1: \documentclass[12pt,letter]{iopart}
2:
3: \usepackage{graphicx,amssymb}
4:
5: \begin{document}
6:
7: \title{Covalent bonding and the nature
8: of band gaps in some half-Heusler compounds}
9:
10: \author{Hem Chandra Kandpal\dag, Claudia Felser\dag,\\ and Ram Seshadri\ddag}
11:
12: \address{\ddag Institut f\"ur Anorganische Chemie und Analytische Chemie\\
13: Johannes Gutenberg-Universit\"at, Staudinger Weg 9, 55099 Mainz\\
14: felser@uni-mainz.de}
15:
16: \address{\ddag Materials Department and Materials Research Laboratory\\
17: University of California, Santa Barbara CA 93106\\
18: seshadri@mrl.ucsb.edu}
19:
20: \begin{abstract}
21:
22: Half-Heusler compounds \textit{XYZ}, also called semi-Heusler compounds,
23: crystallize in the MgAgAs structure, in the space group $F\bar43m$. We report
24: a systematic examination of band gaps and the nature (covalent or ionic)
25: of bonding in semiconducting 8- and 18- electron half-Heusler compounds
26: through first-principles density functional calculations. We find
27: the most appropriate description of these compounds from the viewpoint of
28: electronic structures is one of a \textit{YZ} zinc blende lattice stuffed
29: by the \textit{X} ion. Simple valence rules are obeyed for bonding
30: in the 8-electron compound. For example, LiMgN can be written
31: Li$^+$ + (MgN)$^-$, and (MgN)$^-$, which is isoelectronic with (SiSi),
32: forms a zinc blende lattice. The 18-electron compounds can similarly be
33: considered as obeying valence rules. A semiconductor
34: such as TiCoSb can be written Ti$^{4+}$ + (CoSb)$^{4-}$; the latter unit
35: is isoelectronic and isostructural with zinc-blende GaSb. For both the
36: 8- and 18-electron compounds, when \textit{X} is fixed as some electropositive
37: cation, the computed band gap varies approximately as the difference in Pauling
38: electronegativities of \textit{Y} and \textit{Z}. What is particularly
39: exciting is that this simple idea of a covalently bonded \textit{YZ} lattice
40: can also be extended to the very important \textit{magnetic\/} half-Heusler
41: phases; we describe these as valence compounds \textit{ie.\/} possessing a
42: band gap at the Fermi energy albeit only in one spin direction. The
43: \textit{local\/} moment in these magnetic compounds resides on the
44: \textit{X} site.
45:
46: \end{abstract}
47:
48: \pacs{
49: 75.50.-y, % Studies of specific magnetic materials
50: 71.20.-b, %Electron density of states and band structure of crystalline solids
51: 75.50.Cc %Other ferromagnetic metals and alloys
52: }
53:
54: \maketitle
55:
56: \section{Introduction}
57:
58: The half-Heusler phases \textit{XYZ}, comprising three interpenetrating
59: \textit{fcc} lattices, constitute an important class of materials with
60: particular regard to their magnetic properties. de Groot\cite{deGroot}
61: and coworkers showed a number of years ago that the half-Heusler compound
62: NiMnSb can be described as a half-metallic ferromagnet, whose computed band
63: structure resembles a metal in one spin direction, and a
64: semiconductor in the other. Since then, and indeed, even prior to that, it has
65: been recognized that the electronic structure and hence properties of Heusler
66: compounds\cite{Kubler} and half-Heusler compounds\cite{Pierre1,Pierre2} are
67: very sensitive to the valence electron count.
68:
69: A number of electronic structural studies have been carried out on the
70: half-Heuslers. We focus here on those studies which systematically address
71: behavior in families of half-Heusler compounds, rather than studies focused
72: on individual ones. From the viewpoint of chemical bonding in these compounds,
73: Whangbo and coworkers\cite{Whangbo} have examined the non-magnetic band
74: structures, using the extended H\"uckel method, of a number of half-Heusler
75: compounds with varying valence electron counts. These authors have recognized
76: that many \textit{XYZ} half-Heuslers can be thought of as comprising an
77: \textit{X}$^{n+}$ ion stuffing a zinc blende \textit{YZ}$^{n-}$ sublattice
78: where the number of valence electrons associated with \textit{YZ}$^{n-}$ are
79: 18 ($d^{10}$ + $s^2$ + $p^6$). 18 electron compounds are therefore closed shell
80: species; non-magnetic and semiconducting. They further suggest that the
81: 17 and 19 electron \textit{XYZ} would undergo a Stoner
82: instability\cite{Stoner} to a ferromagnetic ground state, while the 22
83: electron compounds (typically with Mn$^{3+}$ at the \textit{X} site) should
84: be localized moment ferromagnets. The 22 electrons divide themselves
85: into 13 in the majority spin and 9 in the minority spin direction, resulting in
86: a semiconducting gap (half-metallic behavior) in the minority spin direction.
87: Recently, Galanakis \textit{et al.\/}\cite{Galanakis} have placed this ``18
88: electron'' rule on a more formal footing.
89:
90: Pierre \textit{et al.\/}\cite{Pierre1} were amongst the first to recognize the
91: importance of the valence electron count in the half-Heuslers. In more recent
92: work, Tobola and Pierre\cite{Pierre2} have emphasized the importance of
93: covalency in these compounds. The \textit{Z} element is often a pnictogen
94: (As, Sb or Bi) or some other main group element because only covalent bonding
95: would justify the somewhat open half-Heusler structure. \"O\u{g}\"ut and
96: Rabe\cite{Ogut} have examined the electronic structures of the compounds
97: \textit{X}NiSn with \textit{X} = Ti, Zr, or Hf, and interpreted phase
98: stability and the nature of band gaps. Also important from the approach that
99: we will take here is the work of Wood, Zunger, and de Groot\cite{Wood} who have
100: examined a number of non-magnetic ``stuffed zinc blende'' semiconductors,
101: the so-called Nowotny-Juza compounds\cite{Nowotny,Juza} including
102: half-Heusler phases such as LiZnP, in order to control the nature of the band
103: gap (direct or indirect). Nanda and Dasgupta\cite{Dasgupta} have examined
104: nearly 20 different half-Heusler compounds using the FP-LMTO and LMTO-ASA
105: methods, including a detailed analysis of the bonding and the nature
106: of the band gaps. They argue as we do here, for the very important role played
107: by covalent bonding in these systems. They ascribe half-metallicity to arise
108: in some of the half-Heuslers due to the large \textit{Y}-\textit{Z} covalency,
109: in conjunction with large exchange -splitting due to highly magnetic \textit{X}
110: ions.
111:
112: In this contribution, we focus on chemical bonding in 8, 18 and magnetic
113: half-Heusler compounds, and attempt to relate the electronic structure
114: to simple concepts such as the electronegativity of the component species.
115: We also demonstrate covalency and the local nature of the magnetic moment
116: in the magnetic compounds using real-space descriptors derived from first
117: principles theory.
118:
119: \section{Crystal structures and methods}
120:
121: \begin{figure}
122: \centering \includegraphics[width=6cm]{fig01.eps}
123: \caption{(Color) \textit{XYZ} Half-Heusler crystal structure in the
124: $F\bar43m$ space group. Cyan \textit{Z} atoms are at the origin, orange
125: \textit{X} at $(\frac12, \frac12, \frac12)$ and blue \textit{Y} at
126: $(\frac14, \frac14, \frac14)$. Note the tetrahedral zinc blende (diamondoid)
127: sublattice formed by \textit{Y} and \textit{Z}.}
128: \end{figure}
129:
130: Half-Heusler \textit{XYZ} compounds crystallize in the space group of zinc
131: blende ($F\bar43m$) with a cubic cell parameter near 6.0\,\AA. The least and
132: most electronegative elements are \textit{X} at $(\frac12,\frac12,\frac12)$
133: and \textit{Z} at $(0,0,0)$ forming a rock salt lattice. \textit{Y} are found
134: at $(\frac14,\frac14,\frac14)$ in the centers of tetrahedra formed
135: by \textit{Z}, as well as by \textit{X}. Connecting \textit{Y} and \textit{Z}
136: reveals the stuffed zinc blende lattice of the half-Heusler structure displayed
137: in Figure\,1. There are other, equivalent descriptions of this structure, but
138: this is the one we chose, because it is closest to our description of the
139: chemical bonding.
140:
141: Density functional theory-based electronic structure calculations were
142: performed using the full-potential Linear Augmented Plane Wave (LAPW) code
143: \textsc{Wien2k}\cite{Wien} to optimize cell volumes of the different
144: half-Heusler compounds described here. The electronic structural descriptions
145: made use of Linear Muffin Tin Orbital (LMTO) calculations within the local spin
146: density approximation, as implemented in the \textsc{Stuttgart tb-lmto-asa}
147: program.\cite{StuttgartLMTO} Starting structures for LMTO calculations were
148: obtained from the results of volume optimization using \textsc{Wien2k}. Two
149: important tools have been used to visualize the electronic Structure of these
150: phases. The crystal orbital Hamiltonian population (COHP) \cite{COHP} enables
151: the repartitioning of densities of states into regions which are pairwise
152: bonding, non-bonding, and antibonding. The electron localization function
153: (ELF)\cite{Becke,Silvi} is a real-space indicator of the extent to which
154: electrons are localized, and display a strong Pauli repulsion. The ELF
155: therefore serves to locate bonding and non-bonding electron pairs in the real
156: space of the crystal structure. A real-space bonding analysis of half-Heusler
157: compounds has not, to our knowledge, been previously carried out.
158:
159: \section{Results and discussion}
160:
161: \subsection{8-electron compounds}
162:
163: \begin{table}
164: \centering \begin{tabular}{|l|l|l|l|l|l|l|l|}
165: \hline
166: \multicolumn{8}{|c|}{8 electron compounds with \textit{X} = Li}\\
167: \hline \hline
168: \textit{XYZ} & $a_{\mathrm{Calc.}}$ (\AA) &
169: $a_{\mathrm{Exp. }}$ (\AA) & $\chi_X$ & $\chi_Y$
170: & $\chi_Z$ & Gap (eV) & $B$ (GPa) \\
171: \hline
172: LiMgN & 5.072 & 4.955 & 0.98 & 1.31 & 3.04 & 2.51 & 80.1 \\
173: \hline
174: LiMgP & 6.028 & 6.021 & 0.98 & 1.31 & 2.19 & 1.92 & 49.6 \\
175: \hline
176: LiMgAs & 6.218 & 6.19 & 0.98 & 1.31 & 2.18 & 1.55 & 42.9 \\
177: \hline
178: LiMgBi & 6.803 & 6.74 & 0.98 & 1.31 & 2.02 & 0.64 & 30.3 \\
179: \hline
180: LiZnP & 5.707 & 5.779 & 0.98 & 1.65 & 2.19 & 1.23 & 65.4 \\
181: \hline
182: LiCdP & 6.118 & 6.087 & 0.98 & 1.69 & 2.19 & 0.85 & 52.8 \\
183: \hline
184: LiAlSi & 5.937 & 5.930 & 0.98 & 1.61 & 1.90 & 0.45 & 62.8 \\
185: \hline
186: \end{tabular}
187: \caption{Results of density functional calculations on Li\textit{YZ} phases,
188: with experimental cell parameters for comparison. Cell parameters and
189: bulk moduli are from LAPW calculations, and band gaps from LMTO calculations.}
190: \end{table}
191:
192: We have examined 7 compounds Li\textit{YZ} with the half-Heusler structure
193: using LAPW and LMTO calculations. Results from the calculations
194: are summarized in Table\,1. The calculated cell parameters match very well
195: with experimental cell parameters obtained from standard
196: tabulations.\cite{Pearsons} For LiMgN, the experimental cell parameter
197: is from Kuriyama \textit{et al.}\cite{kuriyama}
198: This suggests that in all cases, the assignment
199: of atomic positions, which is not always a simple matter to determine from
200: x-ray diffraction, is well justified. We discuss these results here in detail.
201:
202: \begin{figure}
203: \centering \includegraphics[width=11cm]{fig02.eps}
204: \caption{LMTO densities of state for a number of different Li-based
205: half-Heusler phases compared with diamond Si.}
206: \end{figure}
207:
208: In the three panels of Figure\,2, we compare the densities of states near the
209: Fermi energy (taken as the top of the valence band, and set as the origin)
210: for 6 different Li\textit{YZ} half-Heusler compounds. In panel (a) of this
211: figure two compounds LiMgN and LiMgBi are compared. Both compounds have a well
212: defined band gap. Replacing N with the heavier pnictogens P, As, (DOS
213: not shown) or Bi found to narrow the band gap due to the increasing
214: band width of both the valence and conduction bands. If the \textit{X} and
215: \textit{Z} ions are held constant, as in LiMgP and LiCdP, we observe that
216: replacing the more ionic Mg by the softer Cd also results in a narrowing of the
217: band gap. In panel (c) we compare LiAlSi with isoelectronic Si in the diamond
218: structure. The electronic structures display a remarkable similarity in the
219: nature and extents of the valence and conduction bands. This strong similarity
220: has been noted previously by Christenson\cite{Christenson} and is fully
221: in keeping with the description of LiAlSi being a Zintl or valence
222: compound,\cite{Zintl} wherein the identical electron counts of (AlSi)$^-$ and
223: (SiSi) in turn imply that the structures would be similar. It should be noted
224: that Mg$_2$Si in the fluorite structure also obeys the same rule, if
225: we recognize it can be recast as Mg$^{2+}$(MgSi)$^{2-}$ with the (MgSi)$^{2-}$
226: crystallizing in a zinc blende lattice. Indeed the electronic structure of
227: MgSi$_2$\cite{Froseth} is quite similar to what we find for LiAlSi.
228: The change in the space group, comparing ($Fm\bar3m$) Mg$_2$Si and
229: ($F\bar43m$) LiAlSi arises because the atoms in the \textit{X} and \textit{Y}
230: sites in the former are identical.
231:
232: \begin{figure}
233: \centering \includegraphics[width=11cm]{fig03.eps}
234: \caption{Crystal orbital Hamiltonian populations (COHPs) for pairwise
235: interactions in diamond Si, LiAlSi, and LiMgN. The dashed lines are
236: integrations of the COHPs.}
237: \end{figure}
238:
239: The similarity in the electronic structures of Si and stuffed
240: zinc blende compounds is further emphasized through an analysis of the Si-Si,
241: the Al-Si, and the Mg-N COHPs of the three different compounds: Si, LiAlSi,
242: and LiMgN, shown in Figure\,3. The dashed line in this figure is an integration
243: of the COHP up to $E_{\rm F}$, yielding a number that is indicative of the
244: strength of the bonding. Not only are the extents of the bonding and
245: antibonding COHPs of Si and LiAlSi very similar, but so is the value of
246: the integrated COHP: near -1.5\,eV \textit{per\/} interaction for Si and
247: for LiAlSi. We interpret this as indicative of very similar extents of
248: covalency in the diamond lattice of Si and the zinc blende sublattice of
249: LiAlSi. This value is slightly reduced in the more polar LiMgN, and the nature
250: of the COHP is different as well.
251:
252: \begin{figure}
253: \centering \includegraphics[width=12cm]{fig04.eps}
254: \caption{(Color) (a), (b), and (c) are electron localization isosurfaces for
255: ELF values of 0.9, 0.9, and 0.825 respectively for the three compounds
256: Si, LiAlSi, and LiMgN. (d), (e), and (f) are isosurfaces of constant charge
257: density at a value of 0.06\,$e$\,\AA$^{-1}$. The isosurfaces are decorated by
258: the value of the electron localization function. The color-bar at the bottom of
259: the figure indicates increasing localization from left (0.0) to right (1.0).
260: The positions of atoms are as in Figure\,1, with \textit{Z} at the origin
261: \textit{etc.\/}.}
262: \end{figure}
263:
264: In Figure\,4, we compare the electron localization function (ELF) for the three
265: compounds, Si, LiAlSi, and LiMgN. Panels (a), (b), and (c) display isosurfaces
266: of the electron localization function for values of 0.90, 0.90, and 0.825
267: respectively. These are high values of localization (the ELF scale as used
268: here \cite{Silvi} runs from 0 through 1) and indicate highly covalent bonding
269: between Si (cyan spheres) in the elemental structure, as well as between Al
270: (blue) and Si (cyan) in LiAlSi, and between Mg (blue) and N (cyan) in LiMgN.
271: The ELF takes on a curious hemispherical shape in LiMgN, reflecting the
272: large electronegativity difference between Mg and N. The blob of localization
273: is also closer to N than it is to Mg. It must be noted that there is no
274: localization around Li (orange spheres) in either LiAlSi, or LiMgN, as seen
275: also from the map of the ELF projected on the (010) plane at the rear
276: of the unit cells. Li behaves effectively like ionic Li$^+$. In panels (d),
277: (e), and (f) of this figure, we display isosurfaces of charge for a value of
278: 0.06\,$e$\,\AA$^{-3}$ within the space of the unit cell. The charge isosurfaces
279: have been decorated (colored) by the ELF. Bonds distort charges from being
280: spherical, so distortions should be interpreted as covalency. Ionic species
281: on the other hand, would have spherical charge around the nucleus. We observe
282: the highly covalent nature of the diamond lattice in Si, and the zinc blende
283: sublattice in LiAlSi. For LiMgN, while the bonding is still covalent, the
284: charge is closer to the more electronegative nitrogen. The fact that the charge
285: is spherical rather than directed along the bond suggests that an anionic
286: description might be equally valid.
287:
288: \begin{figure}
289: \centering \includegraphics[width=7cm]{fig05.eps}
290: \caption{Dependence of the LMTO energy gap of Li\textit{YZ} compounds
291: on the difference in the Pauling electronegativities ($\chi$) of the \textit{Y}
292: and \textit{Z} species.}
293: \end{figure}
294:
295: Having established that at least the \textit{XYZ\/} half-Heuslers with
296: \textit{X} = Li can be written Li$^+$(\textit{YZ})$^-$, we use the data in
297: Table\,1 to examine systematics in the band gaps of these compounds. We find
298: the computed (LMTO) gaps of Li\textit{YZ} to vary approximately as the
299: difference of the Pauling electronegativities of \textit{Y} and \textit{Z}.
300: The larger the difference in electronegativity of the species in the zinc-blende
301: sublattice (\textit{Y} and \textit{Z}), the greater the band gap as a result of
302: band narrowing. The trend cannot be used quantitatively, but only as
303: and indicator of the Phillips-van Vechten-like\cite{Phillips}
304: behavior that is seen in these
305: complex semiconductors. It should be noted that the siting of \textit{Y}
306: in the tetrahedral position simultaneously allows the more polar pair of
307: \textit{X} and \textit{Z} to form a stable rock salt structure. From the
308: viewpoint of lattice energy, this is perhaps the greater stabilizing
309: influence on the half-Heusler structure.
310:
311: \subsection{18-electron compounds}
312:
313: To examine whether similar rules hold for 18-electron half-Heusler compounds,
314: we start by presenting in Table\,2 the results of density functional
315: calculations on a series of \textit{XYZ} compounds where \textit{X} is
316: electropositive yttrium. The zinc blende lattice is formed by a later
317: transition
318: metal \textit{Y}, and a main group element \textit{Z}. All the compounds are
319: semiconductors according to LMTO calculations, with band gaps ranging from
320: 0\,eV for YAuPb to 0.53\,eV for YNiAs. The existence of band gaps allows
321: us to formulate these phases according to the Zintl (or ``extended Zintl'')
322: rule X$^{3+}$(\textit{YZ})$^{3-}$ where (\textit{YZ})$^{3-}$ becomes
323: isoelectronic with a diamond-structure semiconductor such as GaSb. Once again,
324: we find a simple trend in the band gap with the difference in
325: electronegativities of \textit{Y} and \textit{Z}, as seen from Figure\,6.
326:
327: \begin{table}
328: \centering \begin{tabular}{|l|l|l|l|l|l|l|l|}
329: \hline
330: \multicolumn{8}{|c|}{18 electron compounds with \textit{X} = Y}\\
331: \hline
332: \hline
333: \textit{XYZ} & $a_{\mathrm{Calc.}}$ (\AA) &
334: $a_{\mathrm{Exp. }}$ (\AA) & $\chi_X$ & $\chi_Y$
335: & $\chi_Z$ & Gap (eV) & $B$ (GPa) \\
336: \hline
337: YNiAs & 6.104 & 6.171 & 1.22 & 1.91 & 2.18 & 0.53 & 100.0\\
338: \hline
339: YNiSb & 6.350 & 6.312 & 1.22 & 1.91 & 2.05 & 0.28 & 92.8\\
340: \hline
341: YNiBi & 6.475 & 6.411 & 1.22 & 1.91 & 2.02 & 0.13 & 80.9\\
342: \hline
343: YPdSb & 6.599 & 6.527 & 1.22 & 2.20 & 2.05 & 0.16 & 92.0\\
344: \hline
345: YAuPb & 6.842 & 6.729 & 1.22 & 2.54 & 2.33 & 0 & 70.6\\
346: \hline
347: \end{tabular}
348: \caption{Results of density functional calculations for a number of
349: 18-electron compounds with \textit{X} = Y.}
350: \end{table}
351:
352: \begin{figure}
353: \centering \includegraphics[width=7cm]{fig06.eps}
354: \caption{Dependence of the LMTO energy gap of Y\textit{YZ} compounds
355: on the difference in the Pauling electronegativities ($\chi$) of the \textit{Y}
356: and \textit{Z} species.}
357: \end{figure}
358:
359: We proceed to examine in detail, the electronic structure of select 18-electron
360: half-Heusler compounds with different \textit{X}, \textit{Y}, and \textit{Z}
361: elements. Figure\,7(a) displays the LMTO densities of state VFeSb, TiCoSb,
362: and YNiSb, allowing the trends with changing the later transition metal
363: \textit{Y} to emerge. Projections of the densities of state on the different
364: atomic levels (not displayed) reveal that the valence band has \textit{Y}
365: $d$ character and \textit{Z} $p$ character. The conduction band has some of the
366: character from these states, but in addition, has empty $d$ states from the
367: \textit{X} atom. This fits with our expectation of the \textit{X} atom being
368: nearly fully ionized (or more accurately, having attained the group valence)
369: and the $d$ shell of the \textit{Y} atom being filled as a result. The gap is
370: largest for TiCoSb, and smallest for YNiSb. When both \textit{X} and \textit{Y}
371: are changed, it is more difficult to seek trends in the gap. The gap in
372: VFeSb is reduced due to V and Fe not being well separated in electronegativity.
373: The gap in YNiSn is reduced because the unoccupied Y 4$d$ states are rather
374: broad, at least within LSDA. What is evident is that compounds with Co on
375: the \textit{Y} site have a strong propensity to maintain ``clean'' gaps as
376: further seen in panel (b) of this figure where the DOS of TiCoSb, VCoSn, and
377: NbCoSn are displayed.
378:
379: \begin{figure}
380: \centering \includegraphics[width=11cm]{fig07.eps}
381: \caption{(Color) LMTO densities of state for the 18-electron half-Heusler
382: compounds (a) VFeSb, TiCoSb, and YNiSb, showing how changing the nature of the
383: \textit{Y} atom affects the band gap, and for (b) TiCoSb, VCoSn, and NbCoSn
384: emphasizing the propensity of Co-based half-Heuslers to posses a ``clean'' gap.
385: The plots have been offset for clarity.}
386: \end{figure}
387:
388: From Figure\,8, we see that even in these transition-metal rich phases,
389: we can perform a bonding analysis based on COHPs. These are shown for all
390: pairwise interactions in the two half-Heuslers VFeSb and TiCoSb. Integrating
391: the COHPs, we find that the strongest bonding interactions are between
392: \textit{Y} and \textit{Z} (Co and Sb, and Fe and Sb) while the interactions
393: between the early and late transition metal (Ti and Co, and V and Fe)
394: are also significant. The compounds are electronically very stable as seen
395: from a complete absence of any antibonding interaction below the top of the
396: valence band. The fact that the 18-electron compounds are clearly valence
397: compounds with the band gap being located between bonding and antibonding
398: levels supports our description of these phases being Zintl-like.
399: The band gaps in VFeSb and TiCoSb would seem to be determined by bonding
400: between \textit{X} and \textit{Y}, so at first sight, it would seem that at
401: least these two transition-metal rich phases should not be described simply
402: as cation-stuffed zinc blendes. However, as we shall see from the ELF analysis,
403: the \textit{localized\/} bonding remains in the zinc blende \textit{YZ}
404: lattice. Weak bonding and antibonding COHPs between \textit{X} and \textit{Z}
405: (Ti and Sb, and V and Sb) support the view that the \textit{XZ} rock salt
406: sublattice is ionic in character.
407:
408: \begin{figure}
409: \centering \includegraphics[width=11cm]{fig08.eps}
410: \caption{Crystal orbital Hamiltonian populations for all pairwise
411: interactions in VFeSb and TiCoSb.}
412: \end{figure}
413:
414: Figure\,9, displays electron localization functions
415: for the two 18-electron half-Heusler compounds
416: for which the COHPs are displayed in Figure\,8, namely TiCoSb and VFeSb.
417: The ELFSs, visualized respectively for values of 0.73 and 0.71 in Figure.\,9(a)
418: and 9(b) are clearly indicative of strongly covalent bonding in the zinc blende
419: sublattice, with TiCoSb displaying the greater tendency to covalent bonding.
420: As we observed in the 8-electron compounds, localization isosurface is closer
421: to the more electronegative Sb atoms. It must be pointed out here that the
422: ELF is notoriously difficult to apply in $d$ electron system, and the fact
423: that the localization emerges so clearly here is compelling evidence for
424: dealing with these systems as if they were valence compounds with strongly
425: covalent bonding.
426:
427: The valence charge densities displayed in Figure\,9(c) and 9(d) are very
428: distinct from what was seen for the 8-electron compounds, because the filled
429: $d$ shell on \textit{Y} forms a large nearly spherical blobs around
430: that atom, visualized for a charge density of 0.06\,$e$\,\AA$^{-3}$. We find
431: that these blobs of charge are pulled out into four strongly localized (as
432: seen from the coloring) lobes arranged tetrahedrally and facing \textit{Z}.
433: Interestingly, in VFeSb, there is also some $d$-like localization around V,
434: and there is some little localization seen on the isosurface between V and Fe.
435: The smaller electronegativity difference between V and Fe, when
436: compared with the electronegativity difference between Ti and Co, in
437: conjunction with the nature of COHPs displayed in Figure\,8 leads us to
438: point this out as the origin for the smaller band gap of VFeSb. Another
439: argument that one can proffer is that TiCoSb has greater polar intermetallic
440: character,\cite{Brewer,Calhorda,Abdon} with strong covalent Ti-Co bonds.
441: V and Fe are closer together in electronegativity, and the
442: high formal charge state (V$^{5+}$) is much more covalent. This results
443: in a smaller band gap in VFeSb.
444:
445: In the next subsection, we will point out that these simple ideas of covalent
446: bonding can be carried over to the important magnetic half-Heusler compounds.
447:
448: \begin{figure}
449: \centering \includegraphics[width=8cm]{fig09.eps}
450: \caption{(Color) (a) and (b) are ELF isosurfaces for TiCoSb and VFeSb for
451: values of 0.73 and 0.71 respectively. (c) and (d) are charge densities
452: for a value of 0.06\,$e$\,\AA$^{-3}$, decorated by the ELF.}
453: \end{figure}
454:
455: \subsection{Magnetic compounds}
456:
457: \begin{table}
458: \centering \begin{tabular}{|l|l|l|l|l|l|}
459: \hline
460: \multicolumn{6}{|c|}{\textit{X}CoSb}\\
461: \hline
462: \hline
463: \textit{X} & $a_{\rm{Calc.}}$ & $a_{\rm{Exp.}}$ & $B$ (GPa) & $n_V$ & $M$ \\
464: \hline
465: Sc & 6.095 & N.A. & 110.5 & 17 & 0 \\
466: \hline
467: Ti & 5.888 & 5.884 & 151.5 & 18 & 0 \\
468: \hline
469: V & 5.823 & 5.802 & 150.8 & 19 & 1 \\
470: \hline
471: Cr & 5.820 & N.A. & 135.4 & 20 & 2 \\
472: \hline
473: Mn & 5.810 & 5.875 & 139.1 & 21 & 3 \\
474: \hline
475: \end{tabular}
476: \caption{Optimized (LAPW) and experimental cell parameters for the
477: the half-Heusler compounds \textit{X}CoSb. The computed bulk moduli are also
478: indicated. $n_V$ is the number of valence electrons and $M$ is the
479: magnetic moment obtained from LMTO calculations.}
480: \end{table}
481:
482: \begin{figure}
483: \centering \includegraphics[width=11cm]{fig10.eps}
484: \caption{Densities of state of the half-Heusler compounds \textit{X}CoSb,
485: plotted separately in the two spin directions. The number of valence
486: electron ($n_V$) in each formula unit are also indicated.}
487: \end{figure}
488:
489: The results of LAPW optimization of \textit{X}CoSb compounds are presented in
490: Table\,3, with experimental cell parameters presented for comparison for the
491: known compounds (\textit{X} = Ti, V, and Mn). The table also presents the
492: computed (LAPW) bulk modulus, which is seen to go through a maximum for TiCoSb,
493: associated, as we will observe in the COHPs, with a completely filled
494: \textit{bonding\/} valence band and an empty, \textit{antibonding\/}
495: conduction band. Figure\,10 displays densities of state for \textit{X}CoSb
496: phases with \textit{X} = Sc, Ti, V, Cr, and Mn. The number of valence
497: electrons $n_V$ per formula unit are indicated within each panel.
498: The DOS are plotted in the two spin directions in each panel, even for the
499: non-magnetic compounds. The compound ScCoSb is not known, and neither is
500: CrCoSb. The non-existence of CrCoSb could be associated with the high peak in
501: the densities of state at $E_{\rm F}$ for this compound. The (hypothetical)
502: compound ScCoSb is not magnetic within LMTO-LSDA because the bands are too
503: broad for the Stoner criterion to be fulfilled.
504:
505: All compounds except ScCoSb obey the Slater-Pauling rules for half-Heuslers,
506: $M = n_V - 18$. TiCoSb with $n_V$ = 18 is a non-magnetic semiconductor,
507: and the calculated moments (Table\,3) for VCoSb, CrCoSb, and MnCoSb are
508: precisely 1, 2, and 3\,$\mu_{\rm B}$. This means that these three compounds
509: are half-metals, as seen from the densities of states in Figure\,10.
510: In particular, CrCoSb, and MnCoSb have ``clean'' gaps in the minority spin
511: direction. With increasing $n_V$, we notice that the $d$ states above the
512: Fermi energy, which are derived from the electropositive \textit{X} atom
513: drop down with respect to the filled $d$ states on Co both because of their
514: partial filling as well as because of the well known tendency of transition
515: metal $d$ levels to be stabilized in energy on going across the $d$
516: series.\cite{Jobic,Zaanen} One of the consequences of the narrowing
517: of the $d$ separation between \textit{Y} and \textit{X} is that for larger
518: valence electron counts than 21 or 22 (found when \textit{Y} = Ni),
519: half-Heuslers become unstable with respect to other structure types.
520:
521: \begin{figure}
522: \centering \includegraphics[width=11cm]{fig11.eps}
523: \caption{(Color) COHPs of VCoSb and MnCoSb in the two spin directions. In the
524: absence of spin-orbit coupling, majority and minority spin states do not
525: interact.}
526: \end{figure}
527:
528: In the different panels of Figure\,11, we display COHPs for
529: \textit{X}-Co, \textit{X}-Sb, and Co-Sb interactions (\textit{X} = V or Mn).
530: These are half-metallic ferromagnets with 19 (VCoSb) or 21 (MnCoSb)
531: so the COHPs are spin-resolved. While the \textit{X}-Sb interaction within the
532: rock-salt sublattice is clearly negligible, both \textit{X}-Co and Co-Sb are
533: seen to be important. The \textit{X}-Co interaction is seen to be highly
534: spin-polarized. The origin of the half-metallicity is revealed by the clear
535: separation of antibonding \textit{majority\/} states which cross the Fermi
536: energy from the antibonding \textit{minority\/} states which are separated
537: by a gap equal to the exchange energy. This allows us to make the following
538: generalization: The 18-electron half-Heuslers are the most stable phases,
539: with well separated bonding and antibonding states. Additional electrons (more
540: than 18) must go into antibonding states and these are split by
541: spin-polarization and separated into majority and minority states. While
542: these compounds are intrinsically less stable than the 18-electron
543: compounds, they maximize their stability by ensuring that minority
544: antibonding states remain unoccupied. The Co-Sb interaction in these two
545: compounds is seen to be strongly covalent, but not very much affected
546: by spin-polarization. There remains a clear separation of bonding from
547: antibonding states as we had observed in the 18-electron semiconductors.
548:
549: \begin{figure}
550: \centering \includegraphics[width=12cm]{fig12.eps}
551: \caption{(Color) (a) and (d) are ELF isosurfaces for VCoSb and MnCoSb for
552: ELF values of 0.71. (b) and (e) are charge densities for a value of
553: 0.06\,$e$\,\AA$^{-3}$, decorated by the ELF. (c) and (f) are isosurfaces of
554: constant spin of value 0.05\,spins\,\AA$^{-3}$.}
555: \end{figure}
556:
557: Real-space visualizations of the electronic structure in Figure\,12 reveal that
558: even in the magnetic compounds \textit{X}CoSb with \textit{X} = V or Mn,
559: the ELFs are strongly localized on the bonds of the zinc blende CoSb network.
560: As could be anticipated from the similarities in the COHPs, there is almost
561: no change in the Co-Sb localization pattern on going from 19-electron
562: VCoSb to 21 electron MnCoSb [Figure\,12(a) and (d)]. There is a strongly
563: localized region slightly closer to the more electronegative Sb atoms in
564: both these compounds. Again, the charge density decorated by the ELF [(b) and
565: (e)] confirm this localization. The $d$ electron density around Co is spherical
566: apart from the four lobes facing Sb. What is interesting is that the
567: magnetic moment, as visualized from an isosurface of constant spin density
568: is clearly located on the stuffing \textit{X} atom in both in VCoSb and in
569: MnCoSb, as seen in Figure\,12(c) and (f). The magnetic half-Heuslers can
570: therefore be regarded as zinc-blende lattices of a late transition metal
571: and a main group element, stuffed by relatively electropositive magnetic ions.
572: Despite the presence of magnetic \textit{X}$^{n+}$ transition metal ions,
573: the \textit{YZ}$^{n-}$ network can still be described in simple valence terms.
574: This is reminiscent of compounds prepared by Kauzlarich and
575: coworkers\cite{Kauzlarich1, Kauzlarich2} wherein magnetic ions such as
576: Mn$^{3+}$ are found to behave like electropositive cations such as Al$^{3+}$
577: which donate charge to a closed shell anionic sublattice.
578:
579: \subsection{MnNiSb}
580:
581: \begin{figure}
582: \centering \includegraphics[width=11cm]{fig13.eps}
583: \caption{Band structure of MnNiSb decorated by the indicated orbital
584: contributions from transition metal $d$ states.}
585: \end{figure}
586:
587: \begin{figure}
588: \centering \includegraphics[width=11cm]{fig14.eps}
589: \caption{(Color) (a) ELF isosurfaces of MnNiSb for an ELF value of 0.73. (b)
590: is the charge density isosurface for a value of 0.055 $e$\,\AA$^{-3}$,
591: decorated by the ELF. (c) is an isosurfaces of constant spin density
592: corresponding to 0.05 spins \AA$^{-3}$.}
593: \end{figure}
594:
595: For completion, we discuss in this subsection, the electronic structure
596: of the canonical 22-electron half-Heusler compound, MnNiSb.\cite{deGroot}
597: As we observed from the COHP of MnCoSb, we expect for MnNiSb that
598: the minority gap will be formed from metal $d$ states, and this gap will fall
599: within the larger gap associated with the zinc blende (NiSb) sublattice.
600: The panels of figure\,13(a-d) display band structure of MnNISb in the
601: so-called fatband representation\cite{Jepsen} where the bands are decorated
602: with widths proportional to various specific orbital contributions;
603: In the different panels, the $d$ orbitals of Mn and Ni are indicated in the
604: different spin directions. From the band structures, it is evident that the
605: magnetism resides largely on Mn. The valence band has Ni $d$ character in
606: both spin direction, but only majority Mn $d$ states. The conduction band has
607: minority Mn $d$ states. The disperse majority band which traverses the Fermi
608: energy arises due to covalent bonding between majority Ni $d$ states and
609: majority Mn $d$ states with some intermediation by Sb $p$. Indeed, we find
610: that he disperse band going from $\Gamma$ to L and $\Gamma$ to X is retained
611: even when the calculation is performed in MnNiSb where the Sb atoms are
612: replaced by empty spheres.
613:
614: Figure\,14(a) and (b) display the ELF and charge of MnNiSb and confirm the
615: picture of covalent bonding in the zinc blende network of this structure.
616: Figure\,13(c) is the spin density, and is evidently almost completely localized
617: on Mn in this compound as suggested by the band structure.
618:
619: \ack
620: {Work at Mainz was supported by the \textit{Deutsche Forschungsgemeinschaft\/}
621: through projects FG 559 and the SPP 1166. RS gratefully acknowledges the
622: National Science Foundation for support through a Career Award
623: (NSF-DMR\,0449354).}
624:
625: \section*{References}
626:
627: \begin{thebibliography}{99}
628:
629: \bibitem{deGroot}
630: de Groot R A, Mueller F M, van Engen P G and Buschow K H J
631: 1983 \textit{Phys. Rev. Lett.\/} \textbf{50} 2024
632:
633: \bibitem{Kubler}
634: K\"ubler J, Williams A R and Sommers C B 1983 \textit{Phys. Rev. B\/}
635: \textbf{28} 1745
636:
637: \bibitem{Pierre1}
638: Pierre J, Skolozdra R V, Tobola J, Kaprzyk S, Hordequin C, Kouacou M A,
639: Karla I, Currat R and Leli\'evre-Berna E 1997 \textit{J. Alloys Compounds\/}
640: \textbf{262-263} 101
641:
642: \bibitem{Pierre2}
643: Tobola J and Pierre J 2000 \textit{J. Alloys Compounds\/} \textbf{296} 243
644:
645: \bibitem{Whangbo}
646: Jung D, Koo H-J and Whangbo M-H 2000 \textit{J. Mol. Struc. Theochem\/}
647: \textbf{527} 113
648:
649: \bibitem{Stoner}
650: Stoner E C 1938 \textit{Proc. R. Soc. Lond A\/} \textbf{165} 372
651:
652: \bibitem{Galanakis}
653: Galanakis I, Dederichs P H and Papanikolaou N 2002 \textit{Phys. Rev. B}
654: \textbf{66} 134428
655:
656: \bibitem{Ogut}
657: \"O{\u g}\"ut S and Rabe K M 1995 \textit{Phys. Rev. B\/} \textbf{51} 10443
658:
659: \bibitem{Wood}
660: Wood D M, Zunger A and de Groot R 1985 \textit{Phys. Rev. B\/}
661: \textbf{31} 2570
662:
663: \bibitem{Nowotny}
664: Nowotny H and Bachmayer K 1950 \textit{Monatsh. Chem.\/} \textbf{81} 488
665:
666: \bibitem{Juza}
667: Juza R and Hund F 1948 \textit{Z. Anorg. Chem.\/} \textbf{257} 1
668:
669: \bibitem{Dasgupta} Nanda B R K and Dasgupta I 2003 \textit{J. Phys. Condens.
670: Matter\/} \textbf{15} 7307
671:
672: \bibitem{Wien}
673: Madsen G K H, Blaha P, Schwarz K, Sj\"ostedt E and N\"ordstr\"om L,
674: 2001 \textit{Phys. Rev. B\/} \textbf{64} 195134;
675: see http://www.wien2k.at
676:
677: \bibitem{StuttgartLMTO}
678: Andersen O K 1975 \textit{Phys. Rev. B\/} \textbf{12} 3060;
679: Jepsen O and Andersen OK 2000 STUTTGART TB-LMTO-ASA Program version 47,
680: MPI f\"ur Festk\"orperforschung, Stuttgart, Germany
681:
682: \bibitem{COHP}
683: Dronskowski R and Bl\"ochl P E 1993 \textit{J. Phys. Chem.\/}
684: \textbf{97} 8617
685:
686: \bibitem{Becke}
687: Becke A D and Edgecombe K E 1990 \textit{J. Chem. Phys.\/} \textbf{92} 5397
688:
689: \bibitem{Silvi}
690: Silvi B and Savin A 1994 \textit{Nature\/} \textbf{371} 683
691:
692: \bibitem{Pearsons}
693: Villars P and Calvert L D 1996, Pearson's Handbook of Crystallographic
694: Data for Intermetallic Phases, 2nd Ed. ASM International, Materials Park, Ohio.
695:
696: \bibitem{kuriyama}
697: Kuriyama K, Nagasawa K and Kushida K 2002 \textit{J. Cryst. Growth\/}
698: \textbf{237} 2019
699:
700: \bibitem{Christenson}
701: Christenson N E 1985 \textit{Phys. Rev. B} \textbf{32} 6490
702:
703: \bibitem{Zintl}
704: Zintl E 1939 \textit{Angew. Chem.\/} \textbf{52} 1
705:
706: \bibitem{Froseth}
707: Fr\o seth A G, H\o ier R, Derlet P M, Andersen S J and Marioara C D
708: 2003 \textit{Phys. Rev. B\/} \textbf{67} 224106
709:
710: \bibitem{Phillips}
711: Phillips J C and van Vechten J A 1970 \textit{Phys. Rev. B} \textbf{2} 2147
712:
713: \bibitem{Brewer} Brewer L and Wengert PR 1973 \textit{Metall. Trans.\/}
714: \textbf{4} 83
715:
716: \bibitem{Calhorda}
717: Calhorda M J and Hoffmann R 1988 \textit{Inorg. Chem.\/} \textbf{27} 4679
718:
719: \bibitem{Abdon}
720: Abdon R L and Hughbanks T 1995 \textit{J. Am. Chem. Soc.\/} \textbf{117} 10035
721:
722: \bibitem{Jobic}
723: Jobic S, Brec R and Rouxel J 1992 \textit{J. Alloys Compounds\/}
724: \textbf{178} 253
725:
726: \bibitem{Zaanen}
727: Zaanen J, Sawatzky G A and Allen J W 1985
728: \textit{Phys. Rev. Lett\/} \textbf{55} 418
729:
730: \bibitem{Kauzlarich1}
731: Kuramoto T Y, Kauzlarich S M and Webb D J 1992 \textit{Chem. Mater.\/}
732: \textbf{4} 435
733:
734: \bibitem{Kauzlarich2}
735: Rehr A, Kuramoto T Y, Kauzlarich S M, Del Castillo J and Webb D J
736: 1994 \textit{Chem. Mater.\/} \textbf{6} 93
737:
738: \bibitem{Jepsen}
739: Jepsen O and Andersen O K 1995 \textit{Z Phys B\/} \textbf{97} 35
740:
741: \end{thebibliography}
742:
743: \end{document}
744: