Quantum Monte Carlo Simulations of Two-Dimensional Repulsive Fermi Gases With Population Imbalance
Quantum Monte Carlo Simulations of Two-Dimensional Repulsive Fermi Gases With Population Imbalance
Quantum Monte Carlo Simulations of Two-Dimensional Repulsive Fermi Gases With Population Imbalance
population imbalance
larons. To fully characterize this expansion, we determine the polarons’ effective mass and their
coupling parameter, complementing previous studies on their chemical potential. Furthermore, we
extract the magnetic susceptibility from low-imbalance data, finding only small deviations from the
mean-field prediction. While the mean-field theory predicts a direct transition from a paramagnetic
to a fully ferromagnetic phase, our diffusion Monte Carlo results suggest that the partially ferro-
magnetic phase is stable in a narrow interval of the interaction parameter. This finding calls for
further analyses on the effects due to the fixed-node constraint.
X X ~2 2 X
-0.1 H=− ∇iσ + v(ri↑ i↓ ) ; (1)
i =1
2m i ,i
σ=↑,↓ σ ↑ ↓
-0.15
here, m is the particle mass for both components and ~ is
-0.2 HD the reduced Planck constant. The indices i↑ = 1, . . . , N↑
SD and i↓ = 1, . . . , N↓ label atoms of the two components,
-0.25
2nd order hereafter referred to as spin-up and spin-down particles.
-0.3 Fully polarized
The distance between opposite-spin fermions is ri↑ i↓ =
0 0.1 0.2 0.3 0.4 0.5 0.6
kFa2D ri↑ − ri↓ . The total number of fermions is N = N↑ +N↓ ,
and the polarization is defined as P = (N↑ − N↓ )/N .
The particles move is a square box of size L with pe-
FIG. 1. (Color online) Equation of state at zero population
riodic boundary conditions. The total density is thus
imbalance, i.e., P = 0. The dimensionless correlation energy
(e − eMF )/eFG is plotted as a function of the dimensionless
n = N/L2 = n↑ + n↓ , where the partial densities are
interaction parameter kF a2D . e is the ground-state energy nσ = Nσ /L2 . The latter allow defining the Fermi ener-
per particle, eMF is the mean-field result (see Eq. (6)), and gies of the component σ: EFσ = (~kFσ )2 /2m, √ where the
eFG is the energy of a balanced ideal Fermi gas. (Red) points corresponding Fermi wavevectors are kFσ = 4πnσ . v(r)
correspond to the hard-disk intercomponent potential (HD), is a short-range potential that describes the intercom-
and (green) empty squares correspond to the soft-disk (SD) ponent interactions. We consider two model potentials.
potential with range R = 2a2D . The number of particles is The first is the hard-disk (HD) model: v(r) = +∞ if
N = 98. The continuous (black) curve indicates the second- r < a2D and zero otherwise. The disk diameter coincides
order equation of state (see Eq. (8)). The long-dash (purple) with the 2D s-wave scattering length a2D . The second
curve indicates the energy of the fully imbalanced ideal Fermi one is the soft-disk (SD) potential: v(r) = V0 if r < R
gas. The dot-dash curves are guides to the eye. Here and
and zero otherwise, where V0 > 0 is the potential inten-
in all figures, the errorbars are smaller than the symbol size
when not visible. sity. In this case, the scattering length is related to the
disk diameter R by the relation:
1 I0 (K0 R)
a2D = R exp − , (2)
K0 R I00 (K0 R)
where K0 = mV0 /~2 and I0 (x) is the modified Bessel
function of the first kind and I00 (x) its derivative. In this
rameter. Furthermore, from low-imbalance data, we de- Article, we set R and V0 so that R = 2a2D . This al-
termine the magnetic susceptibility. These quantities al- lows us to analyze the possible role played by details of
low us to estimate the critical interaction strength for the the model potential beyond the s-wave scattering length.
Stoner ferromagnetic instability. The transition from a Notice that, due to the logarithmic dependence of the 2D
paramagnetic to a partially ferromagnetic ground state is scattering amplitude on energy, in the literature various
signaled by the divergence of the susceptibility. The sta- definitions of the 2D scattering length have been used.
bility region of the fully ferromagnetic phase is identified We stick to the notation that is most natural for hard-
from the polaron chemical potential. As we discuss, while disks – namely, a2D corresponds to the disk diameter –
the mean-field theory predicts a direct transition from a which was also used in Refs. [47, 48, 50, 53, 64]. An
paramagnetic to a fully ferromagnetic phase [36, 62, 63], alternative definition, which is often used when consid-
our QMC results suggest that a partially ferromagnetic ering attractive interactions [43, 56, 59], sets the scat-
phase is stable in a narrow intermediate window of the tering length equal to b, where the dimer binding en-
interaction parameter. ergy is |B | = ~2 /mb2 . The relation between the two
definitions is b = a2D eγ /2, where γ ∼ = 0.577 is Euler-
The rest of the Article is organized as follows: the Mascheroni’s constant, so that the coupling constant
model Hamiltonian and our computational method are ln (kF b) = ln (kF a2D ) + γ − ln 2 ' ln (kF a2D ) − 0.12. In
described in Section II. Some known results from pertur- cold-atom experiments, 2D systems are created by lim-
bative expansions are reviewed in Section III. Our QMC iting the particle motion in one direction to zero-point
results for the zero-temperature equation of state, for oscillations using a strong confining potential. The effec-
the polaron’s properties, and the analysis of the onset tive 2D scattering length for dilute gases is determined
of ferromagnetism are reported in Section IV. Section V by solving the scattering problem in the presence of such
provides a summary of the main findings and discusses strong confinement, integrating over virtual excitations
some future perspectives. induced by the short-range interaction [65, 66].
3
e/eFG
quastic Hamiltonians, this algorithm provides unbiased 1.6
estimates of the energy, provided that possible biases 1.4
due to the finite time step and the finite random-walker
population are reduced below the statistical uncertainty. 1.2
In order to circumvent the negative-sign problem, which 1.0
affects fermionic simulations in dimensions D > 1, the
fixed-node constraint is introduced. It consists in impos- (b)
ing that the nodal surface of the many-body wavefunction 2.0
is the same as that of a suitably chosen trial wavefunction
ψT . The predicted energies are rigorous variational up-
e/eFG
per bounds and are very close to the exact ground state
energy if the nodes of ψT are good approximations of the 1.9
ground-state nodal surface. We choose trial wavefunc- N = 50
tions of the Jastrow-Slater type, defined as: N = 74
1.8 N = 98
Y N = 122
ψT (r1 , ..., rN ) = D↑ (N↑ )D↓ (N↓ ) f (ri↑ i↓ ) , (3)
0 0.2 0.4 0.6 0.8 1
i↑ ,i↓
P
where D↑(↓) denotes the Slater determinant of single-
particle plane waves for the spin-up (spin-down) parti-
FIG. 2. (Color online) Equation of state of imbalanced
cles. The Jastrow correlation term f (r) is taken to be Fermi gases. The energy per particle e/eFG is plotted as a
the solution of the s-wave radial Schrödinger equation de- function of the population imbalance P . eFG is the energy of
scribing two-particle scattering with the potential v(r). the balanced ideal Fermi gas. Different symbols correspond
The scattering energy is set so that f 0 (r = L/2) = 0 (for to different particle numbers N . Different datasets corre-
more details, see Ref. [64]). Since f (r) > 0, the nodal spond to different interaction parameters kF a2D , increasing
surface is determined by the Slater determinants. from bottom to top. Panel (a) includes data for kF a2D ∼ =
Beyond the aforementioned possible biases, the QMC 0, 0.0022, 0.0332, 0.111, 0.222, 0.332, 0.410, 0.443, 0.487, 0.554.
results might be affected by finite-size effects. To reduce Panel (b) is a zoom in the vertical axis, including only
them, we correct the QMC energies using the finite-size data for kF a2D ∼ = 0.222, 0.332, 0.410, 0.443, 0.487. The (red)
continuous curves represent the quadratic fitting functions
correction corresponding to noninteracting gases with the
defined in Eq. (11). The (brown) dashed curves represent the
same partial densities. This correction is rescaled accord- Landau-Pomeranchuk functional, namely, Eq. (10), which is
ing to the particle’s effective mass m∗0 of the interact- applicable in the large polarization regime.
ing system, hleading to the following correction
i formula:
↑ ↓
E → E − Eid (N↑ , N↓ ) − N↑ eFG − N↓ eFG m/m∗0 [67],
where eσFG ≡ EFG σ
/Nσ = EFσ /2 is the energy per parti-
cle of the ideal fully imbalanced Fermi gas of compo-
nent σ, and id refers to ideal gases. Notice that we high-imbalance regime, since the interaction effects on
use the same approximation for the effective mass of the majority component are strongly reduced, arguably
both components. When extracting the magnetic sus- leading to an effective mass closer to the particle mass
ceptibility from low-imbalance data (and when display- m.
ing the QMC results in Fig. 2), we use the estimate from The polaron effective mass can be determined in DMC
second-order perturbation theory for balanced gases [68]: simulations from the imaginary-time diffusion coefficient
m∗0 /m = 1 + 2/ ln2 c0 na22D , where c0 = πe2γ /2 ∼
= 4.98. of a spin-down impurity in a spin-up Fermi sea, as [69,
See the discussion in Section III for the choice of c0 . To 70]:
account for the uncertainty in the validity of the pertur-
bative effective mass, half of the difference between the
above correction and the one obtained with m∗0 = m is m∗ m D 2
E
summed in quadrature to the statistical uncertainty. The = lim (r↓ (τ ) − r↓ (0)) , (4)
m τ →∞ 2τ ~
above energy correction is implicit in the definition of
the polaron chemical potential (see Section IV). When
extracting the polaron coupling parameter from high- where r↓ (τ ) is the impurity position at imaginary time τ
imbalance data, we compute the energy correction with and the angular brackets indicate the Monte Carlo aver-
m∗0 = m. This choice is indeed more appropriate for the age.
4
A
HD
Fermi system this results in an energy-dependent cou- 1 SD
pling g̃(E) = (4π~2 /m)/ ln[Ea /(µ↑ + µ↓ + E)], where HD (Bombin et al.)
Ea = 4~2 /ma22D e2γ and µσ is the chemical potential of 0.5
MF
the spin-σ component. Differently from the 3D case, the µ↑/e↑FG
zero-energy limit E → 0 displays a significant density
0
dependence, albeit logarithmic, through the chemical po- 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
tentials. By approximating the latter with the Fermi en- ↑
kF a2D
ergies EFσ we obtain g̃ = (4π~2 /m)/| ln(c0 na22D )|, where,
as already mentioned in Section II, c0 = πe2γ /2 ∼ = 4.98.
Notice that only the total density n = n↑ + n↓ appears, FIG. 3. (Color online) Chemical potential at zero concen-
tration of the repulsive polaron A as a function of the inter-
even in the imbalanced case, due to the linearity of the
action parameter kF↑ a2D , in units of the energy per particle
Fermi energy with the 2D density. In the literature, it is √
of the fully imbalanced ideal Fermi gas e↑FG . kF↑ = 4πn↑ is
sometimes set c0 = 1 [72], which is correct within loga- the Fermi wavevector of the fully polarized ideal Fermi gas.
rithmic accuracy in the weak-coupling regime. However, (Blue) squares correspond to the hard-disk (HD) potential.
we observe that our choice for c0 allows for much more (Red) triangles correspond to the soft-disk (SD) potential.
accurate perturbative expressions, when compared to the The number of majority-spin particles is N↑ = 61. The con-
nonperturbative results, consistently with [53, 54, 68]. tinuous (brown) curve represents the mean-field prediction
At the mean-field level, the energy density ε = E/V in Eq. (9). The thin horizontal (black) line indicates the chemi-
a volume V = L2 is the sum of the kinetic contributions cal potential of the majority component µ↑ . The empty (blue)
of the two components, plus an interaction term which squares represent the results from Ref. [54].
is given by the coupling constant times the number of
possible pairs, leading to:
εMF (n↑ , n↓ ) = e↑FG n↑ + e↓FG n↓ + g̃n↑ n↓ . (5) IV. RESULTS
This standard expression is valid both for balanced and
imbalanced systems [72]. In the balanced case (n↑ = The zero-temperature equation of state for balanced
n↓ = n/2), it readily yields the energy per particle: populations (corresponding to the polarization P = 0)
has been investigated in Ref. [53]. In Fig. 1, our QMC
eMF (n) ≡ εMF (n/2, n/2)/n = eFG (1 + 2g) , (6) results for the energy per particle e = E/N are plotted
where eFG = EF /2 = π~2 n/2m is the energy per parti- as a function of the dimensionless
√ interaction parame-
cle of the ideal balanced Fermi gas, and we defined the ter kF a2D , where kF = 2πn is the Fermi wavevector
dimensionless expansion parameter of the balanced ideal Fermi gas. To better visualize the
interaction effects, we subtract the mean-field prediction
1
g= . (7) Eq. (6). We also show the perturbative second-order re-
| ln(c0 na22D )| sult Eq. (8) with the continuous black curve. Evidently,
The second-order expansion in g for the balanced case is this second-order expansion is valid only for relatively
also known [68, 73, 74]. It reads weak interactions kF a2D . 0.1 [53]. At relatively strong
interactions kF a2D ' 0.45, the energy of the balanced
e2nd (n) = eFG [1 + 2g + (3 − 4 ln 2)g 2 ]. (8) gas overcomes the one of the fully polarized configura-
The coefficient of the second-order term is fixed by the tion (corresponding to P = 1). This implies that the
choice of c0 . paramagnetic phase is unstable [17]. In this regime, the
The single-polaron chemical potential is the energy of a results for the HD and the SD potentials deviate by less
single spin-down impurity in a Fermi sea of the majority than 1%. This indicates that the onset of ferromagnetism
spin-up component. At the mean-field level, it can be is essentially universal in terms of the 2D s-wave scatter-
derived from Eq. (5), and, in dimensionless form, it reads ing length, while other details of the interaction potential
[54]: play a marginal role.
Precisely locating the transition from the paramag-
AMF = L2 εMF (n↑ , 1/L2 ) − εMF (n↑ , 0) /e↑FG = 4g ,
netic to a ferromagnetic ground-state requires simulating
(9) imbalanced components. In Fig. 2, the energy per par-
where we eventually took the thermodynamic limit and ticle e is plotted as a function of the polarization P , for
g is calculated at n = n↑ . several values of the interaction parameter kF a2D . One
5
m/m*
0.8
↑
m
ELP = EFG 1 + Ax + ∗ x2 + F x2 , (10) 0.7
QMC
m fit
0.6 Ngampruetikorn et al.
where x = N↓ /N↑ is the concentration of the minority Schmidt et al.
component, here identified with the spin-down particles. 0.5 Experiment
The first term on the right hand side is the kinetic en- 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
ergy of the majority component; the second term is pro- ↑
portional to the dimensionless chemical potential at zero kF a2D
concentration of the polarons A(kF↑ a2D ); the third term
represents the kinetic energy of the polaron gas, and it FIG. 4. (Color online) Inverse effective mass m/m∗ as a
is fixed by the polaron effective mass m∗ (kF↑ a2D ). The function of the interaction parameter kF↑ a2D . m is the particle
fourth term represents the energy contribution due to mass. (Blue) squares represent the QMC results for the HD
potential. The continuous (blue) line is a Padé fitting func-
correlations among polarons, and its magnitude is fixed
tion (see text). The dashed (purple) curve and the dot-dash
by the dimensionless coupling parameter F (kF↑ a2D ). A (brown) curve represent two previous theoretical predictions
quadratic scaling with the concentration, namely F x2 , is for upper-branch polarons, extracted from Refs. [55] and [56],
found to accurately describe the QMC data. If the expo- respectively. The (red) circles represent the experimental re-
nent of x is used as a fitting parameter, the results are sults, extracted from Ref. [43].
compatible with the quadratic Ansatz. In fact, this is the
same scaling of the 3D case [3, 41]. An additional loga-
rithmic factor might apply to the 2D case, but it would
not be noticeable on the available range of concentra- polaron effective mass m∗ is determined from QMC sim-
tion values. It is worth noticing that, in two dimensions, ulations with a single spin-down impurity via Eq. (4).
both the third and the fourth terms scale with the second The results for the HD potential are shown in Fig. 4.
power of the concentration x, as opposed to the 3D case, Previous studies reported corresponding QMC data for
m 5/3
where the former scales as m ∗x [3]. dipolar interactions and for soft-core potentials with large
In Fig. 2, Eq. (10) is compared to the QMC datasets effective range [54, 75]. Notably, the effective mass in-
for different (fixed) values of kF a2D , for varying Pq. The creases up to m∗ ≈ 1.35m in the regime where ferro-
following conversion formulas are used: kF↑ = kF 2 magnetism occurs. In three dimensions, the correspond-
1+x ,
↑
ing effect is significantly smaller, with the effective mass
EFG = 2EFG /(1 + x)2 , and x = 1−P 1+P . The po- reaching m∗ ≈ 1.1m [27]. This suggests that correlation
laron chemical potential is determined from QMC sim- effects are more relevant in two dimensions. In Fig. 4,
ulations as the fixed-volume energy difference A = we compare our QMC predictions against two previous
[E(N↑ , 1) − E(N↑ , 0)] /e↑FG , where E(N↑ , N↓ ) is the en- theories for the upper branch of resonant attractive in-
ergy of a gas with N↑ spin-up and N↓ spin-down parti- teractions. The first is the Nozières-Schmitt-Rink calcu-
cles. Our results are shown in Fig. 3. They are com- lation of Ref. [56], the second is the calculation based
pared with the QMC data from Ref. [54], finding excel- on a particle-hole variational wave-function performed
lent agreement. The mean-field prediction of Eq. (9) is in Ref. [55]. While good agreement is found for mod-
also shown. It appears to be accurate only in the regime erate interactions kF↑ . 0.4, in the strongly-interacting
kF↑ a2D . 0.3. When A exceeds the chemical potential regime the two previous theories appear to overestimate
of the majority component µ↑ = EF↑ , a state with two the effective mass compared to the QMC data. This dis-
fully separated domains, each hosting one component crepancy might be attributed to the different approxima-
only, is thermodynamically stable [16]. This criterion al- tions in the compared theories, or to intrinsic differences
lows one to pinpoint the onset of full ferromagnetism at between the repulsive model potentials adopted in our
kF↑ a2D ' 0.69, corresponding to kF a2D ' 0.49 for a gas QMC simulations and the upper-branch models adopted
with globally balanced populations. Again, the small de- in the two previous theories. The experimental results
viations between the HD and the SD results indicate the of Ref. [43] are also shown in Fig. 4. They instead un-
marginal role played by nonuniversal details beyond the derestimate the polaron effective mass compared to the
s-wave scattering length. If one uses the mean-field result QMC prediction. As argued in Ref. [43], the experiment
Eq. (9), the condition for the stability of the fully ferro- might not be describable by a purely 2D model, possibly
magnetic state reads AMF = 4g > 2. This corresponds to explaining this discrepancy (see also Ref. [76]).
the critical interaction parameter kF a2D ∼ = 0.413. The We determine the polaron coupling parameter F by
6
fitting the Landau-Pomeranchuk functional Eq. (10) to a weakly first-order transition, with two competing min-
the HD QMC data in the high-imbalance regime, as il- ima in the e(P ) curve – one at P = 0 and the other at
lustrated in Fig. 2. For this fitting procedure, a pre- finite P – separated by an extremely shallow maximum.
determined parametrization of the polaron chemical po- According to the MF theory, the divergence occurs when
tential and of the effective mass is used. The first is g = 1/2. Interestingly, this critical point coincides with
obtained from the HD data in Fig. 3, which are well de- the MF prediction for the onset of full ferromagnetism
scribed by the following polynomial A(kF↑ a2D ) = 4g + discussed above. Therefore, the MF theory predicts a
aA g 2 + bA g 3 + cA g 4 , with coefficients aA = 1.52(5), direct transition from the paramagnetic to the fully fer-
bA = −6.9(2), and cA = 4.2(2). Here, the expansion pa- romagnetic phase [36, 63]. To locate the transition point
rameter g, defined in Eq. (7), is computed at n = n↑ = from the QMC data, we perform a fit in the critical re-
−γ
kF↑2 /4π. Notice that this expansion is consistent with gion with the scaling law χ ∝ kF a2D − kF acrit
2D , where
the mean-field result Eq. (9) in the weakly-interacting γ = 1 is the susceptibility critical exponent for the fer-
limit. The inverse effective mass is parametrized as: romagnetic transition in metallic systems [79]. This fit
m ↑ 2 2 is represented by the dashed segment in Fig. 5. The
m∗ (kF a2D ) = (1 + am g )/(1 + bm g ), with am = 3.2(6) ∼
and bm = 5.4(7) (see Fig. 4). It is interesting to observe best-fit parameter kF acrit
2D = 0.44 represents an estimate
∗
that this parametrization is consistent with m 2 of the critical interaction strength. This value is siz-
m ' 1+2g
in the weakly-interacting limit, analogously to the per- ably smaller than the critical point for the onset of full
turbative result in the balanced case. To parametrize ferromagnetism (predicted from QMC results) discussed
the polaron coupling parameter, the polynomial function above, namely, kF a2D ' 0.49. Therefore, according to
F (kF↑ a2D ) = aF g 2 + bF g 3 is found to be particularly the QMC data, a partially ferromagnetic phase is sta-
accurate. Indeed, with the optimal fitting parameters ble in the narrow window 0.44 . kF a2D . 0.49. This
aF = 6.2(1) and bF = −4.6(4), the Landau-Pomeranchuk statement should be taken with caution. As discussed
functional accurately describes all high-imbalance data in Section II, the QMC predictions are affected by the
before the ferromagnetic transition [77]. Notice that fixed-node constraint. It is possible that more accurate
these values imply that the polaron coupling parameter nodal surfaces based on, e.g., backflow correlations [60] or
Pfaffian wavefunctions [80, 81], would provide lower vari-
is as large as, e.g., F ≈ 0.43 for kF↑ a2D = 0.3, indicating
ational upper bounds for balanced populations, leading
the relevance of interpolaron correlations. A polynomial
to a direct paramagnetic to fully-ferromagnetic transi-
expansion of the polaron coupling constant F was devel-
tion, as predicted by the MF theory. It is worth mention-
oped in Ref. [41] for 3D attractive Fermi gases using the
ing that beyond mean-field effects were found to open a
variational scheme of Ref. [78]. Our result could serve as
narrow partially ferromagnetic window also in Ref. [36].
a benchmark for analogous studies for 2D systems.
However, that study considered a density-independent
The stability region of the paramagnetic ground-state
coupling constant, which applies to quasi-2D traps in the
can be identified by analyzing the magnetic susceptibil-
h 2 i−1 weakly-interacting regime, where the 3D s-wave scatter-
ity χ = n1 ∂P ∂ e
2 of balanced gases. We extract χ by ing length is much smaller than the cloud size in the
fitting the low-polarization energies with quadratic func- confined direction [82]. It is worth emphasizing that, in
tions written in the form: two dimensions, QMC simulations locate the ferromag-
netic transition at larger interaction strength compared
χ0 2 to the mean-field prediction, as opposed to the 3D case
e(P ) = eFG a + P , (11)
χ where the mean-field theory overestimates the critical in-
teraction strength.
where e(P ) is the energy per particle at polarization P ,
a and χ are the fitting parameters and χ0 = n/(2eFG ) The 2D Stoner ferromagnetic instability has been stud-
is the susceptibility of the 2D balanced ideal Fermi gas. ied also in the 2D Hubbard model with infinite on-site
The inverse susceptibility for the HD potential is shown repulsion [83]. This reference employed QMC algorithms
in Fig. 5. These estimates are averaged over different and predicted a sharp transition from the paramagnetic
fitting windows, extending to different maximum polar- to the fully ferromagnetic phase as the density increases,
izations from P ' 0.10 to the maximum values displayed with a narrow intermediate region where the polarization
in Fig. 2. Notably, we find good agreement with the state could not be unambiguously determined. These
mean-field prediction, which we derive from Eq. (5): QMC results were described in the framework of an
infinite-order phase transition [84]. This is characterized
χMF = χ0 (1 − 2g)−1 . (12) by the vanishing of all coefficients of the polynomial ex-
pansion of e(P ), leading to the flattening of the curve at
Small deviations occur only close to the divergence point. the transition point. This scenario implies a discontinu-
This divergence signals the transition from the param- ous jump of the polarization from P = 0 to the saturation
agnetic to a ferromagnetic phase. This criterion corre- value, as in first-order transitions, but without hysteresis.
sponds to a second-order transition. Our QMC data are The susceptibility exponent in the paramagnetic phase is,
indeed consistent with the second-order scenario. How- again, γ = 1 [84]. Similarly to the results of Ref. [83],
ever, from the numerics one cannot rigorously rule out also our QMC data display a flattening of the e(P ) curve
7
[2] R. A. Duine and A. H. MacDonald, Itinerant ferromag- [20] X. Cui and H. Zhai, Stability of a fully magnetized ferro-
netism in an ultracold atom Fermi gas, Phys. Rev. Lett. magnetic state in repulsively interacting ultracold Fermi
95, 230403 (2005). gases, Phys. Rev. A 81, 041602 (2010).
[3] C. Lobo, A. Recati, S. Giorgini, and S. Stringari, Normal [21] F. Arias de Saavedra, F. Mazzanti, J. Boronat, and
state of a polarized Fermi gas at unitarity, Phys. Rev. A. Polls, Ferromagnetic transition of a two-component
Lett. 97, 200403 (2006). Fermi gas of hard spheres, Phys. Rev. A 85, 033615
[4] S. Pilati and S. Giorgini, Phase separation in a polarized (2012).
Fermi gas at zero temperature, Phys. Rev. Lett. 100, [22] P. Massignan and G. Bruun, Repulsive polarons and itin-
030401 (2008). erant ferromagnetism in strongly polarized Fermi gases,
[5] G. Bertaina and S. Giorgini, Density profiles of polarized Eur. Phys. J. D 65, 83 (2011).
Fermi gases confined in harmonic traps, Phys. Rev. A 79, [23] P. Massignan, M. Zaccanti, and G. M. Bruun, Polarons,
013616 (2009). dressed molecules and itinerant ferromagnetism in ultra-
[6] Y.-i. Shin, C. H. Schunck, A. Schirotzek, and W. Ket- cold Fermi gases, Rep. Prog. Phys. 77, 034401 (2014).
terle, Phase diagram of a two-component Fermi gas with [24] D. Pekker, M. Babadi, R. Sensarma, N. Zinner, L. Pollet,
resonant interactions, Nature 451, 689 (2008). M. W. Zwierlein, and E. Demler, Competition between
[7] S. Nascimbène, N. Navon, K. J. Jiang, L. Tarruell, M. Te- pairing and ferromagnetic instabilities in ultracold Fermi
ichmann, J. McKeever, F. Chevy, and C. Salomon, Col- gases near Feshbach resonances, Phys. Rev. Lett. 106,
lective oscillations of an imbalanced Fermi gas: Axial 050402 (2011).
compression modes and polaron effective mass, Phys. [25] Y.-R. Lee, M.-S. Heo, J.-H. Choi, T. T. Wang, C. A.
Rev. Lett. 103, 170402 (2009). Christensen, T. M. Rvachov, and W. Ketterle, Compress-
[8] A. Schirotzek, C.-H. Wu, A. Sommer, and M. W. Zwier- ibility of an ultracold Fermi gas with repulsive interac-
lein, Observation of Fermi polarons in a tunable Fermi tions, Phys. Rev. A 85, 063615 (2012).
liquid of ultracold atoms, Phys. Rev. Lett. 102, 230402 [26] C. Sanner, E. J. Su, W. Huang, A. Keshet, J. Gillen,
(2009). and W. Ketterle, Correlations and pair formation in a
[9] Y.-i. Shin, A. Schirotzek, C. H. Schunck, and W. Ketterle, repulsively interacting Fermi gas, Phys. Rev. Lett. 108,
Realization of a strongly interacting Bose-Fermi mixture 240404 (2012).
from a two-component Fermi gas, Phys. Rev. Lett. 101, [27] P. N. Ma, S. Pilati, M. Troyer, and X. Dai, Density func-
070404 (2008). tional theory for atomic Fermi gases, Nat. Phys. 8, 601
[10] B. A. Olsen, M. C. Revelle, J. A. Fry, D. E. Sheehy, and (2012).
R. G. Hulet, Phase diagram of a strongly interacting spin- [28] S. Pilati, I. Zintchenko, and M. Troyer, Ferromagnetism
imbalanced Fermi gas, Phys. Rev. A 92, 063616 (2015). of a repulsive atomic Fermi gas in an optical lattice:
[11] S. Nascimbène, N. Navon, S. Pilati, F. Chevy, S. Giorgini, a quantum Monte Carlo study, Phys. Rev. Lett. 112,
A. Georges, and C. Salomon, Fermi-liquid behavior of the 015301 (2014).
normal phase of a strongly interacting gas of cold atoms, [29] S. Pilati and E. Fratini, Ferromagnetism in a repulsive
Phys. Rev. Lett. 106, 215303 (2011). atomic Fermi gas with correlated disorder, Phys. Rev. A
[12] E. C. Stoner, LXXX. Atomic moments in ferromagnetic 93, 051604(R) (2016).
metals and alloys with non-ferromagnetic elements, Lon- [30] C. W. von Keyserlingk and G. J. Conduit, Itinerant fer-
don, Edinburgh Dublin Philos. Mag. J. Sci. 15, 1018 romagnetism in an interacting Fermi gas with mass im-
(1933). balance, Phys. Rev. A 83, 053625 (2011).
[13] G. Valtolina, F. Scazza, A. Amico, A. Burchianti, A. Re- [31] X. Cui and T.-L. Ho, Phase separation in mixtures of
cati, T. Enss, M. Inguscio, M. Zaccanti, and G. Roati, Ex- repulsive Fermi gases driven by mass difference, Phys.
ploring the ferromagnetic behaviour of a repulsive Fermi Rev. Lett. 110, 165302 (2013).
gas through spin dynamics, Nat. Phys. 13, 704 (2017). [32] E. Fratini and S. Pilati, Zero-temperature equation of
[14] G.-B. Jo, Y.-R. Lee, J.-H. Choi, C. A. Christensen, T. H. state and phase diagram of repulsive fermionic mixtures,
Kim, J. H. Thywissen, D. E. Pritchard, and W. Ket- Phys. Rev. A 90, 023605 (2014).
terle, Itinerant ferromagnetism in a Fermi gas of ultracold [33] D. V. Kurlov, S. I. Matveenko, V. Gritsev, and G. V.
atoms, Science 325, 1521 (2009). Shlyapnikov, One-dimensional two-component fermions
[15] G. J. Conduit, A. G. Green, and B. D. Simons, Inho- with contact even-wave repulsion and SU(2)-symmetry-
mogeneous phase formation on the border of itinerant breaking near-resonant odd-wave attraction, Phys. Rev.
ferromagnetism, Phys. Rev. Lett. 103, 207201 (2009). A 99, 043631 (2019).
[16] S. Pilati, G. Bertaina, S. Giorgini, and M. Troyer, Itin- [34] Y. Jiang, D. V. Kurlov, X.-W. Guan, F. Schreck, and
erant ferromagnetism of a repulsive atomic Fermi gas: G. V. Shlyapnikov, Itinerant ferromagnetism in one-
A quantum Monte Carlo study, Phys. Rev. Lett. 105, dimensional two-component Fermi gases, Phys. Rev. A
030405 (2010). 94, 011601(R) (2016).
[17] S.-Y. Chang, M. Randeria, and N. Trivedi, Ferromag- [35] M. Singh, S. Pilati, and G. Orso, Itinerant ferro-
netism in the upper branch of the Feshbach resonance magnetism in the repulsive Hubbard chain with spin-
and the hard-sphere Fermi gas, Proc. Natl. Acad. Sci. anisotropic odd-wave attraction, Phys. Rev. A 102,
U.S.A. 108, 51 (2011). 053301 (2020).
[18] G. J. Conduit and B. D. Simons, Itinerant ferromag- [36] G. J. Conduit, Itinerant ferromagnetism in a two-
netism in an atomic Fermi gas: Influence of population dimensional atomic gas, Phys. Rev. A 82, 043604 (2010).
imbalance, Phys. Rev. A 79, 053606 (2009). [37] G. J. Conduit, Quantum Monte Carlo study of the
[19] P. Massignan, Z. Yu, and G. M. Bruun, Itinerant fer- two-dimensional ferromagnet, Phys. Rev. B 87, 184414
romagnetism in a polarized two-component Fermi gas, (2013).
Phys. Rev. Lett. 110, 230401 (2013).
9
[38] X. Cui and T.-L. Ho, Ground-state ferromagnetic tran- [57] M. M. Parish and J. Levinsen, Highly polarized Fermi
sition in strongly repulsive one-dimensional Fermi gases, gases in two dimensions, Phys. Rev. A 87, 033616 (2013).
Phys. Rev. A 89, 023611 (2014). [58] J. Vlietinck, J. Ryckebusch, and K. Van Houcke, Dia-
[39] P. O. Bugnion and G. J. Conduit, Ferromagnetic spin grammatic Monte Carlo study of the Fermi polaron in
correlations in a few-fermion system, Phys. Rev. A 87, two dimensions, Phys. Rev. B 89, 085119 (2014).
060502(R) (2013). [59] H. S. Adlong, W. E. Liu, F. Scazza, M. Zaccanti, N. D.
[40] S. E. Gharashi and D. Blume, Correlations of the up- Oppong, S. Fölling, M. M. Parish, and J. Levinsen,
per branch of 1D harmonically trapped two-component Quasiparticle lifetime of the repulsive Fermi polaron,
Fermi gases, Phys. Rev. Lett. 111, 045302 (2013). Phys. Rev. Lett. 125, 133401 (2020).
[41] C. Mora and F. Chevy, Normal phase of an imbalanced [60] T. Comparin, R. Bombı́n, M. Holzmann, F. Mazzanti,
Fermi gas, Phys. Rev. Lett. 104, 230402 (2010). J. Boronat, and S. Giorgini, Two-dimensional mixture of
[42] C. Kohstall, M. Zaccanti, M. Jag, A. Trenkwalder, dipolar fermions: Equation of state and magnetic phases,
P. Massignan, G. M. Bruun, F. Schreck, and R. Grimm, Phys. Rev. A 99, 043609 (2019).
Metastability and coherence of repulsive polarons in a [61] B. L. Hammond, W. A. Lester, and P. J.
strongly interacting Fermi mixture, Nature 485, 615 Reynolds, Monte Carlo methods in ab initio quan-
(2012). tum chemistry (WORLD SCIENTIFIC, 1994)
[43] M. Koschorreck, D. Pertot, E. Vogt, B. Fröhlich, M. Feld, https://www.worldscientific.com/doi/pdf/10.1142/1170.
and M. Köhl, Attractive and repulsive Fermi polarons in [62] V. Penna and L. Salasnich, Itinerant ferromagnetism of
two dimensions, Nature 485, 619 (2012). two-dimensional repulsive fermions with Rabi coupling,
[44] N. Darkwah Oppong, L. Riegger, O. Bettermann, New J, of Phys. 19, 043018 (2017).
M. Höfer, J. Levinsen, M. M. Parish, I. Bloch, and [63] A. Ambrosetti, G. Lombardi, L. Salasnich, P. L. Sil-
S. Fölling, Observation of coherent multiorbital polarons vestrelli, and F. Toigo, Polarization of a quasi-two-
in a two-dimensional Fermi gas, Phys. Rev. Lett. 122, dimensional repulsive Fermi gas with Rashba spin-orbit
193604 (2019). coupling: A variational study, Phys. Rev. A 90, 043614
[45] L. He, Finite range and upper branch effects on itin- (2014).
erant ferromagnetism in repulsive Fermi gases: Bethe– [64] S. Pilati, J. Boronat, J. Casulleras, and S. Giorgini,
Goldstone ladder resummation approach, Ann. Phys. Quantum Monte Carlo simulation of a two-dimensional
(N.Y.) 351, 477 (2014). Bose gas, Phys. Rev. A 71, 023605 (2005).
[46] L. He and X.-G. Huang, Nonperturbative effects on the [65] D. S. Petrov, M. Holzmann, and G. V. Shlyapnikov, Bose-
ferromagnetic transition in repulsive fermi gases, Phys. einstein condensation in quasi-2d trapped gases, Phys.
Rev. A 85, 043624 (2012). Rev. Lett. 84, 2551 (2000).
[47] G. Bertaina and S. Giorgini, BCS-BEC Crossover in [66] D. S. Petrov and G. V. Shlyapnikov, Interatomic colli-
a Two-Dimensional Fermi Gas, Phys. Rev. Lett. 106, sions in a tightly confined Bose gas, Phys. Rev. A 64,
110403 (2011). 012706 (2001).
[48] A. Galea, H. Dawkins, S. Gandolfi, and A. Gezerlis, [67] C. Lin, F. H. Zong, and D. M. Ceperley, Twist-averaged
Diffusion Monte Carlo study of strongly interacting boundary conditions in continuum quantum Monte Carlo
two-dimensional Fermi gases, Phys. Rev. A 93, 023602 algorithms, Phys. Rev. E 64, 016702 (2001).
(2016). [68] J. R. Engelbrecht, M. Randeria, and L. Zhang, Landau f
[49] M. Bauer, M. M. Parish, and T. Enss, Universal Equation function for the dilute Fermi gas in two dimensions, Phys.
of State and Pseudogap in the Two-Dimensional Fermi Rev. B 45, 10135 (1992).
Gas, Physical Review Letters 112, 135302 (2014). [69] J. Boronat and J. Casulleras, Quantum Monte Carlo
[50] H. Shi, S. Chiesa, and S. Zhang, Ground-state properties study of static properties of one 3 He atom in superfluid
4
of strongly interacting Fermi gases in two dimensions, He, Phys. Rev. B 59, 8844 (1999).
Phys. Rev. A 92, 033603 (2015). [70] L. A. Peña Ardila and S. Giorgini, Impurity in a Bose-
[51] T. Zielinski, B. Ross, and A. Gezerlis, Pairing in two- Einstein condensate: Study of the attractive and repul-
dimensional Fermi gases with a coordinate-space poten- sive branch using quantum Monte Carlo methods, Phys.
tial, Phys. Rev. A 101, 033601 (2020). Rev. A 92, 033612 (2015).
[52] E. R. Anderson and J. E. Drut, Pressure, Compress- [71] A. A. Abrikosov, I. Dzyaloshinskii, L. P. Gorkov, and
ibility, and Contact of the Two-Dimensional Attractive R. A. Silverman, Methods of quantum field theory in sta-
Fermi Gas, Phys. Rev. Lett. 115, 115301 (2015). tistical physics (Dover, New York, NY, 1975).
[53] G. Bertaina, Two-dimensional short-range interacting at- [72] E. H. Lieb, R. Seiringer, and J. P. Solovej, Ground-state
tractive and repulsive Fermi gases at zero temperature, energy of the low-density Fermi gas, Phys. Rev. A 71,
Eur. Phys. J Spec. Top. 217, 153 (2013). 053605 (2005).
[54] R. Bombı́n, T. Comparin, G. Bertaina, F. Mazzanti, [73] P. Bloom, Two-dimensional Fermi gas, Phys. Rev. B 12,
S. Giorgini, and J. Boronat, Two-dimensional repulsive 125 (1975).
Fermi polarons with short- and long-range interactions, [74] J. R. Engelbrecht and M. Randeria, Low-density repul-
Phys. Rev. A 100, 023608 (2019). sive Fermi gas in two dimensions: Bound-pair excita-
[55] V. Ngampruetikorn, J. Levinsen, and M. M. Parish, Re- tions and Fermi-liquid behavior, Phys. Rev. B 45, 12419
pulsive polarons in two-dimensional Fermi gases, EPL 98, (1992).
30005 (2012). [75] R. Bombı́n, V. Cikojević, J. Sánchez-Baena, and
[56] R. Schmidt, T. Enss, V. Pietilä, and E. Demler, Fermi J. Boronat, Finite-range effects in the two-dimensional
polarons in two dimensions, Phys. Rev. A 85, 021602 repulsive Fermi polaron (2020), arXiv:2008.10510 [cond-
(2012). mat.quant-gas].
10
[76] P. Dyke, K. Fenech, T. Peppler, M. G. Lingham, [84] L. Benguigui, Critical point of infinite type, Phys. Rev.
S. Hoinka, W. Zhang, S.-G. Peng, B. Mulkerin, H. Hu, B 16, 1266 (1977).
X.-J. Liu, and C. J. Vale, Criteria for two-dimensional [85] N. Navon, S. Nascimbène, F. Chevy, and C. Salomon,
kinematics in an interacting Fermi gas, Phys. Rev. A 93, The equation of state of a low-temperature Fermi gas
011603 (2016). with tunable interactions, Science 328, 729 (2010).
[77] In Fig. 2, one might notice some minuscule discrepancies. [86] B. Fröhlich, M. Feld, E. Vogt, M. Koschorreck, W. Zw-
They are due to the use, in that figure, of the finite-size erger, and M. Köhl, Radio-frequency spectroscopy of a
correction with m∗0 from second-order perturbations the- strongly interacting two-dimensional Fermi gas, Phys.
ory. Actually, in the analysis of the Landau-Pomeranchuk Rev. Lett. 106, 105301 (2011).
functional we set m∗0 = m, which is more appropriate for [87] A. Recati and S. Stringari, Spin fluctuations, susceptibil-
the high-imbalance regime. See discussion in Section II. ity, and the dipole oscillation of a nearly ferromagnetic
[78] R. Combescot, A. Recati, C. Lobo, and F. Chevy, Normal Fermi gas, Phys. Rev. Lett. 106, 080402 (2011).
state of highly polarized Fermi gases: Simple many-body [88] M. Koschorreck, D. Pertot, E. Vogt, and M. Köhl, Uni-
approaches, Phys. Rev. Lett. 98, 180402 (2007). versal spin dynamics in two-dimensional Fermi gases,
[79] T. R. Kirkpatrick and D. Belitz, Exponent relations at Nat. Phys. 9, 405 (2013).
quantum phase transitions with applications to metallic [89] C. Luciuk, S. Smale, F. Böttcher, H. Sharum, B. A.
quantum ferromagnets, Phys. Rev. B 91, 214407 (2015). Olsen, S. Trotzky, T. Enss, and J. H. Thywissen, Obser-
[80] M. Bajdich, L. Mitas, G. Drobný, L. K. Wagner, vation of Quantum-Limited Spin Transport in Strongly
and K. E. Schmidt, Pfaffian pairing wave functions in Interacting Two-Dimensional Fermi Gases, Phys. Rev.
electronic-structure quantum Monte Carlo simulations, Lett. 118, 130405 (2017).
Phys. Rev. Lett. 96, 130201 (2006). [90] C. Wellenhofer, C. Drischler, and A. Schwenk, Dilute
[81] C. Genovese, T. Shirakawa, K. Nakano, and Fermi gas at fourth order in effective field theory, Phys.
S. Sorella, General correlated geminal Ansatz for Lett. B 802, 135247 (2020).
electronic structure calculations: Exploiting Pfaf- [91] C. Wellenhofer, D. R. Phillips, and A. Schwenk, From
fians in place of determinants, J. Chem. The- weak to strong: Constrained extrapolation of pertur-
ory Comput. 16, 6114 (2020), pMID: 32804497, bation series with applications to dilute Fermi systems,
https://doi.org/10.1021/acs.jctc.0c00165. Phys. Rev. Research 2, 043372 (2020).
[82] R. K. Bhaduri, S. M. Reimann, S. Viefers, A. G. Choud- [92] C. Wellenhofer, C. Drischler, and A. Schwenk, Effec-
hury, and M. K. Srivastava, The effect of interactions on tive field theory for dilute Fermi systems at fourth order
Bose-Einstein condensation in a quasi two-dimensional (2021), arXiv:2102.05966 [cond-mat.quant-gas].
harmonic trap, J. Phys. B 33, 3895 (2000). [93] A. V. Chubukov, Kohn-Luttinger effect and the instabil-
[83] G. Carleo, S. Moroni, F. Becca, and S. Baroni, Itinerant ity of a two-dimensional repulsive Fermi liquid at T=0,
ferromagnetic phase of the Hubbard model, Phys. Rev. Phys. Rev. B 48, 1097 (1993).
B 83, 060411 (2011). [94] S. Pilati, G. Orso, and G. Bertaina, Data for: Quan-
tum Monte Carlo simulations of two- dimensional repul-
sive Fermi gases with population imbalance, 10.5281/zen-
odo.4631946 (2021).