1: \documentstyle[prl,aps,epsfig,multicol]{revtex}
2:
3: \begin{document}
4: \title{Intercalant-Driven Superconductivity in YbC$_{6}$ and CaC$_{6}$}
5: \author{I.I. Mazin}
6: \address{Center for Computational Materials Science,\\
7: Naval Research Laboratory, Washington, DC 20375}
8: \date{\today }
9: \maketitle
10:
11: \begin{abstract}
12: Recently deiscovered superconductivity in YbC$_6$ and CaC$_6$ at temperatures
13: substantially higher than previously known for intercalated graphites,
14: raised several new questions: (1) Is the mechanism considerably different
15: from the previously known intercalated graphites? (2) If superconductivity
16: is conventional, what are the relevant phonons? (3) Given extreme similarity
17: between YbC$_6$ and CaCa$_6$, why their critical temperatures are so different?
18: We address these questions on the basis of first-principles calculations and
19: conclude that
20: coupling with intercalant phonons is likely to be the main force for
21: superconductivity in YbC$_6$ and CaC$_6$, but not in alkaline-intercalated
22: compounds, and explain the difference in $T_c$ by the ``isotope effect'' due
23: to the difference in Yb and Ca atomic masses.
24: \end{abstract}
25:
26: \pacs{74.25.Jb,74.70.Ad}
27:
28: \begin{multicols}{2}
29:
30: Recent discovery of relatively high temperature superconductivity in
31: graphite intercalated compounds (GIC) YbC$_{6}$ and CaC$_{6}$\cite{exp} of
32: 6.5 and 11.5 K, respectively, the highest among GIC, has renewed theoretical
33: interest in superconductivity in GIC\cite{teor,we}. In particular, it
34: inspired Csanyi $et$ $al$\cite{teor} to analyze four superconducting and
35: three non-superconducting GIC in order to elucidate common trends and get
36: more insight into the mechanism of superconductivity. They discovered an
37: interesting
38: empirical correlation between the occupation of the only 3D band in the
39: system, and the appearence of superconductivity, and, using this
40: observation, they argued that superconductivity in all GIC is
41: electronic by origin, intermediate bosons being probably excitons or acoustic
42: plasmons. This calls for revising the conventional wisdom that
43: superconductivity in GIC is conventional by nature and mostly due to carbon
44: phonons.
45:
46: In this Letter we shall argue, using first principle calculations
47: and experimental data \cite{exp,belash}, that while the standart picture of
48: electron-phonon coupling mainly with the C modes is probably in doubt, at
49: least in these two compounds, superconductivity is likely to arise from the
50: inetercalant vibrations, and not from electronic excitations. In this sense,
51: YbC$_{6}$ and CaC$_{6}$ are somewhat close to another high-T$_{c}$ (18K)
52: transition metal - carbon superconductor, Y$_{2}$C$_{3},$ where
53: superconductivity seems to be related to Y phonons\cite{y2c3}
54:
55: Our analysis is based on highly accurate all electron fully relativistic
56: LAPW calculations \cite{WIEN}. LDA+U correction was applied to the
57: f-electrons in Yb, to account for Hubbard correlations. Details of the
58: calculations for YbC$_{6}$ are described elsewhere\cite{we}. Calculations
59: for CaC$_{6}$ and for other materials discussed below
60: were performed in the same setup as for YbC$_{6},$ but without
61: LDA+U and spin-orbit corrections. For the purpose of comparison, we also
62: performed similar calculations for Li GIC: LiC$_{6}$ and LiC$_{3}.$
63:
64: Let us first discuss the viability of the electronic mechanism scenario\cite%
65: {teor}. This conjecture is based on four assumptions: (1) the 3D free
66: electron like band crosses the Fermi level in all superconducting GIC and
67: is fully empty in all nonsuperconducting ones; (2) this band is not related
68: to intercalant $s$ or $p$ states, but is formed by free electrons propagating in
69: the interstitial space; (3) this band is much weaker coupled with the phonons than
70: the other bands, and (4) such band stracture is advantageous for the excitonic
71: "sandwich" mechanism\cite{sand} or for the acoustic plasmons mechanism\cite{ap}.
72:
73: The first
74: assumption is correct for many, but, apparently, not all GIC. For instance, in LiC$_3$, in
75: pseudotential calculations of Ref. \cite{teor}, the band in question touches
76: the Fermi level. In our fully converged all-electron calculations with a
77: fine k-mesh (13x13x10) this band was 0.2 eV above the Fermi level (Fig. \ref%
78: {bandsLi}). Yet, according to the experiment, superconductivity was observed
79: in this compound\cite{belash}, although because of low temperature and broad
80: transition the authors failed to give the exact number for $T_{c}.$ On the
81: other hand, Eu in EuC$_{6}$ is known to be divalent\cite{SM}, just as Yb or Ca,
82: and forms exactly the same crystal structure, yet the
83: material is not superconducting \cite{MarkE}. Eu in EuC$_{6}$ is magnetic,
84: but, if Eu electrons are not involved in superconductivity,
85: the long coherence length in GIC would have prevented magnetic
86: pair-breaking, as long the material remains well ordered
87: antiferromagnetically ($cf.$, for instance, superconducting antiferromagnetic
88: Chevrel phases).
89: \begin{figure}[tbp]
90: \centerline{\epsfig{file=Li3Bands.eps,width=0.95\linewidth,angle=0,clip=}}
91: \vspace{0.3cm}
92: \caption{LAPW band structure of LiC$_3$.
93: The left panel shows the partial Li character, and the right panel interstitial
94: character. Note uniform participation of the interstitial states in all bands,
95: and selective participation of the Li states in the free electron like band.
96: (color online) }
97: \label{bandsLi}
98: \end{figure}
99:
100:
101: The second assumption is somewhat philosophical,
102: because it is hard to tag an itinerant free-electron like
103: band as an interstitial or as an $sp$ band of
104: an alkali metal. However, decomposition of the wave function of this band
105: shows (for instance,
106: in case of LiC$_{3}$, displayed in Fig. \ref{bandsLi}), that while interstitial
107: plane wave states have the same weight in this band as in the other, 2D
108: states, Li $s$ and $p$ (mostly $p_{z})$ orbitals participate nearly
109: exclusively in this band, and provide much more share of the total weight
110: than the volume occupied by the Li MT spheres. By the standard band theory
111: parlance, this identify them as at least substantially Li-derived.
112: As an independent test, we
113: performed calculations for a hypothetical compound in which the Li atom
114: were replaced by a free electron, and found that the 3D band dispersion
115: changed enormously (Fig.\ref{empty}).
116: \begin{figure}[tbp]
117: \centerline{\epsfig{file=emptycomp.eps,width=0.95\linewidth,angle=0,clip=}}
118: \vspace{0.3cm}
119: \caption{Band structure of LiC$_3$ (thick blue lines) and $e$C$_3$
120: (thin red lines).
121: Note that the most affected band is the free electron like one, specifically,
122: its in-plane dispersion.
123: (color online) }
124: \label{empty}
125: \end{figure}
126:
127: The validity of the third assumption can be tested by direct calculations:
128: one can evaluate the electron-phonon matrix elements at a particular
129: high-symmetry point in the Brillouin zone with a specific phonon by applying
130: a frozen displacement and looking at the induced band splittings. We
131: employed this technique to compute the coupling at the point half-way between
132: $\Gamma$ and A with the ``breathing'' Li phonon, that is, the one
133: corresponding to
134: a breathing displacement of Li along $c$. The results for LiC$_6$
135: are shown in Fig. \ref{phonsLi}. One can see that within $\pm$ 5 eV of the Fermi
136: level the free-electron like band in the one with the {\it
137: strongest } electron-phonon coupling.
138: In view of the noticed in Ref. \cite{teor} high
139: sensitivity of this band to interplane distance, one should also expect
140: a strong coupling with the buckling C modes, but we did not test this numerically.
141:
142: Turning to the "sandwich" mechanism\cite{sand}, we
143: observe that the original papers were strongly based on the idea that the
144: electonic excitations reside in a {\it dielectric }layer (otherwise metallic
145: screening prevents exciton formation), while the interlayer band is
146: {\it metallic} in the well superconducting GIC, as observed in Ref. \cite{teor}.
147: Finally, acoustic plasmons
148: would form in this band either if its effective mass were much heavier than
149: in the other bands, or if it were 2D. Neither condition holds.
150: \begin{figure}[tbp]
151: \centerline{\epsfig{file=phon.eps,width=0.95\linewidth,angle=0,clip=}}
152: \vspace{0.3cm}
153: \caption{LAPW band structure of LiC$_3$. The radii of the filled
154: circles half-way between $\Gamma$ and A are proportional to the
155: electron-phonon interaction matrix elements of the correcponding
156: band with the Li breathing mode.
157: Note that the most affected band is the free electron like one.
158: }
159: \label{phonsLi}
160: \end{figure}
161:
162: Since electronic superconductivity appears to be rather unlikely, we need to find
163: another mechanism. It was conjectured in Ref. \cite{we} that
164: superconductivity in YbC$_{6}$ is largely due to Yb phonons, in analogy with
165: Y$_{2}$C$_{3}$\cite{y2c3}. Comparison between YbC$_{6}$ and CaC$_{6}$ lends
166: additional support to this scenario. Indeed, a detail examination of the two
167: band structures (Fig. \ref{Ca-Yb})
168: finds
169: practically no difference for all but one band in the
170: vicinity of the Fermi level, including the 3D
171: interlayer band. There is some effect of additional hybridization
172: with the f-states on the lowest unoccupied state at the M point,
173: which, however, affects one band out of six, and does not change
174: the density of states near the Fermi level
175: (Fig.\ref{DOS}).
176: If superconductivity were not related to the intercalant
177: atom, one would expect the critical temperature to change only slightly,
178: a situation
179: analogous to YBa$_{2}$Cu$_{3}$O$_{7},$ where Y can be substituted by any
180: trivalent rare-earth with $T_{c}$ changing within a few per cent only. On
181: the contrary, critical temperature of CaC$_{6}$ is 1.77 of that of YbC$_{6}.
182: $
183: \begin{figure}[tbp]
184: \centerline{\epsfig{file=Ca-Yb.eps,width=0.95\linewidth,angle=0,clip=}}
185: \vspace{0.3cm}
186: \caption{Band structure of CaC$_6$ near the Fermi level,
187: compared with that of YbC$_6$. }
188: \label{Ca-Yb}
189: \end{figure}
190: \begin{figure}[tbp]
191: \centerline{\epsfig{file=dos.eps,width=0.95\linewidth,angle=0,clip=}}
192: \vspace{0.3cm}
193: \caption{Density of states of CaC$_6$ near the Fermi level,
194: compared with that of YbC$_6$. (color online) }
195: \label{DOS}
196: \end{figure}
197:
198:
199:
200:
201: At this point we observe that $\sqrt{M_{Yb}/M_{Ca}},$ where $M_{Ca(Yb)}$ is
202: the atomic mass of Ca(Yb), is 2.08. This means that ``isotope effect'' on $%
203: T_{c},$ due to substitution of Yb by Ca, is 1.77/2.08=0.85 of the ``full''
204: isotope effect if superconductivity were entirely due to Yb/Ca modes, and
205: no other differences between the two materials was relevant for superconductivity.
206: Recalling that partial isotope effects in binaries are scaled with partial
207: coupling constants, we find that $\lambda _{R}/(\lambda _{R}+\lambda
208: _{C})\approx 0.85,$ (here $R$ stands for either Ca or Yb), that is, 15\% of
209: the electron-phonon coupling comes from C, and the rest from Ca/Yb. It is
210: more curious than important that the rough estimate given in Ref. \cite{y2c3}
211: for Y$_{2}$C$_{3}$ was 10\% of total coupling coming from C-C phonons, and the
212: rest from
213: pure Y or mixed Y-C modes, in an interesting agreement
214: with the above estimate for YbC$_{6}.$
215:
216: To summarize, we propose that unusually high for intercalated graphites
217: critical temperatures in CaC$_{6}$ and YbC$_{6}$ are mainly due to
218: substantial participation of the intercalant electronic states at the Fermi
219: level, and, as a consequence, sizeable coupling with soft intercalant modes.
220: It remains unclear to what extent the same mechanism is present in other, low $%
221: T_{c}$ GIC, such as KC$_{x},$ LiC$_{x}$ and NaC$_{x}.$ Although their
222: electronic structure shares some similarities with YbC$_{6}/$CaC$_{6},$ it
223: is substantially different, especially regarding intercalant states. It
224: seems unlikely that intercalants in the former are involved in
225: superconductivity nearly as strong as in the latter.
226:
227: Finally, let us discuss what experiments can test the proposed scenario.
228: Measuring isotope effect on Ca is predicted to yield an exponent of the
229: order 0.4, and that on C of 0.1 or less. Another prediction is that mixed
230: intercalation of Ca and Yb should produce samples whose $T_{c}$ scales with
231: concentration as the average logarithmic phonon frequency, that is, as $%
232: T_{Ca}^{x}T_{Yb}^{(1-x)},$ where $x$ is the Ca concentration. An interesting
233: question is, what would be a result of partial substitution of Ca with Mg or
234: Sr? Their ionic radii are substantially different from those of Ca or Yb
235: (which are practically the same in hexagonal coordination). A moderate
236: substitution with, say, Mg will reduce the interplanar distance, thus making
237: Ca-C force constants larger and the coupling constant with electrons for Ca
238: modes smaller. On the other hand, Mg ions themself will sit a pore relatively
239: large for their ionic radius, and thus will have smaller force constant,
240: leading to some increase of $\lambda .$ The third effect is that the
241: corresponding Mg modes will have higher frequency for the same force
242: constants because of smaller mass. If the first two effect approximally
243: cancel each other, co-doping with Mg may be a route to even higher $T_{c}.$
244: Obviously, more experimental and computational work is required to clarify
245: this issue.
246:
247: This was supported by the Office of Naval Research.
248:
249: \begin{references}
250: \bibitem{exp} T. E. Weller, M. Ellerby, S.S. Saxena, R. P. Smith, and N.T.
251: Skipper, cond-mat/0503570, submitted to Nature.
252:
253: \bibitem{teor} G. Cs\'{a}nyi, P. B. Littlewood, A. H. Nevidomskyy, C. J.
254: Pickard, and B. D. Simons, cond-mat/0503569, submitted to Nature.
255:
256: \bibitem{we} I.I. Mazin and S.L. Molodtsov, cond-mat/0503650, submitted
257: to Phys. Rev. B.
258:
259: \bibitem{belash} I.T. Belash, O.V. Zharikov, and A.V. Palnichenko, Synt.
260: Metals, {\bf 34}, 455 (1989); {\it ibid, }{\bf 36,} 283 (302).
261: \bibitem{y2c3} D.J. Singh and I. I. Mazin, Phys. Rev. {\bf B70}, 052504
262: (2004)
263:
264: \bibitem{WIEN} P. Blaha, K. Schwarz, G. K. H. Madsen,D. Kvasnicka and J.
265: Luitz,
266: \emph{Wien2k}, 2002, an Augmented Plane Wave + Local Orbitals Program for
267: Calculating Crystal Properties (Karlheinz Schwarz, Techn. Universitat Wien,
268: Austria), ISBN 3-9501031-1-2.
269:
270: \bibitem{sand} G.F. Zharkov and Y. A. Uspenski, ZhETF {\bf 65}, 1460 (1973)
271: [JETP, {\bf 34}, 1132 (1972)]; {\it ibid.} {\bf 65}, 2511 (1973), D.
272: Allender, J. Bray and J. Bardeen, Phys. Rev. {\bf B7}, 1020 (1973)
273:
274: \bibitem{ap} B.T. Geilikman, Sov. Phys. Uspekhi, {\bf 88}, 327 (1966) [note
275: that a sign error in that paper led to overly optimistic predictions; see
276: Ref.]; E.A. Pashitski, ZhETF, {\bf 64}, 2387 (1968); H. Fr\"{o}hlich, Proc.
277: Phys. Soc. {\bf C1}, 544 (1968).
278:
279: \bibitem{SM} S. L. Molodtsov, C. Laubschat, M. Richter, Th. Gantz, and A. M.
280: Shikin, Phys. Rev. {\bf B53}, 16621 (1996)
281:
282: \bibitem{MarkE} M. Ellerby, private communication.
283: \end{references}
284:
285: \end{multicols}
286:
287: \end{document}
288: