1: % INJ
2:
3: \documentclass[epj]{svjour}
4: %\documentclass[epj,referee]{svjour}
5:
6: \usepackage{graphicx}
7:
8:
9:
10: \renewcommand{\d}{\mathop{\rm d}}
11:
12: \begin{document}
13:
14: \title{Symmetry breaking and restoring under high pressure:\\
15: the amazing behaviour of the ``simple'' alkali metals}
16:
17: \author{G. G. N. Angilella \and F. Siringo \and R. Pucci}
18:
19: \titlerunning{Friedel oscillations in alkalis at high pressure}
20:
21: \authorrunning{G. G. N. Angilella \emph{et al.}}
22:
23: \institute{Dipartimento di Fisica e Astronomia, Universit\`a di
24: Catania, and Istituto Nazionale per la Fisica della Materia, UdR
25: di Catania,\\ Via S. Sofia, 64, I-95123 Catania, Italy}
26:
27:
28: \date{\today}
29:
30: \abstract{%
31: We argue that an ionic lattice surrounded by a Fermi liquid changes phase
32: several times under pressure, oscillating between the symmetric
33: phase and a low-symmetry dimerized structure, as a consequence of
34: Friedel oscillations in the pair potential.
35: Phase oscillations explain the tendency towards dimerization which has
36: been recently reported for the light alkali metals under high
37: pressure.
38: Moreover, a restoring of the symmetric phase is predicted for such
39: elements at an even higher density.
40: \PACS{%
41: {71.20.Dg}{Alkali and alkaline earth metals}
42: \and
43: {62.50.+p}{High-pressure and shock-wave effects in solids and liquids}
44: \and
45: {64.70.Kb}{Solid-solid transitions}}}
46:
47: \maketitle
48:
49: \section{Introduction}
50:
51: Some years ago we predicted \cite{Siringo:97} that under high pressure
52: the light alkali metals should be unstable
53: towards less symmetric phases.
54: We also argued that the lowering of symmetry could give rise to a
55: metal-insulator transition.
56: We did not attempt to predict the structure of the high pressure
57: phase, and unfortunately any such theoretical prediction is quite
58: difficult, because of the small energy differences which separate
59: most phases.
60: Our result followed the early suggestion that electron correlation in
61: the alkalis could give rise to broken symmetry of the charge
62: density wave (CDW) type, albeit incommensurate with the lattice
63: \cite{Overhauser:85}.
64: The role of core repulsion in stabilizing the high pressure phases of
65: the alkali had been also emphasized by McMahan
66: \emph{et al.} \cite{Moriarty:82,McMahan:84}.
67:
68: The prediction of a symmetry lowering was quite unexpected, as it is
69: usually believed that an increase of pressure should give rise to
70: an increase in the symmetry of the system.
71: Actually, under very high pressure the most likely state of matter is
72: a uniform plasma, so that any system is expected to raise its
73: symmetry at a sufficiently high pressure.
74: However, many ``simple'' metals are still far from such a limiting
75: behaviour, and their path towards the ultimate metallization may
76: contain oscillations of their degree of symmetry.
77:
78: More recently, both theoretical calculations \cite{Neaton:99} and
79: experimental findings \cite{Hanfland:99,Hanfland:00,Fortov:99} have
80: confirmed the tendency of the light alkali metals to lower their
81: symmetry under high pressure.
82: Low symmetry phases have been observed for Li \cite{Hanfland:00} and
83: Na \cite{Hanfland:private} (see also Ref.~\cite{Neaton:01}), whereas
84: the eventual occurrence of a
85: metal-insulator transition has not been observed yet \cite{Fortov:99}.
86: According to recent calculations \cite{Christensen:01}, both Li and Na
87: show a tendency towards the formation of atomic pairs, and a
88: dimerized $oC$8 structure would be the most stable phase above
89: 165~GPa for Li, and above 220~GPa for Na.
90:
91: The physical reason for dimerization is not evident.
92: \emph{Ab initio} electronic structure calculations \cite{Rousseau:00}
93: indicated a tendency towards ``distance alternation'', due to a
94: sizeable overlap of $p\pi$ orbitals in the interstitial regions.
95: On the other hand, it has been noticed \cite{Christensen:01} that the increase of $s$-$p$
96: hybridization could give rise to a low coordination number.
97: However, even the fully dimerized phase is far from any standard
98: covalent solid: the electron density is uniformly spread and almost
99: constant, while the first and second neighbours distances are
100: comparable.
101: Such a dimerized phase is better described as a charge density wave in a high
102: density metal rather than a molecular solid.
103: Now, at zero pressure both Li and Na are already well described by a
104: simple degenerate Fermi liquid where the kinetic energy is the most
105: relevant energy term \cite{AM}.
106: At high density all the interactions should become smaller and smaller
107: compared to the kinetic energy, and if all the energy terms are
108: assumed to be monotonic functions of density, then nothing relevant
109: is expected to happen.
110: On the other hand, the high density limit is where the Fermi liquid
111: model should work better, thus it remains to be explained why a
112: Fermi liquid should be unstable towards a charge density wave.
113:
114: In this paper we address such a problem and show that an ionic lattice,
115: surrounded by a Fermi liquid, oscillates between a simple metal and a
116: charge density wave, several times with increasing density.
117: We start in Sec.~\ref{sec:structural} by reviewing the evaluation of
118: the structural energy of a solid lattice within linear response theory.
119: In Sec.~\ref{sec:alkali}, we later discuss the case of lithium and
120: other alkali metals under high pressure, where the presence of
121: Friedel oscillations in the screened pair interactions can justify
122: the occurrence of an instability towards a broken-symmetry phase.
123: We eventually summarize in Sec.~\ref{sec:conclusions}.
124:
125: \section{Structural energy of solids close to a phase instability}
126: \label{sec:structural}
127:
128: The basic idea is foreshadowed in the seminal works of Pettifor
129: \emph{et al.}
130: \cite{Pettifor:70,Pettifor:84}, who discussed the possible
131: existence of phase oscillations.
132: The physical motivation is related to Friedel oscillations
133: \cite{Friedel:54}, which characterize the screened interactions and
134: give rise to a non-monotonic oscillating behaviour of the
135: structural energy terms.
136: A dimerization of the lattice can be regarded as a broken-symmetry
137: phase described by an order parameter $y$ which vanishes in the
138: symmetric phase.
139: In most cases, $y$ can be taken to be the difference between first and
140: second neighbours distances, and $y=0$ yields the monoatomic
141: lattice.
142: The total structural energy of the system must be an even function
143: of $y$ and can be expanded as $U(y) \approx U(0) + \frac{1}{2}
144: U^{\prime\prime} (0) y^2$.
145: Its main contribution comes from screened pair interactions which
146: oscillate with a wavevector $q=2k_{\rm F}$, where $k_{\rm F}$ is
147: the Fermi wavevector (Friedel oscillations).
148: Thus the sign of the second derivative $U^{\prime\prime} (0)$ is
149: dominated by the sign of the second derivative of the pair
150: interaction with respect to the pair distance.
151: But an oscillating pair interaction yields oscillating derivatives,
152: so that $U^{\prime\prime} (0)$ is expected to change sign as the
153: pair distance decreases.
154: A positive sign $U^{\prime\prime} (0)>0$ corresponds to having a
155: minimum for $y=0$ (symmetric phase); a
156: negative sign $U^{\prime\prime} (0)<0$ corresponds to having a
157: relative maximum for $y=0$, thus indicating that
158: the symmetric phase is unstable towards a dimerized phase.
159: Moreover, we expect that with increasing density the sign of
160: $U^{\prime\prime} (0)$ could change
161: several times, thus giving rise to phase oscillations between a
162: symmetric and a dimerized lattice.
163:
164: In order to make the above idea more quantitative, let us specialize
165: to the light alkali metals.
166: We start by expressing the total energy per atom of a metal lattice in
167: real-space formulation as \cite{Hafner:87}:
168: \begin{equation}
169: U = U_0 + \frac{1}{2} \sum_{i\neq j} \Phi (R_{ij} ),
170: \label{eq:totenergy}
171: \end{equation}
172: where the pair potential $\Phi (R_{ij} )$ measures the interaction
173: between two ions located at sites $i$ and $j$ in the lattice,
174: $R_{ij}$ being their mutual distance, and $U_0$ summarizes all
175: the contributions independent of the lattice structure.
176: Within second order local pseudopotential theory, the pair potential
177: can be written as \cite{Pettifor:84,Hafner:86}:
178: \begin{equation}
179: \Phi (R) = \frac{Z^2}{R} \left( 1 + \frac{2}{\pi} \int_0^\infty h(q)
180: v^2 (q) \frac{\sin qR}{q} \d q \right),
181: \label{eq:pairpot}
182: \end{equation}
183: where $Z$ is the atomic number,
184: \begin{equation}
185: h(q) = - \frac{\kappa^2}{q^2} \frac{\chi (q)}{\epsilon (q)}
186: = - \kappa^2 \frac{\chi (q)}{q^2 + \kappa^2 [1-G(q)]\chi (q)} ,
187: \end{equation}
188: with $\epsilon(q)$ the dielectric screening function, $\kappa^2 = 4
189: k_{\rm F} / \pi$ is the Thomas-Fermi screening parameter, $\chi(q)$
190: is the normalized Lindhard susceptibility [$\chi(0)=1$], and $G(q)$
191: takes into account for exchange and correlation corrections (local
192: field corrections) to the electron-electron interaction
193: \cite{Ichimaru:81}.
194: In the following, we shall take $Z=1$ for simplicity, which is a good approximation
195: for the light alkali at ambient conditions.
196: In Eq.~(\ref{eq:pairpot}) we assume the `empty core' model for the
197: ionic pseudopotential \cite{Ashcroft:66},
198: \begin{equation}
199: v(q) = \cos q R_c ,
200: \label{eq:pseudopot}
201: \end{equation}
202: where $R_c$ defines the radius of the atomic core.
203: The core radius $R_c$ is usually obtained by fitting
204: Eq.~\ref{eq:pseudopot} against the value of the band gap $2|v({\bf
205: g})|$.
206: We have checked that Eq.~(\ref{eq:pseudopot}) is also consistent with
207: the calculated band structure of lithium both in the bcc
208: \cite{Ching:74} and in the fcc structure \cite{Boettger:85}.
209: However, different sources of experimental data result in slightly
210: different values of $R_c$ for a given element \cite{Cohen:70}.
211: Moreover, the value of $R_c$ can be effectively altered by chemical
212: substitution (e.g. in alloys).
213: Therefore, in the following we will regard $R_c$ as a
214: parameter ranging within given bounds for each element of
215: interest.
216: The other independent parameter of the model is the electron
217: spacing $r_s$, defined as $4\pi r_s^3 / 3 = N/V$, where $N/V$ is
218: the conduction electron density.
219: Such parameter enters Eq.~(\ref{eq:pairpot}) through the Fermi
220: momentum $k_{\rm F}$ in the Thomas-Fermi parameter $\kappa^2$, and
221: can be used as a measure of pressure, with $r_s$ decreasing as
222: pressure increases.
223: Moreover, for a given lattice structure, all inter-site distances
224: $R_{ij}$ in Eq.~(\ref{eq:totenergy}) scale with $r_s$.
225:
226: As a result of the singular behaviour of $h(q)$ at $q=2k_{\rm F}$,
227: which in turn arises from the logarithmic discontinuity in the
228: derivative of the Lindhard function $\chi(q)$, the Fourier
229: transform in Eq.~(\ref{eq:pairpot}) endows the pair potential
230: $\Phi(R)$ with an oscillating behaviour, with a characteristic
231: length $\sim \pi/k_{\rm F}$ (Friedel oscillations, see
232: Fig.~\ref{fig:pairpot}, inset).
233: The analytical properties of the pair potential $\Phi(R)$ have
234: been analyzed further in Ref.~\cite{Hafner:86}.
235: The actual value of the total energy $U$ in Eq.~(\ref{eq:totenergy})
236: then depends on whether the distances of nearest and farther neighbours in
237: the lattice lay close to maxima or minima in the plot of $\Phi(R)$
238: (Fig.~\ref{fig:pairpot}).
239: For a fixed value of the pseudopotential parameter $R_c$, such
240: distances can be actually shifted to lower values by decreasing
241: $r_s$, i.e. by means of an applied pressure.
242: On the basis of such considerations, the stability of
243: the crystal structures of several elements under pressure has been
244: discussed in the past \cite{Hafner:83}.
245:
246: \begin{figure}
247: \centering
248: \includegraphics[height=0.9\columnwidth,angle=-90]{neighbours_rs.ps}
249: \caption{%
250: Scaled pair potential $R\Phi(R)$ versus interatomic distance $R$.
251: \underline{\sl Main plot:} $R_c = 1.76$~a.u., $r_s = 1.5 -
252: 3.326$~a.u., with thicker line corresponding to $r_s = 3.326$~a.u.
253: Symbols correspond to nearest and farther neighbours locations in the
254: bcc (open circles), fcc (full circles), and hcp (triangles)
255: lattices, respectively.
256: \underline{\sl Inset:} Same as main plot for $R_c = 1.76$~a.u., $r_s =
257: 1.5$~a.u., but now with distances scaled by $k_{\rm F} /\pi$.
258: Notice that minima in $R\Phi(R)$ occur close to integer values of
259: $k_{\rm F} R/\pi$ (Friedel oscillations).
260: }
261: \label{fig:pairpot}
262: \end{figure}
263:
264:
265:
266: \section{The case of lithium and other light alkali metals under high
267: pressure}
268: \label{sec:alkali}
269:
270: The phase diagram of the alkali at low pressure has been extensively
271: addressed both experimentally and theoretically, with lithium
272: crystallizing in the 9R phase at zero temperature and pressure
273: \cite{Boettger:85,Lin:86,Holzapfel:96,Liu:99}.
274: At high pressure ($\sim 39$~GPa), a new structural phase (Pearson
275: symbol $cI$16) has been recently detected \cite{Hanfland:00}.
276: Hanfland \emph{et al.} \cite{Hanfland:private} have recently observed
277: bcc~$\to$~fcc~$\to$~$cI$16 transitions in Na.
278: Thus we can regard the $cI$16 phase as our starting point for the
279: following discussion on alkali metals.
280: Such a high-pressure phase is characterized by a bcc primitive cell,
281: with an 8-atom basis \cite{Christensen:01}, and can be thought of
282: as a distorted bcc phase, with a distortion parameter $x=0.045 -
283: 0.060$ \cite{Hanfland:00}.
284: The $cI$16 phase formally reduces to the bcc structure (`supercell' with
285: eight usual cubic cells) for $x=0$ \cite{Christensen:01}.
286: In the undistorted bcc phase, the lattice is composed of two
287: interpenetrating cubic sublattices, $A$ and $B$, say.
288: Such classification applies to the $cI$16 structure as well, after
289: distortion from the parent bcc lattice, but now with a basis of
290: four atoms for each sublattice.
291: The $cI$16 phase by itself is not dimerized, and it has been recently
292: predicted to be even superconducting \cite{Christensen:01a,Shimizu:02a}.
293: We would like to
294: test its stability with respect to a dimerized phase obtained
295: by rigidly shifting the two sublattices each other
296: of a tiny amount $y$ along the $(111)$ direction.
297: Here, $y$
298: plays the role of the order parameter discussed above.
299: For a fixed value of the distortion parameter $x$, one can then think
300: of expanding the total energy per atom Eq.~(\ref{eq:totenergy}) in
301: powers of our `dimerization' parameter $y$,
302: \begin{equation}
303: U[x,y] = U_0 + U_{y} [x,0] y + \frac{1}{2} U_{yy} [x,0] y^2 +
304: O(y^3 ),
305: \end{equation}
306: where $U_y = \partial U /\partial y$ etc.
307: Due to crystal symmetry, it can be proved analytically that $U_{y}
308: [0,0] = 0$ exactly, whereas we numerically checked that $|U_y [x,0]
309: /U_{yy} [x,0] | \ll 1$, for $x\ll 1$ \cite{note1}.
310: Therefore, an indication of instability towards `dimerization' is
311: provided by the condition $U_{yy} [x,0] < 0$.
312: From Eqs.~(\ref{eq:totenergy}) and (\ref{eq:pairpot}), $U_{yy} [x,0]$
313: can be expressed as
314: \begin{eqnarray}
315: U_{yy} [x,0] &=& \frac{1}{4} \sum_{{\bf n}\mu\nu}
316: \left[
317: \left( \Phi^{\prime\prime} (R^0_{{\bf n}\mu\nu} ) - \frac{1}{R^0_{{\bf
318: n}\mu\nu}} \Phi^\prime (R^0_{{\bf n}\mu\nu} ) \right)
319: \right.\times \nonumber\\
320: &&
321: \left.
322: \left( \frac{\partial R^0_{{\bf n}\mu\nu}}{\partial y} \right)^2 +
323: \frac{3a^2}{R^0_{{\bf n}\mu\nu}} \Phi^\prime (R^0_{{\bf n}\mu\nu} )
324: \right],
325: \label{eq:derivata}
326: \end{eqnarray}
327: where $\vec{n}$ labels lattice points in the primitive bcc lattice,
328: with lattice parameter $a$, $\mu$ and $\nu$ label the basis
329: vectors in sublattice $A$ and $B$, respectively
330: \cite{Christensen:01}, and $R^0_{{\bf
331: n}\mu\nu}$ denotes the mutual distances between lattice points for
332: $y=0$.
333: Eq.~(\ref{eq:derivata}) has been evaluated on a finite lattice, large
334: enough to reach full convergence.
335: Figure~\ref{fig:phasediag} displays our numerical results for $U_{yy}
336: [x,0]$ as a function of parameters $(R_c , r_s )$, for $x=0$
337: (undistorted bcc phase) and $x=0.05$ (representative value of the
338: high-pressure $cI$16 phase
339: according to Refs.\cite{Hanfland:00,Christensen:01}).
340: \begin{figure}
341: \centering
342: \includegraphics[bb=84 120 443 619,clip,width=0.7\columnwidth]{phasediag.ps}
343: \caption{%
344: Shaded regions in the plane of parameters $(R_c , r_s )$ are
345: characterized by the condition $U_{yy} [x,0] < 0$, which
346: signals an instability towards dimerization.
347: The regions corresponding to instability towards dimerization in the
348: undistorted bcc phase ($x=0$) and in the $cI$16 phase ($x=0.05$) are
349: bounded by continuous and broken lines, respectively.
350: Data for $(R_c , r_s )$ pairs for the light alkali metals at ambient
351: pressure have been taken from
352: Refs.~\protect\cite{Hafner:83,Cohen:70}.
353: Different methods of fitting pseudopotentials to the experimental data
354: result in a rather wide range for $R_c$, which is here displayed as
355: an error bar.
356: }
357: \label{fig:phasediag}
358: \end{figure}
359:
360: In the plane of parameters $(R_c, r_s)$ we find regions of instability
361: towards dimerization where $U_{yy}[x,0]<0$
362: (Fig.~\ref{fig:phasediag}, shaded areas).
363: It is remarkable that at ambient pressure all the alkali metals are
364: predicted to be stable in the symmetric phase.
365: As the core radius $R_c$ is related to the band gap, one would in
366: general expect $R_c$ to be a (monotonic) function of the density
367: parameter $r_s$.
368: More generally, a more quantitative analysis would require a non-local
369: pseudopotential, with a density-dependent range in reciprocal
370: space.
371: However, within linear response
372: theory, the long wavelength behaviour of the pseudopotential is
373: what actually matters.
374: Indeed, the first zero of Eq.~\ref{eq:pseudopot} occurs at $q=\pi/2
375: R_c$, which is very close to the smallest reciprocal lattice vector
376: of most metals.
377: Therefore, at the level of approximation implied by the present
378: calculation, we can safely neglect the density dependence of $R_c$,
379: without qualitatively alter our main conclusions.
380: In this way, the effect of applied pressure on the phase point
381: corresponding to each alkali metal in Fig.~\ref{fig:phasediag} can
382: be tracked as a vertical shift at constant $R_c$, with increasing
383: density.
384: However, the error bars for $R_c$ under normal conditions grossly provide
385: order of magnitude boundaries for the generally expected variation
386: of $R_c$ with density.
387:
388: A sign change in $U_{yy}$ for a given alkali metal then corresponds to
389: a critical density.
390: The uncertainty on $R_c$ lends a wide window for the critical
391: densities of Li and Na: the largest values (still compatible with
392: the core radius data of Refs.~\cite{Hafner:83,Cohen:70}) are $r_s
393: \sim 2 - 2.2$~a.u., which are not too far from the values observed by
394: Hanfland \emph{et al.} \cite{Hanfland:00} for the onset of a dimerized
395: phase for Li ($r_s\approx 2.1$~a.u. and $P=165$~GPa) and predicted by
396: Christensen and Novikov \cite{Christensen:01} for the onset of a
397: dimerized phase for Na ($r_s\approx 2.3$~a.u. and $P=220$~GPa).
398: However, we caution that such a critical density occurs for $r_s \sim
399: R_c$, \emph{i.e.} close to the limits of applicability of the
400: empty-core pseudopotential approximation, thus suggesting that more
401: refined pseudopotentials should be employed in more realistic
402: calculations.
403: The phase diagram of Fig.~\ref{fig:phasediag} also suggests that, with
404: a further increase of density, both Li and Na should drop below the
405: instability area, and a second transition would eventually restore
406: the symmetry of the lattice.
407: This oscillating behaviour, from a symmetric phase to a less symmetric
408: one and back to a restored symmetric phase, seems to be a general
409: effect due to the presence of Friedel oscillations in the
410: interatomic distance dependence of the pair potential.
411: Such oscillations in turn are a consequence of the existence of a
412: Fermi surface at $k=k_{\rm F}$.
413: Thus the tendency towards the formation of atomic pairs would be a
414: signature of the Fermi liquid behaviour of the alkali metals.
415:
416: Our method relies on linear response theory, which contains the
417: Thomas-Fermi approximation as a long wavelength limit
418: \cite{note:Hafner}.
419: Both such methods are known to be reliable if the density gradients
420: are not too strong.
421: In large atoms, the Thomas-Fermi method yields results which only
422: differ by 8\% from Hartree-Fock calculations, in space regions
423: where the density gradient ${\nabla \rho/\rho}\approx 2.5/{a_0}$
424: ($a_0$ is the Bohr radius) \cite{note:Slater}.
425: Here, the typical density gradients are ${\nabla \rho/\rho}<
426: 0.5/{a_0}$, as reported by Ref.~\cite{Christensen:01} for the high
427: pressure phase of Li at 165~GPa.
428: Thus, we expect linear response to be reliable in the high pressure
429: range, where eventual second order corrections are very small (for
430: these gradients even the simple Thomas-Fermi approximation would
431: deviate less than 1\% from Hartree-Fock calculations).
432:
433: \section{Conclusions}
434: \label{sec:conclusions}
435:
436: In summary, we have shown that a distorted bcc lattice ($cI16$ phase),
437: surrounded by
438: a Fermi liquid, undergoes several structural transitions under high
439: pressure, oscillating between a symmetric phase and a
440: broken-symmetry dimerized phase.
441: These phase oscillations seem to be a direct consequence of the non
442: monotonic behaviour of the pair potential which is characterized by
443: a Friedel wavelength $\pi/k_{\rm F}$ in the presence of a sharp Fermi
444: sphere.
445: Such findings could be relevant for the understanding of the tendency
446: towards the formation of atomic pairs, which has been recently
447: reported for the light alkali metals.
448: We furthermore predict that a restoring of symmetry should take place
449: at some stage under higher pressure, which would be signalled by a
450: reentrant metallic character.
451:
452:
453: \begin{acknowledgement}
454: We thank N. W. Ashcroft, P. Ballone, N. H. March, G. Piccitto,
455: P. S. Riseborough, K. Syassen for useful discussions and
456: correspondence.
457: \end{acknowledgement}
458:
459:
460: \bibliographystyle{mprsty}
461:
462: \bibliography{a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u,v,w,x,y,z,zzproceedings,Angilella,notes}
463:
464: \end{document}
465:
466: