponding sample. Although no crystalline structures are apparent in either the micrograph or diffraction pattern, the presence of the broad diffuse ring in the electron diffraction pattern indicated the beginning of short range ordering, i.e. the onset of either alloy formation or the crystallization of gold and/or silicon. After 20 min exposure [Fig. 2(c)], several clearly discernable developments were noted. As the sample was exposed for longer periods, the number and size of the crystals formed was found to increase until they began to impinge. Upon 22 min exposure [Fig. 2(d)], these crystals had grown to the extent that they now consumed the surrounding "amorphous" matrix. At this point, large, nearly transparent crystals began to form that were subsequently identified as silicon. After strong beam heating (all apertures removed, spot size No. 1), the structures containing 11.3 ~ fringes were found to diminish in size and number as the growth of silicon crystals continued. After 30 min, none of the crystals containing 11.3/~ fringes were seen. To investigate further the identity of these crystals, electron micrographs were made using the JEOL 4000-EX of electron beam heated films previously studied in the JEOL 200-CX (Fig. 3). Region " A " (Fig. 3) was determined to be silicon in the (111) orientation. This crystalline region was initially identified by the close packed array of silicon atoms and subsequently confirmed by optical diffraction. Areas B, C, and D were found to be Au2Si in the (570), (110) and (110) orientations respectively. The presence of pure crystalline silicon and the areas of light contrast adjacent to regions undergoing the formation of Au2Si suggest the reaction sequence: (2 Au + 2 Si) co-deposited film Au2Si + Si (excess). In this way randomly distributed Au and Si atoms in the co-deposited film combine to form crystalline Au2Si with the concomitant exclusion of excess Si from the region. Regions of pure excess silicon could result from depletive diffusion of either silicon or gold

182

ROBISON et al.:

HIGH RESOLUTION STUDY OF Au-Si ALLOYS

Fig. 2. High resolution electron micrograph of Au-Si film between periods of electron beam heating, (a) electron diffraction pattern and (b) image of sample film after 1.0 min exposure, (c) after 20 min exposure, (d) after 22 min exposure, (e) after 25 min exposure, and (f) after 30 min exposure.

ROBISON et al.:

HIGH RESOLUTION STUDY OF Au Si ALLOYS

183

3

W

§§m

Fig. 3. High resolution micrograph and optical diffraction patterns showing formation of (A) silicon [111], (B) Au2Si [570], (C) Au2Si [880], and (D) Au2Si [220] adjacent to region undergoing reaction.

from the original equal molar mixture as Au2Si crystallizes. Evidence has previously been given that Si is more mobile under the circumstances of these experiments [9, 10]. The light regions between C, B and D consist primarily of silicon left behind a s A u 2 Si grew from the 50-50 mixture. This silicon rich material has not yet become crystalline.

Figure 4 also shows Au2Si in different orientations within the sample. The composition, Au2Si, is assigned as reported in the ASTM file card [14]. Although there is insufficient evidence to establish a genetic relationship between these crystalline regions, the variety of orientations indicates their formation from nuclei formed in the bulk rather than at the film

184

ROBISON et al.:

HIGH RESOLUTION STUDY OF Au-Si ALLOYS

Fig. 4. High-resolution micrograph and corresponding optical diffraction patterns of Au2Si in the (A) [135], (B) [3, 5, 15], (C) [785], and (D) [734] orientations. surfaces. According to Refs [4, 14] Au2Si has a f.c.c. structure with a0 = 19.503/~. Comparison of the present work shows that all the reflections agree within the limits of experimental error with those obtained by X-ray diffraction of these bulk samples (Table 1). When comparisons were made of other reported Au/Si compounds only the 11.3/~ d-value was found to match and this corresponded only to the (001)

reflection of the orthorhombic structure reported by Anderson [4]. Samples used in XRD determinations confirmed the TEM observations and provided additional evidence of the growth mechanism of Au2Si in films supported on (100) NaC1 surfaces (see Fig. 5). Normally, in bulk "powder" diffraction patterns, the highest intensity peak of Au2Si is the (331) reflection

ROBISON et al.:

5

HIGH RESOLUTION STUDY OF Au-Si ALLOYS

15

25

35

45

185

50

Fig. 5. Thin film X-ray diffraction patterns of gold-silicon films heated at (a) (room temperature), (b) 100°C, (c) 150°C, (d) 200°C and (e) 250°C for 5 min.

/ ..... l I L_,,. ..... / _j..,.._:,__~t'k.~ - _L---___-_..L " ....

5

15

,,_

~. '

. . . .

25

35

45

50

Fig. 6. Thin film X-ray diffraction pattern of gold silicon films 1 month later.

186

ROBISON et al.: HIGH RESOLUTION STUDY OF Au-Si ALLOYS Table 1. Comparison of d-values of structures observed in co-deposited Au/Si thin films with reported and refined crystal lattices Au2Si(f.c.c.) hkl 1/Io Ref. [14] This study 111 60 11.3 11.15 300 -6.41 311 40 5.92 5.89 222 -5.692 331 60 4.48 -511 60 3.75 3.79 440 40b 3.45 -531 20 3.30 3.38 533 40b 2.98 3.08 622 40b 2.94 2.95 551 40b 2.77 -731 40 2.54 -800 40b 2.43 -820 50 2.36 -660 40b 2.30 -662 100 2.25 -840 60b 2.18 2.16 911 80 2.135 -664 20b 2.10 -Mean % error for optical diffraction measurement is about 2%.

[14]. This was also the case in filr0s heated less than 150°C and indicates r a n d o m nucleation and growth within the film. F o r films heated at higher temperatures but below 250°C, the growth in the ( 1 1 1 ) orientation predominates. This is due partly to the close match between the (111) Au2Si (d = 11.29/~) and the (100) NaCI (2d = 11.2804/~). The enhanced growth of Au2Si in the (111 ) orientation may also be related to surface diffusion of gold and silicon. Samples heated above 200°C were found to decompose to crystalline gold. The lack of any discernable silicon phases (by X R D ) is attributed to either the low X-ray scattering from the lighter elements or a lack of crystallinity in the Si. These samples were stored at r o o m temperature for one m o n t h under vacuum and X-ray determinations were repeated as described above (see Fig. 6). With the exception of the unheated sample, all indications of Au2Si were missing, having decomposed during that period. The only pattern present in the sample was that of crystalline gold. O f particular interest is the apparent similarity of structures observed here using low temperature codeposited elements and those prepared by rapidly quenching (liquisol quench) high temperature melts to low temperatures with subsequent warming to r o o m temperature [3, 5-7]. The experiments reported here suggest that t h e alloy is formed only at low temperatures and decomposes rapidly in thin films at temperatures above 100°C. The samples used in the current study represent an unusual form of m a t t e r - elements which are essentially quenched from the gas phase to form an intimate mixture. Assuming the phase diagram . (Fig. 1) is correct, the elements in samples thus prepared should segregate into a eutec-

tic mixture of pure phases. This final phase separation, however, required extensive electron bombardment over a period of 30 min. The formation of Au2Si in a metastable state appears most likely, however, stability at quite low temperatures cannot be ruled out.

CONCLUSIONS

These studies show that homogeneous films of Au and Si are stable with respect to low intensity electron beam heating. Larger irradiation doses were required to promote reaction. The main reaction observed was the formation of Au2Si. As a result of the reaction, excess Si has been observed to segregate and recrystallize in the (111 ) orientation. Crystallographic data obtained from these samples by optical and electron diffraction methods correspond closely with that reported to be Au2Si by Suryanarayana in the A S T M X-ray powder diffraction file [14]. Continued heating of the film results in the decomposition of Au2Si to its elements which also crystallize. Subsequent X-ray diffraction analysis of thicker films confirms these observations as well as the lattice parameter determinations. The observations in this study suggest a similarity between samples prepared by co-deposition in thin film and those prepared by a rapid quench from melts. Acknowledgements--We are grateful to the NSF for research support through Grant DMR-8820017, and for access to the Center for High Resolution Electron Microscopy funded by the NSF through DMR-8611609. REFERENCES

1. C. diCapua, Rend. Accad. Naz. Lineci 29, I l l (1920). 2. E. Virouroux, Ann. Chim. Phys. 12, 170 (1897). 3. H. L. Luo, W. Klement and T. R. Anatharama, Trans. Indian Inst. Metals 18, 214 (1965). 4. Bull. Alloy Phase Diagr. 2, 359 (1981). 5. G. A. Andersen, J. L. Bestel, A. A. Johnson and B. Post, Mater. Sci. Engng 7, 83 (1971). 6. C. Suryanarayana and T. R. Anatharama, Mater. Sci. Engng 13, 73 (1974). 7. T. R. Anantharama, H. L. Luo and W. Klement, Nature 210, 1040 (1966). 8. O. Predecki, B. C. Giessen and J. J. Grant, Trans. metall. Soc. A.LM.E. 233, 1438 (1965). 9. A. K. Green and E. Bauer, J. appl. Phys. 47, 1284 (1976). 10. A. Hiraki, E. Lugujio and J. W. Mayer, g. appl. Phys. 43, 3643 (1972). 11. K. Oura and T. Hanawa, Surf Sei. 82, 202 (1979). 12. A. Cros et al., Thin Solid Films I00, 17-24 (1983). 13. S. R. Herd, K. Y. Ahn and K. N. Tu, Thin SolidFilms 104, 197 (1983). 14. ASTM Powder Diffraction File; (JCPDS) card no. 26-724. 15. CRYSTALS Program by G. E. Spinnler and M. T. Otten for the Center for Solid State Science, Arizona State Univ. (1986).