1: \documentstyle[12pt,aps,epsfig,tabularx]{revtex}
2: \begin{document}
3: \title{Formation of magnetic characteristics \\
4: and hyperfine fields in metal-metalloid alloys}
5: \author{A. K. Arzhnikov and L. V. Dobysheva}
6: \address{Physical-Technical Institute, Ural Branch of Russian Academy
7: of Sciences, Izhevsk, Russia}
8: \maketitle
9:
10: \begin{abstract}
11: This work deals with the analysis of peculiarities of formation of the
12: hyperfine fields (HFF) at the Fe nuclei in disordered alloys metal-
13: metalloid using the "first-principles" calculations. Some
14: phenomenological models and justification of their usage for the
15: interpretation of the experimental HFF distributions are discussed.
16: \end{abstract}
17:
18: \pacs{75.50.Bb, 71.15.Ap}
19:
20: Keywords: hyperfine field, magnetic moment, disordered alloy
21:
22: \vskip .5cm
23: 1. Introduction
24:
25: The development of technologies for preparing materials for modern
26: electronics requires a detailed knowledge of different local physical
27: quantities at the atomic level. In the experimental physics the
28: possibility to measure the local characteristics at such a scale are
29: very limited. Among few existing methods, of special accuracy are the
30: nuclear ones (for example, the Mossbauer spectroscopy) that measure the
31: hyperfine magnetic fields (HFF) at nuclei. It is widely believed that
32: these techniques can reflect some chemical and topological features of
33: the atomic surroundings of the excited nuclei. Indeed, in numerous
34: studies the spectra are interpreted in terms of phenomenological
35: models considering only the nearest atomic environment (see e.g. [1-3]).
36: There is no doubt that such a description is often useful and efficient.
37: We think, however, that there are some cases when such a simplified
38: approach is not justified. Moreover, for the spectra of transition
39: metals and alloys on their basis, the successful description within the
40: framework of such models is rather an exception than a rule, and every
41: time additional arguments are needed to justify these limitations.
42:
43: In this work we analyze the peculiarities of formation of HFF in
44: disordered alloys of Fe and metalloid using the "first-principles" \
45: calculations. One may consider an HFF as a sum of three main terms:
46: $H^{core}$, $H^{val}$ and $H^{orb}$ which are the core, valence and
47: orbital contributions, respectively. This standard division is based on
48: the physical features that determine the magnitude of these
49: contributions, $H^{core}$ and $H^{orb}$ being mostly dependent on the
50: topological and chemical peculiarities of the nearest neighboring only,
51: whereas $H^{val}$ is significantly influenced by the atoms of the
52: second and more distant spheres. At the same time, in disordered alloys
53: the effect of the distant spheres on $H^{val}$ may be considerably
54: suppressed due to the great localization of the valence electrons.
55: Owing to this, in some cases [1-3] the HFF distributions can be
56: successfully described by phenomenological models of the
57: Jaccarino-Walker type where the magnitudes of the local magnetic
58: moments (LMM) and HFF are taken to be dependent on the number of
59: metalloid atoms in the nearest neighboring. Often such descriptions are
60: surprisingly accurate and efficient. One must bear in mind, however,
61: that the phenomenological models have fitting parameters, and there are
62: various examples in the history of sciences when the accuracy and
63: efficiency were illusory, being only a result of manipulation with the
64: number and magnitudes of the parameters.
65:
66: To avoid such a trap, one must realize how to remove the contradiction
67: between the localized character of the description of HFF and the
68: itinerant behavior of the d- and sp- electrons in metals. At present,
69: the solid state theory based on the "first-principles" calculations
70: allows one not only to justify the use of some phenomenological models,
71: but also to make them more reliable by defining the parameters from the
72: "first-principles" calculations.
73:
74: In this paper, we analyze the behavior of $H^{core}$, $H^{val}$ and
75: $H^{orb}$ using the "first-principles" \ calculations performed for the
76: ordered alloys $Fe_{100-c}Me_c$ ($Me=Si, Sn$), namely: $Fe_{31}Si$,
77: $Fe_{15}Si$, $Fe_{31}Sn$, $Fe_{15}Sn$, $Fe_{26}Si_{6}$, $Fe_{25}Si_{7}$,
78: $Fe_{24}Si_{8}$, $Fe_{23}Si_{9}$ and $Fe_{22}Si_{10}$, and experimental
79: data [4]. The calculations were carried out by the full-potential
80: linearized augmented plane wave method (FLAPW) using the WIEN-97 program
81: package [5]. Some results are presented in Table 1.
82:
83: The systems were simulated on a bcc lattice which in the disordered
84: alloys under consideration is retained within a wide concentration range
85: [1]. The lattice parameters were chosen in accordance with the
86: experimental values of the alloys with corresponding concentrations
87: ($c=3.125, 6.25, 18.75, 21.875, 25, 28.125$ and $31.25 at.\%$). It
88: should be mentioned that even at small concentrations the bcc lattice is
89: somewhat distorted due to the repulsion/attraction by the Sn/Si atom
90: magnetic of the surrounding Fe atoms. As shown in our paper [6], the
91: changes in characteristics because of this relaxation are insignificant.
92: Though the results in Table 1 were obtained with allowance for this
93: relaxation, we do not discuss it here.
94:
95: \vskip .5cm
96: 2. The local magnetic moment
97:
98: Hereinafter, the local magnetic moment (LMM) at Fe atom $i$ means the
99: spin magnetic moment over the muffin-tin (MT) sphere, $M_i^d$. The
100: average magnetic moment over the unit cell is $M=\sum_i
101: (M_i^d+M_i^{orb}) + M^{int}$, where $M_i^{orb}$ is the orbital magnetic
102: moment, $M^{int}$ is the spin magnetic moment outside the MT spheres.
103: The main contribution to the LMM comes from the d-electrons that are
104: almost entirely inside the MT-sphere, whereas $M_{int}$ is formed by the
105: s- and p- electrons and has a small negative value as compared to the
106: LMM. We neglect $M^{orb}$, as its variations with concentration and the
107: metalloid atoms arrangement are small and amount to less than 1 \% [7].
108: The calculated values are two times less than the experimental ones. We
109: believe that the actual changes of $M^{orb}$ do not exceed twice the
110: calculated ones. It should be noted that the neglect of $H^{orb}$
111: requires some caution: in what follows we shall show that the changes
112: in $H^{orb}$ may be quite significant amounting to 10 $\div$ 15 \% of
113: the average HFF. Mention the principal features of the LMM formation.
114:
115: A. The main peculiarities of the magnetic moment depend weakly on the
116: metalloid type and, as shown in [8], is governed by the lattice
117: parameter, a. At equal concentrations, the lattice parameter of the Sn
118: alloy is greater than that of the Si alloy, so the magnetic moment in
119: the first case is greater.
120:
121: B. The LMM of the Fe atom closest to the impurity is of the smallest among
122: other values. As noted in [8], this difference is defined by
123: the competition between two mechanisms: the LMM reduction due to flattening
124: of the d-band because of the s-d hybridization that is the strongest near
125: the impurity, and the LMM increase due to narrowing of the d-band because of
126: a decrease of the wave function overlap. There also exists third mechanism
127: of the LMM reduction due to the difference in the impurity-potential
128: screening by d-electrons (the difference in pushing out the impurity levels
129: by the bands with spin up and down) [7].
130:
131: C. The LMMs are concentration dependent. So the LMM of the Fe atom
132: closest to the Si atom is 2.181 $\mu_B$ at c=3.125 at.\% ($Fe_{31}Si$), and
133: 2.262 $\mu_B$ at c=6.25 at.\% ($Fe_{15}Si$) (Table 1).
134:
135: At the same time, our "first-principles" calculations and the
136: mathematical treatment [9-10] testify that, to an accuracy of $10 \div
137: 15 \%$, the LMMs in disordered alloys are independent of concentration
138: and governed by the number of nearest metalloid atoms only.
139:
140: D. It is natural to expect that, with a large number of nearest metalloid
141: atoms the LMM will considerably decrease down to zero, due strong s-d
142: hybridization [10].
143:
144: \vskip .5cm
145: 3. Hyperfine magnetic fields at nuclei
146:
147: The program package WIEN-97 allows one to calculate the interaction
148: between the nucleus magnetic moment and the spin and orbital magnetic
149: moments of the electron subsystem. The spin dipole contributions are
150: small (~2 $\div$ 3 kGs), they are suppressed due to the symmetry
151: relations, and the main contribution comes from the spin polarization at
152: the nucleus (Fermi-contact interaction), $H^{core}+H^{val}$, and the
153: orbital magnetic moment. For the core-electron polarization the simple
154: relation $H^{core}_i =\gamma^s Md_i$ is satisfied, where $\gamma^s$ does
155: not depend to a high accuracy on neither the metalloid type nor its
156: concentration, and is determined only by the approximation for the
157: exchange-correlation potential. Here we use the GGA approximation [11]
158: which gives $\gamma^s\approx 123 kGs/\mu_B$. Due to the simple relation
159: between the LMM and $H^{core}$ and owing to the possibility of the
160: rough description of the LMM as a function of neighboring metalloids
161: number discussed in Section 2 point C, if
162: $H^{val}$ is not too large one can use the Jaccarino-Walker-type
163: phenomenological models while interpreting the HFF distributions of
164: disordered metal- metalloid alloys, but the accuracy in this case
165: cannot be high.
166:
167: The proportionality $H^{orb}_i =\gamma^{orb} M_i^{orb}$ is fulfilled in
168: a somewhat worse way, but still rather satisfactorily. Note, however,
169: that $\gamma^{orb}$ is positive and about five times larger in magnitude
170: than $\gamma^s$. Hence, if the changes of $M_i^{orb}$ affect the
171: magnetic moment only slightly, the changes in $H_i^{orb}$ with
172: allowance for the actual magnitude of $M^{orb}$ may amount to 20 kGs
173: even at low concentrations (see Table 1).
174:
175: $H_i^{val}$ behaves in a more complicated way. This is primarily
176: associated with strong delocalization of the s- and p-like electrons
177: that interfere at sites with different magnetic and charge properties,
178: and therefore the $H_i^{val}$ behavior cannot be in fact quantitatively
179: predicted in disordered systems. However, we succeeded in revealing some
180: qualitative regularities supported by experimental evidence. We analyzed
181: this contribution using the simple functional relation
182: $$
183: H_i^{val}=F_0 M_i+ \sum_j F_j N_j,\eqno(1)
184: $$
185: where $F_0$ is a constant describing the
186: polarization of the valence electrons of the Fe atom at site $i$ by its
187: own magnetic moment, $F_j$ are factors describing the change of
188: polarization at atom $i$ by the metalloid atom in sphere $j$, $\sum_j$
189: is summation over all coordination spheres ($j>0$), $N_j$ is the number of
190: metalloid atoms in sphere $j$. Fig.1 shows the values of $F_j$ for the
191: calculations of different sets of alloys. The first curve was obtained
192: by solving the equation system Eqn. (1) by the least-squares method with
193: calculated values $H_i^{val}$ and $M_j$ for the alloys $Fe_{26}Si_{6}$,
194: $Fe_{25}Si_{7}$, $Fe_{24}Si_{8}$, $Fe_{23}Si_{9}$ and $Fe_{22}Si_{10}$,
195: the second curve was calculated for the data for the alloys $Fe_{31}Si$
196: and $Fe_{15}Si$, the third curve was received for $Fe_{31}Sn$ and
197: $Fe_{15}Sn$ alloys. One can see that $F_j$ is of oscillating nature and
198: depends on the metalloid type and concentration. This dependence reveals
199: itself mostly at close distances, which leads to an amazing
200: conclusion: $H^{val}$ may be considered, to some extent, as a linear
201: superposition of the changes in spin polarization by distant atoms,
202: whereas the polarization by close atoms can hardly be regarded as
203: linear because of a significant re-distribution of the charge density
204: which is not linear.
205:
206: \begin{figure}[t]
207: \epsfxsize=15cm
208: \centerline{\epsfbox{fig1.eps}}
209: \caption{The parameters of the model (1), $F_j$
210: ($kGs$), as a function of sphere number}
211:
212: \end{figure}
213:
214: Thus, it is clear why the use of the simplest RKKY (Ruderman- Kasuya-
215: Yosida- Kittel) interaction for the description of HFF resulted
216: sometimes in reasonable conclusions for superlattice. First, for the
217: realistic sp-electron densities, this interaction has an oscillation
218: period close to our result ($2 \div 3$ interatomic distances, Fig.1).
219: Second, the spin polarization by the distant metalloid atoms may be
220: considered additive. In the superlattices (e.g. in the $DO_3$ type
221: ordered alloy $Fe_{75\pm c}Si_{25\mp c}$), the possibility of a
222: configuration of impurity (here, by an impurity is meant an atom of $Fe$
223: or $Si$ which appears at a "wrong" \ site as compared to the totally
224: ordered $Fe_3Si$ alloy) atoms arrangement around the $Fe$ atom with
225: large number of the impurity atoms in the second and third spheres may
226: be high even at low concentration of the impurity atoms. In the light
227: of the above, the models where the spin polarization is considered as
228: an additive contribution from the RKKY polarization by the first
229: sphere, are doubtful.
230:
231: So far, our statements were mainly based on the "first-principles"
232: calculations of the ordered alloys, while the disordered one were
233: discussed. It should be mentioned that the problem of quantitative
234: calculations of HFFs in disordered alloys remains unsolved, that is why
235: we can examine the effect of disorder only qualitatively, using the
236: model systems.
237:
238: In our paper [8], the effect of disorder was investigated by
239: averaging the RKKY interaction whose parameters were determined from the
240: "first-principles" calculations. Now we realize that such an approach
241: should be used with caution. However, the result of this paper [8]
242: that the decrease of the HFF magnitude should be larger than the
243: decrease in the magnetic moment with disordering (i.e. with the
244: increase in metalloid concentration), holds true if the behavior of the
245: polarization by the third and most distant spheres being averaged in
246: disordered alloys, is similar to that of Fig.1 (the positive
247: polarization by the third sphere).
248:
249: Note that the decrease in $H_i / M_i$ is also determined by the
250: increase in the positive contribution of $H^{orb}$ with the increase of
251: the disorder degree [12]. It is difficult to separate these
252: contributions and to answer unambiguously why the change of this ratio
253: is experimentally observed.
254:
255: \vskip .5cm
256: Acknowledgements
257:
258: One of the authors (A.K.A.) is thankful to the Organizing Committee of
259: the conference EMRS-2001 and Prof. H.Dreysse for the financial support.
260: The present work was supported by Russian Foundation for Fundamental
261: Researches, Grant No 00-02-17355.
262:
263: \vskip .5cm
264: References
265:
266: [1] E.P. Yelsukov, Phys. Met. Metallogr. 76 (1993) 451.
267:
268: [2] N.N. Delyagin et al, Journal Eksp. i Teoret. Phys. 116 (1999) N 1(7),
269: 130.
270:
271: [3] B. Rodmacq et al, Phys. Rev. B 21 (1980) 1911.
272:
273: [4] E.P. Yelsukov, Phys. Met. Metallogr. 85 (1998) 307.
274:
275: [5] P. Blaha, K. Schwarz, J. Luitz. WIEN97, A Full Potential Linearized
276: Augmented Plane Wave Package for Calculating Crystal Properties (Karlheinz
277: Schwarz, Techn. Universit"at Wien, Austria). (1999) ISBN 3- 9501031-0-4.
278:
279: [6] A.K. Arzhnikov, L.V. Dobysheva, Phys. Rev. B 62 (2000) 5324.
280:
281: [7] A.K. Arzhnikov, L.V. Dobysheva, ArXiv: cond-mat /0101033 (2000).
282:
283: [8] A.K. Arzhnikov, L.V. Dobysheva, F. Brouers, Phys. Sol. St. 42
284: (2000) 89.
285:
286: [9] A.K. Arzhnikov, L.V. Dobysheva, E.P. Elsukov, A.V. Zagainov, JETP 83
287: (1996) 623.
288:
289: [10] A.K. Arzhnikov, L.V. Dobysheva, JMMM 117 (1992) 87.
290:
291: [11] J.P.Perdew et al, Phys. Rev. Lett. 77 (1996) 3865.
292:
293: [12] A.K. Arzhnikov, L.V. Dobysheva, Izvestiya Akademii Nauk.
294: Ser.Fiz. 65 (2001) 985. (in Russian).
295:
296: \newpage
297: \begin{table}
298: \caption{The results of the calculations. Configuration of impurities
299: in the Fe-atom environment, [nm...], denotes the number of metalloid atoms
300: in the first (n), second (m) etc. spheres. $N_{Fe}$ is the number of such
301: Fe atoms in the unit cell. }
302: \vskip .4cm
303: \begin{tabular}{ccccccc}
304: \hline
305: &$[nm...]$, $N_{Fe}$ & $M^d$, $\mu_B$ & $H^{core}$,kGs
306: & $H^{val}$,kGs & $M^{orb}$, $\mu_B$ & $H^{orb}$,kGs \\
307: \hline
308: $Fe_{31}Sn$, & [1000], 8& 2.246 & -277.50 & -37.59 & .0489 & 27.94\\
309: a=21.804a.u. & [0003], 8& 2.507 & -310.35 & -30.94 & .0483 & 26.41\\
310: & [0100], 6& 2.432 & -300.17 & -34.84 & .0482 & 26.62\\
311: & [0020], 6& 2.425 & -299.79 & -22.39 & .0479 & 26.28\\
312: & [0000], 2& 2.391 & -295.13 & -34.34 & .0441 & 23.66\\
313: & [0000], 1& 2.423 & -299.68 & -24.17 & .0470 & 25.15\\
314: \hline
315: $Fe_{15}Sn$, & [1003], 8& 2.442 & -301.01 & -20.25 & .0548 & 29.77\\
316: 2a=21.985a.u.& [0200], 3& 2.530 & -312.64 & -29.97 & .0535 & 29.08\\
317: & [0040], 3& 2.542 & -314.55 & -2.46 & .0550 & 29.38\\
318: & [0000], 1& 2.390 & -294.33 & -29.97 & .0450 & 23.32\\
319: \hline
320: $Fe_{31}Si$, & [1000], 8& 2.181 & -269.44 & -38.68 & .0497 & 28.27\\
321: a=21.604 a.u.& [0003], 8& 2.413 & -297.89 & -35.15 & .0471 & 25.55\\
322: & [0100], 6& 2.343 & -288.51 & -31.31 & .0446 & 24.46\\
323: & [0020], 6& 2.385 & -294.06 & -16.92 & .0468 & 25.32\\
324: & [0000], 2& 2.489 & -308.13 & -19.72 & .0485 & 26.44\\
325: & [0000], 1& 2.389 & -294.81 & -13.56 & .0451 & 24.28\\
326: \hline
327: $Fe_{15}Si$, & [1003], 8& 2.262 & -277.20 & -29.11 & .0594 & 32.38\\
328: 2a=21.585a.u.& [0200], 3& 2.353 & -289.52 & -21.64 & .0456 & 24.44\\
329: & [0040], 3& 2.432 & -298.87 & 5.49 & .0509 & 27.31\\
330: & [0000], 1& 2.536 & -313.82 & -13.89 & .0496 & 26.83\\
331: \end{tabular}
332: \end{table}
333:
334: \end{document}
335: