File: SECT03.WM of Tape: Various/ETH/s10-diss
(Source file text)
- 25 - @ka III. ANISOTROPIC RELAXATION TIMES . ___________________________________ 1) Structural and free electron data. ____________________________________ @ke @ba Several general quantities always crop up in calculations concerning Fermi-surfaces or involving pseudo-potentials. The proper low-temperature values are not always used and we find it convenient to include these constants for future reference. @be @ba The periodic boundary conditions in a crystal show that the density in k-space is given by: 2 V(cryst)/(2&p)^3^. From this it follows that the Fermi-wavenumber k\F0\, corresponding to the free electron sphere is: @be @ea 3&p^2^ Z s 3&p^2^ Z (3.1) k\F0\ = [---------]^1/3^ = [-------]^1/3^ V(cryst) &O\o\ @ee @ba where Z is the valence of the considered atom, s the number of atoms per unit cell, V(Cryst) the volume of the unit cell and &O\o\ the atomic volume. @be @ba Because all calculations are done in k-space, it is of advantage to work in so called computational units and one choses one of the unit reciprocal lattice vectors 2 &p/a as the new unit length. With x(vol)= &O\o\s/a^3^ we get: @be @ea cu 3 Z s (3.2) k\F\ = (------------)^1/3^ in comp. units 8 &p x(vol) @ee - 26 - @ba The x(vol) constant is a characteristic of each crystal structure, i.e. for FCT x=c/a, HEX: x=SQRT(3)c/(2a), HCP: x=SQRT(2), RHOMB: x=(1-COS(&a)) SQRT(1+2COS(&a)) @be @ba One would like to avoid the use of atomic units, but it just happens that most of the literature abounds in Rydbergs, Bohr radii, etc. For reasons of completeness we give the definitions and conversion factors:(au=atomic units,cu=computational units) @be @ka Quantity Unit Factor Unit ______________________________________________________ Mass m\e\ 9.1091 10^-28^ grams Length h'^2^/me^2^ .52917 Angstroem =a0 =Bohr radius Time h'^3^/me^4^ 2.419 10^-17^ seconds Energy(a.u.) me^4^/h'^2^ 4.3594 10^-18^ Joule 27.2105 eV " (Rydberg) me^4^/2h'^2^ 2.1797 10^-18^ Joule 13.6052 eV Resistivity at. units 21.74 &m&O-cm/100At% @ke @ba We now have: @be @ea au au cu k\F\ = ( 2 &p / a ) k\F\ @ee @ba and the Fermi energy in computational units is usually defined as: @be @ea cu cu E\F\ = (k\F\ )^2^ @ee - 27 - @ba Table 3.1 gives a collection of lattice spacings and atomic volumes in a.u. for the solvents and solutes of interest (no values are given for the transition metal solutes because we can not treat them with the pseudopotential formalism ). Where available the low temperature lattice spacings have been used, otherwise we extrapolated the room temperature l.c.'s to 0K with the aid of thermal expansion data. @be @ka _________________________________________________________ | | | | | | | | | Metal | Struct| a | c | b | | | | and | and | in | in | or | c/a | &O | |valence| states| a.u. | a.u. | Angle | | o | | | | | | | | | _________________________________________________________ | | | | | | | | | Bi/5 | RH/2 | 8.9319| | 57.314| | 236.36| | | | | | | | | |*Si/4 | DIA/8 |10.2628| | | | 135.11| | | | | | | | | |*Ge/4 | DIA/8 |10.6910| | | | 152.74| | | | | | | | | | Sn/4 | BCT/4 | 10.985| 5.9636| | .5429 | 179.91| | | | | | | | | | Pb/4 | FCC/4 | 9.2874| | | | 200.27| | | | | | | | | | Al/3 | FCC/4 | 7.6192| | | | 110.58| | | | | | | | | | Ga/3 | ORH/8 | 8.5324| 14.422| 8.4814| 1.6903| 130.46| | | | | | | | | | In/3 | FCT/4 | 8.6038| 9.3254| | 1.0839| 172.58| | | | | | | | | | Tl/3 | HEX/2 | 6.4970|10.3521| | 1.5934| 189.22| | | | | | | | | | Mg/2 | HEX/2 | 6.0361| 9.7959| | 1.6229| 154.55| | | | | | | | | | Zn/2 | HEX/2 | 5.0275| 9.1868| | 1.8273| 100.54| | | | | | | | | | Cd/2 | HEX/2 | 5.6103|10.4482| | 1.8624| 142.40| | | | | | | | | | Hg/2 | RH/1 | 5.6434| |70.7433| | 155.18| | | | | | | | | | Li/1 | FCC/4 | 8.3338| | | | 144.70| | | | | | | | | | Cu/1 | FCC/4 | 6.8093| | | | 78.93| | | | | | | | | | Ag/1 | FCC/4 | 7.6892| | | | 113.65| | | | | | | | | _________________________________________________________ Table 3.1: Low temperature structural data @ke - 28 - @ba Table 3.2 gives the followig constants computed with the values of Table 3.1: the free electron Fermi-vector (in c.u.), the Fermi energy in a.u., the reciprocal lattice constant in a.u. and the factor A(cu), which is used to convert dHvA frequencies F to computational areas : Area=A(cu) F[Gigagauss]. The references indicated in Table 3.2 also apply to Table 3.1. @be @ka _________________________________________ | | | | | | | | 2&p | cu | au | | | Metal | ---- | k | E | A(cu) | | | a(au) | F | F | | | | | | | | _________________________________________ | | | | | | | Bi | .7035 | 1.2163| .3661 | 5.4020| Barret (1960) | | | | | | | *Si | .6122 | 1.5632| .4580 | 7.1313| Ben et al.(1962) | | | | | | | *Ge | .5877 | 1.5632| .4220 | 7.7389| Cooper (1960) | | | | | | | Sn | .5720 | 1.5209| .3784 | 8.1703| Lee;Raynor(1954) a) | | | | | | | Pb | .6765 | 1.2407| .3595 | 5.8402| Pearson (1964) b) | | | | | | | Al | .8246 | 1.1273| .4321 | 3.9306| Figgins (1956) c) | | | | | | | Ga | .7364 | 1.1946| .3870 | 4.9293| Barret (1962) | | | | | | | In | .7303 | 1.0974| .3211 | 5.0120| Graham (1955) d) | | | | | | | Tl | .9671 | 0.8036| .3020 | 2.8580| Barret (1958) | | | | | | | Mg |1.0409 | 0.6978| .2638 | 2.4669| Pearson (1964) e) | | | | | | | Zn |1.2498 | 0.6707| .3513 | 1.7114| Lynch (1965) e) | | | | | | | Cd |1.1199 | 0.6665| .2786 | 2.1311| Lynch (1965) e) | | | | | | | Hg |1.1133 | 0.6515| .2631 | 2.1564| Barret (1957) | | | | | | | Li | .7539 | 0.7816| .1736 | 4.7023| Khatkevic (1952) | | | | | | | Cu | .9227 | 0.7816| .2601 | 3.1394| Frohnmeyer (1953) f) | | | | | | | Ag | .8171 | 0.7816| .2040 | 4.0031| Smakula (1955) b) | | | | | | _________________________________________ Table 3.2: Free electron constants. (*: Room temperature values) @ke - 29 - @ka Thermal expansion references: a) White (1964) b) Nix;McNair (1942) c) Extrapolated. d) Collins et al.(1967) e) McCammon;White (1965) f) Nix;McNair (1941) @ke - 30 - @ka 2) Pseudopotentials. ___________________ a). Introduction. _________________ @ke @ba The problem of determining the band structure of a metal is an old one and there are several , more or less succesful, ways of solving it. @be @ba What one always has to do, is to solve the Schroedinger equation: @be @ea (3.3) H &Q\i\ = ( T + V(r) ) &Q\i\ = E\i\ &Q\i\ @ee @ba where T is the kinetic energy, V(r) the self-consistent potential seen by an electron and E\i\ the energy of the i'th state. Now it happens that the potential is "strong" in the sense that it creates low lying bound states, called "core states". If one solves the Schroedinger equation in the k-representation: @be @ea (3.4) det| [ 1/2 ( k - g )^2^ - E ] &d\gg'\ + V\gg'\ | = 0 @ee @ba (here the V\gg'\'s are the Fourier components of V(r) ), we have to go to very large k's (small wavelengths) in order to make the calculation convergent. ( For Al the secular equation would have to be of order 10^6^ x 10^6^ ). The lowest eigenvalue of this solution will correspond to the lowest core state and will not be particularly interesting for the properties of a metal. It is quite clear that such a calculation would be senseless and some means are needed to eliminate the core states. @be - 31 - @ba The fact that the V(r) has a deep potential well, leading to core states, entails that the wave functions have several nodes in the core region. One can now prescribe a core radius R\c\, such that the core states (wiggles) are all included in a sphere of radius R\c\. One can calculate the wave function at R\c\ for the free atom and join these solutions to the solution of the Schroedinger equation outside this sphere. One can now show that the original V(r) can be replaced by another poten- tial V\ps\(r) called the pseudopotential (PP). The solutions of the Schroedinger equation using this potential are much smoother and correspond to the outer valence states. Now the lowest eigenvalue belongs to the lowest valence state. @be @ba Another way of looking at this problem is by using the scattering theory approach. An incident wave |k> falling on an atom will emerge with an angular distribution: @be @ea (3.5) 4&p/k $S (2l+1) [ exp(2i&h\l\) - 1] P\l\(cos &j) @ee @ba The phase angle can be written: @be @ea (3.6) &h\l\ = p\l\ &p + &d\l\ |&d\l\| <= &p/2 @ee @ba The constant p\l\ corresponds to the number of bound states, but does not have any influence on the scattered wave. A pseudopotential is now a potential which produces only phase shifts of &d\l\; the bound states have been removed. We shall be using these two aspects in our further discussion: the second aspect will be helpful in discussing scattering properties in relation with the character of the wave packets (s,p,d, character) @be - 32 - @ka b). The empty core or Ashcroft(1966) potential. _______________________________________________ @ke @ba This is the simplest of all pseudopotentials and uses the "cancellation theorem" (Heine(1970)) to set the depth of the well exactly to zero: @be @ea V\ash\^ion^(r) = 0 for r<R\c\ (3.7) = -Z/r for r>R\c\ @ee @ba Here R\c\ is some model radius adjusted for best fit with experimental data and Z is the valence of the metal. The Fourier transform of a potential is: @be @ea (3.8) v(q) = 1/&O\o\ $I V^ion^(r) exp(-iqr) d^3^r @ee thus, @ea (3.9) v\ash\ = - [ 4&pZ/ ( q^2^&e(q) &O\o\] cos( qR\c\) @ee @ba In reduced q's: x=q/2k\F\ @be @ea k\F\ (3.10) v\ash\ (x) = - ------------ cos( 2 x R\c\ k\F\) 3 &p x^2^ &e(x) @ee @ba where &e(x) is the dielectric screening function without exchange: @be @ea k\F\ (3.11) &e(x) = 1 + ---------------- f(x) 3 &p x^2^ (2/3 E\F\) @ee @ea with f(x) the Lindhardt screening function: 1 - x^2^ 1 + x (3.12) f(x) = 1/2 + ------ ln | ----- | 4 x 1 - x @ee - 33 - @ba It is seen that in the limit of x going to zero this potential goes to -2/3E\F\ (the deformation potential), as every local pseudopotential should. @be @ba We have fitted this PP to experimental Fermi surface data, where available, and we give the obtained R\c\'s and the sources used in Table 3.3. @be @ba These pseudopotentials are useful for qualitative discussions, but it is difficult to see how they are influenced when used for solute calculations. In the next paragraph we develop a much more realistic model for this situation. @be @ka _________________________________________________ | | | Li Mg Al Si | | 1.73 1.30 1.15 1.02 | | | | Cu Zn Ga Ge | | 0.81 1.11 1.06 0.98 | | | | Ag Cd In Sn | | 1.04 1.25 1.09 1.11 | | | | Hg Tl Pb Bi | | 1.27 1.11 1.00 0.98 | | | _________________________________________________ Table 3.3: R\c\'s for Ashcroft PP. (in a.u.) Zn,Al,Mg,Pb,In Sorbello;Griessen (1973) Bi,Tl,Li Cohen;Heine (1974) Cu,Ag Ashcroft;Langreth (1967) Si,Ge Brust (1964) Sn Stafleu;de Vroomen (1967) Ga Reed (1969) Hg Brandt;Rayne (1966) Cd Stark;Falicov (1967) @ke - 34 - @ka c). The Shaw potential. _______________________ @ke @ba The pseudopotentials of the Heine-Abarenkov-Shaw type are so called "model potentials" and were first given by Animalu and Heine(1965) and later on Shaw(1968) showed that it was possible to optimize them in a certain way. These potentials are called model potentials to distinguish them from ab initio pseudopotentials and also because they make use of empirical information concerning the valence electron energy levels. @be @ba As already mentioned, the problem lies in knowing the wavefunction of the Z times ionized atom outside some model radius R\l\.(the R\l\'s may be different for different angular momenta l=0,1,2;see Fig. 3.1) These wavefunctions will naturally be energy dependent and one has to evaluate them at the appropriate energy E\F\^ion^ with respect to the free ion. Now we know that the solutions to the hydrogenlike Schroedinger equation are only finite for energies at the term values (n,l) of the ion in question and can be represented by Whittaker functions of the first kind for the quantum numbers n,l. (Shaw(1968b)). @be @ba The logarithmic derivative at the radius R\l\ of the solution n,l can then be calculated and a solution corresponding to the interior square well of depth A\x\ can be made to fit this derivative. The resultant A\x\ will then specify the depth of the well A\l\ for a certain l. @be - 35 - @fa FIGURE 1 Figure 3.1:The optimized model potential; the square wells fit exactly in the Coulomb part Z/r and are of different depths. @fe @ba Shaw(1968,1968b) has shown that an optimized form of the potential (optimized in the sense that the wavefunction is as smooth as possible) results if the condition A\l\=Z/R\l\ is fulfilled. He also shows that it is not neccessary to model wave-functions higher than l\0\, where l\0\ is the highest angular momentum of the core-wavefunctions.(e.g. l\0\=0 for Li). He has given tables for relating the A\l\ with the term values E\l\. We have reproduced these tables in a graphical form in Fig. 3.2, where ln(A\l\/Z^2^) is given in function of ln(-E\l\/Z^2^). @be - 36 - @ba The only, but major, problem remaining is the inter- or extrapolation of the A\l\(n)'s to the energy E\F\^ion^. Shaw has shown, empirically, that a linear extrapolation is reasonable provided that the d-bands are far enough in energy fron the conduction band. We have found that a linear interpolation in our ln-ln plot is often superiour and a logarithmic extrapolation will tend to flatten out at low energies provided the absolute value of the slope is smaller than 1. (which is the case for the metals we considered). There is no a priori reason to make a linear extrapolation and the fact that our interpolation is often better, makes us feel confident that our method is comparable or even better than the linear method. Our interpolation formula is thus as follows: @be @ea (3.13) A/ Z^2^ = &a ( -E / Z^2^ )^&b^ @ee @ba Some of the elements we studied did not allow optimization, viz: Cu,Ag and the transition metals Ti,V,Cr and Mn. In these metals the d-band is too close to the conduction band and the usual hydrogen-like term scheme breaks down. Animalu(1973) has given a whole series of model potentials, which includes the last mentioned, but the effect of scattering resonances seems rather difficult to take into account. @be - 37 - @ba The metals Zn,Cd and Hg are an intermediate case, where the influence of the d-bands just starts to be noticeable. This series of elements has been studied by Evans(1970) by incorporating the quantum defect method in Shaw's formalism. We have determined our constants &a and &b from his values of A\l\ and $dA\l\/$dE at E\F\^ion^ for Zn,Cd,Hg and also for Pb,In,Tl. In Table 3.4 we give our fitted values of &a and &b for future reference. @be @fa FIGURE 3.2 careful: 0 1 2 regions not indicated Figure 3.2: ln-ln plot of Shaw's tables. Region 1 corresponds to the usual one node so- lution, region 0 to exponentially de- caying solutions. @fe - 38 - @ka d). Fermi energy for solutes. _____________________________ @ke @ba The problem remains of the E\F\^ion^ at which one should evaluate the A\l\'s. This problem has been discussed by Animalu and Heine(1965) and their method has been adopted Shaw,Evans, etc. In solving the Schroedinger equation (3) for an ion in a metal the potential energy term can be written as follows: @be @ea (3.14) V(r) = V\ion\(r) + V\rest\(r) @ee @ba where V\ion\ is the potential for a single isolated ion and V\rest\ a potential due to the rest of the system, principally the conduction electrons in the neighbourhood of the ion when immersed in the metal. If V\rest\ can be considered more or less constant in the model sphere, the energy at which to solve for the free ion will be: @be @ea (3.15) E\F\^ion^ = E\F\^vac^ - V\rest\ @ee @ba The value of the absolute energy E\F\^vac^ can be obtained by considering the following expression for the cohesive energy: (Kittel(1963)p.93,Seitz(1940),Animalu;Heine(1965)) @be @ea (3.16) -|B| = |I| + E\HF\ + E\c\ + E\WS\ - E\xc\^cb^ - E\se\^cb^ @ee - 39 - @ka _________________________________________________________ | | | | | | | | | Metal | s | p | d | | | &a | &b | &a | &b | &a | &b | | | | | | | | | _________________________________________________________ | | | | | | | | | Bi | .2548 | .3640 | .2623 | .3508 | .0563 |-.0872 | | | | | | | | | | Si | .2125 | .2022 | .1681 | .0211 | -- | -- | | | | | | | | | | Ge | .2296 | .2286 | .1837 | .0728 | .0927 |-.1462 | | | | | | | | | | Sn | .2377 | .2918 | .1915 | .1546 | .1141 | .0271 | | | | | | | | | | Pb | .2602 | .3130 | .1751 | .1373 | .0435 |-.2786 | | | | | | | | | | Al | .2243 | .1703 | .1961 | .0289 | -- | -- | | | | | | | | | | Ga | .2605 | .1992 | .2125 | .0313 | .1464 |-.0421 | | | | | | | | | | In | .2640 | .2499 | .1860 | .0544 | .0438 |-.4070 | | | | | | | | | | Tl | .3021 | .2833 | .2434 | .1588 | .0455 |-.3445 | | | | | | | | | | Mg | .2534 | .1318 | .2427 | .0241 | -- | -- | | | | | | | | | | Zn | .3544 | .1884 | .4832 | .1820 | .1904 | .0084 | | | | | | | | | | Cd | .3511 | .2288 | .3725 | .1560 | .1903 | .0076 | | | | | | | | | | Hg | .4185 | .2528 | .4584 | .1991 | .3507 | .1935 | | | | | | | | | | Li | .3707 | .0726 | -- | -- | -- | -- | | | | | | | | | _________________________________________________________ Table 3.4: Values of &a and &b. @ke @ba This equation is set up per electron (not per atom!) and the quantities have the following meaning: @be - 40 - @ka B = absolute value of cohesive energy per electron I = absolute value of (first Z ionization energies)/Z E\HF\ = Hartree-Fock energy of free-electron gas = (3/5) E\F\ + E\x\ (E\x\=exchange energy) E\c\ = correlation energy of free-electron gas E\WS\ = Wigner-Seitz bottom of band energy E\xc\^cb^ = exchange/correlation of conduction band E\se\^cb^ = self-energy of conduction band electrons @ke @ba The last two terms had to be subtracted, otherwise they would be counted twice in E\HF\ and E\WS\.The quantity E\F\^vac^ we want is the energy from the vacuum to the bottom of the conduction band(=E\WS\) plus the energy from the bottom to the top of the conduction band(=E\F\).Rewriting the above equation 16 we get: @be @ea (3.17) E\F\^vac^ = -|B| - |I| + (2/5)E\F\+ E\se\^cb^+ E\xc\^cb^- E\x\- E\c\ @ee @ba The term V\rest\ can be expressed as: V\rest\=V\pot\+E\xc\^cb^, where V\pot\ is the position dependent potential of an electron interacting with Z electrons in the cell with radius r\o\ and E\xc\^cb^ the exchange and correlation interaction with these same electrons.The potential V\pot\ can be obtained by integrating over the cell and then averaging over a sphere of radius R\l\: @be @ea Z 3Z R\l\ (3.18) V\pot\ = ---- [ 3 - (r/r\o\)^2^] =^~^ ---- [ 1 - (----)^2^] 2 r\o\ 2 r\o\ 2 r\o\ @ee - 41 - @ba We can now write a final expression for E\F\^ion^ which will be used as the energy to be used for extrapolating for the A\l\'s. This energy will, itself, depend on A\l\ through the term V\pot\ and we will have to solve self-consistently for each A\l\. @be @ea (3.19) E\F\^ion^ = -|B| - |I| + .2 k\F\^2^ + .6 Z/r\o\ + .239 k\F\ +.0575 - .0155 ln(1.9192/k\F\) - V\pot\ @ee @ba In this equation the explicit expressions for the different energies have been inserted, k\F\ is the usual Fermi-momentum and r\o\ the radius of the Wigner-Seitz sphere.All the energies are given in atomic units.It is to be noted that the term E\xc\^cb^ dropped out in subtracting E\F\^vac^ and V\rest\. We calculated the energies E\F\^ion^ self-consistently, for the elements in question, for each l. The average of the energies for s,p,d was then used to extrapolate the A\l\'s with the aid of equation 3.13.The different quantities of interest, for the pure elements, are given in Table 3.5. (The quantity E\F\^*^ is the constant part of E\F\^ion^, i.e. E\F\^ion^+ V\pot\) @be @ba The situation of an impurity in an Al or In matrix, in the context of a Shaw potential, has not been treated elsewhere. Our assumptions are, first of all, that the Fermi energy of the host matrix is unchanged for small concentrations of the impurity atoms, thus, the term E\F\^*^ will have to be taken for the host matrix. Secondly, the terms in V\rest\ will change in a way which will depend on the properties of the immersed impurity atom. We propose the following reasonable assumptions: @be - 42 - @ba a) the radius r\o\ of the Wigner-Seitz cell, used in the calculation of V\pot\, will be taken equal to the radius of the impurity including any changes of volume due to distortion,b) the the valence Z used in V\pot\ will be the impurity valence (the screening effectively neutralizes the impurity), c) the term E\xc\^cb^ will be given the same value as that of the host. @be @ka _________________________________________________________ | | | | | | | | | | * | ion | | | | | | Metal | E | -E | A | A | A | Z | | | F | F | 0 | 1 | 2 | | | | | | | | | | _________________________________________________________ | | | | | | | | | Bi | .053 | 1.727 | 2.41 | 2.57 | 1.78 | 5 | | | | | | | | | | Si | .219 | 1.526 | 2.11 | 2.56 | --- | 4 | | | | | | | | | | Ge | .169 | 1.507 | 2.14 | 2.47 | 2.09 | 4 | | | | | | | | | | Sn | .205 | 1.348 | 1.85 | 2.09 | 1.71 | 4 | | | | | | | | | | Pb | .140 | 1.343 | 1.92 | 1.99 | 1.39 | 4 | | | | | | | | | | Al | .352 | .988 | 1.39 | 1.66 | --- | 3 | | | | | | | | | | Ga | .247 | 1.051 | 1.53 | 1.79 | 1.44 | 3 | | | | | | | | | | In | .207 | .936 | 1.35 | 1.48 | .99 | 3 | | | | | | | | | | Tl | .142 | .968 | 1.45 | 1.54 | .88 | 3 | | | | | | | | | | Mg | .237 | .550 | .78 | .93 | --- | 2 | | | | | | | | | | Zn | .273 | .630 | 1.00 | 1.38 | .75 | 2 | | | | | | | | | | Cd | .206 | .610 | .91 | 1.11 | .75 | 2 | | | | | | | | | | Hg | .131 | .702 | 1.08 | 1.30 | 1.00 | 2 | | | | | | | | | | Li | .175 | .185 | .33 | --- | --- | 1 | | | | | | | | | _________________________________________________________ Table 3.5: Final parameters for optimized potential in pure elements. @ke - 43 - @ka e). Results: Shaw potential parameters for solutes in Al and In. _______________________________________________________________ @ke @ba The results for impurities in Al and In are given in Tables 3.6 and 3.7., where the values of A\l\ and R\l\ are the input parameters for the final calculation of the pseudopotential. @be @ba The values for the cohesive energy B were taken from a recent compilation which appears in Kittel (1971). The ionization energies I came from the tables of Moore (1949), the same source was used for the atomic energy levels needed for the determination of our constants &a and &b. @be @ka _________________________________________________ | | | | | | | | | ion | | | | Evans| | Solute| -E\F\ | A\0\ | A\1\ | A\2\ | R\2\ | | | | | | | | _________________________________________________ | | | | | | | | Si | 1.542 | 2.12 | 2.56 | --- | | | | | | | | | | Ge | 1.438 | 2.12 | 2.47 | 2.11 | | | | | | | | | | Ga | .979 | 1.51 | 1.78 | 1.45 | | | | | | | | | | Mg | .464 | .76 | .92 | --- | | | | | | | | | | Zn | .535 | .97 | 1.34 | .75 | 3.0 | | | | | | | | | Cd | .550 | .89 | 1.09 | .75 | 3.0 | | | | | | | | _________________________________________________ Table 3.6: Final parameters for Al solutes. @ke - 44 - @ka _________________________________________________ | | | | | | | | | ion | | | | Evans| | Solute| -E\F\ | A\0\ | A\1\ | A\2\ | R\2\ | | | | | | | | _________________________________________________ | | | | | | | | Bi | 1.606 | 2.34 | 2.50 | 1.79 | | | | | | | | | | Sn | 1.347 | 1.85 | 2.09 | 1.71 | | | | | | | | | | Pb | 1.272 | 1.88 | 1.98 | 1.41 | | | | | | | | | | Ga | 1.065 | 1.53 | 1.79 | 1.44 | | | | | | | | | | Tl | .912 | 1.42 | 1.52 | .90 | | | | | | | | | | Mg | .577 | .78 | .93 | --- | | | | | | | | | | Zn | .661 | 1.01 | 1.39 | .75 | 3.0 | | | | | | | | | Cd | .589 | .91 | 1.10 | .75 | 3.0 | | | | | | | | | Hg | .623 | 1.05 | 1.27 | .98 | 3.4 | | | | | | | | | Li | .147 | .32 | --- | --- | | | | | | | | | _________________________________________________ Table 3.7: Final parameters for In solutes. @ke @ba The values of R\2\ indicated for some elements in the last two tables are the values given by Evans(1970) for those elements he was not able to optimize. As he indicated, this failure was due to the inclusion of the d-core states in the quantum-defect method. All other R\l\'s can be obtained by the relation: R\l\=Z/A\l\. @be - 45 - @ka f). Calculation of Shaw potential. __________________________________ @ke @ba From these values of the A\l\'s we have calculated the pseudopotentials, for our selection of impurities, in Al and In. Shaw (1968,1968B) shows that his optimized pseudopotential can be written in the following form: @be @ea l\o\ Z Z (3.20) V\o\ = - - - $S &J(R\l\ - r) [A\l\ - -] Pr\l\ r r l=0 @ee @ba where &J is a step function and Pr\l\ an operator which projects out the l'th component of the wave function. The fact that we are only interested in potentials representing the scattering effects of impurities allows us to neglect the refinements of non-local screening etc., which are necessary for calculations involving either the total energy or the effective ion-ion interaction. We thus use essentially the screening function as indicated in equations (3.11) and (3.12). The unscreened potential in function of x=q/(2k\F\) can be split up in two parts, the first is a local contribution B(q), the second a non-local term f(k\F\,k\F\+q); the sum of these is then locally screened. @be @ea 4&pZg 4&p &m\c\R\c\^3^ Zh&a\eff\ (3.21) B(q) = - ---- - -- [ ----- + ------ ] [j\o\(qR\c\)+j\2\(qR\c\)] &O\o\q^2^ &O\o\ 3 q^2^ @ee - 46 - @ba the expression for the non-local part is: @be @ea l\o\ 4&p (3.22) f(k\F\,k\F\+q) = - -- $S (2l+1)P\l\(cos&j)A\l\R\l\^3^ I(R\l\,k,q) &O\o\ l=0 @ee @ba with @be @ea 1 (3.23) I (R\l\,k,q) = $I x (x-1) j\l\((k\F\+q)R\l\x) j\l\(k\F\R\l\x) dx 0 @ee @ba In the equation for B(q) the first term comes from the pure Coulomb part of the potential, the second term contains contributions from the correlation and orthogonalization with the conduction band of the host material. Zg and Zh are the valences of the guest and host material, &m\c\ a quantity analogous to the term E\c\ appearing in eqs. 3.16-19, R\c\ is the ionic radius, &a\eff\ the value defined by Animalu; Heine (1965) for the orthogonalization hole (note that we used the R\c\ of the guest atom, but Zh for the valence), j\l\ and P\l\ the Bessel and Legendre functions of order l.The expression for the non-local part is obtained by expanding the plane waves in spherical coordinates and then evaluating the matrix elements. @be @ba The screening is done with the following expression: @be @ea (3.24) &e^~^ (x) = [ &e(x) - 1 ] [ 1 - f(x) ] + 1 f (x) = x^2^/ [ 2x^2^+ .5 + 1/&pk\F\] @ee - 47 - @ba where &e(x) is from eq. 3.11 and f(x) is a correction for exchange and correlation due to Sham(1965). @be @ba We hope that our new evaluation of scattering potentials will give a more reliable estimate of resistivities, Dingle temperatures, etc. We will use these results further on when discussing the results of Hall effect measurements and Wejgaard(1975) has also used them for comparison to our de Haas-van Alphen experiments. @be @ka 3) Scattering. ______________ a). Transition probabilities. _____________________________ @ke @ba As indicated in the preceding paragraph, the potential felt by an electron in a metal can be represented by a relativily weak pseudopotential. Let v(r) be the PP of a single ion, then the potential energy of an electron at the position r is given by: @be @ea (3.25) V(r) = $S v ( | r - r\j\ | ) j @ee @ba where the r\j\'s are the positions of the individual ions. We do not require any particular arrangement of the ions for the moment. If the self-consistent screening of the electrons is already included in the v(r)'s, then the V(r) will also be a screened potential. @be @ba To obtain the wave-function representing this electron one has to solve the time dependent Schroedinger equation: @be - 48 - @ea (3.26) ( T + V(r) ) &F = ih' $d&F / $dt @ee @ba For the scattering problem one asks for the evolution of an electron in state |k> into a state |k+q>. Expanding &F in plane waves: @be @ea (3.27) &F(r,t) = $S a\q\(r) |k+q> exp( i&o\k+q\t ) @ee @ba where &o\k+q\ is the frequency corresponding to the energy of the state |k+q>. Setting a\0\ =1, a\q\=0 for q =/ 0 at the time 0, and substituting in eq. 3.26, on obtains in first order perturbation: @be @ea . (3.28) i h' a\q\ exp( -i&o\k+q\t ) = <k+q|V|k> exp( -i&o\k\t ) @ee @ba where the matrix elements are defined as: @be @ea (3.29) <k+q|V|k> = 1/&O $I exp(-i(k+q)r) V(r) exp(ikr) d&t &O @ee @ba Integrating this with respect to time and using a\q\(0)=0, one obtains the probability that an electron will be found in the state |k+q> at the time t: @be @ea |<k+q|V|k>|^2^ sin^2^[(&o\k\-&o\k+q\) t/2] (3.30) a\q\^*^a\q\ = 4 ------------ ------------------- h'^2^ (&o\k\-&o\k+q\)^2^ @ee @ba Recognizing that the factor sin^2^(&at)/&pt&a^2^ behaves like a &d function for long times, we obtain the usual Golden Rule no:1 for the transition probability: @be - 49 - @ea (3.31) P\k,k+q\ = (2&p/h') * |<k+q|V|k>|^2^ * n(E) @ee @ba where n(E) is the density of final states. It is to be noted that this derivation is a first order perturbation result and the transitions are supposed to be due to a quasi-periodic perturbation. This will not be exactly the case for isolated impurities which are randomly distributed in an otherwise regular host crystal. @be @ba Now we have to find out more about V(r): consider a binary solid solution containing host ions h and guest ions g. Then (see eq.3.25) the potential W of the dilute alloy will be: @be @ea (3.32) W(r) = $S v^h^(r-r\h\) + $S v^g^(r-r\g\) h g @ee @ba for the matrix elements we get using the definition (3.29): @be @ea (3.33) <k+q|W|k> = 1/N $S exp(-iqr\h\)<k+q|v^h^|k> h + 1/N $S exp(-iqr\g\)<k+q|v^g^|k> g @ee @ba If we assume that the impurity atoms are substitutional, i.e. if they are located on host atom sites, we can write: @be @ea (3.34) <k+q|W|k> = 1/N $S exp(-iqr\h,g\)<k+q|v^h^|k> h,g + 1/N $S exp(-iqr\g\)<k+q|v^g^-v^h^|k> g @ee - 50 - @ea (3.35) <k+q|W|k> = S(q)<k+q|v^h^|k> + S'(q)<k+q|v^g^-v^h^|k> @ee @ba The S(q) is the structure factor of the average lattice, which, in the case of substitutional impurities, is equal to the pure structure factor. This structure factor is non-zero only for reciprocal lattice vector values q\o\ of q.This is an infinitesimal set of all the q's playing a role in scattering processes and all it does is determine the shape of the Fermi- surface and the character of the wave functions on this same surface. When q=/q\o\, S(q) is zero and we are left with: @be @ea (3.36) <k+q|W|k> = S'\i\(q) <k+q|v^g^|k> + S'\v\(q) <k+q|v^h^|k> @ee @ba Here S'(q) is the structure factor of the randomly distributed guest atoms. It is easily seen that S'\v\(q)=-exp(-iqr\o\)/N for a vacancy and S'\i\(q)=exp(-iqr\o\)/N for an interstitial at the position r\o\. @be @ka b). Lattice distortions. ________________________ @ke @ba There is one complication in giving an exact expression for S'\v\(q): an impurity introduced in the host lattice induces a local distortion of its surrounding. This effect is difficult to take into account, but several authors have given a treatment of this problem. Eshelby(1954,1956) has shown that in the limit of continuum elasticity the distortion can be written as @be @ea _ (3.37) u(r) = A r / r^3^ @ee - 51 - @ba and that the constant A can be obtained from X-ray data in the following way: @be @ea &Da 4 (3.38) -- = - &p&hA&g a 3 @ee @ba where &Da/a is the change in X-ray lattice constant, &h the concentration of defects per unit volume and &g=3(1-&s)/(1+&s), where &s is the Poisson ratio for the host material. (&s=.353 for Al, &s=.425 for In) Hence, the relative change in volume around an impurity is: @be @ea (1+&s) 1 da 4&pA (3.39) ----- - -- = --- (1-&s) a dC &O\o\ @ee @ba where c is the relative atomic concentration and &O\o\ the host atomic volume. Now S'\v\(q) becomes: @be @ea (3.40) S'\v\(q) = 1/N $S exp(-iq(r\i\+u\i\) @ee @ea __ 4&pA &O\o\qr\i\ (3.41) = 1/N $S exp(-iqr\i\) - --- $S ----- exp(-iqr\i\) &O\o\ 4&pr\i\^3^ i @ee @ba The first term evaluated for a vacancy, gives exp(-iqr\o\)/N and the second sum can be shown to give (Harrison,1966): @be @ea q&O\o\ j\1\(q|r\i\-r\o\|) (3.42) g(q) exp(-iqr\o\)/N , g(q) = --- $S ------------ 4&p |r\i\-r\o\|^2^ @ee - 52 - @ba Hence we are left with: @be @ea 4&pA (3.43) S'\v\(q) = - (1 + --- g(q) ) exp(-iqr\o\)/N &O\o\ @ee @ba The term in brackets represents the effective volume of the vacancy in the limit of long wavelengths. This change in volume has been used by Blatt(1957) to define an effective valence Z (1+&DV/&O\o\) for the host. The function g(q) was evaluated and tabulated by Shaw(1968b). @be @ba The assumption that the distortion can be represented by an elastic law like in eq.3.37, has been questioned by several authors: Benedek; Baratoff(1971), Flocken; Hardy(1970), Beal- Monod; Kohn(1968). They calculate the ion displacement from the force constants between the ions and, thus, obtain anisotropic displacements. Although the treatment of anisotropic scattering centers seems rather difficult to take into account, it would be interesting to include these nearest neighbour displacements in defining an effective scattering potential of the distortion cluster. We will see in the next sections that the homovalent impurities give rather bad results, when compared to the experiments. It is just in these cases that the influence of distortions could be very important. @be - 53 - @ka c). Phase shifts.(Results) __________________________ @ke @ba The final form of the scattering potential W of an impurity in a host lattice is as follows: (see eqs. 3.36 and 3.43) @be @ea (3.44) <k+q|W|k> = 1/N exp(-iqr\i\) <k+q|w|k> 4&pA <k+q|w|k> = <k+q|v^g^|k> - [1 + --- g(q) ] <k+q|v^h^|k> &O\o\ @ee @ba Some typical results of our computed <k+q|w|k> are shown in Fig.3.3. For later discussions it will be convenient to parametrize these potentials with the phase shifts obtained in the Born approximation and calculated with the help of eq.3.45 (see Harrison(1970) p.186) @be @ea 3&pZh (3.45) &d\l\^BA^ = - ---- $I P\l\(cos &j) <k+q|w|k> d(cos &j) 8E\F\ @ee @ba In Tables 3.8 and 3.9 we give the resulting phase shifts for Al and In (including the pure metal and the vacancy impurity) along with the effective valence Z^*^ of the impurity obtained from the Friedel sum rule: @be @ea 4 * 2 (3.46) Z = - $S (2l + 1) &d\l\ &p l=0 @ee - 54 - @fa FIGURE 3.3 Figure 3.3:Selected scattering potentials <k+q|w|k> for impurities in Al and In. @fe - 55 - @ka _________________________________________________________ | | | | | | | | | | | | | | | * | | | &d | &d | &d | &d | &d | Z | | | 0 | 1 | 2 | 3 | 4 | | | | | | | | | | _________________________________________________________ | | | | | | | | | Al | .773 | .769 | .241 | .052 | .008 | 3.01 | | | | | | | | | | AlVac | .042 | .108 | .075 | .062 | .032 | 0.93 | | | | | | | | | | AlSi | .629 | .266 | .063 | .027 | .011 | 1.29 | | | | | | | | | | AlGe | .618 | .197 | .011 | -.004 | -.007 | 0.75 | | | | | | | | | | AlGa | .177 | -.027 | -.049 | -.014 | -.007 | -0.19 | | | | | | | | | | AlMg | -.515 | -.296 | -.076 | -.037 | -.016 | -1.39 | | | | | | | | | | AlZn | -.145 | -.163 | -.157 | .007 | .004 | -0.85 | | | | | | | | | | AlCd | -.196 | -.103 | -.090 | .047 | .022 | -0.27 | | | | | | | | | _________________________________________________________ @ke Table 3.8: Phase shifts and Z^*^ for Al. @ka _________________________________________________________ | | | | | | | | | | | | | | | * | | | &d | &d | &d | &d | &d | Z | | | 0 | 1 | 2 | 3 | 4 | | | | | | | | | | _________________________________________________________ | | | | | | | | | In | 1.116 | .788 | .158 | .055 | .006 | 3.00 | | | | | | | | | | InBi | .819 | .383 | .054 | -.032 | -.014 | 1.20 | | | | | | | | | | InSn | .395 | .220 | .082 | -.010 | -.004 | 0.87 | | | | | | | | | | InPb | .465 | .158 | .018 | -.027 | -.010 | 0.48 | | | | | | | | | | InGa | .247 | .029 | .070 | .024 | .010 | 0.60 | | | | | | | | | | InTl | .123 | -.016 | -.059 | -.007 | -.003 | -0.19 | | | | | | | | | | InMg | -.606 | -.239 | .006 | -.024 | -.003 | -0.95 | | | | | | | | | | InZn | -.179 | -.123 | .017 | .057 | .021 | 0.08 | | | | | | | | | | InCd | -.325 | -.187 | -.031 | .004 | .004 | -0.62 | | | | | | | | | | InHg | -.097 | -.173 | -.114 | .022 | .007 | -0.62 | | | | | | | | | _________________________________________________________ Table 3.9: Phase shifts and Z^*^ for In. @ke