IB Physica B 208&209 (1995) 367-368
Structure of amorphous metal-metalloid alloys B.T. Williams a, S.J. Gurman a'*, J.C. Amiss b aDepartment of Physics, University of Leicester, Leicester LEI 7RH, UK b BMS, Sheffield Hallam University, SheffieM SI I WB, UK
Abstract EXAFS data has been obtained for both K absorption edges of amorphous Gel _xTi~ samples prepared in thin film form by RF sputtering over the composition range 0 < x < 0.7. The data have been analysed both conventionallyand by means of Reverse Monte Carlo simulation. All samples show a degree of chemical order. The coordination numbers, both partial and total, vary significantly across the composition range. SAXS data show that no samples are phase separated but contain voids on the 10-20 A scale. The structural information obtained from the EXAFS data is used to aid our understanding of the metal-insulator transition exhibited by these alloys.
2. Experimental and results
Amorphous alloys of transition metals with group IV metalloid atoms show mechanical, electrical and magnetic properties which make them of interest in a technical as well as fundamental sense. Thin film amorphous samples of such alloys can be produced over a wide composition range by means of RF sputtering. Such alloys exhibit a metal-insulator transition (MIT) at between 15 and 25% metal content. We present here an application of the EXAFS technique to obtain short range order information on the radial distribution functions of a series of amorphous Gel -xTix alloys with x = 0.12 to 0.70 in order to aid our understanding of the MIT. We have also used the SAXS technique, sensitive to medium range features, to detect any phase separation present in our samples: both Si-Ni  and G e - F e  alloys are known to show phase segregation on the 30 ,A scale.
EXAFS data were taken at Daresbury SRS on station 7.1. A silicon (1 1 1) monochromator was used with harmonic rejection set at 50%. Beam currents during data taking were between 200 and 300 mA. The data were calibrated and background subtracted in the usual way and analysed using EXCURV92 and also a method new to our group, a Reverse Monte Carlo (RMC) method . Each has useful advantages over the other. (i) EXCURV92 fits a theoretical structure factor to the data by iterating the parameters of a theoretical EXAFS function. Interatomic distances can be fitted to within an accuracy of 1% because of the sensitivity of the least squares parameter fitting. Coordination numbers however can only be determined to within 10%. The theoretical EXAFS function assumes a Gaussian distance distribution so will miss any long-R tails which may be present. (ii) RMC models the sample as a three dimensional array of particles by adjusting atomic positions until the structure factor of the artificial network matches the data.
* Corresponding author.
0921-4526/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved SSDI 0 9 2 1 - 4 5 2 6 ( 9 4 ) 0 0 8 5 2 - 3
B. 1~ Williams et al./ Physica B 208&209 (1995) 367-368
Table 1 Gel xTix partial coordinations _+ 0.5 x 0.12 0.22 0.22 0.33 0.42 0.51 0.51 0.70
Ea E Ra E E E R E
4.2 3.8 4.6 2.0 2.7 6.1 2.7
(0.4) (1.2) 1.3 1.4 2.5 3.1 3.3 6.4
2.7 4.3 4.6 3.5 4.0 3.7 3.3 (2.7)
-0.9 ....... -2.3 1.5
aE: EXCURV92; R: RMC. Bracketed values were calculated using the consistency condition on the number of Ge-Ti bonds.
This method makes no a priori assumptions about the material, only the composition, density and the data itself are entered. Both the Ge and Ti edges are fitted simultaneously. It will however return the most disordered structure consistent with the data, because of the selection method employed to match the structure factors. The method also makes heavy demands on computing time and memory space. SAXS data were taken on station 8.2 at Daresbury at 8 keV using a camera length of 1.5 m (Qm~n = 0.02/~- 1).
with increasing metal content as more of the larger Ti atoms compete for space. The T i - G e bond length remained the same throughout the composition range (2.64 ~,), varying only by 0.02/~. The G e - T i bond length however fluctuated more widely but essentially remained constant around 2.68 + 0.05A, consistent with the T i - G e bond length. We could only fit a reliable T i - T i to our data for x = 0.70 using EXCURV92. Two compositions to date, spanning the MIT, have been analysed using the R M C method. The distances calculated are less precise than the E X C U R V 9 2 equivalents because of a real space resolution limit of 0.1/~. These distances are in very good agreement with the E X C U R V 9 2 values, the G e - G e bond being 2.5 A for both compositions, the G e - T i bond 2.6/~ for x = 0.22 and 2.5/~ for x = 0.51. We also obtained a value for the Ti Ti bond length, 2.7 ,A for both compositions. This is 0.21 ~, shorter than the crystal value but consistent with the covalent radii of Ti. By inspection of Table 1 it can be seen that the coordination numbers returned by R M C are in good agreement with E X C U R V 9 2 with the exception of G e - G e at x = 0.51. In this case the R M C value is deemed to be correct as it fits the trend of decreasing coordination number with increasing metal content and sixfold total Ge coordination. An anomalously high D e b y e - W a l l e r factor associated with the E X C U R V 9 2 value serves to reinforce this belief.
3. Discussion The SAXS data show no evidence for phase separation. Guinier plots show all samples contain isolated structures of radius 10-20 ,A. Since these also exist in a-Ge we interpret them as argon gas bubbles introduced during the sputtering process. The fact that the total coordination of Ge at low Ti concentration (x < 0.22) is similar to that of pure a-Ge suggest that the basic tetrahedral structure is still maintained in this composition range. The total coordination of Ge increases with Ti concentration extending over a wide range. It is apparent from Table 1 that Ti coordination increases rapidly with the increase of metal concentration above x = 0.33 to adopt a sixfold structure at x = 0.70. F o r compositions x = 0.12 to 0.33 the G e - G e bond length was found to be essentially the same as in the crystal, 2.44 ,A. This bond shows signs of lengthening
4. Conclusions F r o m our E X C U R V 9 2 results the structural network of sputtered Gel -xTix is reasonably chemically ordered with the G e - T i bond favoured. Distances are constant throughout the composition range and indicate a covalent radius of 1.25 A for Ge and 1.35/~ for Ti.
References  E.A. Davis, S.C. Bayliss, R. Asal and M. Mansor, J. NonCryst. Solids 114 (1989) 465.  R.D. Lorentz, A. Bienenstock and T.I. Morrison, Phys. Rev. B 49 (1994) 3172.  S.J. Gurman and R.L. McGreevy, J. Phys.: Condens. Matter 2 (1990) 9463.