Giant effective g๐‘”gitalic_g-factor due to spin bifurcations in polariton condensates

A. Bochin School of Physics and Engineering, ITMO University, Kronverksky Pr. 49, bldg. A, St. Petersburg, 197101, Russia โ€ƒโ€ƒ I. Chestnov School of Physics and Engineering, ITMO University, Kronverksky Pr. 49, bldg. A, St. Petersburg, 197101, Russia โ€ƒโ€ƒ A. Nalitov Moscow Institute of Physics and Technology, Institutskiy per.,ย 9, Dolgoprudnyi, Moscow Region, Russia 141701
Abstract

We predict giant susceptibility of spin-bifurcating polariton condensates to externally applied permanent magnetic field. In the presence of spin-anisotropic polariton-polariton interactions, the condensate spontaneously takes an elliptically polarised state, whose perturbation dynamics can be interpreted in terms of the presence of strong effective magnetic field significantly surpassing the external one. Surprisingly, this behaviour of the addressed strongly out-of-equilibrium system in the vicinity of a critical point exhibits intriguing analogy with the second-order phase transition. The predicted field-enhancement effect can be utilized for creation of topologically nontrivial states of Bogoliubovโ€™s excitations existing on top of the polariton condensate.

Magneto-optic phenomena lifting degeneracy of photonic modes in the presence of external magnetic field are remarkable manifestations of light-matter coupling. As an example, Zeeman effect for electron excitations in dielectrics or semiconductors causes energy splitting of otherwise degenerate circularly polarised photonic modes, which, in turn, results in magneto-optic Faraday effect or Kerr rotation [1]. The possibility of local time-reversal symmetry breaking for electromagnetic waves due to magneto-optic coupling is crucial for engineering topological photonic states and suppressing backscattering in signal transmission [2, 3].

Although the magneto-optic effect is weak in the optical frequency domain, it can be significantly enhanced in the strong coupling regime, where mixed light-matter quasi-particles, such as exciton-polaritons, emerge. In particular, topological exciton-polariton states were demonstrated in optical cavity lattices subject to strong magnetic fields [4]. In addition, incorporating ferromagnetic materials in strongly coupled optical cavities was proposed recently to reach giant values of the effective g๐‘”gitalic_g-factor, which quantifies the magneto-optic coupling strength [5].

The full potential of strongly coupled systems supporting exciton polaritons is revealed in the nonlinear regime, where the macroscopic coherent states associated with bosonic condensates are formed. In particular, spin-anisotropic polariton-polariton interaction results in Larmor precession of the condensate pseudospin in a self-induced effective magnetic field even in the absence of an externally applied one [7].

In the thermodynamic equilibrium limit, the externally applied field and the effective self-induced one are opposite and even exactly compensate each other below a certain threshold, giving rise to the spin Meissner effect [8]. In contrast, nonequilibrium condensates can develop spin polarisation and thus produce effective magnetic fields spontaneously due to the spin bifurcation mechanism [9]. Similarly, coherently driven nonequilibrium polariton states also exhibit spontaneous spin polarisation and spin multistability [10]. Note that the interplay between the equilibrium effect of locked polarisation direction in the spin Meissner regime and a spontaneous choice of spin polarisation typical to driven-dissipative condensates is still under discussion [11, 12].

Artificial gauge [13] and Zeeman [14] fields emerging in spatially structured optical systems can replace the external magnetic field giving rise to topological photonic states. Similarly, the effective field in spontaneously spin-polarised condensate lattices was shown to result in topologically nontrivial excitation spectra and unidirectional edge states [15]. However, the underlying spin bifurcation mechanism requires a delicate balance and degeneracy of the spin states, rendering this system extremely susceptible to external symmetry breaking factors. This inspires investigation of the possibility to exploit this sensitivity to control the band topological invariants in polariton condensate lattices.

In this work, we show that the effective field of a single spin-bifurcating condensate is not only aligned with the external Zeeman field, but can also significantly surpass it in magnitude. Magnification of the weak Zeeman field with the strong and aligned effective field can be considered as a giant enhancement of the effective g๐‘”gitalic_g-factor. We demonstrate that the proposed enhancement is most pronounced in the vicinity of the critical bifurcation point and identify the range of the optimal parameters.

The system of interest represents a condensate of exciton-polaritons whose coherent state is governed by the driven-dissipative Gross-Pitaevskii equation for a spinor wave function ๐šฟ=(ฮจ+,ฮจโˆ’)โŠบ๐šฟsuperscriptsubscriptฮจsubscriptฮจโŠบ\mathbf{\Psi}=\left(\Psi_{+},\Psi_{-}\right)^{\intercal}bold_ฮจ = ( roman_ฮจ start_POSTSUBSCRIPT + end_POSTSUBSCRIPT , roman_ฮจ start_POSTSUBSCRIPT - end_POSTSUBSCRIPT ) start_POSTSUPERSCRIPT โŠบ end_POSTSUPERSCRIPT:

iโขฮจห™ยฑ=๐‘–subscriptห™ฮจplus-or-minusabsent\displaystyle i\dot{\Psi}_{\pm}=italic_i overห™ start_ARG roman_ฮจ end_ARG start_POSTSUBSCRIPT ยฑ end_POSTSUBSCRIPT = โˆ’\displaystyle-- i2โข(iโขฮ”+Wโˆ’ฮ“โˆ’ฮทโขNpol)โขฮจยฑโˆ’12โข(ฮต+iโขฮณ)โขฮจโˆ“๐‘–2๐‘–ฮ”๐‘Šฮ“๐œ‚subscript๐‘polsubscriptฮจplus-or-minus12๐œ€๐‘–๐›พsubscriptฮจminus-or-plus\displaystyle\frac{i}{2}\left(i\Delta+W-\Gamma-\eta N_{\rm pol}\right)\Psi_{% \pm}-\frac{1}{2}\left(\varepsilon+i\gamma\right)\Psi_{\mp}divide start_ARG italic_i end_ARG start_ARG 2 end_ARG ( italic_i roman_ฮ” + italic_W - roman_ฮ“ - italic_ฮท italic_N start_POSTSUBSCRIPT roman_pol end_POSTSUBSCRIPT ) roman_ฮจ start_POSTSUBSCRIPT ยฑ end_POSTSUBSCRIPT - divide start_ARG 1 end_ARG start_ARG 2 end_ARG ( italic_ฮต + italic_i italic_ฮณ ) roman_ฮจ start_POSTSUBSCRIPT โˆ“ end_POSTSUBSCRIPT (1)
+\displaystyle++ 12โข(ฮฑ1โข|ฮจยฑ|2+ฮฑ2โข|ฮจโˆ“|2)โขฮจยฑ,12subscript๐›ผ1superscriptsubscriptฮจplus-or-minus2subscript๐›ผ2superscriptsubscriptฮจminus-or-plus2subscriptฮจplus-or-minus\displaystyle\frac{1}{2}\left(\alpha_{1}\left|\Psi_{\pm}\right|^{2}+\alpha_{2}% \left|\Psi_{\mp}\right|^{2}\right)\Psi_{\pm},divide start_ARG 1 end_ARG start_ARG 2 end_ARG ( italic_ฮฑ start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT | roman_ฮจ start_POSTSUBSCRIPT ยฑ end_POSTSUBSCRIPT | start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT + italic_ฮฑ start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT | roman_ฮจ start_POSTSUBSCRIPT โˆ“ end_POSTSUBSCRIPT | start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ) roman_ฮจ start_POSTSUBSCRIPT ยฑ end_POSTSUBSCRIPT ,

where ฮจ+(โˆ’)subscriptฮจ\Psi_{+(-)}roman_ฮจ start_POSTSUBSCRIPT + ( - ) end_POSTSUBSCRIPT stands for spin-up (down) projections of the polariton state on the structure growth axis corresponding to the right (left) circular polarisation of the emitted photons. Here ฮ“ฮ“\Gammaroman_ฮ“ is the polariton decay rate, W๐‘ŠWitalic_W is the rate of stimulated scattering from the incoherent particle reservoir, ฮท๐œ‚\etaitalic_ฮท is the strength of the gain saturation proportional to the condensate occupation Npolsubscript๐‘polN_{\rm pol}italic_N start_POSTSUBSCRIPT roman_pol end_POSTSUBSCRIPT, while ฮต๐œ€\varepsilonitalic_ฮต and ฮณ๐›พ\gammaitalic_ฮณ are respectively the energy splitting and the loss-rate difference of the two linearly polarised polariton states. Finally, ฮฑ1โข(2)subscript๐›ผ12\alpha_{1(2)}italic_ฮฑ start_POSTSUBSCRIPT 1 ( 2 ) end_POSTSUBSCRIPT is the interaction constant for polaritons with the same (opposite) spin.

Following Ref.ย [9], we treat the system using the classical pseudospin vector ๐’=1/2โข๐šฟโ€ โข๐ˆโข๐šฟ๐’12superscript๐šฟโ€ ๐ˆ๐šฟ\mathbf{S}=1/2\mathbf{\Psi}^{{\dagger}}\bm{\sigma}\mathbf{\Psi}bold_S = 1 / 2 bold_ฮจ start_POSTSUPERSCRIPT โ€  end_POSTSUPERSCRIPT bold_italic_ฯƒ bold_ฮจ defined via Pauli vector ๐ˆ๐ˆ\bm{\sigma}bold_italic_ฯƒ with the magnitude S๐‘†Sitalic_S proportional to the condensate occupation and the direction characterising its spin state. Note that the pseudospin ๐’๐’\mathbf{S}bold_S is directly connected to the Stokes parameters of the laser emission as S=S0๐‘†subscript๐‘†0S=S_{0}italic_S = italic_S start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT, Sx=S1subscript๐‘†๐‘ฅsubscript๐‘†1S_{x}=S_{1}italic_S start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT = italic_S start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT, Sy=S2subscript๐‘†๐‘ฆsubscript๐‘†2S_{y}=S_{2}italic_S start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT = italic_S start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT, Sz=S3subscript๐‘†๐‘งsubscript๐‘†3S_{z}=S_{3}italic_S start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT = italic_S start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT. The pseudospin dynamics is described with the following system of equations:

Sห™xsubscriptห™๐‘†๐‘ฅ\displaystyle\dot{S}_{x}overห™ start_ARG italic_S end_ARG start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT =\displaystyle== (Wโˆ’ฮทโขSโˆ’ฮ“)โขSxโˆ’ฮณโขSโˆ’(ฮฑโขSz+ฮ”)โขSy,๐‘Š๐œ‚๐‘†ฮ“subscript๐‘†๐‘ฅ๐›พ๐‘†๐›ผsubscript๐‘†๐‘งฮ”subscript๐‘†๐‘ฆ\displaystyle(W-\eta S-\Gamma)S_{x}-\gamma S-(\alpha S_{z}+\Delta)S_{y},( italic_W - italic_ฮท italic_S - roman_ฮ“ ) italic_S start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT - italic_ฮณ italic_S - ( italic_ฮฑ italic_S start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT + roman_ฮ” ) italic_S start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT , (2a)
Sห™ysubscriptห™๐‘†๐‘ฆ\displaystyle\dot{S}_{y}overห™ start_ARG italic_S end_ARG start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT =\displaystyle== (Wโˆ’ฮทโขSโˆ’ฮ“)โขSy+ฮตโขSz+(ฮฑโขSz+ฮ”)โขSx,๐‘Š๐œ‚๐‘†ฮ“subscript๐‘†๐‘ฆ๐œ€subscript๐‘†๐‘ง๐›ผsubscript๐‘†๐‘งฮ”subscript๐‘†๐‘ฅ\displaystyle(W-\eta S-\Gamma)S_{y}+\varepsilon S_{z}+(\alpha S_{z}+\Delta)S_{% x},( italic_W - italic_ฮท italic_S - roman_ฮ“ ) italic_S start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT + italic_ฮต italic_S start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT + ( italic_ฮฑ italic_S start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT + roman_ฮ” ) italic_S start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT , (2b)
Sห™zsubscriptห™๐‘†๐‘ง\displaystyle\dot{S}_{z}overห™ start_ARG italic_S end_ARG start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT =\displaystyle== (Wโˆ’ฮทโขSโˆ’ฮ“)โขSzโˆ’ฮตโขSy.๐‘Š๐œ‚๐‘†ฮ“subscript๐‘†๐‘ง๐œ€subscript๐‘†๐‘ฆ\displaystyle(W-\eta S-\Gamma)S_{z}-\varepsilon S_{y}.( italic_W - italic_ฮท italic_S - roman_ฮ“ ) italic_S start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT - italic_ฮต italic_S start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT . (2c)

In analogy with the classical spin in magnetic field, the pseudospin vector ๐’๐’\mathbf{S}bold_S is precessing about the z๐‘งzitalic_z-axis due to the mixed effect of Zeeman splitting ฮ”ฮ”\Deltaroman_ฮ” from the external out-of-plane magnetic field and a self-induced effective field ฮฑโขSz๐›ผsubscript๐‘†๐‘ง\alpha S_{z}italic_ฮฑ italic_S start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT [6, 7]. The latter is governed by the interaction constant ฮฑ=ฮฑ1โˆ’ฮฑ2>0๐›ผsubscript๐›ผ1subscript๐›ผ20\alpha{=\alpha_{1}-\alpha_{2}}>0italic_ฮฑ = italic_ฮฑ start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT - italic_ฮฑ start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT > 0, which accounts for the strong spin anisotropy of polariton-polariton interactions.

In what follows, we use the dimensionless form of system (2):

sห™xsubscriptห™๐‘ ๐‘ฅ\displaystyle\dot{s}_{x}overห™ start_ARG italic_s end_ARG start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT =\displaystyle== (pโˆ’s)โขsxโˆ’gโขsโˆ’(aโขsz+ฮด)โขsy,๐‘๐‘ subscript๐‘ ๐‘ฅ๐‘”๐‘ ๐‘Žsubscript๐‘ ๐‘ง๐›ฟsubscript๐‘ ๐‘ฆ\displaystyle(p-s)s_{x}-gs-(as_{z}+\delta)s_{y},( italic_p - italic_s ) italic_s start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT - italic_g italic_s - ( italic_a italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT + italic_ฮด ) italic_s start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT , (3a)
sห™ysubscriptห™๐‘ ๐‘ฆ\displaystyle\dot{s}_{y}overห™ start_ARG italic_s end_ARG start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT =\displaystyle== (pโˆ’s)โขsy+sz+(aโขsz+ฮด)โขsx,๐‘๐‘ subscript๐‘ ๐‘ฆsubscript๐‘ ๐‘ง๐‘Žsubscript๐‘ ๐‘ง๐›ฟsubscript๐‘ ๐‘ฅ\displaystyle(p-s)s_{y}+s_{z}+(as_{z}+\delta)s_{x},( italic_p - italic_s ) italic_s start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT + italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT + ( italic_a italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT + italic_ฮด ) italic_s start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT , (3b)
sห™zsubscriptห™๐‘ ๐‘ง\displaystyle\dot{s}_{z}overห™ start_ARG italic_s end_ARG start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT =\displaystyle== (pโˆ’s)โขszโˆ’sy,๐‘๐‘ subscript๐‘ ๐‘งsubscript๐‘ ๐‘ฆ\displaystyle(p-s)s_{z}-s_{y},( italic_p - italic_s ) italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT - italic_s start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT , (3c)

with the effective splitting field ฮด=ฮ”/ฮต๐›ฟฮ”๐œ€\delta=\Delta/\varepsilonitalic_ฮด = roman_ฮ” / italic_ฮต, the dimensionless time ฯ„=tโขฮต๐œ๐‘ก๐œ€\tau=t\varepsilonitalic_ฯ„ = italic_t italic_ฮต and a couple of dimensionless parameters a=ฮฑ/ฮท๐‘Ž๐›ผ๐œ‚a=\alpha/\etaitalic_a = italic_ฮฑ / italic_ฮท, g=ฮณ/ฮต๐‘”๐›พ๐œ€g=\gamma/\varepsilonitalic_g = italic_ฮณ / italic_ฮต. We also define the effective pump strength p=(Wโˆ’ฮ“)/ฮต๐‘๐‘Šฮ“๐œ€p=(W-\Gamma)/\varepsilonitalic_p = ( italic_W - roman_ฮ“ ) / italic_ฮต and the normalized Stokes vector ๐ฌ=ฮทโข๐’/ฮต๐ฌ๐œ‚๐’๐œ€\mathbf{s}=\eta\mathbf{S}/\varepsilonbold_s = italic_ฮท bold_S / italic_ฮต.

In the absence of an external magnetic field, ฮด=0๐›ฟ0\delta=0italic_ฮด = 0, a couple of linearly polarised (sz=sy=0subscript๐‘ ๐‘งsubscript๐‘ ๐‘ฆ0s_{z}=s_{y}=0italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT = italic_s start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT = 0) trivial solutions of system (3) are characterised by the magnitudes sยฑโข(p)=pยฑgsubscript๐‘ plus-or-minus๐‘plus-or-minus๐‘๐‘”s_{\pm}(p)=p\pm gitalic_s start_POSTSUBSCRIPT ยฑ end_POSTSUBSCRIPT ( italic_p ) = italic_p ยฑ italic_g. These states are pinned to the x๐‘ฅxitalic_x-axis, sx=โˆ“sยฑ(0)subscript๐‘ ๐‘ฅminus-or-plussuperscriptsubscript๐‘ plus-or-minus0s_{x}=\mp s_{\pm}^{(0)}italic_s start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT = โˆ“ italic_s start_POSTSUBSCRIPT ยฑ end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT. At g>0๐‘”0g>0italic_g > 0 (the upper linearly polarised mode dissipates stronger than the lower one), the sโˆ’subscript๐‘ s_{-}italic_s start_POSTSUBSCRIPT - end_POSTSUBSCRIPT-state is dynamically unstable against weak perturbations according to the regular Lyapunov stability analysis. The more populated state s+subscript๐‘ s_{+}italic_s start_POSTSUBSCRIPT + end_POSTSUBSCRIPT is stable below the critical pum** pc=(1โˆ’aโขg+g2)/asubscript๐‘๐‘1๐‘Ž๐‘”superscript๐‘”2๐‘Žp_{c}=(1-ag+g^{2})/aitalic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT = ( 1 - italic_a italic_g + italic_g start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ) / italic_a, characterised by the condensate population sc=(g2+1)/asubscript๐‘ ๐‘superscript๐‘”21๐‘Žs_{c}=(g^{2}+1)/aitalic_s start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT = ( italic_g start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT + 1 ) / italic_a.

At p=pc๐‘subscript๐‘๐‘p=p_{c}italic_p = italic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT, the s+subscript๐‘ s_{+}italic_s start_POSTSUBSCRIPT + end_POSTSUBSCRIPT-state is destabilized via a pitchfork bifurcation, which has a supercritical character if a>gโˆ’gโˆ’1๐‘Ž๐‘”superscript๐‘”1a>g-g^{-1}italic_a > italic_g - italic_g start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT and a subcritical one otherwise (see the phase diagram in Fig.ย 1(a)). This gives rise to a couple of nontrivial elliptically polarised (szโ‰ 0subscript๐‘ ๐‘ง0s_{z}\neq 0italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT โ‰  0) states with

sx=(pโˆ’s)โขs/g,sy=(pโˆ’s)โขsz,formulae-sequencesubscript๐‘ ๐‘ฅ๐‘๐‘ ๐‘ ๐‘”subscript๐‘ ๐‘ฆ๐‘๐‘ subscript๐‘ ๐‘ง\displaystyle s_{x}=(p-s)s/g,\>\ \ \ s_{y}=(p-s)s_{z},italic_s start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT = ( italic_p - italic_s ) italic_s / italic_g , italic_s start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT = ( italic_p - italic_s ) italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT , (4a)
sz=ยฑsgโขg2โˆ’(pโˆ’s)21+(pโˆ’s)2,subscript๐‘ ๐‘งplus-or-minus๐‘ ๐‘”superscript๐‘”2superscript๐‘๐‘ 21superscript๐‘๐‘ 2\displaystyle s_{z}=\pm{s\over g}\sqrt{g^{2}-(p-s)^{2}\over 1+(p-s)^{2}},italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT = ยฑ divide start_ARG italic_s end_ARG start_ARG italic_g end_ARG square-root start_ARG divide start_ARG italic_g start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT - ( italic_p - italic_s ) start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT end_ARG start_ARG 1 + ( italic_p - italic_s ) start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT end_ARG end_ARG , (4b)

and the condensate population s๐‘ sitalic_s given by the positive root of the quadratic equation

(agโˆ’1)โขs2โˆ’(agโˆ’2)โขpโขsโˆ’(p2+1)=0.๐‘Ž๐‘”1superscript๐‘ 2๐‘Ž๐‘”2๐‘๐‘ superscript๐‘210\left({a\over g}-1\right)s^{2}-\left({a\over g}-2\right)ps-(p^{2}+1)=0.( divide start_ARG italic_a end_ARG start_ARG italic_g end_ARG - 1 ) italic_s start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT - ( divide start_ARG italic_a end_ARG start_ARG italic_g end_ARG - 2 ) italic_p italic_s - ( italic_p start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT + 1 ) = 0 . (5)

Note that at g<0๐‘”0g<0italic_g < 0, the symmetry-breaking bifurcation does not alter stability properties of sยฑsubscript๐‘ plus-or-minuss_{\pm}italic_s start_POSTSUBSCRIPT ยฑ end_POSTSUBSCRIPT-states. In this regime, the more populated state sโˆ’subscript๐‘ s_{-}italic_s start_POSTSUBSCRIPT - end_POSTSUBSCRIPT always remains the only stable linearly polarised solution. In what follows, we focus on the case g>0๐‘”0g>0italic_g > 0, corresponding to experimental conditions [9].

Refer to caption
Figure 1: Fig. 1. (a) The phase diagram of the spontaneous symmetry breaking pitchfork bifurcation at ฮด=0๐›ฟ0\delta=0italic_ฮด = 0. Pseudospin projection szsubscript๐‘ ๐‘งs_{z}italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT (left axis) and field enhancement parameter ฯ‡๐œ’\chiitalic_ฯ‡ (right axis) in the supercritical regime at g=0.1๐‘”0.1g=0.1italic_g = 0.1 and a=2.28๐‘Ž2.28a=2.28italic_a = 2.28 [9]. The solid lines correspond to stable states while dashing shows dynamically unstable solutions. (c) The same as on panel (b) but in the subcritical regime at g=3๐‘”3g=3italic_g = 3 and a=1๐‘Ž1a=1italic_a = 1. The dotted blue curve corresponds to the field-enhancement ฯ‡๐œ’\chiitalic_ฯ‡ of the upper stable elliptically polarised state which appears in the fold bifurcation.

To describe the impact of the external magnetic field, we start with the limit of perturbatively weakย fieldย ฮด๐›ฟ\deltaitalic_ฮด. In the first approximation, stationary solutions acquire a correction ๐ฌ(๐Ÿ)=(sx(1),sy(1),sz(1))โŠบsuperscript๐ฌ1superscriptsuperscriptsubscript๐‘ ๐‘ฅ1superscriptsubscript๐‘ ๐‘ฆ1superscriptsubscript๐‘ ๐‘ง1โŠบ\mathbf{s^{(1)}}=\left(s_{x}^{(1)},s_{y}^{(1)},s_{z}^{(1)}\right)^{\intercal}bold_s start_POSTSUPERSCRIPT ( bold_1 ) end_POSTSUPERSCRIPT = ( italic_s start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 1 ) end_POSTSUPERSCRIPT , italic_s start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 1 ) end_POSTSUPERSCRIPT , italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 1 ) end_POSTSUPERSCRIPT ) start_POSTSUPERSCRIPT โŠบ end_POSTSUPERSCRIPT linear in ฮด๐›ฟ\deltaitalic_ฮด:

๐ฌ(1)=ฮดร—[Jโข(๐ฌ(0))]โˆ’1โข(sy(0),โˆ’sx(0),0)โŠบ,superscript๐ฌ1๐›ฟsuperscriptdelimited-[]๐ฝsuperscript๐ฌ01superscriptsuperscriptsubscript๐‘ ๐‘ฆ0superscriptsubscript๐‘ ๐‘ฅ00โŠบ\mathbf{s}^{(1)}=\delta\times\left[J\left(\mathbf{s}^{(0)}\right)\right]^{-1}% \left(s_{y}^{(0)},-s_{x}^{(0)},0\right)^{\intercal},bold_s start_POSTSUPERSCRIPT ( 1 ) end_POSTSUPERSCRIPT = italic_ฮด ร— [ italic_J ( bold_s start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT ) ] start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ( italic_s start_POSTSUBSCRIPT italic_y end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT , - italic_s start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT , 0 ) start_POSTSUPERSCRIPT โŠบ end_POSTSUPERSCRIPT , (6)

where Jโข(๐ฌ(๐ŸŽ))๐ฝsuperscript๐ฌ0J(\mathbf{\mathbf{s}^{(0)}})italic_J ( bold_s start_POSTSUPERSCRIPT ( bold_0 ) end_POSTSUPERSCRIPT ) is the Jacobi matrix of the system (3) calculated at the unperturbed fixed point solutionย ๐ฌ(0)superscript๐ฌ0\mathbf{s}^{(0)}bold_s start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT.

In the case of the linearly polarised fixed point s+subscript๐‘ s_{+}italic_s start_POSTSUBSCRIPT + end_POSTSUBSCRIPT with sx(0)=โˆ’(p+g)superscriptsubscript๐‘ ๐‘ฅ0๐‘๐‘”s_{x}^{(0)}=-(p+g)italic_s start_POSTSUBSCRIPT italic_x end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT = - ( italic_p + italic_g ) which is stable below the critical pum** pcsubscript๐‘๐‘p_{c}italic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT, the correction (6) provides a non-zero interaction-driven field aโขsz(1)=ฮดโขs(0)/(pcโˆ’p)๐‘Žsuperscriptsubscript๐‘ ๐‘ง1/๐›ฟsuperscript๐‘ 0subscript๐‘๐‘๐‘as_{z}^{(1)}=\delta s^{(0)}\left/\left(p_{c}-p\right)\right.italic_a italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 1 ) end_POSTSUPERSCRIPT = italic_ฮด italic_s start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT / ( italic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT - italic_p ). The condensate response can thus be characterized by the field enhancement parameter ฯ‡=aโขsz(1)/ฮด๐œ’๐‘Žsuperscriptsubscript๐‘ ๐‘ง1๐›ฟ\chi=as_{z}^{(1)}/\deltaitalic_ฯ‡ = italic_a italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 1 ) end_POSTSUPERSCRIPT / italic_ฮด and the effective g๐‘”gitalic_g-factor

geff=1+ฯ‡=scpcโˆ’psubscript๐‘”eff1๐œ’subscript๐‘ ๐‘subscript๐‘๐‘๐‘g_{\rm eff}=1+\chi={s_{c}\over p_{c}-p}italic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT = 1 + italic_ฯ‡ = divide start_ARG italic_s start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT end_ARG start_ARG italic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT - italic_p end_ARG (7)

responsible for the condensate pseudospin precession frequency ฮดโขgeff๐›ฟsubscript๐‘”eff\delta g_{\rm eff}italic_ฮด italic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT according to equations (3).

Since geff>0subscript๐‘”eff0g_{\rm eff}>0italic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT > 0 at p<pc๐‘subscript๐‘๐‘p<p_{c}italic_p < italic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT, the induced field is aligned with the external one. Note that expression (7) diverges at the critical point similarly to the susceptibility behaviour in Landau theory of second order phase transitions, and linearly scales with the condensate population, as shown in Fig.ย 1(b). Moreover, as p๐‘pitalic_p approaches pcsubscript๐‘๐‘p_{c}italic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT from above, pโ†’pc+โ†’๐‘superscriptsubscript๐‘๐‘p\rightarrow p_{c}^{+}italic_p โ†’ italic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT, the field enhancement parameter ฯ‡๐œ’\chiitalic_ฯ‡, which is now governed by the response of the elliptical states (4), has a similar asymptotic behavior. However, it is important to note that geff>0subscript๐‘”eff0g_{\rm eff}>0italic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT > 0 only for the state whose built-in field is aligned with the external one, sz(0)/ฮด>0superscriptsubscript๐‘ ๐‘ง0๐›ฟ0s_{z}^{(0)}/\delta>0italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT / italic_ฮด > 0 (the upper elliptical state in Fig.ย 1(b)). For the anti-aligned state, a real Zeeman splitting reduces the total effective field. That is why we consider the aligned states only. In the supercritical case their response diverges as sc/(2โข|pโˆ’pc|)subscript๐‘ ๐‘2๐‘subscript๐‘๐‘s_{c}/(2|p-p_{c}|)italic_s start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT / ( 2 | italic_p - italic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT | ) near p=pc๐‘subscript๐‘๐‘p=p_{c}italic_p = italic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT. However, in the subcritical regime, dynamically unstable states emerging from the pitchfork exhibit a negative response, aโขsz(1)/ฮด<0๐‘Žsuperscriptsubscript๐‘ ๐‘ง1๐›ฟ0as_{z}^{(1)}/\delta<0italic_a italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 1 ) end_POSTSUPERSCRIPT / italic_ฮด < 0, reducing the effective field magnitude, โ€“ see the red dashed lines in Fig.ย 1(c).

Figures 1(b) and 1(c) show that close to the critical pum**, polariton pseudospin perceives up to hundredfold increase in the Zeeman field magnitude. One of the striking manifestations of this strong field enhancement effect can be observed in the topological properties of Bogoliubov excitations in polariton lattices [15, 16]. Since emergence of nontrivial topological phases typically requires time-reversal symmetry breaking, the strong magnetic field is needed โ€“ up to 5ย T in Ref.ย [4]. However, the elementary excitations emerging in the condensate are also subject to the effective field, created by the condensate pseudospin polarisation and potentially significantly exceeding the external one.

In order to estimate the efficiency of the proposed field-enhancement principle for manipulating topology of bogolons, we proceed with the analysis of the dynamics of single condensate elementary excitations. In particular, we focus on the precession of weakly perturbed stationary pseudospin in the presence of finite magnetic field ฮด๐›ฟ\deltaitalic_ฮด.

Dynamics of the perturbation ๐–˜๐–˜\bm{\mathfrak{s}}bold_fraktur_s of the fixed point state ๐ฌ๐ฌ\mathbf{s}bold_s is governed by the linearized system (3), i.e. its Jacobi matrix, ๐–˜ห™=Jโข(๐ฌ)โข๐–˜ห™๐–˜๐ฝ๐ฌ๐–˜\dot{\bm{\mathfrak{s}}}=J(\mathbf{s})\bm{\mathfrak{s}}overห™ start_ARG bold_fraktur_s end_ARG = italic_J ( bold_s ) bold_fraktur_s. Therefore, its naturally expected that the excitation experiences the same effective field as the pseudospin state, geffโขฮด=aโขsz(0)+ฮดsubscript๐‘”eff๐›ฟ๐‘Žsuperscriptsubscript๐‘ ๐‘ง0๐›ฟg_{\rm eff}\delta=as_{z}^{(0)}+\deltaitalic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT italic_ฮด = italic_a italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT + italic_ฮด. However, due to the intrinsic non-Hermiticity of J๐ฝJitalic_J, the evolution of the weak perturbation strongly differs from the corresponding pseudospin dynamics.

The ๐–˜โข(t)๐–˜๐‘ก\bm{\mathfrak{s}}(t)bold_fraktur_s ( italic_t )-evolution can be given in terms of eigenvalues ฮปisubscript๐œ†๐‘–\lambda_{i}italic_ฮป start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT and eigenvectors ๐ฏisubscript๐ฏ๐‘–\mathbf{v}_{i}bold_v start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT of Jโข(๐ฌ)๐ฝ๐ฌJ(\mathbf{s})italic_J ( bold_s ). Since the Jacobian of (3) is purely real, it has either three real or a single real and two complex conjugate eigenvalues. We are interested in the latter case where the perturbation dynamics reads:

๐ˆ(t)=C1eโˆ’ฮณpโขt[๐ฎ1cos(ฮฉt+ฯ†)โˆ’\displaystyle\bm{\sigma}(t)=C_{1}e^{-\gamma_{p}t}\left[\mathbf{u}_{1}\cos(% \Omega t+\varphi)-\right.bold_italic_ฯƒ ( italic_t ) = italic_C start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT italic_e start_POSTSUPERSCRIPT - italic_ฮณ start_POSTSUBSCRIPT italic_p end_POSTSUBSCRIPT italic_t end_POSTSUPERSCRIPT [ bold_u start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT roman_cos ( roman_ฮฉ italic_t + italic_ฯ† ) -
๐ฎ2sin(ฮฉt+ฯ†)]+C2๐ฏ3eโˆ’ฮ“pโขt.\displaystyle\left.\mathbf{u}_{2}\sin(\Omega t+\varphi)\right]+C_{2}\mathbf{v}% _{3}e^{-\Gamma_{p}t}.bold_u start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT roman_sin ( roman_ฮฉ italic_t + italic_ฯ† ) ] + italic_C start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT bold_v start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT italic_e start_POSTSUPERSCRIPT - roman_ฮ“ start_POSTSUBSCRIPT italic_p end_POSTSUBSCRIPT italic_t end_POSTSUPERSCRIPT . (8)

Here ๐ฏ1=๐ฎ1+iโข๐ฎ2subscript๐ฏ1subscript๐ฎ1๐‘–subscript๐ฎ2\mathbf{v}_{1}=\mathbf{u}_{1}+i\mathbf{u}_{2}bold_v start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT = bold_u start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT + italic_i bold_u start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT is the eigenvector corresponding to either of complex conjugate eigenvalues ฮป1,2=โˆ’ฮณpยฑiโขฮฉsubscript๐œ†12plus-or-minussubscript๐›พ๐‘๐‘–ฮฉ\lambda_{1,2}=-\gamma_{p}\pm i\Omegaitalic_ฮป start_POSTSUBSCRIPT 1 , 2 end_POSTSUBSCRIPT = - italic_ฮณ start_POSTSUBSCRIPT italic_p end_POSTSUBSCRIPT ยฑ italic_i roman_ฮฉ, ๐ฏ3subscript๐ฏ3\mathbf{v}_{3}bold_v start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT is the eigenvector for purely real ฮป3=โˆ’ฮ“psubscript๐œ†3subscriptฮ“๐‘\lambda_{3}=-\Gamma_{p}italic_ฮป start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT = - roman_ฮ“ start_POSTSUBSCRIPT italic_p end_POSTSUBSCRIPT. Coefficients C1,2subscript๐ถ12C_{1,2}italic_C start_POSTSUBSCRIPT 1 , 2 end_POSTSUBSCRIPT and ฯ†๐œ‘\varphiitalic_ฯ† are real.

Refer to caption
Figure 2: Fig. 2. (a) A schematic evolution of the Bogoliubovโ€™s elementary excitation ๐–˜โข(t)๐–˜๐‘ก\bm{\mathfrak{s}}(t)bold_fraktur_s ( italic_t ) shown with the blue line in the pseudospin space. The end-point to which the spiral curls up corresponds to the dynamically stable stationary state ๐ฌ๐ฌ\mathbf{s}bold_s. The pale orange plane contains vectors ๐ฎ1,2subscript๐ฎ12\mathbf{u}_{1,2}bold_u start_POSTSUBSCRIPT 1 , 2 end_POSTSUBSCRIPT while the vector ๐›€๐›€\bm{\Omega}bold_ฮฉ is its normal. (b) Comparison between definitions of the effective magnetic field acting on the Bogoliubovโ€™s elementary excitation no top of the stable pseudospin states at ฮด=0.1๐›ฟ0.1\delta=0.1italic_ฮด = 0.1, a=2.28๐‘Ž2.28a=2.28italic_a = 2.28 and g=0.1๐‘”0.1g=0.1italic_g = 0.1. The p๐‘pitalic_p-dependencies of the precession frequency ฮฉฮฉ\Omegaroman_ฮฉ and the effective magnetic field ฮฉzsubscriptฮฉ๐‘ง{\Omega}_{z}roman_ฮฉ start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT illustrated in the panel (a). The vertical gray line indicates position of the fold bifurcation, โ€“ cf. with Fig.ย 1(c).

The long-term evolution (Giant effective g๐‘”gitalic_g-factor due to spin bifurcations in polariton condensates) of a weak excitation shown in Fig.ย 2(a) represents the dumped rotation with the frequency ฮฉฮฉ\Omegaroman_ฮฉ in the plane determined by ๐ฎ1,2subscript๐ฎ12\mathbf{u}_{1,2}bold_u start_POSTSUBSCRIPT 1 , 2 end_POSTSUBSCRIPT. The normal ๐งโŸ‚(๐ฎ1,๐ฎ2)perpendicular-to๐งsubscript๐ฎ1subscript๐ฎ2\mathbf{n}\perp(\mathbf{u}_{1},\mathbf{u}_{2})bold_n โŸ‚ ( bold_u start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , bold_u start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ) assigns direction to the effective magnetic field ๐›€=๐งโขฮฉ๐›€๐งฮฉ\mathbf{\Omega}=\mathbf{n}\Omegabold_ฮฉ = bold_n roman_ฮฉ whose z๐‘งzitalic_z-component ฮฉzsubscriptฮฉ๐‘ง\Omega_{z}roman_ฮฉ start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT is responsible for time-reversal symmetry breaking for Bogoliubovโ€™s excitations. The axis of pure relaxation ๐ฏ3subscript๐ฏ3\mathbf{v}_{3}bold_v start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT is misaligned with ๐›€๐›€\mathbf{\Omega}bold_ฮฉ in contrast to the problem of a classical spin in magnetic field.

In general, ๐›€๐›€\mathbf{\Omega}bold_ฮฉ differs from the effective field geffโขฮดsubscript๐‘”eff๐›ฟg_{\rm eff}\deltaitalic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT italic_ฮด acting on a pseudospin itself. However, in the limit of strong pum**, the excitation dynamics is dominated by the interaction-driven field aโขsz(0)๐‘Žsuperscriptsubscript๐‘ ๐‘ง0as_{z}^{(0)}italic_a italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT that implies ฮฉzโ‰ˆgeffโขฮดsubscriptฮฉ๐‘งsubscript๐‘”eff๐›ฟ\Omega_{z}\approx g_{\rm eff}\deltaroman_ฮฉ start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT โ‰ˆ italic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT italic_ฮด. In contrast, in the vicinity of the bifurcation point, where the giant enhancement of the external field occurs, this simple asymptotics fails.

The values of ฮฉzsubscriptฮฉ๐‘ง\Omega_{z}roman_ฮฉ start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT and geffโขฮดsubscript๐‘”eff๐›ฟg_{\rm eff}\deltaitalic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT italic_ฮด near the fold (saddle-node) bifurcation are compared in Fig.ย 2(b). Despite significant difference in the involved dynamics, the magnitudes of the effective field for a pseudospin ๐ฌ๐ฌ\mathbf{s}bold_s and for Bogoliubovโ€™s excitations ๐–˜๐–˜\bm{\mathfrak{s}}bold_fraktur_s follow the same trend. In addition, the existing mismatch quickly vanishes with the increase of the real field amplitude ฮด๐›ฟ\deltaitalic_ฮด.

In the case of coupled condensate lattices, the topological gap is expected to open in the Bogoliubovโ€™s excitation spectrum in the presence of symmetry breaking and spin-anisotropic interactions [15]. Moreover, in the effective field approximation, the uniform spin polarisation is equivalent to a conservative polariton Zeeman field and reduces to Hermitian spin-splitting diagonal terms in the Bogoliubov matrix. These terms responsible for topological gap opening are proportional to both the uniform spin polarisation and the strength of spin-anisotropic interaction, rendering the corresponding effective Zeeman field identical to the self-induced field aโขsz๐‘Žsubscript๐‘ ๐‘งas_{z}italic_a italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT in Eqs. (3).

In order to find the field-enhancement strength beyond perturbative approach, we numerically search for stable stationary solutions of systemย (3). The resulting g๐‘”gitalic_g-factor geff=(aโขsz+ฮด)/ฮดsubscript๐‘”eff/๐‘Žsubscript๐‘ ๐‘ง๐›ฟ๐›ฟg_{\rm eff}=\left.\left(as_{z}+\delta\right)\right/\deltaitalic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT = ( italic_a italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT + italic_ฮด ) / italic_ฮด is shown in Fig.ย 3(a) on the (ฮด,p)๐›ฟ๐‘(\delta,p)( italic_ฮด , italic_p ) parameter plane. Even away from the region of fast divergence, polariton condensate is able to amplify external magnetic field up to an order of magnitude.

The best results can be obtained at the weak magnetic field where geffsubscript๐‘”effg_{\rm eff}italic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT diverges according to the perturbation theory. This regime, however, has a significant drawback: at strong pum** which favours large geffsubscript๐‘”effg_{\rm eff}italic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT, polariton condensate exhibits bistability with two stable pseudospin configurations with opposite directions of the interaction-driven effective field. Under non-resonant excitation, the building-up of internal magnetization in the spin-bifurcation event occurs spontaneously with randomly selected direction. In particular, in the field-free case ฮดโ†’0โ†’๐›ฟ0\delta\rightarrow 0italic_ฮด โ†’ 0, the condensate excited above the critical pum** pcsubscript๐‘๐‘p_{c}italic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT occupies either of two elliptically polarised states (4) with equal probabilities [9].

Refer to caption
Figure 3: Fig. 3. (a) Effective g๐‘”gitalic_g-factor as a function of the pump p๐‘pitalic_p and the external magnetic field strength ฮด๐›ฟ\deltaitalic_ฮด. It is assumed that the condensate occupies a stable pseudospin state with aโขsz/ฮด>0๐‘Žsubscript๐‘ ๐‘ง๐›ฟ0as_{z}/\delta>0italic_a italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT / italic_ฮด > 0. (b) The probability ๐’ซ๐’ซ\mathcal{P}caligraphic_P of excitation of the state whose interaction-driven field is co-directed with the external one. A bistable regime with ๐’ซ<1๐’ซ1\mathcal{P}<1caligraphic_P < 1 occurs in the region on the right bottom corner of the (p๐‘pitalic_p,ฮด๐›ฟ\deltaitalic_ฮด) parameter plane. The dash-dotted line indicates position of the fold bifurcation which gives birth to the stable state with anti-aligned effective field. The descending ladder of probability cuts are shown with gray lines which are duplicated in panel (a). Each data point was obtained by a direct numerical solution of (3) with random initial conditions and averaging over 200 realizations.

However, in the broken symmetry case of finite ฮด๐›ฟ\deltaitalic_ฮด, the balance between spin-up and spin-down states shifts towards the state with the aligned built-in magnetic field corresponding to positive geffsubscript๐‘”effg_{\rm eff}italic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT. In the configuration shown in Fig.ย 1(c), it is the upper state with sz>0subscript๐‘ ๐‘ง0s_{z}>0italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT > 0.

The probability ๐’ซ๐’ซ\mathcal{P}caligraphic_P of occupation of the aligned state is shown in Fig.ย 3(b) in the parameter space spanned by p๐‘pitalic_p and ฮด๐›ฟ\deltaitalic_ฮด. Within the domain of ๐’ซ=1๐’ซ1\mathcal{P}=1caligraphic_P = 1, the condensate supports a single stable configuration. Far away from the low-ฮด๐›ฟ\deltaitalic_ฮด limit, this regime provides quite small geffsubscript๐‘”effg_{\rm eff}italic_g start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT about few units, โ€“ see Fig.ย 3(a). Above the critical pum** (dash-dotted line) corresponding to the fold bifurcation shown in Fig.ย 1(c), the probability is continuously declining and converges to ๐’ซ=0.5๐’ซ0.5\mathcal{P}=0.5caligraphic_P = 0.5 as ฮดโ†’0โ†’๐›ฟ0\delta\rightarrow 0italic_ฮด โ†’ 0 or pโ†’โˆžโ†’๐‘p\rightarrow\inftyitalic_p โ†’ โˆž. Therefore, the optimal regime, where strong field enhancement is combined with the high predictability of the condensate magnetization direction, is reached in a close vicinity to the bifurcation conditions.

In conclusion, we summarize the obtained results. Intrinsically non-equilibrium bosonic condensate of exciton polaritons can spontaneously develop strong effective magnetic fields due to the spin bifurcation phenomenon [9]. Since the very first studies [6, 7] this field was associated with the so-called self-induced Larmor precession of the Stokes vector of the light emitted by the condensate. Here we demonstrated that this phenomenon is controllable with an external permanent magnetic field and has a pronounced manifestation at the level of Bogoliubovโ€™s excitations [15]. The nonequilibrium condensate exhibits a very sensitive response to the applied magnetic field providing its strong enhancement. The enhancement strength can be characterised by the effective g๐‘”gitalic_g-factor which exceeds one hundred and strongly depends on the pump intensity and the external field magnitude.

The probabilistic character of the steady condensate spin polarisation and thus the direction of the effective field reduces the available parameter range down to the region near the fold bifurcation of the condensate pseudospin. Outside of this range, in particular, where spin multistability takes place, polariton spin fluctuations are expected to destabilize the condensate at sufficiently high temperatures [17, 18].

In addition, we notice a peculiar connection with the Landau theory of second-order phase transitions. The magnetic-field susceptibility of the condensate diverges near the critical point where a circular polarisation szโ‰ 0subscript๐‘ ๐‘ง0s_{z}\neq 0italic_s start_POSTSUBSCRIPT italic_z end_POSTSUBSCRIPT โ‰  0 appears spontaneously. For the symmetry-broken states, the divergence at p=pc๐‘subscript๐‘๐‘p=p_{c}italic_p = italic_p start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT is twice slower.

The obtained results pave the way to further investigations of the topological properties of Bogoliubovโ€™s excitations in the lattices of driven-dissipative polariton condensates.

Acknowledgements. The work of I.C. (analysis of the non-Hermitian dynamics, text writing) is supported by the Russian Science Foundation Grant No. 22-72-00061. The work of A.N. (analytical calculations, work supervision, text writing) is supported by he Russian Science Foundation under Grant No. 22-12-00144.

References

  • [1] M. Born, E. Wolf, and A. B. Bhatia, Principles of optics Cambridge University Press, Cambridge, 2019).
  • [2] F. D. M. Haldane and S. Raghu, Phys. Rev. Lett. 100, 1 (2008).
  • [3] T. Ozawa, H. M. Price, A. Amo et al., Rev. Mod. Phys. 91, 015006 (2019).
  • [4] S. Klembt, T. H. Harder, O. A. Egorov et al., Nature 562, 552 (2018).
  • [5] T. P. Lyons, D. J. Gillard, C. Leblanc et al., Nat. Photon. 16, 632 (2022).
  • [6] I. Shelykh, G. Malpuech, K. V. Kavokin et al., Phys. Rev. B 70, 115301 (2004).
  • [7] F. P. Laussy, I. A. Shelykh, G. Malpuech et al., Phys. Rev. B 73, 035315 (2006).
  • [8] Y. G. Rubo, A. Kavokin, and I. Shelykh, Phys. Lett. A 358, 227 (2006).
  • [9] H. Ohadi, A. Dreismann, Y. Rubo et al., Phys. Rev. X 5, 031002 (2015).
  • [10] N. A. Gippius, I. A. Shelykh, D. D. Solnyshkov et al., Phys. Rev. Lett. 98, 236401 (2007).
  • [11] Krรณl, R. Mirek, D. Stephan et al., Phys. Rev. B 99, 115318 (2019).
  • [12] K. Sawicki, D. Dovzhenko, Y. Wang et al., Phys. Rev. B 109, 125307 (2024).
  • [13] C. Whittaker, T. Dowling, A. Nalitov et al., Nat. Photon. 15, 193 (2021).
  • [14] M. C. Rechtsman, J. M. Zeuner, Y. Plotnik et al., Nature 496, 196 (2013).
  • [15] H. Sigurdsson, Y. S. Krivosenko, I. V. Iorsh et al., Phys. Rev. B 100, 235444 (2019).
  • [16] S. L. Harrison, A. Nalitov, P. G. Lagoudakis et al., Opt. Mater. Express 13, 2550 (2023).
  • [17] M. M. Glazov, M. A. Semina, E. Ya Sherman et al., Phys. Rev. B 88, 041309, (2013).
  • [18] I. I. Ryzhov, M. M. Glazov, A. V. Kavokin et al., Phys. Rev. B 93, 241307 (2016)