1: \documentclass[prl,twocolumn,showpacs,amsmath]{revtex4}
2: \usepackage{bm}
3: \newcommand{\eps}{\varepsilon}
4: \newcommand{\rmd}{\mathrm d}
5: \newcommand{\rmi}{\mathrm i}
6: \newcommand{\mat}[1]{\tensor{#1}}
7: \newcommand{\bracket}[1]{\langle #1 \rangle}
8: \newcommand{\Bra}[1]{\Bigl \langle #1 \Bigr \rvert}
9: \newcommand{\Ket}[1]{\Bigl \lvert #1 \Bigr \rangle}
10: \newcommand{\ket}[1]{\lvert #1 \rangle}
11: \newcommand{\dpar}[2]{\frac{\partial #1}{\partial #2}}
12: \newcommand{\dtot}[2]{\frac{\rmd #1}{\rmd #2}}
13:
14: \begin{document}
15:
16: \title{Berry phase correction to electron density of states in solids}
17: \author{Di Xiao}
18: \affiliation{%
19: Department of Physics,
20: The University of Texas,
21: Austin, TX 78712-0264, USA}
22: \author{Junren Shi}
23: \affiliation{%
24: Department of Physics,
25: The University of Texas,
26: Austin, TX 78712-0264, USA}
27: \author{Qian Niu}
28: \affiliation{%
29: Department of Physics,
30: The University of Texas,
31: Austin, TX 78712-0264, USA}
32:
33: \date{February 14, 2005}
34:
35: \begin{abstract}
36: Liouville's theorem on the conservation of phase space volume is
37: violated by Berry phase in the semiclassical dynamics of Bloch
38: electrons. This leads to a modification of the phase space density of
39: states, whose significance is discussed in a number of examples: field
40: modification of the Fermi-sea volume, connection to the anomalous Hall
41: effect, and a general formula for orbital magnetization. The
42: effective quantum mechanics of Bloch electrons is also sketched, where
43: the modified density of states plays an essential role.
44: \end{abstract}
45:
46: \pacs{73.43.-f, 72.15.-v, 75.20.-g}
47:
48: \maketitle
49:
50: Semiclassical dynamics of Bloch electrons in external fields has
51: provided a powerful theoretical framework to account for various
52: properties of metals, semiconductors and insulators
53: \cite{ashcroft1976}. In recent years, it has become increasingly
54: clear that essential modification of the semiclassical dynamics is
55: necessary for a proper understanding of a number of phenomena. It was
56: known earlier that global geometric phase effects
57: \cite{berry1984,gphase2003} on Bloch states are very important for
58: insulators in our understanding of the quantum Hall effect
59: \cite{thouless1982}, quantized adiabatic pumps \cite{thouless1983},
60: and electric polarization \cite{kingsmith1993,resta1994}. It was
61: shown \cite{chang1996,sundaram1999} later that geometric phase also
62: modifies the local dynamics of Bloch electrons and thus affects the
63: transport properties of metals and semiconductors. Recently these
64: ideas have been successfully applied to the anomalous Hall effect in
65: ferromagnetic semiconductors and metals \cite{jungwirth2002, fang2003,
66: yao2004,haldane2004}, as well as spin transport
67: \cite{murakami2003,sinova2004}.
68:
69: In this Letter, we reveal a general property of the Berry phase
70: modified semiclassical dynamics which has been overlooked so far: the
71: violation of Liouville's theorem for the conservation of phase space
72: volume. Liouville's theorem was originally established for standard
73: classical Hamiltonian dynamics, and its importance cannot be over
74: emphasized as it serves as a foundation for classical statistical
75: physics. The Berry phase makes, in general, the equations of motion
76: non-canonical~\cite{kuratsuji1988, littlejohn1991, chang1996,
77: sundaram1999, panati2003}, rendering the violation of
78: Liouville's theorem. Nevertheless, we are able to remedy the
79: situation by modifying the density of states in the phase space.
80:
81: This modified phase-space density of states enters naturally in the
82: semiclassical expression for the expectation value of physical
83: quantities, and has profound effects on equilibrium as well as
84: transport properties. We demonstrate this with several examples.
85: First, we consider a Fermi sea of electrons in a weak magnetic field,
86: and show that the Fermi sea volume can be changed linearly by the
87: field. Second, we show how the Berry phase formula for the intrinsic
88: anomalous Hall conductivity may be derived from equilibrium
89: thermodynamics using the St{\v r}eda formula \cite{streda1982}.
90: Third, we provide a general derivation of an orbital-magnetization
91: formula which is convenient for first-principles calculations.
92:
93: In addition, we present an effective quantum mechanics for Bloch
94: electrons in solids by quantizing the semiclassical dynamics with the
95: geometric phase. The density of states enters in a nontrivial manner
96: into the commutators of the phase space coordinates, and relates
97: directly to the minimal uncertainty volume in the phase space.
98:
99: To begin with, we write down the semiclassical equations of motion for
100: a Bloch electron in weak electric and magnetic fields
101: \cite{sundaram1999}
102: \begin{subequations}
103: \label{EOM}
104: \begin{align}
105: \dot{\bm r} &= \frac{1}{\hbar} \dpar{\eps_n(\bm k)}{\bm k}
106: - \dot{\bm k} \times \bm \Omega_n(\bm k) \;, \\
107: \hbar \dot{\bm k} &= -e \bm E(\bm r) - e\dot{\bm r}
108: \times \bm B(\bm r) \;,
109: \end{align}
110: \end{subequations}
111: where $\bm \Omega_n(\bm k)$ is the Berry curvature of electronic Bloch
112: states defined by $\bm \Omega_n(\bm k) = \rmi \bracket{\bm \nabla_{\bm
113: k} u_n(\bm k)|\times|\bm \nabla_{\bm k} u_n(\bm k)}$ with
114: $\ket{u_n(\bm k)}$ being the periodic part of Bloch waves in the $n$th
115: band, $\eps_n(\bm k)$ is the band energy with a correction due to the
116: orbital magnetic moment [see Eq.~\eqref{energy} and above]. For
117: crystals with broken time-reversal symmetry (such as ferromagnetic
118: materials) or spatial inversion symmetry (such as GaAs), the Berry
119: curvature $\bm\Omega_n(\bm k)$ is nonzero.
120:
121: To show the violation of Liouville's theorem, we consider the
122: time evolution of a volume element $\Delta V = \Delta \bm r\Delta \bm
123: k$ in the phase space. The equation of motion for $\Delta V$ is given
124: by $(1/\Delta V) \rmd \Delta V / \rmd t = \bm \nabla_{\bm r} \cdot
125: \dot{\bm r} + \bm \nabla_{\bm k} \cdot \dot{\bm k}$~\cite{reichl1980}.
126: A straightforward but somewhat tedious calculation shows that the
127: right hand side is equal to $-\rmd \ln (1 + e\bm B \cdot \bm
128: \Omega/\hbar)/\rmd t$, which is a total time derivative. Therefore we
129: can solve for the time evolution of the volume element and obtain
130: \begin{equation}
131: \Delta V = \Delta V_0/(1 + e \bm B \cdot \bm \Omega_n / \hbar) \;.
132: \label{volume}
133: \end{equation}
134: The fact that the Berry curvature is generally $\bm k$ dependent (and
135: the magnetic field can also depend on $\bm r$) implies that the phase
136: space volume element changes during time evolution of the state
137: variables $(\bm r, \bm k)$.
138:
139: Nevertheless, we have a remedy to this breakdown of Liouville's
140: theorem. Equation~\eqref{volume} shows that the volume element is a
141: local function of the state variables (through the magnetic field and
142: the Berry curvature) and has nothing to do with the history of time
143: evolution. We can thus introduce a modified density of states
144: \begin{equation}
145: D_n(\bm r, \bm k) = (2\pi)^{-d}
146: (1 + e \bm B \cdot \bm \Omega_n / \hbar) \;,
147: \label{DOS}
148: \end{equation}
149: such that the number of states in the volume element, $D_n(\bm r, \bm
150: k) \Delta V$, remains constant in time, where $d$ is the spatial
151: dimensionality of the system. The prefactor $(2\pi)^{-d}$ is obtained
152: by demanding that the density of states $D_n(\bm r, \bm k)$ reduces to
153: the conventional form when the Berry curvature vanishes. As will be
154: shown later, this density of states corresponds to the minimal quantum
155: uncertainty volume of the state variables. Therefore, it does serve
156: as the semiclassical measure for the number of quantum states per unit
157: volume in the phase space. Based on this understanding, we write the
158: classical phase-space probability density as
159: \begin{equation}
160: \rho_n(\bm r, \bm k, t) = D_n(\bm r, \bm k) f_n(\bm r, \bm k, t)\; ,
161: \label{rho}
162: \end{equation}
163: with $f_n(\bm r, \bm k, t)$ being the occupation number of the state
164: labeled by $(\bm r, \bm k)$. Probability conservation demands that
165: $\rho_n(\bm r, \bm k, t)$ satisfies the continuity equation in phase
166: space. On the other hand, our density of states satisfies $\rmd D_n /
167: \rmd t = -(\bm \nabla_{\bm r} \cdot \dot{\bm r} + \bm \nabla_{\bm k}
168: \cdot \dot{\bm k})D_n$. It then follows that the occupation number
169: introduced above has the desired property of being invariant along the
170: trajectory, i.e., $\rmd f_n /\rmd t=0$~\cite{boltzmann}.
171:
172: We can thus write the real space density of a physical observable
173: $\hat O$ in the form~\cite{culcer2004}
174: \begin{equation}
175: \bar{O}(\bm R) = \sum_n \int \rmd\bm k D_n(\bm r, \bm k)
176: f_n(\bm r, \bm k, t)
177: \langle \hat{O}\delta(\hat{\bm r}-\bm R) \rangle_{\bm r \bm k n}
178: \label{average_general}
179: \end{equation}
180: where $\langle \cdots \rangle_{\bm r \bm k n}$ denotes the expectation
181: value in the wave-packet state centered at ($\bm r$, $\bm k$) with the
182: band index $n$. In the spatially homogeneous case, it reduces to:
183: \begin{equation}
184: \bar{O} = \sum_n \int \rmd\bm k D_n(\bm k)
185: f_n(\bm k) O_n(\bm k) \;,
186: \label{average}
187: \end{equation}
188: where $O_n(\bm k)$ is the expectation value of $\hat O$ in a Bloch
189: state. For simpler notation, we will drop the band index $n$ and
190: assume that the integral over $\bm k$ includes the sum over $n$.
191:
192: We now discuss the magnitude of the correction term $e\bm B \cdot \bm
193: \Omega/\hbar$ to the density of states in Eq.~\eqref{DOS}. The Berry
194: curvature for several materials has been calculated before using
195: first-principles method~\cite{fang2003,yao2004}. Over large regions
196: of the Brillouin zone, its magnitude is on the order of $a^2$ with $a$
197: being the lattice constant. Thus, $e\bm B\cdot\bm\Omega/\hbar\sim
198: eBa^2/\hbar$ is the ratio of the magnetic flux through a unit cell to
199: the magnetic flux quantum, and can be $10^{-2}$ to $10^{-3}$ for a
200: magnetic field of 1 tesla. In the vicinity of some isolated points,
201: the Berry curvature can be several orders of magnitude higher, leading
202: to bigger effects for measurement. In the following, we will present
203: a number of applications of our formula Eq(6).
204:
205: In our first example, we consider the quantity of electron density and
206: show that the Fermi sea volume can be changed linearly by a magnetic
207: field when the Berry curvature is non-zero. Assuming zero temperature
208: and using Eq.~\eqref{DOS}, we have the electron density as
209: \begin{equation}
210: n_e = \int^\mu
211: \frac{\rmd\bm k}{(2\pi)^d}
212: \left(1+\frac{e\bm B \cdot \bm\Omega}{\hbar} \right) \;,
213: \label{density}
214: \end{equation}
215: where the upper limit means that the integral is over states with
216: energies below the chemical potential $\mu$. Noting that the electron
217: density is fixed by the background charge density, we conclude that
218: the Fermi volume must change with the magnetic field. To first
219: order, this change is given by
220: \begin{equation}
221: \delta V_F = - \int^{\mu_0}
222: \rmd\bm k\,\frac{e\bm B\cdot \bm \Omega}{\hbar} \;.
223: \label{FermiV}
224: \end{equation}
225: We note that while Landau levels make the Fermi sea volume oscillate
226: with the field, the effect described above gives an overall shift on
227: average. Such a shift has important implications to those
228: Fermi-surface related behaviors such as transport properties. For
229: instance, in metals, it can induce a magnetoresistance linearly
230: depending on the magnetic field. On the other hand, in band
231: insulators, the $\bm k$ space is limited to the Brillouin zone.
232: Electrons must populate a higher band if $(e/\hbar)\int_\text{BZ}
233: \rmd\bm k\, \bm B\cdot \bm \Omega$ is negative. When this quantity is
234: positive, holes must appear at the top of the valence bands.
235: Discontinuous behavior of physical properties in a magnetic field is
236: therefore expected for band insulators with non-zero integral of the
237: Berry curvatures (Chern numbers).
238:
239: In our second example, we show a connection between our phase space
240: density of states to the intrinsic anomalous Hall effect, which is due
241: to spin-orbit coupling in the band structure of a ferromagnetic
242: crystal. In the context of the quantum Hall effect, St{\v r}eda
243: derived a formula relating the Hall conductivity to the field
244: derivative of the electron density at a fixed chemical potential
245: \cite{streda1982}, $\sigma_{xy}=-e(\partial n_e/\partial B_z)_\mu$.
246: There is a simple justification of this relation by a thermodynamic
247: argument by considering the following adiabatic process in two
248: dimensions. A time dependent magnetic flux generates an electric field
249: with an emf around the boundary of some region; and the Hall current
250: leads to a net flow of electrons across the boundary and thus a change
251: of electron density inside. This argument can be straightforwardly
252: applied to the case of anomalous Hall effect and to three dimensions.
253: By taking the derivative of the electron density~\eqref{density} with
254: respect to $\bm B = B\hat{\bm z}$ at fixed chemical potential, we find
255: that
256: \begin{equation}
257: \sigma_{xy} = -\frac{e^2}{\hbar}
258: \int^\mu \frac{\rmd\bm k}{(2\pi)^d} \Omega_z \;.
259: \label{sigma}
260: \end{equation}
261: This is an intrinsic effect because it is independent of scattering,
262: and thus differs from conventional skew scattering and side jump
263: mechanisms~\cite{jungwirth2002,fang2003,yao2004,haldane2004}.
264:
265: As a third example of application, we now derive a semiclassical
266: formula for orbital magnetization. In the semiclassical picture, a
267: Bloch electron is modeled by a wave packet in a Bloch band, which is
268: found to rotate about its center of mass in general, yielding an
269: intrinsic magnetic moment given by $\bm m(\bm k) = -\rmi (e/2\hbar)
270: \bracket{\bm \nabla_{\bm k}u| \times [\hat{H}_0(\bm k) - \eps_0(\bm
271: k)]|\bm \nabla_{\bm k}u}$, where $\hat{H}_0$ is the
272: Hamiltonian~\cite{m-moment}. In the presence of a weak magnetic field
273: $\bm B$, the electron band structure energy $\eps_0(\bm k)$ (which may
274: already include Zeeman energy from spin magnetization) acquires a
275: correction term from this intrinsic orbital moment
276: \cite{chang1996,sundaram1999}, $\eps(\bm k) = \eps_0(\bm k) - \bm
277: m(\bm k) \cdot \bm B$. For an equilibrium ensemble of electrons, the
278: total orbital magnetization can be found from the total energy, which
279: is given by Eq.~\eqref{average} as,
280: \begin{equation}
281: E = \int^\mu \frac{\rmd\bm k}{(2\pi)^d}
282: \Bigl(1 + \frac{e \bm B \cdot \bm \Omega}{\hbar} \Bigr)
283: \Bigl(\eps_0(\bm k) - \bm m(\bm k) \cdot \bm B \Bigr) \;.
284: \label{energy}
285: \end{equation}
286: Taking differential of $E$ with respect to $\bm B$, we obtain the
287: magnetization at zero magnetic field to be
288: \begin{align}
289: \bm M &= \int^{\mu_0} \frac{\rmd\bm k}{(2\pi)^d} \Bigl(\bm m(\bm k)
290: + \frac{e\bm \Omega}{\hbar}\bigr[\mu_0 - \eps_0(\bm k)\bigr] \Bigr) \\
291: &= \frac{e}{2\hbar} \int^{\mu_0} \frac{\rmd \bm k}{(2\pi)^d} \rmi
292: \Bra{\dpar{u}{\bm k}} \times [2\mu_0 - \eps_0(\bm k) - \hat{H}_0]
293: \Ket{\dpar{u}{\bm k}} \;. \nonumber
294: \label{magnetization}
295: \end{align}
296: In the upper line of the above expression, the first term is the
297: contribution from the intrinsic orbital moment of each Bloch electron,
298: and the second term comes from the explicit field dependence of the
299: density of states and the resulting change in the Fermi volume in
300: Eq.~\eqref{FermiV}. We expect this effect to be important in
301: ferromagnetic materials with strong spin-orbit coupling.
302:
303: Gat and Avron obtained an equivalent result for the special case of
304: Hofstadter model~\cite{gat2003}. Our derivation provides a more
305: general formula that is applicable to other systems. Following the
306: discussions on band insulators in our first example, there will be a
307: discontinuity of the orbital magnetization if the integral of the
308: Berry curvature over the Brillouin zone, or the anomalous Hall
309: conductivity, is non-zero and quantized. Depending on the direction of
310: the field, the chemical potential $\mu_0$ in the above formula should
311: be taken at the top of the valence bands or the bottom of the
312: conduction bands. The size of the discontinuity is given by the
313: quantized anomalous Hall conductivity times $E_g/e$, where $E_g$ is
314: the energy gap. For insulators with zero Chern numbers, the orbital
315: magnetization can be directly evaluated from Wannier functions, with
316: result consistent with our general formula \cite{resta2005}. Our
317: general formula can also be derived from a full quantum mechanical
318: linear response analysis \cite{shi2005}.
319:
320: The central result of this paper, equation~\eqref{DOS}, can be
321: extended to the more general case when Berry curvature includes the
322: components of $\mat{\Omega}^{\bm k \bm r}$ as well as $\mat{
323: \Omega}^{\bm k\bm k}$ and $\mat{\Omega}^{\bm r \bm
324: r}$~\cite{sundaram1999}. In this case, we introduce the Berry
325: curvature in phase space,
326: \begin{equation}
327: \mat{\bm \Omega} = \begin{pmatrix} \mat{\Omega}^{\bm r\bm r} &
328: \mat{\Omega}^{\bm r\bm k} \\ \mat{\Omega}^{\bm k\bm r} &
329: \mat{\Omega}^{\bm k\bm k} \end{pmatrix} \;,
330: \end{equation}
331: where each block is a $3\times3$ matrix; $\mat{\Omega}^{\bm r\bm
332: k} = -(\mat{\Omega}^{\bm k\bm r})^T$. The phase space density
333: of states then reads,
334: \begin{equation}
335: D = (2\pi)^{-d} \sqrt{\det(\mat{\bm\Omega} - \mat{\mathbf J})} \;.
336: \label{DOS-gen}
337: \end{equation}
338: with $\mat{\mathbf J} = \bigl(\begin{smallmatrix} 0 & \mat{\mathrm I}
339: \\- \mat{\mathrm I} & 0\end{smallmatrix}\bigr)$. In the special case
340: of electromagnetic perturbations with $\mat{\Omega}^{\bm k\bm
341: k}_{ab}=\epsilon_{abc}\Omega_c$, $\mat{\Omega}^{\bm r\bm r}_{ab}=- (e/
342: \hbar) \epsilon_{abc} B_c$ and $\mat{\Omega}^{\bm k \bm r}=0$, it
343: reduces to \eqref{DOS}. On the other hand, when either
344: $\mat{\Omega}^{\bm k \bm k}$ or $\mat{\Omega}^{\bm r \bm r}$ vanishes,
345: it has a simpler form
346: \begin{equation}
347: D = (2\pi)^{-d} \det(\mat{\mathrm I}
348: - \mat{\Omega}^{\bm r\bm k}) \;.
349: \end{equation}
350: This result has found application in the study of spin-force induced
351: charge-Hall effect \cite{zhang2004}.
352:
353: Finally, we show how the density of states emerges naturally in the
354: effective quantum mechanics of Bloch electrons. Although our system
355: is not canonical, it can nevertheless be quantized following a
356: standard procedure developed for non-holonomic systems with second
357: class constraints~\cite{senjanovic1976,gitman1990}. First, one
358: redefines the Poisson bracket $\{f, g\}^* = (\partial f / \partial
359: \xi^a) M_{ab} (\partial g / \partial \xi^b)$, where $\xi^a$ are the
360: components of phase space coordinates $\bm \xi \equiv (\bm r,\; \bm
361: k)$ and $\mat{\mathbf M} = (\mat{\bm\Omega} - \mat{\mathbf J})^{-1}$.
362: Our equations of motion~\eqref{EOM} can then be written as
363: $\dot{\xi}^a = \{\xi^a, \eps\}^*$, where the energy $\eps(\bm \xi)$
364: plays the role as the Hamiltonian function. Then, one promotes the
365: Poisson brackets into quantum commutators:
366: \begin{equation}
367: [\hat{\xi}^a, \hat{\xi}^b] = \rmi M_{ab} \;,
368: \label{commutator_gen}
369: \end{equation}
370: where $\hat{\xi}^a$ is the quantum operator corresponding to the phase
371: space coordinates. It then follows that a phase space point
372: acquires a minimal uncertainty volume given by \cite{robertson1934}
373: \begin{equation}
374: \min \Bigl(\prod_a \Delta \xi^a\Bigr) =
375: 2^{-d} \Bigl[\det(\mat{\bm \Omega} - \mat{\mathbf J})\Bigr]^{-1/2} \;.
376: \end{equation}
377: This can be understood as the phase space volume occupied by a single
378: quantum state, therefore Eq.~\eqref{DOS-gen}, which is proportional to
379: the reciprocal of this volume, can naturally be regarded as the
380: semiclassical expression for the number of quantum states per unit
381: volume in the phase space.
382:
383: Equation~\eqref{commutator_gen} presents the effective quantum
384: mechanics of Bloch electrons. As a demonstration for the validity of
385: the quantization scheme as well as the quantum effect of the phase
386: space density of states, we consider a simple toy model of two
387: dimensional electron system with a constant Berry curvature, subjected
388: to a uniform magnetic field. The commutators read,
389: \begin{align}
390: &[\hat{x}, \hat{y}] = \rmi \frac{\Omega}{1 + (e/\hbar)B\Omega}
391: \;, \quad [\hat{k}_x, \hat{k}_y] = -\rmi \frac{(e/\hbar)B}{1 +
392: (e/\hbar)B\Omega} \;, \nonumber \\
393: &[\hat{x}, \hat{k}_x] = [\hat{y}, \hat{k}_y] =
394: \rmi \frac{1}{1 + (e/\hbar)B\Omega} \;.
395: \label{commutator}
396: \end{align}
397: In the absence of the Berry curvature, we reduce the problem to a
398: known case with the familiar nontrivial commutator $[\hat{k}_x,
399: \hat{k}_y] = -\rmi (e/\hbar) B$. In the absence of the $ B$ field,
400: we have the nontrivial commutator $[\hat{x}, \hat{y}] = \rmi \Omega$
401: discussed extensively in the literature on non-commutative geometry.
402: It is interesting to see that in the presence of both fields, we do
403: not just have a combination of these nontrivial commutators.
404: Instead, we have a nontrivial density of states which enters into all
405: of the commutators.
406:
407: Assuming $\eps(\bm k)=\hbar^2 {\bm k}^2/2m$, the system can be solved
408: algebraically to yield the energy spectrum and degeneracy. We found
409: that the spectrum consists of a set of Landau levels with the
410: renormalized cyclotron frequency $ \omega_c = \omega_c^0/[1 +
411: (e/\hbar)B\Omega]$, where $\omega_c^0 = eB/m$ is the usual cyclotron
412: frequency~\cite{horvathy2005}. At the same time, it is more important
413: to note that each Landau level still has the same degeneracy of $eB/h$
414: as in the absence of the Berry curvature. It is known that this
415: degeneracy is directly related to the quantized Hall conductance
416: $e^2/h$ for a filled Landau level~\cite{hall}. Had the density of
417: states not entered in the commutators, the Landau level degeneracy
418: would be modified, violating the topological requirement that the Hall
419: conductance for a filled Landau level is quantized.
420:
421: Before closing, we note that the phase space density of states also
422: enters naturally in the alternative quantization scheme with Feynman
423: path integral. The $S$ matrix is calculated by~\cite{senjanovic1976}
424: \begin{equation}
425: \bracket{\text{out}|S|\text{in}} = \int \prod_t [D(\bm \xi)
426: \rmd \bm \xi]
427: \exp\Bigl[{\frac{\mathrm i}{\hbar} \int L \rmd t}\Bigr] \;.
428: \end{equation}
429: where $L$ is the Lagrangian for our system~\cite{sundaram1999},
430: \begin{equation}
431: L = \frac{1}{2}\dot{\xi}^a J_{ab} \xi^b - \eps(\bm \xi)
432: + \dot{\xi}^a \mathcal A_a(\bm \xi) \;
433: \label{lagrangian}
434: \end{equation}
435: with $\mathcal A_a(\bm \xi) \equiv \rmi \bracket{u(\bm
436: \xi)|\nabla_\alpha u(\bm \xi)}$ being the phase space gauge potentials
437: associated with the Berry curvature field $\mat{\bm\Omega}$.
438:
439: In summary, we have found a Berry phase correction to the phase space
440: density of states for Bloch electrons. This correction emerges
441: naturally in both semiclassical and quantum mechanics of Bloch
442: electrons, and has profound effects on the equilibrium and transport
443: properties. Because of the fundamental change introduced by this
444: correction, it could have important implications on other aspects of
445: condensed matter physics, such as the Fermi liquid theory. For
446: instance, in the presence of a magnetic field, interaction between
447: electrons can change the Fermi sea volume by modifying the Berry
448: curvature and thus the phase space density of states.
449:
450: We acknowledge useful discussions with M.~C.~Chang, R.~Resta,
451: G.~Y.~Guo, Y.~G.~Yao, and D.~Vanderbilt. This work is supported by
452: DOE grant DE-FG03-02ER45958.
453:
454: \bibliography{journal,book,eprint,footnote}
455:
456: \end{document}
457:
458:
459: