1:
2: %\usepackage{graphicx}
3: %\usepackage{dcolumn}
4: %\documentstyle[amssymb,aps,prb,multicol,epsfig]{revtex}
5: %\input{tcilatex}
6: %\input{tcilatex}
7:
8:
9: \documentclass[showpacs,aps,prb,twocolumn,epsfig]{revtex4}
10: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
11: \usepackage{amssymb}
12: \usepackage{graphicx}
13:
14: %TCIDATA{OutputFilter=Latex.dll}
15: %TCIDATA{Version=5.50.0.2953}
16: %TCIDATA{<META NAME="SaveForMode" CONTENT="1">}
17: %TCIDATA{BibliographyScheme=Manual}
18: %TCIDATA{LastRevised=Saturday, October 21, 2006 04:14:11}
19: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
20:
21:
22: \begin{document}
23:
24: \title{
25: Combining the advantages of superconducting MgB$_2$ and CaC$_6$
26: in one material: suggestions from first-principles calculations
27: }
28: \author{ Amy Y. Liu$^1$}
29: \author{ I. I. Mazin$^{1,2}$}
30: \affiliation{$^1$ Department of Physics, Georgetown University, Washington, DC 20057-0995
31: \\
32: $^2$ Center for Computational Materials Science, Naval Research Laboratory,
33: Washington, DC 20375}
34:
35: \begin{abstract}
36: We show that a recently predicted layered phase of lithium monoboride, Li$_2$%
37: B$_2$, combines the key mechanism for strong electron-phonon coupling in MgB$%
38: _2$ (i.e., interaction of covalent B $\sigma$ bands with B bond-stretching
39: modes) with the dominant coupling mechanism in CaC$_6$ (i.e., interaction of
40: free-electron-like interlayer states with soft intercalant modes). Yet,
41: surprisingly, the electron-phonon coupling in Li$_{2}$B$_{2}$ is calculated
42: to be weaker than in either MgB$_2$ or CaC$_6$. We demonstrate that this is
43: due to the accidental absence of B $\pi$ states at the Fermi level in Li$_{2}
44: $B$_{2}$. In MgB$_2$, the $\pi$ electrons play an indirect but important
45: role in strengthening the coupling of $\sigma$ electrons. Doping Li$_{2}$B$%
46: _{2}$ to restore $\pi $ electrons at the Fermi level is expected to lead to
47: a new superconductor that could surpass MgB$_{2}$ in $T_c$.
48: \end{abstract}
49:
50: \date{\today }
51: \pacs{ 74.25.Jb, 74.25.Kc, 74.70.Ad}
52: \maketitle
53:
54: \section{Introduction}
55:
56: The discovery of superconductivity in MgB$_{2}$ at $T_c = 39$ K led to a new
57: paradigm for achieving high critical temperatures in conventional
58: superconductors, namely strong electron-phonon coupling due to metallization
59: of strong covalent bonds.\cite{AP} In MgB$_2$, this is achieved through the
60: interaction of electrons forming covalent B-B bonds in the graphene-like
61: layers with vibrational modes that stretch and compress these
62: bonds.\cite{mgb2rev}. Unfortunately, efforts to find related materials
63: with even higher
64: $T_c$ have not been fruitful. This is basically because the electronic
65: states that form the covalent bonds in MgB$_2$ are quasi two-dimensional,
66: which makes it difficult to enhance their density of states and their
67: coupling to phonons through simple doping.\cite{PickettND} Significant
68: progress may only be possible through the introduction of different types of
69: electronic states at the Fermi level.\cite{PickettND}
70:
71: The recent discovery of superconductivity in CaC$_{6}$ at
72: $T_c = 11.5$ K,\cite{cac6}
73: the highest critical temperature among graphite intercalation compounds
74: (GICs), offers an alternative path to increasing $T_c$. Although CaC$_6$ is
75: structurally similar to MgB$_2$, it appears that the key electrons and
76: phonons responsible for its superconductivity are distinct from those in MgB$%
77: _2$.\cite{CM,cac6rev} In particular, free-electron-like states that are
78: concentrated in the region between graphene layers dominate the
79: electron-phonon interaction through coupling with soft intercalant modes and
80: with bond-bending modes of the graphene sheet. Incipient lattice instabilities
81: associated with the soft intercalant phonon modes, however, may present
82: obstacles to attaining significantly higher $T_c$ in GICs.\cite{Kim,CM}
83:
84: It is intriguing to ask whether the MgB$_2$ and CaC$_6$ mechanisms for
85: strong electron-phonon coupling can be incorporated into a single compound
86: to create a superconductor with the potential to surpass MgB$_2$. Although
87: neither MgB$_2$ nor CaC$_6$ can be easily modified so as to incorporate the
88: key electronic and phonon players from the other material, we show here that
89: it is indeed possible to combine the MgB$_2$ and CaC$_6$ paradigms in a
90: different material. Such a material has a good chance of exceeding both MgB$%
91: _2$ and CaC$_6$ in terms of $T_c$.
92:
93: The material of interest is a hypothetical layered form of lithium
94: monoboride, Li$_2$B$_2$, that recent calculations have predicted to be
95: competitive in energy with known LiB phases.\cite{Curtarolo, Curtarolo2}
96: The B graphene sheets in Li$_2$B$_2$ are nearly identical to those in MgB$_2$%
97: . At the Fermi level, there are $\sigma$ bands derived from B $p_{xy}$
98: orbitals that form covalent bonds within the layers, as in MgB$_2$, and
99: there are interlayer free-electron-like bands, which we shall call $\zeta$
100: bands, as in CaC$_6$. By accident, the B $p_z$-derived $\pi$ states that are
101: present in both MgB$_2$ and CaC$_6$ are missing at the Fermi level in Li$_2$B%
102: $_2$. Based on the similarity of the in-plane B physics with MgB$_2$,
103: Kolmogorov and Curtarolo have suggested that this material could be a good
104: superconductor in the mold of MgB$_2$.\cite{Curtarolo}
105:
106: In this paper, we analyze the electronic structure of Li$_2$B$_2$ and show
107: that the electron-phonon interaction has sizable contributions not only from
108: the $\sigma$ bands that dominate the coupling in MgB$_2$, but also from the
109: interlayer $\zeta$ bands, which are the key players in CaC$_6$.
110: Surprisingly, the total electron-phonon coupling in Li$_2$B$_2$ is found to
111: be weaker than in either of the other two materials. We show that this is
112: precisely because of the missing $\pi$ electrons at the Fermi level. Though
113: the $\pi$ bands in MgB$_2$ and CaC$_6$ give only moderate contributions to
114: the electron-phonon coupling, $\pi$ electrons can indirectly
115: strengthen the coupling of $\sigma$ electrons with B bond-stretching
116: phonon modes through screening. Doping Li$_{2}$B$_{2}$ to add $\pi $ sheets to
117: the Fermi surface would enhance the $\sigma$ contribution to the
118: electron-phonon coupling, as well as add a modest $\pi$ contribution. While
119: a relatively low $T_c$ is predicted for pristine Li$_2$B$_2$, an
120: appropriately doped material could have a $T_c$ higher than even MgB$%
121: _{2}.$
122:
123: \section{Calculational Methods}
124:
125: Our analysis is based on density functional calculations.
126: The electronic structure was calculated using the WIEN2K general
127: potential LAPW code,\cite{WIEN} with the GGA functional in the PBE
128: form.\cite{PBE} Local orbitals were added to relax the remaining linearization
129: error and APW orbitals were used to improve the convergence with respect to
130: the cutoff parameter $RK_{\max }.$ In most calculations $RK_{\max }=7$ was
131: used, except for the convergence tests with $RK_{\max }=8.$ Up to 918
132: irreducible \textbf{k}-points were used (32$\times36\times16$). The phonon
133: spectrum and electron-phonon coupling functions were calculated using the
134: linear-response method\cite{LR} within the local density approximation.
135: Norm-conserving pseudopotentials\cite{TM} were used, with a plane-wave
136: cutoff of 50 Ry. The dynamical matrix was calculated on a grid of $%
137: 6\times6\times2$ phonon wavevectors \textbf{q} in the Brillouin zone and
138: Fourier interpolated to denser meshes for calculation of the phonon density
139: of states (DOS). For calculation of the electron-phonon matrix elements, the
140: electronic states were sampled on grids of $36\times36\times12$ \textbf{k}%
141: -points in the Brillouin zone.
142:
143: \section{Results and Discussion}
144:
145: The structure of the hypothetical layered compound Li$_2$B$_2$, first
146: proposed in Ref. \onlinecite{Curtarolo} (in which it is called MS2), consists of
147: honeycomb B sheets, each surrounded above and below by layers of Li on a
148: triangular lattice. There is a horizontal shift between neighboring Li-B-Li
149: sandwich units, leading to an ABAB... stacking sequence of the Li-B-Li
150: trilayer units. While we find that the optimized structural parameters,
151: particularly the interlayer distances, are sensitive to the details of the
152: electron-ion potential and to the exchange-correlation functional, the band
153: structures calculated within the different approximations are very similar.
154: For the results presented here, we have fixed the structural parameters to
155: the optimized values reported in Ref. \onlinecite{Curtarolo}.
156:
157: \begin{figure}[tb]
158: \includegraphics[width=2.80 in]{bnds.eps} \hspace*{0.03in} %
159: \caption{(Color online) Li$_2$B$_2$ band structure near the Fermi level
160: ($E_F=0$).
161: The $\protect\sigma$, $\protect\pi$, and $\protect\zeta$ bands
162: are plotted with blue, green, and magenta symbols, respectively.}
163: \label{fig: BS}
164: \end{figure}
165:
166: \begin{figure}[tb]
167: \includegraphics[width=1.4 in]{charge-zeta2.eps}
168: \caption{(Color online) Isosurface of Li$_2$B$_2$ charge
169: density corresponding to the $\protect\zeta$ state near the Fermi level,
170: half way between the $\Gamma$ and A points. The top, center, and
171: bottom plane in the figure contain honeycomb sheets of B. The charge
172: density in the center B plane is shown. Two Li layers lie between
173: each pair of B sheets. The $\zeta$ states are primarily
174: concentrated in the interlayer region, but have some B $p_z$ character
175: as well.
176: }
177: \label{fig: charge}
178: \end{figure}
179:
180:
181: The band structure of Li$_{2}$B$_{2}$ near the Fermi level is shown in
182: Fig. \ref{fig: BS}. As expected, the $\sigma$ (blue) and $\pi $
183: (green) states are derived primarily from in-plane B $p_{x,y}$ and
184: out-of-plane B $p_{z}$ orbitals, respectively. The $\zeta $ bands (magenta)
185: have substantial Li $sp$ and B $p_{z}$ character. The charge density of the $%
186: \zeta $ electrons, plotted in Fig.\ref{fig: charge}, is
187: concentrated in the region between Li-B-Li trilayer units, though
188: hybridization with B $p_z$ orbitals is evident. This is similar to the
189: character of $\zeta$ bands in GICs.\cite{Cs,CM,I} The Fermi level in Li$_2$B%
190: $_2$ lies at the crossing of the $\pi $ and $\pi ^{\ast }$ bands at K,
191: reminiscent of pure graphite. Note, however, that in graphite, the
192: coincidence of the Fermi level with the $\pi $-$\pi ^{\ast } $ crossing is
193: mandated by the absence of other bands near $E_F$, while in Li$_2$B$_2$, it
194: is accidental. (Expanding $c/a$ by $\sim 10$\%, for example, moves the $\pi$-%
195: $\pi^*$ crossing away from the Fermi level by more then 0.5 eV.) Compared to
196: pure graphite, the presence of Li atoms between the B-based graphene sheets
197: i) makes the $\pi $ bands more 3D and lowers their energy relative to the $%
198: \sigma $ bands, and ii) lowers the $\zeta $ bands due to the attractive
199: ionic potential in the interlayer region where the electronic charge in the $%
200: \zeta $ bands is concentrated.\cite{Cs,cac6rev} With these relative shifts
201: in the bands, the $\sigma $ bands, which are completely filled in pure
202: graphite, become partially occupied in Li$_{2}$B$_{2}$, while the $\zeta $
203: bands, which are unoccupied in graphite, cross the Fermi level in Li$_{2}$B$%
204: _{2}$. Hence Li$_{2}$B$_{2}$ can be viewed as a self-doped analog to pure
205: graphite, with holes in the $\sigma $ bands exactly compensated by electrons
206: in the $\zeta $ band. In contrast, in CaC$_{6}$ the $\pi ^{\ast }$ band is
207: partially occupied as well.
208:
209: In comparing the band structures of Li$_{2}$B$_{2}$ and MgB$_{2}$, which
210: both have B honeycomb sheets separated by \textquotedblleft intercalant"
211: layers that donate one electron per B site, key differences in the relative
212: energies of the different types of bands can be attributed to the difference
213: in the separation between B layers. With two Li layers between B sheets in Li%
214: $_{2}$B$_{2}$, the B-B interlayer distance in Li$_{2}$B$_{2}$ (5.4 \AA ) is
215: significantly larger than in MgB$_{2}$ (3.5 \AA ). This makes the $p_{xy}$%
216: -derived $\sigma $ bands more 2D in Li$_{2}$B$_{2}$, but has little effect
217: on their in-plane dispersion. As in MgB$_{2}$, these bands give rise to
218: nearly cylindrical hole sheets of the Fermi surface
219: surrounding the $\Gamma $ to A line. As can be seen in the Fermi
220: surface plotted in Fig. \ref{fig: FS},
221: there are two nearly degenerate pairs of $\sigma$ cylinders.
222: The larger
223: separation between B planes also causes the B $p_{z}$-derived $\pi $ bands
224: to be more 2D in Li$_{2}$B$_{2}$ and raises them in energy so that the $\pi
225: ^{\ast }$ band that is partially occupied in MgB$_{2} $ becomes unoccupied
226: (and just touches $E_F$ at certain points in the Brillouin zone). The Fermi
227: surface due to the $\pi $ bands thus disappears in Li$_{2}$B$_{2}$. On the
228: other hand, the free-electron-like $\zeta $ bands, which are unoccupied in
229: MgB$_{2}$, drop in energy when the interlayer distance is increased since
230: there is a larger interlayer volume available for the electrons to
231: occupy.\cite{cac6rev,note} In Li$_{2}$B$_{2}$, the $\zeta $ bands produce a
232: slightly squashed ellipsoidal Fermi surface around $\Gamma $, with a smaller
233: effective mass for motion along the $z$ direction than in the plane.
234: These ellipsoids are huge compared to the
235: small $\zeta$ balls in the CaC$_6$ Fermi surface,\cite{cac6rev} even though
236: CaC$_{6}$ has the most occupied $\zeta $ band among all GICs.\cite{Cs}
237:
238: \begin{figure}[tb]
239: \includegraphics[width=3.2 in]{fs.eps}
240: \caption{(Color online) Fermi surface of Li$_2$B$_2$. The cylindrical sheets
241: centered on the $\Gamma$-to-A line arise from the 2D $%
242: \protect\sigma$ bands, while the ellipsoids are due to free-electron-like
243: $\zeta$ bands.}
244: \label{fig: FS}
245: \end{figure}
246:
247: The topological simplicity of the Fermi surface of MgB$_2$, in which sheets
248: of the Fermi surface corresponding to bands of different character are
249: widely separated in reciprocal space, is lost in Li$_2$B$_2$. Not only do
250: the $\zeta$ ellipsoids cross all four $\sigma$ cylinders, they also cross
251: each other. At all of these crossings, there is only weak hybridization
252: between the bands. The multiple crossings between sheets of the Fermi
253: surface arising from bands of different character, along with the presence
254: of bands that nearly touch the Fermi level, make Brillouin-zone integrations
255: sensitive to \textbf{k}-point sampling. This is especially an issue for
256: calculations of the electron-phonon coupling constants, which involve double
257: integrals over the Fermi surface. The quantitative electron-phonon coupling
258: results presented here likely have larger uncertainties than usual but this
259: does not affect the qualitative picture that emerges.
260:
261: \begin{figure}[t]
262: \includegraphics[width=2.5 in ]{a2F.eps}
263: \caption{Li$_2$B$_2$ phonon density of states $F(\omega)$,
264: electron-phonon spectral function $\protect\alpha^2F(\omega)$,
265: and ratio of $\protect\alpha^2F$ to frequency $%
266: \protect\omega$. The electron-phonon coupling constant $\lambda$, which
267: is proportional to the inverse-frequency-moment of $\alpha^2F$,
268: has significant contributions from both low-frequency and high-frequency modes.}
269: \label{fig: ph}
270: \end{figure}
271:
272: The phonon density of states $F(\omega )$ and electron-phonon spectral
273: function $\alpha ^{2}F(\omega )$ calculated for Li$_{2}$B$_{2}$ are shown in
274: Fig. \ref{fig: ph}. The $E_{2g}$ in-plane bond-stretching B modes that are
275: the key players in superconductivity of MgB$_{2}$ are also important in Li$%
276: _{2}$B$_{2}$, as indicated by the peak in the spectral function near 80 meV.
277: In both materials, these vibrational modes couple strongly to the B $\sigma $
278: bands. Given the similarity of the B in-plane physics in the two materials,
279: one expects similar deformation potentials for the $E_{2g}$ modes.\cite%
280: {Curtarolo} Indeed, our frozen-phonon calculations using LAPW-GGA produce
281: exactly the same deformation potential at $\Gamma :$ a displacement $u$
282: opens a gap $\Delta \varepsilon $ such that $\Delta \varepsilon /u=26$ eV/%
283: \AA , precisely the same as found for MgB$_{2}$.\cite{AP}
284: Furthermore, while it is difficult to separate the electronic DOS in Li$_{2}$%
285: B$_{2}$ into contributions from different bands, as is routinely done in MgB$%
286: _{2},$ the effective masses derived from the dispersion of the $\sigma $
287: bands are within 10\% of those in MgB$_{2}$ (the heavy band is 9\% heavier
288: and the light band 8\% lighter). If the phonon frequencies were the same,
289: one would then expect the net intraband electron-phonon coupling constant
290: for the $\sigma$ bands, $\lambda _{\sigma \sigma }$,
291: to be practically the same as in MgB$_{2}$. Surprisingly,
292: the phonon frequencies differ considerably. At the zone center,
293: the $E_{2g}$ phonon frequency of 81 meV in Li$_{2}$B$_{2}$ is signficantly
294: higher that the corresponding 67 meV in MgB$_{2}$,\cite{mgb2rev,anh} which
295: likely reduces $\lambda _{\sigma \sigma }$ from $\approx 1$ to $\approx 0.6
296: - 0.7.$
297:
298: The hardening of the in-plane B bond-stretching modes in Li$_{2}$B$_{2}$ is
299: related to the loss of $\pi $ electrons at the Fermi level.
300: The same electron-phonon coupling process that
301: gives rise to superconductivity also leads to
302: electronic screening of phonons. For a zone-center phonon
303: mode, the softening due to screening by metallic electrons is
304: given by
305: \begin{equation}
306: \Delta \omega^2 = -4 \omega \langle g^2 \rangle N(0),
307: \label{eq: softening}
308: \end{equation}
309: where $\langle g^2 \rangle $ is a Fermi-surface average of the
310: square of the electron-phonon matrix element $g$, and $N(0)$ is the
311: electronic DOS at the Fermi level.\cite{osv, mgb2rev}
312: Because the matrix element $g$ contains a factor of
313: $1/\sqrt\omega$, the right-hand side of Eq. \ref{eq: softening} is
314: independent of the phonon frequency $\omega$.
315: In MgB$_{2}$, both the $\sigma $ and $\pi $ electrons participate in screening
316: the bare $E_{2g}$ phonons.
317: For the $E_{2g}$ mode in MgB$_2$, it has been estimated that
318: $\omega^2$ is reduced by about 1500 meV$^{2}$ due to screening
319: by $\pi $ electrons, and by about 2000 meV$^{2}$ due to screening by $\sigma
320: $ electrons, and that the unscreened frequency is $\omega_0 \approx 90$
321: meV.\cite{mgb2rev} In Li$_{2}$B$_{2}$, the unscreened frequency and the
322: softening due to $\sigma $ bands should be roughly the same magnitude as in
323: MgB$_2$, which gives a rough estimate for the screened $E_{2g}$ phonon
324: frequency of $\omega = \sqrt{\omega_0^{2}-(\Delta\omega^2)_{\sigma} }
325: \approx 78$ meV. This is consistent with our actual
326: calculations for Li$_{2}$B$_{2},$ which yield a frequency of 81 meV at
327: the zone center.
328: Since only phonons with $q_{xy} < 2k_{F}^{\sigma }$, where $k_{F}^{\sigma } $
329: is the radius of the cylindrical $\sigma $ sheets of the Fermi surface, can
330: couple to $\sigma$ electrons, softening by $\sigma $ electrons
331: is only significant when the in-plane component of the phonon wavevector is
332: less than $ 2k_{F}^{\sigma }$. Indeed, at the zone boundary,
333: our calculations yield a frequency of
334: 92 meV for the lowest $E_{2g}$ branch, consistent with
335: estimates for the unscreened $E_{2g}$ frequency.
336: While Li$_{2}$B$_{2}$ also has $\zeta $ electrons, which
337: are not present in MgB$_{2}$, these are largely localized in the interlayer
338: region and do not couple strongly with the B bond-stretching modes, so they
339: are less effective in screening the $E_{2g}$ phonons.
340: The harder $E_{2g} $ frequencies in Li$_{2}$B%
341: $_{2}$ compared to MgB$_{2}$ can thus be attributed to the absence of
342: screening by $\pi $ electrons.
343:
344: %In CaC$_{6}$, the key players in superconductivity are low-frequency Ca
345: %modes and intermediate-frequency modes involving out-of-plane motion of the
346: %C atoms. These vibrational modes couple strongly to the interlayer $\zeta $
347: %electronic states.
348: In addition to the high-frequency $E_{2g}$ peak in the spectral function,
349: there is a low-frequency peak in $\alpha ^{2}F$ arising from ultrasoft modes
350: that involve rigid shifts of the B planes in the $z$ direction coupled with
351: both in-plane and out-of-plane motion of the Li atoms.
352: The frequency of these ultrasoft modes
353: is sensitive to calculational details, so at $\Gamma $,
354: for example, the
355: harmonic frequency of about 8 meV obtained in our LDA linear-response
356: calculation is somewhat lower than the estimate of 11 meV based on
357: force-constant methods using PAW GGA calculations.\cite{Curtarolo} In any
358: case,
359: these low-frequency normal modes couple with the $\zeta $ bands, similar to
360: the situation in CaC$_6$, where Ca sliding modes interact strongly with
361: interlayer electrons. Although relatively small, the low-frequency peak in $%
362: \alpha ^{2}F$ gives a similar contribution to the isotropic
363: electron-phonon coupling constant $\lambda = 2\int d\omega \alpha
364: ^{2}F/\omega $ as
365: the high-frequency peak (see Fig. \ref{fig: ph}, bottom
366: panel). The total $\lambda$ is calculated to be 0.57, with the low-frequency
367: peak contributing about 0.15, and the $E_{2g}$ peak contributing about 0.2.
368: The logarithmically averaged phonon frequency of $\omega _{\ln }=\exp [{%
369: \lambda }^{-1}\int d\omega \ln (\omega )\alpha ^{2}F(\omega )/\omega ]=39$
370: meV, is well below the average frequency of $\omega _{ave}=59$ meV,
371: reflecting the importance of the ultrasoft modes.
372:
373: From the point of view of $T_c$, the prospects for Li$_2$B$_2$ do not look
374: optimistic compared to MgB$_{2},$ because of both the lack of a $\pi$-band
375: contribution to $\lambda ,$ and, perhaps more importantly, the hardening of
376: the key $E_{2g}$ phonons that are no longer screened by $\pi $ electrons.
377: Neglecting possible multiband effects and using the McMillan
378: equation\cite{McMillan}
379: with a typical Coulomb parameter of $\mu ^{\ast }=0.12, $ we
380: estimate $T_{c}\approx 7$ K. In comparison, the estimated \textit{isotropic}
381: $T_{c}$ in MgB$_{2}$ is 25-30 K. Enhancement of $T_c$ due to interband
382: anisotropy is expected to be weaker in Li$_{2}$B$_{2}$ since the
383: two groups of electrons contribute more equally
384: to the coupling with phonons.
385: But even if it were of the same order as in MgB$_{2},$ that is, $%
386: \sim 30\%$ in the effective $\lambda ,$ it would only increase $T_{c}$ to
387: about 15 K. The bottom line is that Li$_{2}$B$_{2},$ if synthesized, might
388: be competitive with CaC$_{6}$ in terms of $T_c$, but not with MgB$_{2}.$
389:
390: However, that is not the end of the story.
391:
392: While the \textit{undoped} Li$_{2}$B$_{2}$ is isoelectronic with MgB$_{2},$
393: and doping of MgB$_{2}$ is known to unequivocally depress $T_{c},$ the same
394: need not be true for Li$_{2}$B$_{2}.$ Electron doping of MgB$_{2}$
395: lowers the $\sigma $ DOS, and hence the electron-phonon coupling,
396: since the bands are not ideally 2D. (Though attempts have been
397: made, hole doping of MgB$_2$ has not yet been achieved.)
398: In Li$_{2}$B$_{2},$
399: the $\sigma $ bands are much more 2D, so doping should have
400: little effect on their DOS and electron-phonon coupling. On the other hand,
401: electron doping would start filling the $\pi $ bands, increasing
402: their DOS. Assuming rigid bands, doping on either B or Li sites
403: could be used to raise the Fermi level by nearly 1 eV without losing
404: the contribution from the $\sigma $ bands. We estimate that this would
405: increase the $\pi$-band DOS to roughly half the value in MgB$_2$.
406: Doping on the Li sites (with Al, Be, Mg, etc.) should yield even higher
407: $\pi$ DOS, since the extra ionic charge in the Li
408: planes would lower the $\pi $ bands relative to the $\sigma $ bands.
409: Experience shows that it is hard to dope more holes into the $\sigma $ bands
410: than there already are in pure MgB$_{2}$. Since Li$_{2}$B$_{2}$ starts with
411: more holes in the $\sigma $ bands than MgB$_{2}$, it might even be easier to
412: synthesize electron-doped Li$_{2}$B$_{2}$ than the undoped material.
413: Assuming that the $\pi $ bands in doped Li$_{2}$B$_{2}$ couple with the $%
414: E_{2g}$ modes at about the same level as in MgB$_{2},$ this
415: additional coupling will soften the $E_{2g}$ phonons and
416: enhance their coupling with the $\sigma $ bands, without
417: strongly affecting the coupling of $\zeta $ bands with soft modes.
418: This is an extremely favorable combination that is
419: likely to have a higher $T_{c}$ than MgB$_{2}.$ To a lesser extent the same
420: holds if the accidental alignment of the Fermi level with the $\pi $-$\pi
421: ^{\ast }$ crossing is modified by uniaxial pressure.
422:
423:
424: While Li$_{2}$B$_{2}$ and related layered lithium borides proposed
425: by Kolmogorov and Curtarolo\cite{Curtarolo} are only
426: hypothetical materials as yet, their calculations of the
427: Li-B phase diagram have suggested a route for synthesizing these materials
428: using pressure.\cite{Curtarolo2}
429: We have shown that although Li$_{2}$B$_{2}$ seems to
430: combine the key mechanism for strong electron-phonon coupling
431: in MgB$_2$ with the dominant coupling mechanism in
432: CaC$_6$, each of which has the highest superconducting transition
433: temperature in its respective class of materials, the transition
434: temperature of Li$_{2}$B$_{2}$ is expected to be lower
435: because of the lack of the seemingly unimportant $\pi$ electrons
436: at the Fermi level. An appropriately doped version
437: of the material, in which $\pi$ electrons are
438: restored at the Fermi level, however, could result in a promising new
439: superconductor that combines the paradigms of MgB$_2$ and CaC$_6$ in
440: a single material without loss. Further calculations using
441: the virtual crystal approximation or the supercell approximation
442: to explore the quantitative effects of doping on the
443: electron-phonon coupling in this material would be of interest.
444:
445:
446: \acknowledgments
447: We thank L. Boeri, G. Bachelet and O.K. Andersen for
448: useful discussions, and S. Massida and E.K.U. Gross for sharing with us
449: their work on superconductivity in CaC$_6$ prior to publication. One
450: of the authors (I.I.M.) thanks Georgetown University for its
451: hospitality during a sabbatical stay. This work
452: was supported by NSF Grant No. DMR-0210717.
453:
454: \begin{thebibliography}{99}
455: \bibitem{AP} J. M. An and W.E. Pickett, Phys. Rev. Lett. \textbf{86}, 4366
456: (2001).
457:
458: \bibitem{mgb2rev} I.I. Mazin and V.P. Antropov, Physica C, {\bf 385}, 49 (2003).
459:
460: \bibitem{PickettND} W. E. Pickett, cond-mat/0603428.
461:
462: \bibitem{cac6} N. Emery, C. Herold, M. d'Astuto, V. Garcia, Ch. Bellin,
463: J. F. Mareche, P. Lagrange, and G. Loupias,
464: Phys. Rev. Lett. {\bf 95}, 087003 (2005).
465:
466: \bibitem{CM} M. Calandra and F. Mauri, Phys. Rev. Lett. \textbf{95}, 237002
467: (2005).
468:
469: \bibitem{cac6rev} I. I. Mazin,
470: L. Boeri, O.V. Dolgov, A.A. Golubov, G.B. Bachelet, M. Giantomassi,
471: and O.K. Andersen,
472: cond-mat/0606404.
473:
474: \bibitem{Kim} J. S. Kim, L. Boeri, R. K. Kremer, and F. S. Razavi,
475: cond-mat/0603530.
476:
477: %\bibitem{hardy} A. Sanna
478: % G. Profeta, A. Floris, A. Marini, E.K.U. Gross, S. Massidda,
479: %to be published.
480:
481: \bibitem{Curtarolo} A. N. Kolmogorov and S. Curtarolo, Phys. Rev. B \textbf{%
482: 73}, 180501(R) (2006).
483:
484: \bibitem{Curtarolo2} A. N. Kolmogorov and S. Curtarolo, cond-mat/0607654.
485:
486: \bibitem{WIEN} P. Blaha \textit{et al.}, WIEN2k, An Augmented Plane Wave +
487: Local Orbitals Program for Calculating Crystal Properties, Karlheinz
488: Schwarz, Techn. Universit\"{a}t Wien, Austria, 2001
489:
490: \bibitem{PBE} J. P. Perdew, K. Burke, and M. Ernzerhof,
491: Phys. Rev. Lett. \textbf{77}, 3865 (1996).
492:
493: \bibitem{LR} A. A. Quong and B. M. Klein, Phys. Rev. B \textbf{46}, 10734
494: (1992); A. Y. Liu and A. A. Quong, \textit{ibid} \textbf{53}, 7575 (1996).
495:
496: \bibitem{TM} N. Troullier and J. L. Martins, Phys. Rev. B \textbf{43}, 1993
497: (1991).
498:
499: \bibitem{Cs} G. Csanyi, P. B. Littlewood, A. H. Nevidomskyy,
500: C. J. Pickard, and B. D. Simons, Nature Phys. \textbf{1}, 42 (2005).
501:
502: \bibitem{I} I. I. Mazin, Phys. Rev. Lett. \textbf{95}, 227001 (2005)
503:
504: \bibitem{note} In CaC$_{6}$ the $\zeta $ band lies substantially lower than
505: in the isoelectronic LiC$_{3}$ (cf. Ref. \onlinecite{Cs}). This is due entirely to
506: an additional interlayer compression of the $\zeta $ band in LiC$_{3},$ due
507: to a smaller $c/a$ (I.I. Mazin, unpublished).
508:
509: \bibitem{anh} Ref. \onlinecite{Curtarolo} reports a 4\% enhancement of the
510: anharmonic frequency of the $E_{2g}$ phonon in Li$_2$B$_2$ compared to MgB$%
511: _2 $, as opposed to our 21\% enhancement of the harmonic frequency.
512: The frozen-phonon in MgB$_2$ might indeed be more
513: anharmonic due to the smaller Fermi energy of the $\sigma$ holes,
514: but anharmonic calculations are known to be exceptionally sensitive to the
515: k-point sampling. Moreover, it has been shown that the
516: frozen-phonon approach, which includes only one-phonon anharmonic effects,
517: severely overestimates the anharmonicity in MgB$_2$, and
518: that harmonic calculations are
519: closer to the real frequencies [M. Lazzeri, M. Calandra, and F. Mauri,
520: Phys. Rev. B \textbf{68}, 220509(R) (2003)].
521:
522: \bibitem{osv} C. O. Rodriguez, A. I. Liechtenstein, I. I. Mazin, O. Jepsen,
523: O. K. Andersen, and M. Methfessel, Phys. Rev. B {\bf 42}, 2692 (1990).
524:
525: \bibitem{McMillan} W. L. McMillan, Phys. Rev. {\bf 167}, 331 (1968).
526: \end{thebibliography}
527:
528: \end{document}
529: