Dumbbells of five-connected silicon atoms and superconductivity in the binary silicides MSi3 (M = Ca, Y, Lu)

Article abstract of DOI:10.1021/ja3055194

The new metastable binary silicides MSi₃ (M = Ca, Y, Lu) have been synthesized by high-pressure, high-temperature reactions at pressures between 12(2) and 15(2) GPa and temperatures from 900(100) to 1400(150) K. The atomic patterns comprise int

Full text of DOI:10.1021/ja3055194

Communication
pubs.acs.org/JACS
Dumbbells of Five-Connected Silicon Atoms and Superconductivity
in the Binary Silicides MSi3 (M = Ca, Y, Lu)
Ulrich Schwarz,* Aron Wosylus,† Helge Rosner, Walter Schnelle, Alim Ormeci, Katrin Meier,
Alexey Baranov, Michael Nicklas, Susann Leipe, Carola J. Muller, and Yuri Grin
̈
Max-Planck-Institut fur Chemische Physik fester Stoffe, 01187 Dresden, Germany
̈
S
* Supporting Information
ABSTRACT: The new metastable binary silicides MSi3
(M = Ca, Y, Lu) have been synthesized by high-pressure,
high-temperature reactions at pressures between 12(2)
and 15(2) GPa and temperatures from 900(100) to
1400(150) K. The atomic patterns comprise intricate
silicon layers of condensed molecule-like Si2 dimers. The
alkaline-earth element adopts the oxidation state +2, while
the rare-earth and transition metals realize +3. All of the
compounds exhibit BCS-type superconductivity with weak
electron−phonon coupling below critical temperatures of
up to 7 K.
he consequent development of in situ high-pressure
Ttechnology has substantially expanded the map of high-
Figure 1. Crystal structure of CaSi3 and the isotypic trisilicides YSi3
temperature superconductors as much as it has stimulated
profound studies of the underlying principles.1 These phases
often expand the spectrum of atomic arrangements in solids by
providing access to configurations with unusual and normally
enhanced coordination numbers. These local atomic environ-
ments can play a decisive role in numerous physical properties.
For example, the expanded coordination spheres of compacted
phases often provide sufficient dispersion of the bands to
induce metallization. In framework compounds, the combina-
tion of homopolar interactions with significant charge-carrier
concentrations gives rise to the character of a covalent metal, a
scenario most favorable for phonon-mediated high-temperature
superconductivity.
and LuSi3 synthesized at elevated pressures and temperatures. Green
spheres indicate silicon atoms, and blue spheres indicate metal atoms.
The connectivity of the silicon framework is indicated by colored lines:
nearest-neighbor distances d(Si1−Si1) = 239.0(1) pm and dSH(Si2−
Si2) = 240.37(8) pm (pink) and longer near-neighbor contacts d(Si1−
Si2) = 258.39(9) pm and dL(Si2−Si2) = 270.37(6) pm (blue). Twelve
M−Si distances per metal atom cover the range from 307.88(2) to
331.80(5) pm for CaSi3.
well-compatible with covalent single bonds. Four significantly
longer contacts aggregate the isolated dimers into intricate
layers that are separated by the calcium atoms. According to the
results of ab initio band structure computations,4 the robust
structure pattern of silicon tolerates a broad variety of metallic
components. Target-oriented subsequent syntheses yielded
isotypic variants such as LuSi3 and YSi3 (see the SI) under
high-temperature, high-pressure conditions similar to those
used to manufacture the calcium prototype.5
The fivefold connectivity of the silicon atoms impeded the
application of the empirical 8 − N rule (where N is the number
of valence electrons) or the elementary Zintl concept6 for
correlating the electron count with the chemical bonding. Thus,
a relevant picture of covalent interactions required elaborate
quantum-mechanical calculations. Herein, quantities operating
directly in real space, such as the electron localizability indicator
(ELI)7 and the charge density, were combined with k-space
tools such as the electronic density of states (DOS).8 The value
of the ELI-D functional9 is proportional to the electronic
A method for accessing the beneficial effects of compression
at ambient pressure is provided by synthesizing metastable
compounds under extreme conditions.2 In this work, we
investigated the potential of silicon-based motifs by combining
efforts on high-pressure, high-temperature preparation and on
total energy calculations, which yielded new framework
compounds MSi3 housing a large variety of metallic elements
(M = Ca, Y, Lu). From the already extensive family of silicon
networks, we first developed the novel compound CaSi3 by
systematically exploring binary metal−silicon mixtures under
high-pressure, high-temperature conditions. The new silicide
was synthesized from CaSi2 and Si(cF8) (Alpha Aesar,
99.9999%) at pressures between 12(2) and 15(2) GPa and
temperatures from 900(100) to 1400(150) K. The crystal
structure motif is isotypic to that of the recently described
CaGe33 and involves two types of Si2 dumbbell species [Figure
1; also see the Supporting Information (SI)]. Within these
diatomic groups, the refined short interatomic distances are
Received: June 7, 2012
Published: July 30, 2012
© 2012 American Chemical Society
13558
dx.doi.org/10.1021/ja3055194 | J. Am. Chem. Soc. 2012, 134, 13558−13561

Journal of the American Chemical Society
Communication
population required for the formation of same-spin electron
pairs around certain points in space. The distribution of the ELI
is spherically symmetric for noninteracting free atoms. In
contrast, local maxima of the ELI, especially in the valence
region, are fingerprints of covalent bonding.10 In CaSi3, the
shapes of the ELI domains around the centers of the Si
dumbbells (Figure 2b,c) clearly indicate covalent two-center
Figure 3. Density of states with partial contributions of the metal and
the silicon atoms as well as the Fermi surfaces of (a) LuSi3 and (b)
CaSi3. The Fermi velocities (indicated by the coloring from blue to red
for increasing values) evidence a reduction of the anisotropy with
increasing metal contributions in proceeding from CaSi3 to LuSi3.
Figure 2. Chemical bonding in CaSi3. (a) Atomic shapes evaluated
applying the QTAIM technique.11 The integration of the electron
density within these volumes yields the electron populations of the
calcium (18.7 e− and 18.5 e−) and silicon (14.4 e− and 14.5 e−) atoms.
(b) ELI-D distributions within the Si2 dumbbells and around the Ca
atoms, with maxima shown in white revealing covalently bonded
silicon dimers. (c) Isosurfaces of high ELI-D values around the Si1 and
Si2 atoms, visualizing the positions and shapes of the maxima
indicating bonding character for the short (pink) Si−Si contacts.
Additional maxima in the region of the long contacts (blue) provide
evidence for covalent Si−Si interactions involving Ca−Si contribu-
tions.
1.4 for CaSi3 and YSi3 and 2.1 for LuSi3, indicated bulk
superconductivity with critical temperatures (Tc) ranging from
4.50(5) K for CaSi3 to 7.1(1) K for LuSi3 (see the SI). For
comparison, electrical resistance data indicated onset temper-
atures of 4.5(3) K for CaSi3 and 7.0(3) K for LuSi3, which
represents the hitherto only superconducting phase in the Lu−
Si system. The bulk character of the superconductivity was
confirmed for all of the compounds by specific heat data
(Figure 4). The specific heat anomalies associated to the
transitions were determined as the differences Δc between the
superconducting-state data cS(T) and the normal-state data
bonding. Integration of the electron density in the regions
assigned to these homopolar interactions yielded 2.1 and 2.2
electrons for the Si1−Si1 and Si2−Si2 contacts, respectively. In
agreement with the refined interatomic distances, this is
consistent with conventional two-center, two-electron bonds.
Interestingly, similar pair formation was observed previously in
a high-pressure modification of elemental silicon, Si(oC16).12
In the regions opposite to the bonds within the dumbbells,
additional maxima were found (Figure 2b,c). For an isolated Si2
pair, such a feature would represent lone-pair-like behavior, as
c
N,fit(T) in the vicinity of Tc. With Sommerfeld coefficients
(γexp) describing the normal-state electronic contribution to the
10
was shown earlier for discrete molecules such as N2 and
Ge2.13 In CaSi3, however, these characteristic domains visualize
multicenter interactions involving the calcium atoms. This
attribution was supported by the partial DOS contributions
(Figure 3), which revealed hybridization of Si 3p and Ca 3d
orbitals, mainly in some of the bands crossing the Fermi energy
and in the energy range between −2 and −1 eV. Integration of
the electron density within the atomic volumes using the
quantum theory of atoms in molecules (QTAIM) technique11
(Figure 2 a) yielded charges of +1.3 and +1.5 for the calcium
atoms and −0.4 and −0.5 for silicon, respectively. These values
are typical for strongly polar framework compounds (e.g.,
Mg1.4+[B0.7− 14) and are well in line with the electronegativity
]
2
differences of the constituent elements.
Figure 4. Specific heat capacity (plotted as cp/T vs T2) for LuSi3 and
YSi3 (vertically shifted by 15 mJ mol−1 K−2) at different magnetic fields
(values in mT). Extrapolation revealed that suppression of the
superconductivity required upper critical magnetic fields μ0Hc2 of
234(5) and 80(3) mT for LuSi3 and YSi3, respectively. The inset
shows the magnetic susceptibility, which displays shielding [zero-field
cooling (zfc)] and the Meissner effect [field cooling (fc)], indicating
superconductivity.
In agreement with the calculated charge transfer, measure-
ments of the physical properties15 revealed diamagnetism of
MSi3 (M = Ca; Y; Lu), indicating oxidation states of +2 for
calcium and +3 for yttrium and lutetium. At low temperatures,
distinctive anomalies in the magnetic susceptibility signaled
transitions into the superconducting state. In small magnetic
fields, the shielded volume fractions, with uncorrected values of
13559
dx.doi.org/10.1021/ja3055194 | J. Am. Chem. Soc. 2012, 134, 13558−13561

Journal of the American Chemical Society
Communication
specific heat, ratios Δc/γexpTc of 1.41 for YSi3 and 1.50 for LuSi3
were calculated. The resulting numbers are in accordance with
the value of 1.43 predicted by the Bardeen−Cooper−Schrieffer
(BCS) theory for weak electron−phonon coupling.
ACKNOWLEDGMENTS
■
We thank Susann Leipe for high-pressure syntheses; Yurii
Prots, Andreas Czulucki, Irene Margiolaki (ESRF), and
Caroline Curfs (ESRF) for synchrotron X-ray diffraction
experiments; Stefan Hoffmann and Susann Scharsach for
differential thermal analysis measurements; and Dr. Steffen
Wirth for competent discussions and critical reading of the
manuscript.
To identify the specific electronic states associated with
superconductivity, the total and partial DOS were analyzed in
detail (Figure 3; also see the SI). In all of the compounds, the
lower part of the valence band (below −6 eV) is dominated by
bonding silicon states, and the two species Si1 and Si2 behave
rather similarly. For YSi3 and LuSi3, the contribution of the
metal atoms is larger for those parts of the valence band that are
closer to the Fermi level, and at EF it is even larger than the Si
contribution per atom. The hybridization of metal d states near
EF with the Si 3p bands indicates important participation in the
superconductivity. This statement holds to a lesser extent for
the related compound CaSi3. Despite the pronounced similarity
of the network-related silicon bands, the participation of metal-
connected d states impedes a simple rigid-band approach. For
substantial d contributions, the calculated total DOS at EF (1.7
states per eV and formula unit) in conjunction with the
experimental Sommerfeld coefficient γexp allows to estimate the
electron−phonon coupling strength. Within the theory of
conventional electron−phonon-coupled superconductivity (a
short systematic discussion is given in a recent study16), the
resulting λ values of 0.6 for YSi3 and 0.65 for LuSi3 are in line
with sizable coupling that facilitates the formation of Cooper
pairs even for moderate densities of states at the Fermi level.
The topology of the bands and Fermi surfaces (Figure 3)
shows for all of the MSi3 compounds considerable anisotropy,
with the dispersion being significantly stronger within the plane
of the layers than perpendicular to them. The ratios of the
computed Fermi velocities parallel and orthogonal to the
covalent units varied from 2.9 in CaSi3 to 1.4 in LuSi3. With the
rather unique situation that the silicon network remains
essentially the same despite the chemical diversity of the
involved metal atoms, the theoretical findings effectively filter
out the relevance of metal d contributions for the direction
dependence of the physical properties. This identification is
considered an essential component for the effective tailoring of
metal−network interactions in structure families of phonon-
driven high-Tc materials. Moreover, the theory-guided realiza-
tion of BCS-type electron−phonon coupling in these emergent
metal−network ensembles underlines that combining high-
pressure synthesis techniques with predictive theoretical
analysis offers significant potential for the materials chemistry
of metastable covalent metals.
REFERENCES
■
(1) (a) Shimizu, K.; Ishikawa, H.; Takao, D.; Yagi, T.; Amaya, K.
Nature 2002, 419, 597−599. (b) Struzhkin, V. V.; Eremets, M. I.; Gan,
W.; Mao, H.-K.; Hemley, R. J. Science 2002, 298, 1213−1215.
(c) Kasinathan, D.; Kunes, J.; Lazicki, A.; Rosner, H.; Yoo, C. S.;
Scalettar, R. T.; Pickett, W. E. Phys. Rev. Lett. 2006, 96, No. 047004.
(d) Medvedev, S.; McQueen, T. M.; Troyan, I. A.; Palasyuk, T.;
Eremets, M. I.; Cava, R. J.; Naghavi, S.; Casper, F.; Ksenofontov, V.;
Wortmann, G.; Felser, C. Nat. Mater. 2010, 8, 630−633. (e) Buzea, C.;
Robbie, K. Supercond. Sci. Technol. 2005, 18, R1−R8. (f) Schilling, J. S.
Physica C 2005, 460−462, 182−185. (g) Shimizu, K.; Kimura, T.;
Furomoto, S.; Takeda, K.; Kontani, K.; Onuki, Y.; Amaya, K. Nature
2001, 412, 316−318. (h) Yabuuchi, T.; Matsuoka, Y.; Shimizu, K. J.
Phys. Soc. Jpn. 2006, 75, No. 083703. (i) Affronte, M.; Sanfilippo, S.;
Nunez-Regueiro, M.; Laborde, O.; LeFloch, S.; Bordet, P.; Hanfland,
M.; Levi, D.; Palenzona, A.; Olcese, G. L. Physica B 2000, 284−288,
1117−1118.
(2) (a) Fukuoka, H.; Yamanaka, S. Phys. Rev. B 2003, 67,
No. 094501. (b) Fukuoka, H.; Tomomitsu, Y.; Inumaru, K. Inorg.
Chem. 2011, 50, 6372−6377. (c) Fukuoka, H.; Suekuni, K.; Onimaru,
T.; Inumaru, K. Inorg. Chem. 2011, 50, 3901−3906. (d) Wosylus, A.;
Meier, K.; Prots, Yu.; Schnelle, W.; Rosner, H.; Schwarz, U.; Grin, Yu.
Angew. Chem., Int. Ed. 2010, 49, 9002−9006.
(3) Schnelle, W.; Ormeci, A.; Wosylus, A.; Meier, K.; Grin, Yu.;
Schwarz, U. Inorg. Chem. 2012, 51, 5509−5511.
(4) FPLO-9.01-34: Koepernik, K.; Eschrig, H. Phys. Rev. B 1999, 59,
1743−1757.
(5) Precursors with composition M:Si = 1:2 were prepared by
inductive heating of the metals with Si(cF8) (Alpha Aesar, 99.9999%)
in sealed tantalum ampoules. Next, 1:1 mixtures of the disilicides and
Si(cF8) were treated at pressures between 12(2) and 15(2) GPa and
temperatures from 900(100) to 1400(150) K, yielding the trisilicides
MSi3 (M = Ca, Y). The synthesis of LuSi3 was performed using excess
silicon, corresponding to a Lu:Si ratio of 1:5. Pressures were generated
within Walker modules encapsulating MgO octahedra. Elevated
temperatures were realized by resistive heating of graphite sleeves
enclosing h-BN sample crucibles. Sample handling (except for the
high-pressure experiments) was performed in argon-filled glove boxes
(MBraun, H2O, O2 < 0.1 ppm). Crystallographic data: Space group
I4/mmm (No. 139), Z = 8. CaSi3: a = 726.76(4) pm, c = 1135.01(7)
pm; Ca1 (0, 0, 0.17384(7)), Uiso = 99(2) pm2; Ca2 (1/2, 0, 1/4), Uiso
= 113(2) pm2; Si1 (0.3356(1), 0, 0), Uiso = 66(2) pm2; Si2
(0.31399(6), x, 0.10589(5)), Uiso = 121(2) pm2; RP = 0.029, Rwp
=
0.036, Rexp = 0.009. YSi3: a = 723.10(3) pm, c = 1079.2(1) pm; Y1 (0,
0, 0.17050(9)), Uiso = 66(3) pm2; Y2 (1/2, 0, 1/4), Uiso = 63(3) pm2;
Si1 (0.3295(3), 0, 0), Uiso = 102(5) pm2; Si2 (0.3141(1), x,
ASSOCIATED CONTENT
■
S
* Supporting Information
Figures S1−S8 and Tables S1−S4. This material is available free
of charge via the Internet at http://pubs.acs.org.
0.1117(1)), Uiso = 110(4) pm2; RP = 0.083, Rwp = 0.115, Rexp
=
0.101. LuSi3: a = 718.35(8) pm, c = 1047.1(1) pm; Lu1 (0, 0,
0.1683(1)), Uiso = 253(2) pm2; Lu2 (1/2, 0, 1/4), Uiso = 254(2) pm2;
Si1 (0.3309(6), 0, 0), Uiso = 270(10) pm2; Si2 (0.3126(2), x,
0.1167(2)), Uiso = 259(9) pm2; RP = 0.054, Rwp = 0.076, Rexp = 0.037.
Details of data collection and refinement and interatomic distances are
given in the SI.
AUTHOR INFORMATION
■
Corresponding Author
schwarz@cpfs.mpg.de
(6) (a) Hume-Rothery, W. Philos. Mag. 1931, 11, 649−678. (b) Zintl,
E.; Brauer, G. Z. Phys. Chem., Abt. B 1933, 20, 245−271. (c) Zintl, E.
Angew. Chem. 1939, 52, 1−6. (d) Klemm, W.; Fricke, H. Z. Anorg. Allg.
Chem. 1955, 282, 162−168. (e) Busmann, E. Z. Anorg. Allg. Chem.
Present Address
†BASF AG, 67056 Ludwigshafen, Germany.
Notes
1961, 313, 90−106. (f) Schafer, H.; Eisenmann, B.; Muller, W. Angew.
̈
̈
The authors declare no competing financial interest.
Chem., Int. Ed. Engl. 1973, 12, 694−712. (g) von Schnering, H. G.
13560
dx.doi.org/10.1021/ja3055194 | J. Am. Chem. Soc. 2012, 134, 13558−13561

Journal of the American Chemical Society
Communication
Angew. Chem., Int. Ed. Engl. 1981, 20, 33−51. (h) Nesper, R. Prog.
Solid State Chem. 1990, 20, 1−45. (i) Chemistry, Structure, and Bonding
of Zintl Phases and Ions; Kauzlarich, S., Ed.; VCH Publishers: New
York, 1996.
(7) (a) Kohout, M. Int. J. Quantum Chem. 2004, 97, 651−658.
(b) Butovskii, M. V.; Doring, Ch.; Bezugly, V.; Wagner, F. R.; Grin,
̈
Yu.; Kempe, Rh. Nat. Chem 2010, 2, 741−744.
(8) Electronic structure calculations and bonding analyses were
carried out for CaSi3, YSi3, and LuSi3 with density functional codes
employing the local density approximation in scalar relativistic
calculations using the tight-binding linear muffin-tin orbital method
within the atomic sphere approximation (TB-LMTO-ASA17), full-
potential local orbital (FPLO-9.01-344) and APW+lo+LO Elk18
program packages. For the TB-LMTO calculations, corrections for
the overlap of atomic spheres were included.19 Basis sets containing
Ca(4s, 3d), Y(5s, 4d), Lu(6s, 5d, 4f), and Si(3s, 3p) orbitals were
selected, with the Ca(4p), Y(5p, 4f), Lu(6p), and Si(3d) functions
being down-folded. In the Elk calculations, an APW+lo+LO basis set
was employed, and the R × Gkmax parameter was set equal to 8, with
the muffin-tin-sphere radii chosen as 1.164 Å for Ca and 0.9525 Å for
Si. The ELI was evaluated in the ELI-D representation9 using ELI
modules within the TB−LMTO−ASA17 and FPLO20 program
packages. The calculations and topological analyses of the ELI-D
distribution based on the Elk code were done on a discrete grid21 with
mesh size of ∼0.0265 Å.
(9) (a) Kohout, M.; Wagner, F. R.; Grin, Yu. Int. J. Quantum Chem.
2006, 106, 1499−1507. (b) Wagner, F. R.; Bezugly, V.; Kohout, M.;
Grin, Yu. Chem.Eur. J. 2007, 13, 5724−5741.
(10) Wagner, F. R.; Kohout, M.; Grin, Yu. J. Phys. Chem. A 2008,
112, 9814−9828.
(11) Bader, R. F. W. Atoms in Molecules: A Quantum Theory.
Clarendon Press: Oxford, U.K., 1990.
(12) (a) Hanfland, M.; Schwarz, U.; Syassen, K.; Takemura, K. Phys.
Rev. Lett. 1999, 82, 1197−1200. (b) Christensen, N. E.; Novikov, D.
L.; Methfessel, M. Solid State Commun. 1999, 110, 615−619.
(c) Schwarz, U.; Jepsen, O.; Syassen, K. Solid State Commun. 2000,
113, 643−648.
(13) Kohout, M.; Wagner, F. R.; Grin, Yu. Theor. Chem. Acc. 2002,
108, 150−156.
(14) Tsirelson, V.; Stash, A.; Kohout, M.; Rosner, H.; Mori, H.; Sato,
S.; Lee, S.; Yamamoto, A.; Tajima, S.; Grin, Yu. Acta Crystallogr. 2003,
B59, 575−583.
(15) The magnetic susceptibility χ(T) = M(T)/H was measured in
external fields μ0H between 2 mT and 7 T at temperatures of 1.8 K−
400 K in a SQUID magnetometer (MPMS XL-7, Quantum Design).
The electrical resistivity ρ(T) was determined by a four-probe method.
Because of the small size of the samples, the uncertainty in ρ was
estimated to be 20%. For CaSi3, additional measurements (1.8−10
K) at a lower current density (j ≈ 8 × 10−3 A mm−2) were performed
in magnetic fields. Heat capacity data for T > 1.9 K were collected with
a relaxation-type method (PPMS, Quantum Design) in fields μ0H
below and above the upper critical field of the superconducting phases.
Measurements on CaSi3 and important physical properties of MSi3 (M
= Ca, Y, Lu) are compiled in the SI.
(16) Pickett, W. E. Physica C 2008, 468, 126.
(17) Jepsen, O., Burkhardt, A., Andersen, O. K. TB-LMTO-ASA,
version 4.7; Max-Planck-Institut fur Festkorperforschung: Stuttgart,
̈
̈
Germany, 1999.
(18) Elk 2011, version 1.2.20; http://elk.sourceforge.net.
(19) Andersen, O. K. Phys. Rev. B 1975, 12, 3060−3083.
(20) Ormeci, A.; Rosner, H.; Wagner, F. R.; Kohout, M.; Grin, Yu. J.
Phys. Chem. A 2006, 110, 1100−1105.
(21) Kohout, M. DGrid, version 4.6; Radebeul, Germany, 2011.
13561
dx.doi.org/10.1021/ja3055194 | J. Am. Chem. Soc. 2012, 134, 13558−13561