unphysical solutions of the kadanoff-baym equations in linear response

Upload: adi63

Post on 26-Feb-2018

214 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/25/2019 Unphysical Solutions of the Kadanoff-Baym Equations in Linear Response

    1/6

    RAPID COMMUNICATIONS

    PHYSICAL REVIEW B93, 041103(R) (2016)

    On the unphysical solutions of the Kadanoff-Baym equations in linear response:

    Correlation-induced homogeneous density-distribution and attractors

    Adrian Stan*

    Sorbonne Universit es, UPMC Universit e Paris VI, UMR8112, LERMA, F-75005 Paris, France;

    LERMA, Observatoire de Paris, PSL Research University, CNRS, UMR8112, F-75014 Paris, France;

    Laboratoire des Solides Irradies, Ecole Polytechnique, CNRS, CEA-DSM, F-91128 Palaiseau, France;

    and European Theoretical Spectroscopy Facility (ETSF)

    (Received 13 September 2015; revised manuscript received 12 December 2015; published 8 January 2016)

    The Kadanoff-Baym equations, which allow for the calculation of time-dependent expectation values of all

    one-particle observables, are found to yield unphysical electron density dynamics in the linear and nonlinear

    response, for -derivable approximations, irrespective of interaction strength or type. In particular, we show

    that when calculated from the Kadanoff-Baym equations using correlated self-energy approximations, the linear

    response dynamics of isolated electron systems damps to an unphysical homogeneous density-distribution. The

    damping is also present for Hartree or Hartree-Fock self-energies. These surprising results supplement previous

    findings on the nonlinear response, and complement them by showing that the linear response is also plagued

    by unphysical dynamics. Being universal, this additional feature indicates the possible presence of an attractor

    that leads to amplitude death and a subsequent tendency to a homogeneous charge and density distribution.

    This unveils a scenario in which the Kadanoff-Baym dynamics simply breaks down, drastically restricting the

    parameter space for which the method can give physically meaningful insights. In addition to their relevance tothe field of ultrafast electron dynamics in isolated and open systems, these findings may also impact the results

    obtained with the Bethe-Salpeter equation in linear response, due to the well-known equivalency between the

    two methods. This suggests the need for a different approach to the dynamics of quantum systems.

    DOI: 10.1103/PhysRevB.93.041103

    In modern quantum many-body physics, the Kadanoff-

    Baym equations (KBEs) [13] are a crucial component in the

    treatment of strongly and weakly correlated systems far from

    equilibrium [49]. They have the form

    [it h(t)]G(t, t) = (t,t)+

    [G](t,t)G(t ,t)dt, (1)

    whereh(t) is the one-body part of the Hamiltonian, G(t,t) is

    the nonequilibrium Green function, and the kernel (t, t) is

    the self-energy approximation [1,10]. The integration is taken

    over the so-called Keldysh contour [11]. The self-energy

    is a functional of the Green function [G] and, when it

    is obtained from a generating functional as [G] = [G]G

    ,

    it satisfies sum rules and macroscopic conservation laws

    [2,3,1214]. Well-known examples of such self-consistent

    conserving approximations are Hartree-Fock (HF), Second

    Born (2B) [1,15], GW[16], and T-matrix [1,17]. Unlike time-

    dependent density functional theory, the use of KBEs is not

    limited by the adiabatic approximation. This prompted their

    use in calculating the properties of both isolated [4,8,1820]

    and open[57,21]quantum systems, and as a complementary

    method in the theoretical efforts to develop density functionals

    [22]. Within the linear response, the time-propagation of

    KBEs corresponds to solving the Bethe-Salpeter equation

    with advanced kernels [8,23], constitutingfor long-time-

    propagationsa benchmark for the latter. Hence, the KBEs

    simultaneously became the method of choice and the state of

    the art in ultrafast dynamics of correlated electron systems.

    *Corresponding author: [email protected]

    A dent in the hallmark of the KBEs was first made by

    Friesen et al. [19], when they showed that in Hubbard systems,

    the time-dependent density of conserving approximations

    exhibits a so-called correlation damping in the nonlinear

    regime. These findings were later reproduced by Uimonen

    [24], and by Hermannset al.[9]. We can illustrate this feature

    of KBEs by considering a system described by the generalHamiltonian

    H(t) =ij,

    hij(t)ci,ci,+

    1

    2

    ijkl,

    ci,c

    j,ck,cl,, (2)

    where i,j label a complete set of one-particle states, ,

    are the spin indices (in this work we only consider spin-

    unpolarized systems ), and c,c are creation and anni-

    hilation operators, respectively. The one-body part of the

    Hamiltonianhij(t) may have an arbitrary time dependence.

    The two-body part accounts for interactions between the elec-

    trons where, in the case of a Coulomb-like interaction[25,26],

    vijkl = vijiljk withvij= vii = Ufor i = jand vij= U

    2(ij)for i = j and, for the Hubbard model, vij= vii = U for

    i = j andv ij= 0.0 for i = j. In this Rapid Communication

    we will consider both Coulomb and Hubbard systemswith

    on-site energiesh ii = 0 and hoppings hij= 1.0 between the

    neighboring sitesi and j in linear and nonlinear responses.

    All parameters are given in terms of hoppings.

    In the inset of Fig. 1 we reproduce the time-dependent

    density, of a Hubbard dimer, within the 2B approximation, as

    shown in Fig.3(b)of Ref. [19]. It is obvious that the density at

    Site 1, where the strong, steplike perturbation was applied

    [w(t) = w0(t) = E = 5.0], quickly damps to an unphysi-

    cal value. (Note here the upwards trend.) The mechanism

    underlying this situation was labeled correlation-induced

    2469-9950/2016/93(4)/041103(6) 041103-1 2016 American Physical Society

    http://dx.doi.org/10.1103/PhysRevB.93.041103http://dx.doi.org/10.1103/PhysRevB.93.041103
  • 7/25/2019 Unphysical Solutions of the Kadanoff-Baym Equations in Linear Response

    2/6

    RAPID COMMUNICATIONS

    ADRIAN STAN PHYSICAL REVIEW B93, 041103(R) (2016)

    0 25 50 75 100 125 150time (a.u.t.)

    0

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    0.45

    0.5

    Site

    density

    Site 1 (Hubbard dimer; U = 4.0, E=5.0)

    Site 1 (Coulomb dimer; U = 4.0, E=5.0)

    0 1 2 3 4 5time (a.u.t.)

    0

    0.2

    0.4

    0.6

    0.8

    Site

    density

    FIG. 1. Long-time density dynamics at the perturbed site of

    Hubbard (solid line) and Coulomb (dotted line) dimers. Inset: The

    short-time density dynamics of the Hubbard dimer at the perturbed

    Site 1 (legend and axes in the notation of Ref. [19]).

    damping and was attributed to the induction of an infinite

    number of poles by the use of partial-summation schemes and

    the subsequent generation of diagrams to all orders. These

    diagrams are thought to introduce an unsustainable number of

    particles and holes in the finite, isolated system, and act as an

    artificial reservoir to which the system couples, dampening its

    dynamics. In contrast, in an infinite system, these unphysical

    excitations would perfectly cancel[19,20]. As in this case the

    damping is faster with increasing the perturbation strength,

    the findings have restricted the applicability of the KBEs in

    finite, isolated systems to a narrow range of external fields,

    i.e., linear response. It was argued that as the dynamics in

    linear response is described by the Bethe-Salpeter equationwith a kernel /G that has a discrete spectrum, the resulting

    response will be undamped. Here, we show that the damping

    is not a simple matter of coupling to an artificial bath, but

    rather a pathological behavior of the KBEs that manifests in

    linear response as well, and even for uncorrelated self-energy

    approximations.

    Let us note that the results in the inset of Fig. 1, aswellas in

    all calculations in Refs. [19,20], areon a very shorttime-scales,

    i.e., 510 atomic units of time (a.u.t.). This short time-scale

    is insufficient for determining the value of the artificial steady

    state to which the time-dependent density damps. Consider

    the same Hubbard system as in the inset of Fig. 1 and,

    in addition, a Coulomb dimer at half-filling, with the same

    U= 4.0, and a constant perturbation E = 5.0 at Site 1. As

    seen in Fig. 1 (main plot), after the initial fast damping, the

    value of the density at the perturbed site gradually increases up

    to 0.5, and stabilizes. Albeit slower, this also holds true for the

    Coulomb dimer. Not only the time-dependent density suffers

    an amplitude death, but it also shows a tendency towards

    a completely unphysical homogeneous density-distribution

    (HD-D). This homogeneity is achieved for both Coulomb

    and Hubbard interactions, despite the large constant shift of

    Site 1 caused by the field applied all throughout the time-

    propagation. Clearly unphysical, this additional feature points

    out a second unphysical trait of the KBEs in the nonlinear

    regime: For strongly correlated, strongly perturbed systems,

    0 25 50 75 100 125 1500.475

    0.48

    0.485

    0.49

    0.495

    0.5

    0.505

    Site

    density

    Coulomb

    0 25 50 75 100 125 1500.47

    0.48

    0.49

    0.5

    0.51

    0.52

    0.53

    Hubbard

    0 25 50 75 100 125 1500.475

    0.48

    0.485

    0.490.495

    0.5

    0.505

    Site

    densi

    ty

    0 25 50 75 100 125 1500.475

    0.48

    0.485

    0.490.495

    0.5

    0.505

    U = 1.0

    U = 2.0

    U = 3.0

    0 25 50 75 100 125 150

    time (a.u.t.)

    0.47

    0.48

    0.49

    0.5

    0.51

    0.52

    0.53

    Site

    density

    0 25 50 75 100 125 150

    time (a.u.t.)

    0.47

    0.48

    0.49

    0.5

    0.51

    0.52

    0.53

    (a) (b)

    (c) (d)

    (e) (f)

    FIG. 2. Time-propagation up to 150 atomic units of time (a.u.t.):

    (a) Exact resultfor a Coulomb dimerU= 2 with a constant field E =

    0.05. (b) Exact result for a Hubbard dimer U= 2 and a kick ofE =

    0.05 at t= 0. (c), (d) Coulomb and Hubbard dimers, respectively,

    with a constant field ofE = 0.05 within the Second Born self-energy

    approximation (U= 1 dotted line; U= 2 dashed line; U= 3 solid

    line). (e), (f) Coulomb and Hubbard dimers, respectively, with a kick

    ofE = 0.05 applied at t= 0, within the Second Born self-energy

    approximation.

    the time-dependent density converges to a homogeneous

    density-distribution, irrespective of the interaction type. This

    result is extremely surprising and constitutes the first important

    findingof the present work. It is hard to attribute this additional

    feature to the coupling of an artificial reservoir generated by

    partial-summation schemes, as argued in Ref. [19].

    We further wish to determine the maximum strength of

    the external perturbation applied to an interacting, correlatedsystem, such that a conserving self-energy approximation

    could still be deemed good with respect to exact. For this we

    consider the same dimers at half-filling, as in Fig. 1, with the

    difference that we drastically reduce the external perturbations

    by twoorders of magnitude, i.e.,E = 5.0 E = 0.05,falling

    back onto the linear response regime. In this regime, the many-

    body approximations have been shown to perform well, and

    without any visible damping, on short time-propagations [19].

    In Fig. 2 we investigate thetime-dependent densityof Coulomb

    and Hubbard dimers, with increasing the interaction U. The

    twotypes of perturbation, constant and kick, are applied to Site

    1 only. First, in Fig. 2(a),we show the exact time-dependent

    density for a Coulomb dimer with a constant field E = 0.05

    applied during the time-propagation. Figure2(b)features the

    exact time-dependent density of a Hubbard dimer with a kick

    E = 0.05 applied at t= 0. As expected, irrespective of the

    interaction or the applied perturbation, an isolated quantum

    system presents no damping of the density oscillations.

    In what concerns the many-body approximations, in

    Figs. 2(c) and 2(d) we plot, for increasing U, the time-

    dependent densityobtained from the time-propagation of the

    KBEs within the 2B self-energy approximationof a system

    subject to a constant field E = 0.05 applied to Site 1. ForU= 0 the damping is absent (not shown), but as Uincreases,

    for both interactions, a now familiar damping of the density

    oscillations becomes apparent. This damping is accompanied

    041103-2

  • 7/25/2019 Unphysical Solutions of the Kadanoff-Baym Equations in Linear Response

    3/6

    RAPID COMMUNICATIONS

    ON THE UNPHYSICAL SOLUTIONS OF THE KADANOFF-BAYM . . . PHYSICAL REVIEW B93, 041103(R) (2016)

    0 50 100 150 200

    time (a.u.t.)

    0.48

    0.485

    0.49

    0.495

    0.5

    0.505

    0.51

    0.515

    0.52

    Site

    densi

    ty

    Coulomb

    0 50 100 150 200

    time (a.u.t.)

    0.49

    0.492

    0.495

    0.497

    0.5

    0.502

    0.505

    0.507

    0.51

    Site 1 (E =0.00, E =0.05)

    Site 2 (E =0.00, E =0.00)

    Site 1 (E =0.05, E =0.00)

    Site 2 (E =0.00, E =0.00)

    Hubbard

    (a) (b)

    FIG. 3. Time-dependent densitiesat Sites 1 and 2 of Coulomb and

    Hubbard dimers, with U= 2 and constant field E tp applied to Site

    1 during time-propagation (dashed-dotted lines). The same densities

    starting from a ground state with a constant shift Egs = 0.05 applied

    to Site 1, and no further time-dependent perturbation (solid lines).

    by a tendency towards a HD-D, whichbecomes more andmore

    apparent with increasing U and is particularly striking after

    amplitude death. We remind the reader that we are now in the

    linear response regime, while observing the same unphysical

    behavior as in nonlinear one (see Fig. 1). On the time-scale

    used here, the unphysical HD-D obtained for Coulomb (U=

    3) and Hubbard (U 2), shown in Figs. 2(c) and 2(d), is

    identically equal to 0.5. A slight tendency towards this latter

    value can be observed also for weaker interactions. With

    increasingU, both the damping and the tendency mentioned

    above become apparent on an ever shorter time-scale. Thislinear response behavior is reminiscent of the case of strong

    external perturbationsshown in Fig.1and in line with the

    observations in Ref. [27]. In Figs.2(e)and2(f),we show the

    time-dependent density following a kick ofE = 0.05 applied

    at t= 0. We note the same time-scale to amplitude death as

    in the case of a constant perturbation. Overall, no matter the

    perturbation, the density-dynamics of Hubbard dimers exhibits

    a damping with an amplitude half-life approximately three

    times shorter than that of Coulomb dimers. In other words, in

    linear and nonlinear response regimes, for a given interaction,

    the time to amplitude death is inversely proportional with the

    strength of the electron-electron interaction.

    In the surprising amplitude death found in Fig. 2, the

    damping of the oscillations occurs around a mean value that

    changes in time. Even during the damping process, this mean

    value shifts towards an unphysical HD-D. It is interesting

    to determine what this mean value is. We consider again a

    Coulomb and a Hubbard dimer at half-filling, and we focus

    on the case ofU= 2. The results are shown in Fig. 3. In

    the first setup we start by calculating the ground state of

    the dimer (Coulomb or Hubbard) and then we propagate

    in time with a constant field (Etp = 0.05) applied to Site

    1 [see also Figs. 2(c) and 2(d)]. The second setup consists

    of calculating the ground state of the dimer with an applied

    constant field (Egs = 0.05) to Site 1, and then propagate this

    ground state in time without applying any further perturbation.

    0 5 10 15 20 25 30 35 40 45

    (a)

    (b)

    500.488

    0.49

    0.492

    0.494

    0.496

    0.498

    0.5

    Site

    density

    U = 4.0, t = t = 1.0

    U = 4.0, t = 1.0, t = 2.0

    0 5 10 15 20 25 30 35 40 45 50

    time (a.u.t.)

    0.488

    0.49

    0.492

    0.494

    0.496

    0.498

    0.5

    Site

    density

    E = 0. 01

    E = 0. 05

    E = 0.1

    FIG. 4. (a) Time-dependent 2B densities at Site 1

    of a symmetric and asymmetric four-site Hubbard dimer, with

    a constant Etp = 0.05 field applied during time-propagation. (b)

    Time-dependent GWdensities at Site 1 of a four-site Hubbard dimer,

    with constant external fields of different strengths applied to Site 1.

    The time-dependent densities of the first setup oscillate around

    the values of the time-dependent densities of the second setup

    until the amplitude death occurs, while following the same

    trend to HD-D (see Fig. 3). For U= 2, on the time-scale

    shown here, this tendency is much more visible for the

    Hubbard dimeras we noted before, the amplitude half-life

    is approximately three times shorter for Hubbardbut it can

    also be observed for the Coulomb dimer, on this time-scale,

    at a close inspection. The results in Fig. 3 bring in a new

    element: For long-time-propagations, the KBEs are incapable

    of propagating an isolated system, with a frequency-dependent

    self-energy, without converging to an unphysical steady state,irrespective of the interaction used andthe presence or absence

    of a time-dependent field. The opposite of this constitutes a

    common misconception and a widespread assumption in the

    field.

    To investigate if the symmetry of the Hamiltonian plays

    a role in the unphysical features underlined above, we can

    alter the symmetry of the system while increasing the number

    of sites (see Fig. 4). The first system we consider is a

    completely symmetric, four-site Hubbard chain with t= 1.0,U= 4.0. A constant fieldEtp = 0.05 is applied to Site 1 during

    time-propagation. In the second system the symmetry of the

    Hubbard chain withU= 4.0 is altered by making the hopping

    between sites 3 4,t2 = 2.0t1, where t1 = 1.0. The field is

    applied at the same site. The results are shown in Fig. 4(a).

    Although the dynamics of the system is changed due to the

    increased kinetic energy of the electrons between sites 3 4,

    the damping still follows the same pattern and the HD-D is

    reached at a slightly later time. These results also show that

    the unphysical feature is not restricted to two-site dimers.

    Albeit 2B is a correlated approximation from the same

    class of correlated, conserving approximations as GW andT-matrix, it only contains four diagrams, while the others

    result from an infinite summation of bubble or ladder diagrams

    [13,28]. In Fig.4(b)we show that the present findings are not

    limited to 2B, but alsoextend to the conserving approximations

    resulting from infinite summations of a type of diagram, e.g.,

    041103-3

  • 7/25/2019 Unphysical Solutions of the Kadanoff-Baym Equations in Linear Response

    4/6

    RAPID COMMUNICATIONS

    ADRIAN STAN PHYSICAL REVIEW B93, 041103(R) (2016)

    GW. The system considered is a four-site Hubbard dimer

    witht= 1.0,U= 4.0 to which fields of different strengths

    but within linear responsewere applied to Site 1. For this

    case, in Fig. 4(b), we show that the density at Site 1 damps

    and tends to HD-D. It can be seen, from a comparison of

    the case Etp = 0.05 with Fig. 4(a), that in GW the lifetime

    of the excitations is shorter, i.e., HD-D is reached slightly

    faster. We note here that the GW, as well as the T-matrixapproximation, were also found to damp the electron dynamics

    of Hubbard systems in the nonlinear response regime [20],

    although the propagation time was too short for a HD-D to be

    observed (see also Fig. 1 andthe discussion in the text). Hence,

    the unphysical behavior remains unchanged across correlated,

    conserving approximations and regimes, and it is present in

    systems of different sizes.

    We point out here that the numerical proofs in this Rapid

    Communication are not a result of numerical inaccuracies. In

    the Supplemental Material we provide the proper numerical

    checks for particle-number conservation and time-reversal

    invariance[29].

    The convergence to HD-D, across both the nonlinear(Fig. 1) and the linear (Fig. 2 and 3) response regimes,

    points to a pathological trait of the KBEs, reminiscent of

    an unphysical (global) attractor, independent of interaction

    strength, external perturbation, or self-energy approximation.

    Although unexpected, the high nonlinearity of theseequations,

    i.e., complex, integrodifferential equations with nonlinear

    kernels that depend on two times, makes the emergence of such

    behavior less surprising. Whenever conserving approxima-

    tions are employed, the complexity of the self-energy kernels

    of the KBEs can be broken down into two pieces,

    [G(t, t)] = HF[G(t,t)]+c[G(t,t

    )], (3)

    whereHF[G(t,t

    )] is the Hartree-Fock part and c[G(t,t

    )]is the correlation part [15,30]. Diagrammatically, this simply

    means that any correlated conserving approximation, e.g., 2B,GW,T-matrix, contains the HF approximation. Furthermore,HF[G(t,t

    )] can be broken down into two diagrams: Hartree

    (H) and exchange. In this way, by systematically considering

    an ever smaller selection of diagrams, we can gauge the role of

    the kernel in thedamping, andwe choose to directlyinvestigate

    the behavior of KBEs for the uncorrelated approximations

    H and HF. We compute the time-dependent Hartree-Fock

    (TDHF) densities for Coulomb and Hubbard dimers, and

    propagate to800 a.u.t. As inthe case of 2B(see Fig. 5), wealso

    calculate the time-dependent densities produced by a setup of

    a ground state calculated with an applied Egs = 0.01 field to

    Site 1, followed by a time-propagation without any further

    perturbation (see also Fig.3). For both Coulomb and Hubbard

    dimers, the time-dependent densities of the system with a

    constant time-dependent field ofEtp = 0.01 oscillate around,

    and ultimately damp to, the values produced by the setup with

    only Egs. This is similar to what is observed in the case of

    2B. However, in contrast with correlated approximations, no

    tendency towards HD-D is present. Another difference is that

    the amplitude half-life is comparable for both interactions.

    Simplifying thekernel even more leads to the same damping

    pattern. In the insets of Fig. 5 we show the time-dependent

    Hartree (TDH) densities for the same systems, but withU= 2.5. The density dynamics is damped also in this case.

    0 200 400 600 8000.497

    0.498

    0.499

    0.50.501

    0.502

    0.503

    0 200 400 600 8000.498

    0.499

    0.5

    0.501

    0.502

    0 200 400 600 800

    time (a.u.t.)

    0.497

    0.498

    0.499

    0.5

    0.501

    0.502

    0.503

    Site

    density

    Site 1 (E =0.00, E =0.01)

    Site 2 (E =0.00, E =0.00)

    Site 1 (E =0.01, E =0.00)

    Site 2 (E =0.00, E =0.00)

    Coulomb

    0 200 400 600 800

    time (a.u.t.)

    0.498

    0.499

    0.5

    0.501

    0.502

    Hubbard

    (a) (b)

    FIG. 5. (a) Time-dependent Hartree-Fock densities of a Coulomb

    dimer (U= 5.0) starting from the ground state, and from a ground

    state with an applied field. Inset: Time-dependent Hartree densities

    of a Coulomb dimer (U= 2.5). (b) Time-dependent Hartree-Fock

    densities of a Hubbard dimer (U= 5.0) starting from a ground stateand from a ground state with an applied field. Inset: Time-dependent

    Hartree densities of a Hubbard dimer (U= 2.5).

    It is worth mentioning here that for Hubbard systems, the

    densities within TDHF with 2Ushould be exactly the same as

    those resulting from TDH withU (this provides an additional

    consistency numerical check). This is indeed the case, as can

    be seen from Fig. 5(b) and the corresponding inset. Hence,

    when used in the KBEs, and the propagation times are large,

    the H and HF self-energy approximations lead to an amplitude

    death of the time-dependent densities, without the tendency to

    a HD-D, in this type of system and regime[31]. On a shorter

    time-scale, the unphysical damping is much harder to observein this case. As the damping also occurs in H and HF, we can

    hypothesize about the existence of an attractor in KBEs.

    A system is said to have an attractor if, for a wide range of

    initial conditions, its dynamics tends towards the same set of

    numerical values [32]. A dynamical systemthat hasreached (or

    is in the vicinity of) a stable fixed point will tend back towards

    this point if a further perturbation is applied.We investigate this

    aspect for a correlated system that has reached an unphysical

    HD-D. In Fig. 6 we show the time-dependent density of a

    Coulomb dimer withU= 3.0, following a kick ofE = 0.01att= 0. When the density ofboth sites is close to0.5, we applied

    a constant field ofE = 0.05 to Site 1. The resulting dynamics

    is quickly damped (in about 30 a.u.t.). This emphasizes thestability of the unphysical solution at 0.5, i.e., the HD-D, and

    is consistent with the dynamical behavior of a system found in

    the vicinity of a fix point.In this Rapid Communication we showed that the KBEs

    yield, for -derivable self-energies, an unphysical behaviorof the electron density-dynamics in both linear and nonlinearresponses. While in the nonlinear response the damping toan unphysical steady state has been known for some time[19,20], we supplemented their findings by pointing out thatthe steady state is a peculiar homogeneous density-distributionand by showing that the previous findings in the nonlinearregime also apply to systems with long-range Coulombinteractions. Furthermore, we showed that the same unphysical

    041103-4

  • 7/25/2019 Unphysical Solutions of the Kadanoff-Baym Equations in Linear Response

    5/6

    RAPID COMMUNICATIONS

    ON THE UNPHYSICAL SOLUTIONS OF THE KADANOFF-BAYM . . . PHYSICAL REVIEW B93, 041103(R) (2016)

    0 25 50 75 100 125 150

    time (a.u.t.)

    0.496

    0.498

    0.5

    0.502

    0.504

    Sited

    ensity

    Site 1

    Site 2

    FIG. 6. Time-dependentdensities of a Coulomb dimer (U= 3.0).

    Att= 0, a deltalike perturbation ofE = 0.01 was applied to Site 1.

    At t= 90, a constant field ofE = 0.05 was switched on at the

    same site.

    behavior is found within 2B and GW, in linear response,for sufficiently long propagation times, irrespective of theinteraction (Hubbard or Coulomb), and in both weakly andstrongly correlated systems. Also, theT-matrix approximationand conserving subsets of diagrams show the same unphysical

    traits[27]. We shelve for later a more detailed analysis. Fora simpler kernel, e.g.,H[G] or HF[G], the Kadanoff-Baymdynamics still damps, but lacks the tendency towards a HD-Dshown by correlated approximations in these types of systems.

    Taken altogether, the findings in Refs. [19,20] and hereinshow that the unphysical behavior is universal, i.e., across allregimes, and points to the existence of attractors in KBEs.

    We investigated this hypothesis by showing that, once theunphysical steady state is reached, perturbing it will lead toan even faster damping. This emphasizes the stability of theunphysical solution.

    Because in the linear response regime the time-propagationof the KBEs corresponds to solving the Bethe-Salpeterequation with advanced kernels [8,23], the numerical evidencehere suggests that a self-consistent Bethe-Salpeter approachmay suffer from similar pathologies. Whether or not this is thecase remains an openquestion. However, these findings furtherlimit the applicability of the KBEs to very weakly interactingsystems in linear response and to systems coupled to externalleads, e.g., for which the lifetime of the excitations is shorter

    than the artificial damping[20]. This work adds to the recentfindings of other unphysical solutions in many-body perturba-tion theory[3337] and suggests the need for a fundamentallydifferent approach to quantum many-body systems.

    The author acknowledges funding by the Academy ofFinland under Grant No. 140327/2010, andwould like to thankAna Rotili and Lucia Reining for useful discussions.

    [1] L. P. Kadanoff and G. Baym, Quantum Statistical Mechanics

    (Benjamin, New York, 1962).

    [2] G. Baym and L. P. Kadanoff, Conservation laws and correlationfunctions,Phys. Rev.124,287(1961).

    [3] G. Baym, Self-Consistent approximations in many-body sys-

    tems,Phys. Rev.127,1391(1962).

    [4] N. E. Dahlen and R. van Leeuwen, Solving the Kadanoff-Baym

    Equations for Inhomogeneous Systems: Application to Atoms

    and Molecules,Phys. Rev. Lett.98,153004(2007).

    [5] P. Myohanen, A. Stan, G. Stefanucci, and R. van Leeuwen,

    A many-body approach to quantum transport dynamics: Initial

    correlations and memory effects, Europhys. Lett. 84, 67001

    (2008).

    [6] K. S. Thygesen and A. Rubio, Conserving GW scheme for

    nonequilibrium quantum transport in molecular contacts, Phys.

    Rev. B77,115333(2008).[7] A.-M. Uimonen, E. Khosravi, A. Stan, G. Stefanucci, S. Kurth,

    R. van Leeuwen, and E. K. U. Gross, Comparative study of

    many-body perturbation theory and time-dependent density

    functional theory in the out-of-equilibrium Anderson model,

    Phys. Rev. B84,115103(2011).

    [8] N.Sakkinen,M. Manninen, and R. vanLeeuwen, The Kadanoff-

    Baym approach to double excitations in finite systems,New J.

    Phys.14,013032(2012).

    [9] S. Hermanns, N. Schlunzen, and M. Bonitz, Hubbard nanoclus-

    ters far from equilibrium,Phys. Rev. B90,125111(2014).

    [10] P. C. Martin and J. Schwinger, Theory of Many-Particle

    Systems. I,Phys. Rev.115,1342(1959).

    [11] O. V. Konstantinov and V. I. Perel, A Diagram technique for

    evaluating transport quantities, Zh. Eksp. Teor. Fiz. 39, 197

    (1960) [Sov. Phys. JETP 12, 142 (1961)].[12] J. M. Luttinger and J. C. Ward, Ground-state energy of a many-

    fermion system. II,Phys. Rev.118,1417(1960).

    [13] U. von Barth, N. E. Dahlen, R. van Leeuwen, and G. Ste-

    fanucci, Conserving approximations in time-dependent density

    functional theory,Phys. Rev. B72,235109(2005).

    [14] R. van Leeuwen, N. E. Dahlen, and A. Stan, Total energies

    from variational functionals of the Green function and the

    renormalized four-point vertex, Phys. Rev. B 74, 195105

    (2006).

    [15] A. Stan, N. E. Dahlen, and R. van Leeuwen, Time propagation

    of the Kadanoff-Baym equations for inhomogeneous systems,

    J. Chem. Phys.130,224101(2009).

    [16] L. Hedin, New method for calculating the one-particle Greensfunctionwith application to the electron-gas problem, Phys. Rev.

    139,A796(1965).

    [17] P. Danielewicz, Quantum theory of nonequilibrium processes,

    I,Ann. Phys. (N.Y.)152,239(1984).

    [18] N. E. Dahlen, R. van Leeuwen, and A. Stan, Propagating the

    Kadanoff-Baym equations for atoms and molecules, J. Phys.:

    Conf. Ser.35,340(2006).

    [19] M. Puig von Friesen, C. Verdozzi, and C.-O. Almbladh, Suc-

    cesses and Failures of Kadanoff-Baym Dynamics in Hubbard

    Nanoclusters,Phys. Rev. Lett.103,176404(2009).

    [20] M. Puig von Friesen, C. Verdozzi, and C.-O. Almbladh,

    Kadanoff-Baym dynamics of Hubbard clusters: Performance of

    041103-5

    http://dx.doi.org/10.1103/PhysRev.124.287http://dx.doi.org/10.1103/PhysRev.124.287http://dx.doi.org/10.1103/PhysRev.124.287http://dx.doi.org/10.1103/PhysRev.124.287http://dx.doi.org/10.1103/PhysRev.124.287http://dx.doi.org/10.1103/PhysRev.124.287http://dx.doi.org/10.1103/PhysRev.127.1391http://dx.doi.org/10.1103/PhysRev.127.1391http://dx.doi.org/10.1103/PhysRev.127.1391http://dx.doi.org/10.1103/PhysRev.127.1391http://dx.doi.org/10.1103/PhysRev.127.1391http://dx.doi.org/10.1103/PhysRev.127.1391http://dx.doi.org/10.1103/PhysRevLett.98.153004http://dx.doi.org/10.1103/PhysRevLett.98.153004http://dx.doi.org/10.1103/PhysRevLett.98.153004http://dx.doi.org/10.1103/PhysRevLett.98.153004http://dx.doi.org/10.1103/PhysRevLett.98.153004http://dx.doi.org/10.1103/PhysRevLett.98.153004http://dx.doi.org/10.1209/0295-5075/84/67001http://dx.doi.org/10.1209/0295-5075/84/67001http://dx.doi.org/10.1209/0295-5075/84/67001http://dx.doi.org/10.1209/0295-5075/84/67001http://dx.doi.org/10.1209/0295-5075/84/67001http://dx.doi.org/10.1209/0295-5075/84/67001http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.84.115103http://dx.doi.org/10.1103/PhysRevB.84.115103http://dx.doi.org/10.1103/PhysRevB.84.115103http://dx.doi.org/10.1103/PhysRevB.84.115103http://dx.doi.org/10.1103/PhysRevB.84.115103http://dx.doi.org/10.1103/PhysRevB.84.115103http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1103/PhysRevB.90.125111http://dx.doi.org/10.1103/PhysRevB.90.125111http://dx.doi.org/10.1103/PhysRevB.90.125111http://dx.doi.org/10.1103/PhysRevB.90.125111http://dx.doi.org/10.1103/PhysRevB.90.125111http://dx.doi.org/10.1103/PhysRev.115.1342http://dx.doi.org/10.1103/PhysRev.115.1342http://dx.doi.org/10.1103/PhysRev.115.1342http://dx.doi.org/10.1103/PhysRev.115.1342http://dx.doi.org/10.1103/PhysRev.115.1342http://dx.doi.org/10.1103/PhysRev.115.1342http://dx.doi.org/10.1103/PhysRev.118.1417http://dx.doi.org/10.1103/PhysRev.118.1417http://dx.doi.org/10.1103/PhysRev.118.1417http://dx.doi.org/10.1103/PhysRev.118.1417http://dx.doi.org/10.1103/PhysRev.118.1417http://dx.doi.org/10.1103/PhysRevB.72.235109http://dx.doi.org/10.1103/PhysRevB.72.235109http://dx.doi.org/10.1103/PhysRevB.72.235109http://dx.doi.org/10.1103/PhysRevB.72.235109http://dx.doi.org/10.1103/PhysRevB.72.235109http://dx.doi.org/10.1103/PhysRevB.72.235109http://dx.doi.org/10.1103/PhysRevB.74.195105http://dx.doi.org/10.1103/PhysRevB.74.195105http://dx.doi.org/10.1103/PhysRevB.74.195105http://dx.doi.org/10.1103/PhysRevB.74.195105http://dx.doi.org/10.1103/PhysRevB.74.195105http://dx.doi.org/10.1103/PhysRevB.74.195105http://dx.doi.org/10.1063/1.3127247http://dx.doi.org/10.1063/1.3127247http://dx.doi.org/10.1063/1.3127247http://dx.doi.org/10.1063/1.3127247http://dx.doi.org/10.1063/1.3127247http://dx.doi.org/10.1063/1.3127247http://dx.doi.org/10.1103/PhysRev.139.A796http://dx.doi.org/10.1103/PhysRev.139.A796http://dx.doi.org/10.1103/PhysRev.139.A796http://dx.doi.org/10.1103/PhysRev.139.A796http://dx.doi.org/10.1103/PhysRev.139.A796http://dx.doi.org/10.1103/PhysRev.139.A796http://dx.doi.org/10.1016/0003-4916(84)90092-7http://dx.doi.org/10.1016/0003-4916(84)90092-7http://dx.doi.org/10.1016/0003-4916(84)90092-7http://dx.doi.org/10.1016/0003-4916(84)90092-7http://dx.doi.org/10.1016/0003-4916(84)90092-7http://dx.doi.org/10.1016/0003-4916(84)90092-7http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1103/PhysRevLett.103.176404http://dx.doi.org/10.1103/PhysRevLett.103.176404http://dx.doi.org/10.1103/PhysRevLett.103.176404http://dx.doi.org/10.1103/PhysRevLett.103.176404http://dx.doi.org/10.1103/PhysRevLett.103.176404http://dx.doi.org/10.1103/PhysRevLett.103.176404http://dx.doi.org/10.1103/PhysRevLett.103.176404http://dx.doi.org/10.1103/PhysRevLett.103.176404http://dx.doi.org/10.1103/PhysRevLett.103.176404http://dx.doi.org/10.1103/PhysRevLett.103.176404http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1088/1742-6596/35/1/031http://dx.doi.org/10.1016/0003-4916(84)90092-7http://dx.doi.org/10.1016/0003-4916(84)90092-7http://dx.doi.org/10.1016/0003-4916(84)90092-7http://dx.doi.org/10.1016/0003-4916(84)90092-7http://dx.doi.org/10.1103/PhysRev.139.A796http://dx.doi.org/10.1103/PhysRev.139.A796http://dx.doi.org/10.1103/PhysRev.139.A796http://dx.doi.org/10.1103/PhysRev.139.A796http://dx.doi.org/10.1063/1.3127247http://dx.doi.org/10.1063/1.3127247http://dx.doi.org/10.1063/1.3127247http://dx.doi.org/10.1063/1.3127247http://dx.doi.org/10.1103/PhysRevB.74.195105http://dx.doi.org/10.1103/PhysRevB.74.195105http://dx.doi.org/10.1103/PhysRevB.74.195105http://dx.doi.org/10.1103/PhysRevB.74.195105http://dx.doi.org/10.1103/PhysRevB.72.235109http://dx.doi.org/10.1103/PhysRevB.72.235109http://dx.doi.org/10.1103/PhysRevB.72.235109http://dx.doi.org/10.1103/PhysRevB.72.235109http://dx.doi.org/10.1103/PhysRev.118.1417http://dx.doi.org/10.1103/PhysRev.118.1417http://dx.doi.org/10.1103/PhysRev.118.1417http://dx.doi.org/10.1103/PhysRev.118.1417http://dx.doi.org/10.1103/PhysRev.115.1342http://dx.doi.org/10.1103/PhysRev.115.1342http://dx.doi.org/10.1103/PhysRev.115.1342http://dx.doi.org/10.1103/PhysRev.115.1342http://dx.doi.org/10.1103/PhysRevB.90.125111http://dx.doi.org/10.1103/PhysRevB.90.125111http://dx.doi.org/10.1103/PhysRevB.90.125111http://dx.doi.org/10.1103/PhysRevB.90.125111http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1088/1367-2630/14/1/013032http://dx.doi.org/10.1103/PhysRevB.84.115103http://dx.doi.org/10.1103/PhysRevB.84.115103http://dx.doi.org/10.1103/PhysRevB.84.115103http://dx.doi.org/10.1103/PhysRevB.84.115103http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1103/PhysRevB.77.115333http://dx.doi.org/10.1209/0295-5075/84/67001http://dx.doi.org/10.1209/0295-5075/84/67001http://dx.doi.org/10.1209/0295-5075/84/67001http://dx.doi.org/10.1209/0295-5075/84/67001http://dx.doi.org/10.1103/PhysRevLett.98.153004http://dx.doi.org/10.1103/PhysRevLett.98.153004http://dx.doi.org/10.1103/PhysRevLett.98.153004http://dx.doi.org/10.1103/PhysRevLett.98.153004http://dx.doi.org/10.1103/PhysRev.127.1391http://dx.doi.org/10.1103/PhysRev.127.1391http://dx.doi.org/10.1103/PhysRev.127.1391http://dx.doi.org/10.1103/PhysRev.127.1391http://dx.doi.org/10.1103/PhysRev.124.287http://dx.doi.org/10.1103/PhysRev.124.287http://dx.doi.org/10.1103/PhysRev.124.287http://dx.doi.org/10.1103/PhysRev.124.287
  • 7/25/2019 Unphysical Solutions of the Kadanoff-Baym Equations in Linear Response

    6/6

    RAPID COMMUNICATIONS

    ADRIAN STAN PHYSICAL REVIEW B93, 041103(R) (2016)

    many-body schemes, correlation-induced damping and multiple

    steady and quasi-steady states,Phys. Rev. B82,155108(2010).

    [21] P. Myohanen, A. Stan, G. Stefanucci, and R. van Leeuwen,

    Kadanoff-Baym approach to quantum transport through inter-

    acting nanoscale systems: From the transient to the steady-state

    regime,Phys. Rev. B80,115107(2009).

    [22] R. van Leeuwen, The Sham-Schluter Equation in Time-

    DependentDensity-Functional Theory, Phys. Rev. Lett. 76, 3610

    (1996).

    [23] N.-H. Kwong and M. Bonitz, Real-Time Kadanoff-Baym

    Approach to Plasma Oscillations in a Correlated Electron Gas,

    Phys. Rev. Lett.84,1768(2000).

    [24] Anna-Maija Maaret Uimonen (unpublished, 2009).

    [25] K. Ohno, Some remarks on the Pariser-Parr-Pople method,

    Theor. Chim. Acta2,219(1964).

    [26] J. Linderberg and Y. Ohrn, Derivationand analysisof the Pariser-

    Parr-Pople model,J. Chem. Phys.49,716(1968).

    [27] A. Stan, Properties of the Screened Interaction in Finite

    Systems, at 16th ETSF Workshop on Electronic Excita-

    tions, Turin, Italy (unpublished, 2011), http://etsf.grenoble.

    cnrs.fr/events/book_abstracts.pdf.[28] A. L. Fetter and J. D. Walecka, Quantum Theory of Many-

    Particle Systems(McGraw-Hill, New York, 1971).

    [29]See Supplemental Material at http://link.aps.org/supplemental/

    10.1103/PhysRevB.93.041103 w hich includes Refs.

    [9,19,20,24].

    [30] A. Stan, N. E. Dahlen, and R. van Leeuwen, Levels of self-

    consistency in the GW approximation, J. Chem. Phys. 130,

    114105(2009).

    [31] The TDHF densities, calculated from the KBEs, show a similar

    damping pattern also in the case of strong perturbations. We

    defer a detailed discussion for elsewhere.

    [32] S. H. Strogatz, Nonlinear Dynamics and Chaos (Perseus,

    Cambridge, MA, 1994).

    [33] J. A. Berger, P. Romaniello, F. Tandetzky, B. S.

    Mendoza, C. Brouder, and L. Reining, Solution to the

    many-body problem in one point, New J. Phys. 16, 113025

    (2014).

    [34] F. Tandetzky, J. K. Dewhurst, S. Sharma, and E. K. U. Gross,

    Multiplicity of solutions to GW-type approximations, Phys. Rev.

    B92,115125(2015).

    [35] E. Kozik, M. Ferrero, and A. Georges, Nonexistence of the

    Luttinger-Ward Functional and Misleading Convergence of

    Skeleton Diagrammatic Series for Hubbard-like Models, Phys.

    Rev. Lett.114,156402(2015).

    [36] R. Rossi and F. Werner, Skeleton series and multivaluedness

    of the self-energy functional in zero space-time dimensions, J.Phys. A48,485202(2015).

    [37] A. Stan, P. Romaniello, S. Rigamonti, L. Reining, and

    J. A. Berger, Unphysical and physical solutions in many-body

    theories: From weak to strong correlation, New J. Phys. 17,

    093045(2015).

    041103-6

    http://dx.doi.org/10.1103/PhysRevB.82.155108http://dx.doi.org/10.1103/PhysRevB.82.155108http://dx.doi.org/10.1103/PhysRevB.82.155108http://dx.doi.org/10.1103/PhysRevB.82.155108http://dx.doi.org/10.1103/PhysRevB.82.155108http://dx.doi.org/10.1103/PhysRevB.82.155108http://dx.doi.org/10.1103/PhysRevB.80.115107http://dx.doi.org/10.1103/PhysRevB.80.115107http://dx.doi.org/10.1103/PhysRevB.80.115107http://dx.doi.org/10.1103/PhysRevB.80.115107http://dx.doi.org/10.1103/PhysRevB.80.115107http://dx.doi.org/10.1103/PhysRevB.80.115107http://dx.doi.org/10.1103/PhysRevLett.76.3610http://dx.doi.org/10.1103/PhysRevLett.76.3610http://dx.doi.org/10.1103/PhysRevLett.76.3610http://dx.doi.org/10.1103/PhysRevLett.76.3610http://dx.doi.org/10.1103/PhysRevLett.76.3610http://dx.doi.org/10.1103/PhysRevLett.76.3610http://dx.doi.org/10.1103/PhysRevLett.84.1768http://dx.doi.org/10.1103/PhysRevLett.84.1768http://dx.doi.org/10.1103/PhysRevLett.84.1768http://dx.doi.org/10.1103/PhysRevLett.84.1768http://dx.doi.org/10.1103/PhysRevLett.84.1768http://dx.doi.org/10.1007/BF00528281http://dx.doi.org/10.1007/BF00528281http://dx.doi.org/10.1007/BF00528281http://dx.doi.org/10.1007/BF00528281http://dx.doi.org/10.1007/BF00528281http://dx.doi.org/10.1063/1.1670129http://dx.doi.org/10.1063/1.1670129http://dx.doi.org/10.1063/1.1670129http://dx.doi.org/10.1063/1.1670129http://dx.doi.org/10.1063/1.1670129http://dx.doi.org/10.1063/1.1670129http://dx.doi.org/10.1063/1.1670129http://etsf.grenoble.cnrs.fr/events/book_abstracts.pdfhttp://etsf.grenoble.cnrs.fr/events/book_abstracts.pdfhttp://etsf.grenoble.cnrs.fr/events/book_abstracts.pdfhttp://etsf.grenoble.cnrs.fr/events/book_abstracts.pdfhttp://link.aps.org/supplemental/10.1103/PhysRevB.93.041103http://link.aps.org/supplemental/10.1103/PhysRevB.93.041103http://link.aps.org/supplemental/10.1103/PhysRevB.93.041103http://link.aps.org/supplemental/10.1103/PhysRevB.93.041103http://dx.doi.org/10.1063/1.3089567http://dx.doi.org/10.1063/1.3089567http://dx.doi.org/10.1063/1.3089567http://dx.doi.org/10.1063/1.3089567http://dx.doi.org/10.1063/1.3089567http://dx.doi.org/10.1088/1367-2630/16/11/113025http://dx.doi.org/10.1088/1367-2630/16/11/113025http://dx.doi.org/10.1088/1367-2630/16/11/113025http://dx.doi.org/10.1088/1367-2630/16/11/113025http://dx.doi.org/10.1088/1367-2630/16/11/113025http://dx.doi.org/10.1088/1367-2630/16/11/113025http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1367-2630/17/9/093045http://dx.doi.org/10.1088/1367-2630/17/9/093045http://dx.doi.org/10.1088/1367-2630/17/9/093045http://dx.doi.org/10.1088/1367-2630/17/9/093045http://dx.doi.org/10.1088/1367-2630/17/9/093045http://dx.doi.org/10.1088/1367-2630/17/9/093045http://dx.doi.org/10.1088/1367-2630/17/9/093045http://dx.doi.org/10.1088/1367-2630/17/9/093045http://dx.doi.org/10.1088/1367-2630/17/9/093045http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1088/1751-8113/48/48/485202http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevLett.114.156402http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1103/PhysRevB.92.115125http://dx.doi.org/10.1088/1367-2630/16/11/113025http://dx.doi.org/10.1088/1367-2630/16/11/113025http://dx.doi.org/10.1088/1367-2630/16/11/113025http://dx.doi.org/10.1088/1367-2630/16/11/113025http://dx.doi.org/10.1063/1.3089567http://dx.doi.org/10.1063/1.3089567http://dx.doi.org/10.1063/1.3089567http://dx.doi.org/10.1063/1.3089567http://link.aps.org/supplemental/10.1103/PhysRevB.93.041103http://etsf.grenoble.cnrs.fr/events/book_abstracts.pdfhttp://dx.doi.org/10.1063/1.1670129http://dx.doi.org/10.1063/1.1670129http://dx.doi.org/10.1063/1.1670129http://dx.doi.org/10.1063/1.1670129http://dx.doi.org/10.1007/BF00528281http://dx.doi.org/10.1007/BF00528281http://dx.doi.org/10.1007/BF00528281http://dx.doi.org/10.1007/BF00528281http://dx.doi.org/10.1103/PhysRevLett.84.1768http://dx.doi.org/10.1103/PhysRevLett.84.1768http://dx.doi.org/10.1103/PhysRevLett.84.1768http://dx.doi.org/10.1103/PhysRevLett.84.1768http://dx.doi.org/10.1103/PhysRevLett.76.3610http://dx.doi.org/10.1103/PhysRevLett.76.3610http://dx.doi.org/10.1103/PhysRevLett.76.3610http://dx.doi.org/10.1103/PhysRevLett.76.3610http://dx.doi.org/10.1103/PhysRevB.80.115107http://dx.doi.org/10.1103/PhysRevB.80.115107http://dx.doi.org/10.1103/PhysRevB.80.115107http://dx.doi.org/10.1103/PhysRevB.80.115107http://dx.doi.org/10.1103/PhysRevB.82.155108http://dx.doi.org/10.1103/PhysRevB.82.155108http://dx.doi.org/10.1103/PhysRevB.82.155108http://dx.doi.org/10.1103/PhysRevB.82.155108