Download arXiv:0905.2946v1 [cond-mat.str-el] 18 May 2009

Topological order in paired states of fermions in two-dimensions with breaking of
parity and time-reversal symmetries
Noah Bray-Ali,1 Letian Ding,1 and Stephan Haas1
arXiv:0905.2946v1 [cond-mat.str-el] 18 May 2009
1
Department of Physics and Astronomy, University of Southern California, Los Angeles, CA 90089
(Dated: May 18, 2009)
We numerically evaluate the entanglement spectrum (singular value decomposition of the wavefunction) of paired states of fermions in two dimensions that break parity and time-reversal symmetries, focusing on the spin-polarized px + ipy case. The entanglement spectrum of the weak-pairing
(BCS) phase contains a Majorana zero mode, indicating non-Abelian topological order. In contrast,
for the strong-pairing (BEC) phase, we find no such mode, suggesting Abelian topological order.
In both phases, the leading correction to the area law behavior of the entanglement entropy has a
geometric origin, while at the quantum phase transition, our large-scale numerical results indicate a
universal, logarithmic correction to the area law. We find that the entanglement spectrum detects
topological order in the ground-state wavefunction more robustly than the entanglement entropy
for states of paired fermions.
PACS numbers: 74.20.Rp;03.67.Pp;71.10.Pm;74.90.+n
Introduction.– Two-dimensional fermion systems with
pairing that breaks parity and time-reversal symmetries come in a variety of forms including quantum hall
fluids,[1, 2] superfluids,[3] superconductors,[4] and condensates of cold atoms near a Feshbach resonance.[5, 6]
For spin-polarized fermions, the simplest pairing order
parameter that breaks these symmetries, ∆p ∝ px + ipy ,
depends on the relative momentum p of the fermions
in a pair. For momentum independent, s-wave pairing, a smooth cross-over occurs from weak-pairing (BCS)
to strong-pairing (BEC). In the px + ipy case, the two
phases have different topological order and are separated by a quantum phase transition.[7] Recent proposals for fault-tolerant quantum computation using twodimensional fermion systems with px + ipy pairing require the system to be in the weak-pairing phase.[8] This
motivates us to investigate the topological order of the
weak-pairing and strong-pairing phases using ideas from
quantum information theory.
The entanglement spectrum[9] and the entanglement
entropy[10] contain information about the universal
properties of a quantum state. We define them by dividing the system into a block A with feature size L and
an environment B, and then performing a Schmidt decomposition,
X 1
e− 2 ξi |ψiA i ⊗ |ψiB i.
(1)
|ψi =
i
Here, the orthonormal sets of states {|ψiA i}, {|ψiB i} span
A and B. The entanglement
P spectrum {ξi } gives the entanglement entropy S = i ξi e−ξi .
Recently, Li and Haldane gave numerical evidence that
the entanglement spectrum probes topologically protected edge excitations.[9] The result can be understood
by considering a physical process that cuts into two parts,
A and B, a system with bulk excitation gap E0 and gapless edge excitation spectrum. If we perform the cut at
some rate, Γ ≪ E0 , adiabatic with respect to the bulk excitation gap, then the possible outcomes in region A are
the Schmidt eigenvectors {|ψiA i}, with the entanglement
spectrum {ξi } measuring the likelihood of each possibility. Now, even if one proceeds slowly with respect to the
gap E0 , avoiding bulk excitations, the gapless edge modes
are still likely to be excited. The entanglement spectrum
detects these topologically protected edge modes, providing insight into the form of topological order in the
bulk.
In this Letter, we report the first large-scale numerical calculations of the entanglement entropy and spectrum of two-dimensional fermion systems with px + ipy
pairing. We find that the entanglement entropy exhibits
universal critical behavior in the vicinity of the quantum phase transition separating weak-pairing and strongpairing. Further, we find that the entanglement spectrum
qualitatively distinguishes the topological order occurring in the two phases. In particular, we find that the
low-lying spectrum in the weak-pairing phase contains
a chiral, gapless fermion excitation. The weak-pairing
phase is known to have a chiral, gapless Majorana edge
mode.[7] This mode is intimately related to the Majorana
zero mode that appears in vortex cores and gives vortices
non-Abelian statistics.[7, 11]
We reduce the problem of evaluating the entanglement
spectrum and entanglement entropy to diagonalizing a
quadratic entanglement Hamiltonian.[12] This approach
does not include fluctuations of the pairing order parameter, and, hence, we do not expect to observe a universal,
topological term in the entanglement entropy[13, 14] in
either the weak-pairing or strong-pairing phase,[15] despite the fact that both phases have non-trivial quantum
dimension D = 2. Instead, we find that the entanglement spectrum detects topological order in the groundstate wavefunction more robustly than the entanglement
entropy for states of paired fermions.
2
Pairing
Hamiltonian.–
The
following
BCS
Hamiltonian[16] serves as a minimal model for a
single band of spin-polarized fermions with px + ipy
pairing on a square lattice:
X
X c†r cr
c†r cr′ + c†r′ cr + 2λ
H=− t
hr,r ′ i
80
Weak-Pairing
0
1.0
=1.0
0.5
=3.0
40
=2.5
=2.0
0.0
0
1
2
3
4
=1.5
hr,r ′ i
=1.0
′
We consider only nearest-neighbor hr, r i hopping t and
pairing γr,r′ interactions. The hopping strength t and
coupling λ are taken to be real and positive, without
loss of generality. The pairing interaction γr,r′ breaks
both time-reversal and parity symmetries: γr,r+x̂ =
−γr,r−x̂ = iγr,r+ŷ = −iγr,r−ŷ = iγ. Here, γ is real and
x̂, ŷ are the primitive translation vectors of the square
lattice. We use periodic boundary conditions in our numerical calculations.
The pairing Hamiltonian (2) is quadratic and can be
solved exactly using a Bogoliubov transformation,[17]
yielding the phase diagram shown in the inset of Fig. 1.[7]
The critical line at λc = 2t, separates the weak-pairing
(BCS) phase from the strong-pairing (BEC) phase.
Both phases have a spectral gap E0 = t|λ − λc | to
bulk excitations show in the inset of Fig. 1 and determined by minimizing
the Bogoliubov quasi-particle
q
2
ξp2 + |∆p | . The pairing order padispersion:Ep =
rameter ∆p = 2γ(sin px + i sin py ) transforms under the
symmetries of the square lattice in the same way as an
ℓ = 1, ℓz = 1 spherical harmonic. At small p, we expand ∆p ∝ px + ipy , and see the px + ipy pairing explicitly. Similarly, at small p, the single-particle kinetic
energy ξp = −2t(cos px + cos py ) + 2λ, takes the form
ξp = p2 /2m∗ − µ, with effective mass m∗ = 1/2t and
µ = 4t − 2λ. The weak-pairing phase λ < λc corresponds
to µ > 0, while strong-pairing λ > λc corresponds to
µ < 0.[7]
Entanglement Hamiltonian.–The two-point correlation
functions provide a complete description of the ground
state of the quadratic Hamiltonian (2), and allow an efficient numerical evaluation of the Schmidt decomposition
(1).[12] In fact, the Schmidt decomposition of the pairing Hamiltonian ground-state reduces to diagonalizing
the following entanglement Hamiltonian He which acts
on the sites of the block A:
X
X
He =
Fr,r′ c†r c†r′ + h.c. .(3)
Cr,r′ c†r cr′ + h.c. +
r,r ′
(BEC)
E
60
S
(2)
Strong-Pairing
(BCS)
1.5
r
X ∗
,
γr,r′ c†r c†r′ + γr,r
−
′ cr ′ cr
2.0
r,r ′
Here, in contrast to (2), the hopping parameters Cr,r′ =
hc†r cr′ i and pairing parametersFr,r′ = hcr cr′ i extend
beyond nearest-neighbors and are given by the twopoint correlation functions in the ground state of the
pairing Hamiltonian (2). The entanglement Hamiltonian is quadratic, and can be exactly solved by numerically performing a Bogoliubov transformation to the
20
0
10
20
L
30
40
50
FIG. 1: (Color online) Entanglement entropy S between a
square of side length L and its environment as a function
of λ at fixed pairing strength γ = 1.0. (Inset) The zerotemperature phase diagram of two-dimensional fermions with
px + ipy pairing and plot of the bulk spectral gap E0 . The
phase boundary between weak-pairing and strong-pairing is
the vertical γ-independent line at λc = 2t. The spectral
gap vanishes at the critical coupling and grows linearly with
|λ − λc |. Data points indicate the parameters chosen in our
numerical calculations (t=1).
quasi-particle operators αn , for n = ±1, ±2, . . . ± NA ,
where, NA is the number of sites in the block A.[17] In
terms of the quasi-particles,
P the entanglement Hamiltonian has the form He = n>0 f (ǫn )(α†n αn + 21 ), where,
f (ǫ) = (eǫ + 1)−1 is the Fermi function and the quasiparticle eigenvalues {ǫn } generate the entanglement spectrum. In P
particular, the entanglement entropy is given
by S = − n f (ǫn ) log f (ǫn ).
Results.– Fig. 1 shows the entanglement entropy S as a
function of the block size L. We consider various coupling
strengths λ that sweep through the quantum phase transition, as shown in the inset. The entropy grows linearly
with L for this two-dimensional system, consistent with
an area law SL = aL + . . . . Fermion systems generically
violate the area law if they have a Fermi surface of gapless excitations.[18, 19, 20] Remarkably, we observe the
area law even at the quantum phase transition where the
bulk excitation gap vanishes at a single point in momentum space. Other two-dimensional fermion systems with
point nodes obey the area law.[20, 21] We present the
first numerical evidence for area law behavior in a gapless fermion system that breaks parity and time-reversal
symmetries. In the gapped phases, rigorous theoretical
arguments[22] suggest that an area law must hold. The
results in Fig. 1 are the first large-scale numerical test of
this argument for fermion systems that break parity and
time-reversal symmetries.
Using these large-scale numerical results, we are able
to extract the leading correction to the area law ∆S =
3
3
1.5
=1.0
=3.0
(a)
A
=2.5
=2.0
1.0
=1.0
S
S
1
0.5
0
0.0
10
20
30
10
40 50
L
20
30
40
50
L
=2.04
(c)
(d)
=2.08
=2.16
3
4
4
2
2
6
=1.5
cr
=1.0
6
=2.0
2
=1.5
sq
A
=2.5
6
(b)
=3.0
Block Enegy Spectrum
=1.0
4
=1.5
=1.0
2
=1.0
cr
cr
=2.5
S / S
S / S
=3.0
=2.00
2
=2.01
=2.02
10
20
30
L
40
50
0
0
sq
sq
3
10
20
30
40
50
L
FIG. 2: (Color online) Leading correction term ∆S to the
area law for a square (a) and cross-shaped (b) partition as a
function of block size L. (Inset) geometry of the partitions.
Notice the linear scale for ∆S and the logarithmic scale for L
in both (a) and (b). Solid lines are guides to the eye. Ratio
∆cr /∆sq of the leading correction terms from (a) and (b)
as function of block size L: (c) within the weak-pairing and
strong-pairing phases; (d) approaching the quantum phase
transition from the strong-pairing regime.
−3(S − aL). We plot the size dependence of the leading
correction ∆Ssq for the square shaped partition shown in
Fig. 2(a) and for the cross-shaped partition ∆Scr shown
in Fig. 2(b). For both geometries, the leading correction grows at the critical point with L, without sign of
saturation. By contrast, in the weak-pairing and strongpairing phases, the leading correction saturates to an L
independent value as L → ∞.
We interpret the growth at the critical point as a logarithmic divergence, of the form S = aL − b log L + . . . .
The constant of the logarithmic growth b clearly depends
on the partition geometry, as can be seen by comparing
the data for the square and cross in Fig. 2(a) and (b).
However, for a given geometry, for example the square,
we observe approximately the same coefficient b ≈ .13
at all points along the phase boundary between weakpairing and strong-pairing (data not shown). We emphasize that two-dimensional fermion systems with px + ipy
pairing obey the area law: the logarithmic divergence
appears as the leading, additive correction. Universal,
logarithmic corrections to the area law occur at other
two-dimensional quantum phase transitions.[23, 24] Nonuniversal, logarithmic corrections to the area law arise
throughout the nodal phase of two-dimensional fermion
systems with time-reversal invariant pairing.[25]
The leading correction to the area law shown in
Fig. 2(a) and 2(b) clearly depends on the geometry of
the partition. In Fig. 2(c) and (d), we analyze this de-
0
/2
(a)
3 /2
=2.0
0
/2
(b)
3 /2
(c)
=3.0
0
/2
3 /2
Angular Momentum
Angular Momentum
Angular Momentum
FIG. 3: (Color online) Low-lying quasi-particle entanglement
spectrum {ǫn } (a) in the weak-pairing phase, (b) at the quantum phase transition, and (c) in the strong-pairing phase with
fixed pairing strength γ = 1.0. We divide the spectrum into
four sectors based on the discrete angular momentum of the
quasi-particle wavefunction.
pendence by plotting the ratio of the leading correction
∆Scr /∆Ssq for the two partition geometries show in the
insets of Fig. 2(a) and (b). Notice that the cross has
twelve corners, while the square has four corners. In
both strong-pairing and weak-pairing phases (Fig. 2c),
the ratio ∆Scr /∆Ssq → 3 approaches a constant in the
limit of large block size L. Remarkably, the ratio of the
leading correction term to the area law equals the ratio of the number of corners of the two partitions. We
have examined other geometries and find the behavior
∆S = cnc , where, nc is the number of corners and c is
a positive coefficient. A similar corner effect has been
observed analytically in a wide class of topologically ordered phases.[26]
As we approach the critical point from the strongpairing phase, Fig. 2(d), the ratio takes longer to saturate as a function of L. At the critical point, λc = 2.0,
the data appear to saturate at a value that is not given
simply by the ratio of corners. This should be contrasted
with the geometry dependence of the logarithmic correction to the area law observed for two-dimensional conformal quantum critical points.[23] In that case, each corner with angle θ contributes a correction to the area law
2
cθ
∆S = 24π
(1 − πθ2 ) log L, where, c is the central charge of
the conformal field theory describing equal-time, spatial
correlations.[23] Conformal quantum critical points have
dynamical expondent z = 2, and are in a different universality class than the quantum phase transition in px + ipy
paired fermions with dynamical exponent z = 1. We do
not expect, and indeed do not numerically observe, the
geometric dependence predicted for conformal quantum
critical points.
4
Minimum Block Energy
10
=1.0
=1.25
1
=1.5
=1.75
=1.92
=1.96
=1.98
=1.99
=2.0
=2.25
0.1
=2.5
=2.75
=3.0
0.01
0.1
1
1/L
FIG. 4: (Color online) Finite-size scaling of minimimum block
energy level ǫ0 plotted on a log-log scale at fixed pairing amplitude γ = 1.0. In the weak-pairing phase λ < 2.0, the
dashed lines are best fits to the scaling form ǫ0 ∼ 1/L.
To detect topological order, we turn to the entanglement spectrum shown in Fig. 3. We label the quasiparticle eigenvalues {ǫn } by discrete angular momenta
ℓ = 0, π2 , π, 3π
2 , describing the transformation properties
under the point group of the square lattice. Under a rotation by π2 , the quasi-particle wavefunction acquires phase
factors eiℓ = 1, +i, −1, −i. The low-lying spectrum contains a dispersive mode whose block energy increases with
angular momentum in the weak-pairing phase Fig 3(a)
and at the quantum phase transition, Fig 3(b). We
contrast this with the featureless low-lying spectrum in
the strong-pairing phase, Fig. 3(c). Switching to p − ip
pairing (data not shown), the dispersive mode in the
weak-pairing phase and at the quantum phase transition
switches direction: its block energy increases with decreasing angular momentum. We interpret this low-lying
feature as a chiral excitation mode, since it has a velocity ∆ǫ/∆ℓ that reverses sign under time-reversal. In the
weak-pairing phase, the minimum block energy ǫ0 ≈ 0
appears to be gapless in Fig. 3(a). In contrast, the spectrum in the strong-pairing phase appears to be gapped.
We show in Fig. 4, the finite-size scaling of the minimimum block energy level ǫ0 plotted on a log-log scale at
fixed pairing amplitude γ = 1.0. In the strong-pairing
phase, the minimum block energy ǫ0 tends to a constant as L → ∞. By contrast, in the weak-pairing
phase λ < 2.0, the minimum block energy drops to zero
ǫ0 ∼ 1/L, indicating the presence of a zero mode in the
entanglement spectrum in the limit L → ∞. At the
quantum phase transition, λc = 2, the finite-size scaling
of the minimum block energy is not constant, but de-
creases much more slowly than the scaling ǫ0 ∼ 1/L seen
in the weak-pairing phase.
Conclusion– In this Letter, we study topological order
in paired states of fermions with parity and time-reversal
symmetry breaking. Large-scale numerical calculations
of the entanglement spectrum and entanglement entropy
reveal universal behavior, including a divergent correction to the area law at the quantum phase transition
separating weak-pairing and strong-pairing phases. We
find evidence for a chiral, gapless Majorana fermion excitation in the entanglement spectrum of the weak-pairing
phase, and contrast this with the gapped spectrum in the
strong-pairing phase. A variety of topological phases can
be described by a pairing Hamiltonian that neglects order parameter fluctuations. We suggest that large-scale
numerical calculations of the entanglement spectrum are
a robust way to detect topological order in the groundstate wavefunction of such phases.
NBA acknowledges the 2008 Boulder Summer School
and NCTS for their hospitality during the completion of
this work. Computational facilities have been generously
provided by HPCC at USC. We are grateful for fruitful
discussions with A. Feguin, M.P.A. Fisher, A. Kitaev,
F.D.M Haldane, Z. Nussinov, K. Raman, and P. Zanardi.
[1] R. L. Willett, J. P. Eisenstein, D. C. Tsui, A. C. Gossard,
and J. H. English, Phys. Rev. Lett. 59, 1776 (1987).
[2] G. Moore and N. Read, Nucl. Phys. B 360, 362 (1991).
[3] D. D. Osheroff, R. C. Richardson, and D. M. Lee, Phys.
Rev. Lett. 28, 885 (1972).
[4] T. M. Rice and M. Sigrist, J. Phys. Condens. Matter 7,
L643 (1995).
[5] V. Gurarie, L. Radzihovsky, and A. V. Andreev, Phys.
Rev. Lett. 94, 230403 (2005).
[6] C.-H. Cheng and S.-K. Yip, Phys. Rev. Lett. 95, 070404
(2005).
[7] N. Read and D. Green, Phys. Rev. B 61, 10267 (2000).
[8] S. Tewari, S. Das Sarma, C. Nayak, C. Zhang, and
P. Zoller, Phys. Rev. Lett. 98, 010506 (2007).
[9] H. Li and F. D. M. Haldane, Phys. Rev. Lett. 101, 010504
(2008).
[10] M. Nielsen and I. Chuang, Quantum Computation and
Quantum Information (Cambridge, 2000), p. 510.
[11] D. A. Ivanov, Phys. Rev. Lett. 86, 268 (2001).
[12] M.-C. Chung and I. Peschel, Phys. Rev. B 64, 064412
(2001).
[13] A. Kitaev and J. Preskill, Phys. Rev. Lett. 96, 110404
(2006).
[14] M. Levin and X.-G. Wen, Phys. Rev. Lett. 96, 110405
(2006).
[15] Z. Nussinov and G. Ortiz, Ann. Phys. (N.Y.) 324, 977
(2009).
[16] J. Bardeen, L. N. Cooper, and J. R. Schrieffer, Phys.
Rev. 108, 1175 (1957).
[17] J.-P. Blaizot and G. Ripka, Quantum Theory of FiniteSystems (MIT Press, 1986), pp. 34–38, 101–103.
[18] M. M. Wolf, Phys. Rev. Lett. 96, 010404 (2006).
5
[19] D. Gioev and I. Klich, Phys. Rev. Lett. 96, 100503
(2006).
[20] W. Li, L. Ding, R. Yu, T. Roscilde, and S. Haas, Phys.
Rev. B 74, 073103 (2006).
[21] T.Barthel, M. Chung, and U. Schollwock, Phys. Rev. A
74, 022329 (2006).
[22] M. M. Wolf, F. Verstraete, M. B. Hastings, and J. I.
Cirac, Phys. Rev. Lett. 100, 070502 (2008).
[23] E. Fradkin and J. E. Moore, Phys. Rev. Lett. 97, 050404
(2006).
[24] R. Yu, H. Saleur, and S. Haas, Phys. Rev. B 77, 140402
(2008).
[25] L. Ding, N. Bray-Ali, R. Yu, and S. Haas, Phys. Rev.
Lett. 100, 215701 (2008).
[26] S. Papanikolaou, K. S. Raman, and E. Fradkin, Phys.
Rev. B 76, 224421 (2007).