Arxiv:1809.08749V3 [Quant-Ph] 5 Dec 2018 a General Guiding Principle in Building the Theory of Fun- Damental Interactions (See, E.G., [1,3])

Resolution of Gauge Ambiguities in Ultrastrong-Coupling Cavity QED

Omar Di Stefano,1 Alessio Settineri,2 Vincenzo Macr`ı,1 Luigi Garziano,1 Roberto Stassi,1 Salvatore Savasta,1, 2, ∗ and Franco Nori1, 3 1Theoretical Quantum Physics Laboratory, RIKEN Cluster for Pioneering Research, Wako-shi, Saitama 351-0198, Japan 2Dipartimento di Scienze Matematiche e Informatiche, Scienze Fisiche e Scienze della Terra, Universitàdi Messina, I-98166 Messina, Italy 3Physics Department, The University of Michigan, Ann Arbor, Michigan 48109-1040, USA Gauge invariance is the cornerstone of modern quantum field theory [1–4]. Recently, it has been shown that the quantum Rabi model, describing the dipolar coupling between a two-level atom and a quantized electromagnetic field, violates this principle [5–7]. This widely used model describes a plethora of quantum systems and physical processes under different interaction regimes [8,9]. In the ultrastrong coupling regime, it provides predictions which drastically depend on the chosen gauge. This failure is attributed to the finite-level truncation of the matter system. We show that a careful application of the gauge principle is able to restore gauge invariance even for extreme light- matter interaction regimes. The resulting quantum Rabi Hamiltonian in the Coulomb gauge differs significantly from the standard model and provides the same physical results obtained by using the dipole gauge. It contains field operators to all orders that cannot be neglected when the coupling strength is high. These results shed light on subtleties of gauge invariance in the nonperturbative and extreme interaction regimes, which are now experimentally accessible, and solve all the long-lasting controversies arising from gauge ambiguities in the quantum Rabi and Dicke models [5, 10–18].

The ultrastrong-coupling (USC) between an effective The coupling to the field, obtained performing the re- two-level system (TLS) and the electromagnetic field has placement ∂µ → Dµ in the free Lagrangian, is called the been realized in several solid state systems [8,9]. In “minimal coupling”. this regime of quantum light-matter interaction, going It has been shown by several authors [46–50] that ap- beyond weak and strong coupling, the coupling strength proximate models for light-matter interactions derived in becomes comparable to the transition frequencies in the different gauges may lead to different predictions. It has system. Recently, circuit QED experiments involving a also been shown that the convergence of the two-photon single LC-oscillator mode coupled to a flux qubit su- transition rate depends significantly on the choice of the perconducting quantum circuit have realized coupling gauge [51]. strengths larger then the system transition frequencies When the light-matter interaction becomes very [19, 20]. This extreme interaction regime has been de- strong, different gauges can lead to drastically different noted deep strong coupling (DSC). In these interaction predictions, leading to controversies [5, 10–18]. For ex- regimes, several properties of coupled light-matter sys- ample, in the case of several TLSs interacting with a tems change drastically, opening the way to a wealth of single mode of an optical resonator [52], different gauges new intriguing physical effects (see, e.g. [21–35]) which may even lead to very different predictions, such as the may offer opportunities for the development of quantum presence or the absence of a quantum phase transition. technologies [36–44]. One important conclusion that can be drawn from The form of the electron-photon interaction is gauge these controversies is that, once the light-matter coupling dependent (see, e.g., [45]). However, all physical results becomes non-perturbative, the validity of the two-level must be independent of this choice. Gauge invariance is approximation for the atomic dipoles depends explicitly on the choice of gauge. In particular, it has been shown

arXiv:1809.08749v3 [quant-ph] 5 Dec 2018 a general guiding principle in building the theory of fundamental interactions (see, e.g., [1,3]). Let us consider, [6] that in the electric dipole gauge (the multipolar gauge for example, a particle field whose action is invariant un- in the dipole approximation) the two-level approxima- der a global phase change [U(1) invariance]. If this phase tion can be performed as long as the Rabi frequency re- is allowed to depend on the space-time coordinate x, its mains much smaller than the energies of the off-resonant action is not invariant because the coordinate-dependent higher-energy levels. However, it can dramatically fail in phase φ(x) does not commute with the derivatives (corre- the Coulomb gauge, even for systems with an extremely sponding to the four-momentum) in the action. The sym- anharmonic spectrum [6]. This unexpected result is par- metry can be restored replacing these derivatives with ticularly unsatisfactory, since the general procedure to covariant derivatives: Dµ = (∂µ + iqAµ), where q is the derive the multipolar gauge consists of using the min- charge parameter and Aµ is the gauge potential. If the imal coupling replacement first, and then applying the transformation of the particle field ψ → eiφ(x) is associ- Coulomb gauge [45]. ated to a gauge transformation of the four-vector poten- In a recent work [7], considering arbitrary gauges for tial Aµ → Aµ −∂µθ, the local U(1) symmetry is restored. the interaction between a TLS and the electromagnetic 2

field, it has been shown that physical predictions in the tracting great interest, but also provides a general insight USC and DSC regimes strongly depend on the gauge on gauge invariance in extreme interaction regimes. We choice. The gauge which provides the more accurate re- find that the usual strategy, which consists of taking into sults seems to depend on the detuning between the cavity account the nonlocality of the atomic potential, perform- mode resonance and the atomic transition frequency. ing the minimal coupling replacement and developing From all of these previous studies, it is clear that ap- the resulting interaction Hamiltonian up to second order proximations in the description of the matter system, in the vector potential, fails when the coupling strength as e.g., a finite-level truncation, ruin the gauge invari- reaches a significant fraction of the resonance frequencies ance of the theory. In 1970, Ref. [48] pointed out that of the system. We demonstrate that these gauge ambi- gauge ambiguities in the calculation of atomic oscilla- guities can be eliminated for arbitrary coupling strengths tor strengths, originate from the occurrence of nonlocal only by taking into account the approximation-induced potentials determined by the approximation procedures. nonlocality and keeping the resulting interaction Hamil- Since a nonlocal potential in the coordinate representa- tonian to all orders in the vector potential Aˆ . The re- tion is an integral operator, it does not commute with sults presented here solve all the long-lasting controver- the coordinate operator. Indeed, it is easy to show that sies arising from gauge ambiguities in the quantum Rabi it can be expressed as a local momentum-dependent op- and Dicke models. erator V (ˆr, pˆ). This affects the interaction of light with quantum systems described by approximate Hamiltoni- ans. Specifically, in order to introduce the coupling of the RESULTS matter system with the electromagnetic field, the minimal replacement rule pˆ → Πˆ = pˆ − Aˆ (r, t) (Aˆ is the Nonlocal potentials vector potential) has to be applied not only to the kinetic energy terms, but also to the nonlocal potentials in In order to understand why local potentials become the effective Hamiltonian of the particles in the system. nonlocal when the Hilbert space is truncated, let us con- By applying such a procedure, the ambiguities in the sider a one-dimensional potential Vˆ . In the position ba- calculation of approximate matrix elements for electric sis, it can written as dipole transitions can be removed [48]. Moreover, taking Z 0 0 0 into account the nonlocality of the approximate poten- Vˆ = dxdx hx|Vˆ |x i |xihx | (1) tial, two-photon transition rates involving Wannier exci- tons in semiconductors become gauge invariant [49]. Also If the potential is local, its matrix elements can be ˆ 0 0 0 the microscopic quantum theory of excitonic polaritons is written as hx|V |x i ≡ V (x, x ) = W (x)δ(x − x ). If affected by the presence of nonlocal potentials. The ap- |ni is a complete orthonormal basis, the matrix ele- 0 0 plication of the standard minimal coupling replacement ments can be expressed as V (x, x ) = W (x)δ(x − x ) = P ∗ 0 leads to a total Hamiltonian which, besides the usual Aˆ ·pˆ n,n0 Wn,n0 ψn0 (x )ψn(x), where we defined ψn(x) ≡ term, displays an additional diamagnetic term e2Aˆ 2/2m, hx|ni. Notice that the Dirac delta function can be re- where e and m are the electron charge and mass respec- constructed only by keeping all the infinite vectors of the tively. If a limited number of exciton levels are explic- basis. Hence, any truncation of the complete basis can itly included in the model, the long wavelength solution transform a local potential into a nonlocal one. If only of the polariton dispersion cannot be recovered without two states are included, e.g., the two lowest energy levels, introducing ad hoc the Thomas-Reiche-Kuhn sum rule we obtain and truncating the summation consistently [53]. This 0 ∗ 0 ∗ 0 V (x, x ) = W1,0 [ψ (x )ψ1(x) + ψ (x )ψ0(x)] (2) procedure presents some ambiguity and appears to be 0 1 rather artificial. However, if the nonlocality of the ap- where, for simplicity, we assumed parity symmetry and proximate potential is taken into account and the result- real matrix elements. It is evident that adding the ing additional terms are included, up to second order in two terms in Eq.(2), which are products of two smooth the vector potential, the correct dispersion relation is au- wavefunctions it is not possible to reproduce the Dirac- tomatically recovered [54, 55]. In summary, the concept delta function and this will result into a potential with of approximation-induced nonlocal potentials, together an high degree of spatial nonlocality. It has been with an expansion of the interaction Hamiltonian up to shown by several authors [48–50] that a nonlocal poten- second order in the vector potential, was able to overcome tial can be expressed as a momentum-dependent oper- a number of gauge ambiguities. ator V (ˆr, pˆ). Indeed, by using the translation operator Here we investigate if this strategy can work in the ψ(x0) = exp [i(x0 − x)ˆp]ψ(x), where pˆ is the momentum maximally truncated Hilbert space provided by a TLS, operator, we obtain and in the nonperturbative regimes of cavity QED. This Z investigation is relevant not only in order to remove gauge 0 0 0 ambiguities in quantum optical systems which are at- V (x, x )ψ(x )dx = V (x, pˆ)ψ(x) . (3) 3

6 The dipole gauge 6 (a)

We consider a single electric dipole with charge q cou- 4 pled to a single mode of the electromagnetic field. The 4 field mode is described by a harmonic oscillator with bare frequency ωc and annihilation and creation operators aˆ and aˆ†. In one dimension, the dipole can be modeled as 2 an effective particle of mass m in a potential V (x). Its 2 2 unperturbed Hamiltonian is Hˆ0 =p ˆ /(2m) + Vˆ (x). Ap- plying the dipole approximation, the total Hamiltonian 0 Resonant in the dipole gauge is 0 2 2 6 (b) ˆ † ˆ q A0ωc 2 † 6 HD = ~ωcaˆ aˆ + H0 + xˆ + iqωcxAˆ 0(â − aˆ) , (4) ~ where A0 is the zero point fluctuation amplitude of the † 4 vector potential Aˆ = A0(â +a ˆ ). 4

If the two lowest-energy levels of the effective quantum Energy levels particle are well separated from the higher energy levels, 2 as in the case of flux qubits [19], and if the detuning 2 δ = ω10 − ωc (ω10 is the transition frequency of the two lowest energy levels) is much smaller than the detunings Detuned with respect to other transitions, it is possible to truncate 0 the Hilbert space considering only the two lowest-energy 00 1 2 3 0 2 levels. In this case, the unperturbed particle Hamiltonian light-matter coupling ˆ reduces to H0 = ~ω10σˆz/2, and the total Hamiltonian becomes Figure 1. Comparison of the energy spectra as a function of the normalized coupling η = g /ω , obtained from the ω D c ˆ † ~ 10 † quantum Rabi Hamiltonians in the dipole gauge (Hˆ ), in HD = ~ωcaˆ aˆ + σˆz + i~gD(â − aˆ)ˆσx (5) D 2 ˆ0 the Coulomb gauge (standard derivation, HC ), and in the where g = ω A d / and d ≡ qh1|xˆ|0i is the dipole Coulomb gauge taking into account the presence of nonlocal D c 0 10 ~ 10 ˆ matrix element. In Eq. (5) we dropped the constant potentials (HC ). The plots in (a) have been obtained using a 2 2 detuning δ = 0; the plots in (b) using δ = 0.5ωc. term C = d10ωcA0/~, which results from the expecta- tion value hi|xˆ2|ji by using the TLS identity operator ˆ I2 = |0ih0| + |1ih1|. This term could be calculated more Projecting Hˆ in a two-level space, we obtain accurately by including it in the particle potential, be- C fore the diagonalization. However, we made the choice of ω q2A2 Hˆ0 = ω aˆ†aˆ+ ~ 10 σˆ + g σ (â† +â)+ 0 (â† +â)2 , considering the interaction terms only after the Hilbert C ~ c 2 z ~ C y 2m space truncation. De Bernardis et al. [6] have shown (7) that, for an effective particle displaying the two lowest- where gC = gD ω10/ωc. The diamagnetic term energy levels well separated from the higher energy levels q2Aˆ2/(2m) can be absorbed by using a Bogoliubov trans- (ω21 gD), the two level approximation provides results formation involving only the photon operators [6, 19]. which are in agreement with those obtained using the In Fig.1 we plot the energy differences ( En − E0) for ˆ ˆ0 ˆ full Hamiltonian in Eq. (4), even for extreme coupling the lowest eigenstates of HD, HC , and HC [see Eq. (10)], strengths (gD ω10). as a function of the normalized coupling η = gD/ωc. Fig- ure 1a has been obtained at zero detuning (δ = 0), while we used δ = 0.5 ωc for Fig. 1b. The comparison in Fig.1 Derivation of the quantum Rabi model in the shows that, for very small values of the coupling, the Coulomb gauge eigenvalues of the different Hamiltonians reproduce the expected behaviour. However, already at moderate cou- In the Coulomb gauge, the Hamiltonian describing the pling strengths, η ∼ 0.1, there are significant deviations interaction of an effective quantum particle with the elec- in the predicted energies, in agreement with the results tromagnetic field can be obtained applying the substitu- obtained in Ref. [6]. For values of η 0.5 the differences ˆ ˆ & tion pˆ → pˆ − qA to the particle Hamiltonian H0. The become dramatic. The analysis carried out in Ref. [6] well-known result is clarifies that the disagreement between the eigenvalues 2 2 ˆ ˆ0 q A of HD and HC is due to the failure of the Rabi model to Hˆ = ω aˆ†aˆ+Hˆ −iqpAˆ (â† +â)+ 0 (â† +â)2 . (6) C ~ c 0 0 2m provide correct results beyond small coupling values in 4

ˆ0 the Coulomb gauge ( HC ). 6 However, this derivation of the quantum Rabi model (a) in the Coulomb gauge does not take into account that, in the presence of a truncated Hilbert space, the particle 4 potential can loose its locality Vˆ (ˆx) → Vˆ 0(ˆx, pˆ). In this case, in order to preserve gauge invariance, one has to ˆ apply the substitution pˆ → pˆ − qA also to the potential. 2 In principle, this procedure can give rise to additional terms in the interaction Hamiltonian to all orders in the vector potential. These higher order terms are expected Resonant to be negligible for small normalized couplings η, but can 0 6 become relevant at higher coupling strengths. (b) As shown in detail in the Supplementary Note1, using some general operator theorems [56], it is possible to apply the minimal coupling replacement to both the ki- 4 netic energy and the nonlocal potential of the eﬀective Hamiltonian of a quantum particle by employing a uni- Energy levels tary transformation [54]: 2

† HˆC = UˆHˆ0Uˆ + Hˆph , (8) Detuned ˆ † where Hph = ~ωcaˆ aˆ. In the dipole approximation, the 0 0.1 0.2 0.3 0.4 unitary operator is light-matter1 coupling " # qxˆAˆ Uˆ = exp i . (9) Figure 2. Comparison of the energy spectra as a function ~ of the normalized coupling η = gD/ωc, obtained from the quantum Rabi Hamiltonians in the dipole gauge (HˆD), in ˆ0 If the two-level approximation is used, in order to the Coulomb gauge (standard derivation, HC ), and in the include the interaction, one has to consistently project Coulomb gauge taking into account the presence of nonlocal ˆ Eq. (8) in the two-level space before performing the uni- potentials (HC ). The plots in (a) have been obtained using a detuning δ = 0; the plots in (b) using δ = 0.5ω . tary transformation of Hˆ0: c

ˆ ˆ ˆ ˆ† ˆ HC = UH0U + Hph , (10) potentials, even without knowing them explicitly. Notic- ing that the operator Uˆ corresponds to a spin rotation where the projected unitary operator is along the x axis, we obtain † Uˆ = exp iησˆx(â +a ˆ ) . (11) ˆ † ~ω10 HC = ~ωcaˆ aˆ + Note that the standard procedure consists of applying 2 (13) † † the minimal coupling replacement to the exact matter σˆz cos 2η(â +a ˆ ) +σ ˆy sin 2η(â +a ˆ ) . Hamiltonian Hˆ0 as, e.g., in Eq. (8), and then to project the resulting Hamiltonian in the truncated Hilbert space: This is the correct Hamiltonian in the Coulomb gauge of the quantum Rabi model. The price one has to pay for pˆ2 preserving the gauge principle in such a truncated space Hˆ0 = PÛˆHˆ Uˆ †Pˆ + Hˆ = Pˆ Uˆ Uˆ † + Vˆ Pˆ + Hˆ , C 0 ph 2m ph is that the resulting Hamiltonian will contain field opera- (12) tors at all orders. The generalization to multimode fields where Pˆ is the projection operator for the truncated [57–59] is straightforward. When the normalized interac- Hilbert space, and we used the relation UˆVˆ Uˆ † = Vˆ , tion strength is sufficiently weak, it is possible to expand valid when the potential is local. Note that, in con- the trigonometric functions up to the linear or quadratic trast to Eq. (10), Eq. (12) contains a nonlocal potential terms. However, when gD becomes comparable with ωc PˆVˆ Pˆ to which the gauge principle has not been applied. (i.e., the USC regime), such expansion is not sufficient. This analysis explains why the standard approach fails When approaching, or reaching, the deep strong coupling when applied to a matter system described in a trun- regime (η & 1) the field terms have to be included at cated Hilbert space. all orders. Figure (1) shows that the numerically cal- Of course, this procedure can be easily generalized to culated energy spectrum of HˆC coincides with that of any truncated Hilbert space. It is remarkable that it HˆD. Hence we can conclude that, when the presence of takes automatically into account the presence of nonlocal nonlocal potentials, owing to the two-level truncation, is 5 properly taken into account, it is possible to obtain correct Resolution of gauge ambiguities results also in the Coulomb gauge. ˆ ˆ Equation (10) can be easily generalized beyond the The Hamiltonians HD and HC are related by a gauge single-mode approximation. Expanding Eq. (10) up to transformation [45], which can be expressed by the uni- ˆ ˆ ˆ ˆ† the second order in the potential, we obtain tary transformation HC = GHDG . In the dipole approximation Gˆ = Uˆ [see Eq. (9)]. Applying this gauge transformation, it is easy to obtain the quantum Rabi 2 ˆ(2) † ~ω10 † ~gC † 2 Hamiltonian in the Coulomb gauge Eq. (6) from Eq. (4). HC = ~ωcaˆ aˆ+ σˆz+~gC σˆy(â+â )+ σˆz(â+â ) . ˆ 2 ω10 Of course, the inverse relationship also holds: HD = † (14) Gˆ HˆC Gˆ. ˆ0 This equation differs from the standard quantum Rabi As discussed above, it has been shown that HC (the ˆ0 Hamiltonian in the Coulomb gauge HC for the diamag- two-level approximation of HˆC ) gives rise to wrong spec- netic term, which now displays a different coefficient and tra, thus ruining gauge invariance. In the previous sub- depends on σˆz. By introducing the Thomas-Reiche-Kuhn section, we have demonstrated that the correct quantum ˆ0 sum rule in the diamagnetic term of HC and applying the Rabi Hamiltonian in the Coulomb gauge, based on the ˆ0 ˆ(2) two-level truncation [7], HC turns into HC . However, complete introduction of the gauge-preserving replace- this procedure is quite arbitrary. Figure (2) compares ment pˆ → Πˆ = pˆ −Aˆ (ˆr, t), is HˆC in Eq. (13). Numerical ˆ0 ˆ(2) ˆ results in Fig.1 show that Hˆ gives correct spectra, thus the energy spectra of HC , HC , and HC . In particular, C we observe that for values of the normalized coupling be- providing clear indications that the procedure developed ˆ(2) ˆ0 here restores gauge invariance in TLSs. low 0.1 the spectra of both HC (red lines) and HC (blue dashed lines) are in quite good agreement with the ex- We now present an analytical demonstration of the (2) gauge invariance of a TLS coupled to the electromag- act results (expecially those of Hˆ ) given by Hˆ (black C C netic field. We show that gauge invariance in a truncated dotted lines). Increasing the coupling, Hˆ0 gives spec- C Hilbert space can be fully preserved if the gauge trans- tra which rapidly become very different from the exact formation is consistently performed. We start from H , ˆ(2) D energy levels. The spectra of HC provide an improved which was shown to be a very good approximation of the approximation, which, for the lowest few levels, is ac- full Hamiltonain HD, and apply the gauge transforma- ceptable for η . 0.2. The range of values of η where the tion projecting the unitary operator Gˆ = Uˆ in the two- approximate models provide acceptable spectra reduces † level space: GˆHˆDGˆ , where Gˆ = Uˆ. The result of this uni- at nonzero detuning [Fig. (2) (b)]. tary transformation should be HˆD → HˆC . Noticing that In order to understand how many powers of the pho- the operator Uˆ corresponds to a spin rotation along the ton operators have to be included in HˆC to obtain correct x axis, and using the Baker-Campbell-Hausdorff lemma, spectra at higher η, we compared in Fig. 3 the approx- it is easy to obtain imate spectra, calculated from different n-order Taylor (n) ˆ ˆ † † ~ω0 † expansions HC of HC , with the exact spectra (the eigen- GˆHˆ Gˆ = ω aˆ aˆ + σˆ cos 2η(â +a ˆ ) D ~ c 2 z values of HˆC ). The results are interesting: for n = 3 (15) † ˆ there is already a significant improvement (with respect +σ ˆy sin 2η(â +a ˆ ) = HC . to n = 2), up to η . 0.25. However, the spectra become † completely wrong already at η ∼ 0.3. Accuracy improves This equation shows that indeed GˆHˆDGˆ = HˆC . This for n = 10, but only up to η . 0.25. For n = 200, there result demonstrates that gauge invariance is preserved in is an excellent agreement, but only for η . 1.3. These a two-level truncated space, if the gauge transformation results show that for values of η larger than 1 (DSC), a ˆ ˆ0 is applied consistently, and if we use HC instead of HC . very large n is needed to get the correct spectra. How- Following Ref. [7], it is possible to employ a formu- ever, further increasing η requires the inclusion of more lation in which the gauge freedom is contained within a and more terms in the expansion. This confirms the non- single real continuous parameter α, which determines the perturbative spatial nonlocality which occurs when heavily gauge through a function Xα. The general gauge trans- truncating the particle’s Hilbert space. formation in the dipole approximation is generated by a Specific results for circuit-QED systems can be ob- unitary transformation determined by the unitary oper- ˆ h ˆ i ˆ ˆ tained analogously. For example, the full Hamiltonian ators Uα = exp iXα , where Xα = αqxÂ/~. The values of a fluxonium capacitively coupled to an LC oscillator α = 0, 1 specify the dipole and the Coulomb gauge, re- circuit [60] (corresponding to the charge gauge) can be spectively. The two-level projected unitary operator is † obtained through an analogous minimal coupling replace- Uˆα = exp [iχˆα], where χˆα = α(gD/ωc)ˆσx(â +a ˆ ). The ment. As shown in the Supplementary Note2, the re- correct α-gauge Hamiltonian for a TLS is thus sulting total Hamiltonian for the two-level model in the ˆ(α) ˆ ˆ ˆ† charge gauge is very similar to Eq. (10). H = UαHDUα . (16) 6

5 n=2 n=3 n=10

0 0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0 0.1 0.2 0.3 s 5 n=40 n=200 n=

2 Energy level

0 0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.3 0.6 0.9 1.2 1.5 0 1 2 3 light-matter coupling

Figure 3. Energy spectra as a function of the normalized coupling η, obtained from the quantum Rabi Hamiltonians in the Coulomb gauge taking into account the presence of nonlocal potentials (HˆC ). The panel shows the comparison between the exact (containing all terms) (black continuous curve, see also the bottom-right panel) and approximated energy levels (red curves). The calculations are performed at diﬀerent order of approximation. The value n indicates the order of the Taylor expansion used in the approximated Hamiltonian.

We obtain presence of nonlocal potentials, this can be done by ap- ˆ(α) † † plying the unitary transformation H = ~ωcaˆ aˆ − i(1 − α)gD(â − aˆ )σx (17) ω HˆN = Uˆ HˆN U † + Hˆ , (19) +~ 0 σˆ cos 2αη(â +a ˆ†) +σ ˆ sin 2αη(â +a ˆ†) . C N 0 N ph 2 z y where Since, the transformation in Eq. (16) is unitary, the h † i Hamiltonian (17) will have the same energy spectra of UˆN = exp i2η(â +a ˆ)Jˆx (20) (α=0) (α=1) HˆD = Hˆ and of HˆC = Hˆ for any value of α. This eliminates any gauge ambiguity of the quantum Rabi ˆ h P ˆ i is the projection of UN = exp iq k xˆkA/~ in a re- model. duced Hilbert space, which is the tensor product of N two-dimensional Hilbert spaces. We obtain Derivation of the Dicke model in the Coulomb gauge ˆN † n ˆ † HC = ~ωcaˆ aˆ + ~ω10 Jz cos 4η(â +a ˆ) We now extend the results obtained above for a single (21) two-level dipole (Rabi) to the multi-dipole case (Dicke) † o + Jˆy sin 4η(â +a ˆ) . [5, 10, 61]. The standard Dicke Hamiltonian in the Coulomb gauge can be written as This is the correct Hamiltonian in the Coulomb gauge ˆ0N † ˆ for the Dicke model. Its gauge invariance can be easily HC = ~ωcaˆ aˆ + ~ω10Jz 2 2 (18) proved following the same procedure shown in the previ- † ˆ q A0 † 2 + 2~gc aˆ +a ˆ Jy + j aˆ +a ˆ , ous subsection. This result eliminates any gauge ambi- m guity [5, 14, 15] of the Dicke model. where the number N of dipoles determines the effective The Hamiltonian in Eq. (21) is equivalent to the one main angular momentum quantum number j = N/2, and in Eq. (45) of Ref. [5] (if direct dipole-dipole interactions ˆ PN (k) 2Ji = k=1 σî . are not included). The latter has been obtained apply- Similarly to the quantum Rabi Hamiltonian, it is possi- ing a polaron transformation to the Dicke Hamiltonian ble to introduce the coupling between the matter Hamil- in the dipole gauge. This unitary transformation, as ob- N ˆ tonian (in this case H0 = ~ω10Jz) and the electromag- served by the authors [5], is equivalent to a gauge trans- netic field, by the minimal coupling replacement. In the formation and hence gives rise to a Dicke Hamiltonian 7 in the Coulomb gauge. However, as pointed out there, tentials. A further interesting development could be the Ref. [5] does not solve the gauge ambiguities of the Rabi application of this analysis to many-body quantum sys- and Dicke models (see also Ref. [5,6]). Indeed, their di- tems strongly interacting with the electromagnetic field rect derivation of the Dicke Hamiltonian in the Coulomb [62]. gauge was obtained employing the standard minimal coupling substitution, instead of the correct generalized one We acknowledge conversations with S. De Liberato and proposed here [Eq. (19]. P. Rabl. FN is supported in part by the: MURI Cen- ter for Dynamic Magneto-Optics via the Air Force Of- fice of Scientific Research (AFOSR) (FA9550-14-1-0040), DISCUSSION Army Research Office (ARO) (Grant No. 73315PH), Asian Office of Aerospace Research and Development Approximate models for light-matter interactions de- (AOARD) (Grant No. FA2386-18-1-4045), Japan Sci- rived in different gauges have produced drastically dif- ence and Technology Agency (JST) (the ImPACT pro- ferent predictions when the light-matter interaction gram and CREST Grant No. JPMJCR1676), Japan So- strength reaches high values, such as the presence or ciety for the Promotion of Science (JSPS) (JSPS-RFBR absence of a quantum phase transition. We succeeded Grant No. 17-52-50023, and JSPS-FWO Grant No. to identify the origin of these gauge ambiguities recently VS.059.18N), RIKEN-AIST Challenge Research Fund, pointed out. and the John Templeton Foundation. Specifically, we found that these ambiguities originate The authors declare no competing interests from the presence of potentials depending on both position and momentum (nonlocal potentials) in the effective Hamltonians describing the matter systems. This is SUPPLEMENTARY NOTE 1: GENERALIZED an unavoidable consequence of the usual approximations MINIMAL COUPLING REPLACEMENT (like Hilbert space truncation). In these cases, gauge invariance of the light-matter Hamiltonian can be pre- In this section, by using some general operator theo- served, even in the presence of extreme light-matter inter- rems [56], we show how to implement the minimal cou- actions, only by applying the minimal coupling replace- pling replacement on a generic operator O(ˆx, pˆ) by a uni- ment to both the kinetic and the potential terms of the tary transformation [54]. Given two noncommuting op- effective Hamiltonian. A careful application of the gauge erators αˆ and βˆ and a parameter µ we want to calculate principle generates a quantum Rabi Hamiltonian in the eµβÔ(ˆα)e−µβˆ. The function O(ˆα) can be expanded in a Coulomb gauge, which is significantly different from the power series: standard quantum Rabi model. It contains field oper- X ators to all orders that cannot be neglected when the O(ˆα) = c αˆn . (22) coupling strength is high. The derivation presented here n n is based on the fundamental minimal coupling replacement, and thus is not limited to the Coulomb gauge. We Using Eq. (22), we have have also derived arbitrary gauge descriptions (depending on a real parameter α) of cavity-QED Hamiltonians for µβˆ −µβˆ X µβˆ n −µβˆ e O(ˆα)e = cne αˆ e . (23) TLSs. These results resolve all the controversies arising n from gauge ambiguities in the quantum Rabi and Dicke models. Observing that This analysis bring a deeper understanding of the sub- µβˆ n −µβˆ µβˆ −µβˆ µβˆ −µβˆ tleties of the gauge principle, which cannot be ignored e αˆ e =e αeˆ e αeˆ ˆ ˆ ˆ ˆn in the nonperturbative and extreme interaction regimes, . . . eµβαeˆ −µβ = eµβαeˆ −µβ . which are now experimentally accessible. The lesson that can be learned is very general: when introducing interac- We have: tions according to the gauge principle, any approximation n in the description of a quantum system requires a gener- µβˆ −µβˆ X µβˆ −µβˆ µβˆ −µβˆ e O(ˆα)e = cn e αeˆ = O(e αeˆ ) . alization of the minimal coupling replacement in order to n avoid inconsistencies and completely wrong predictions. (24) We presented a general method for the derivation of We now apply Eq. (24) to eiχ(ˆx)/~O(ˆx, pˆ)e−iχ(ˆx)/~. For gauge-invariant light-matter Hamiltonians in truncated the sake of simplicity, we consider here the 1D case. Gen- Hilbert spaces. This approach can be extended to study eralization to 3D is straightforward. We obtain ultrastrong and deepstrong light-matter interactions beyond the dipole approximation [50], where the multipolar eiχˆ(x)/~O(x, pˆ)e−iχˆ(x)/~ = O(x, eiχˆ(x)/~peˆ −iχ(x)/~) . gauge [45] is also affected by the presence of nonlocal po- (25) 8

Then, by using the Baker-Campbell-Hausdorff Lemma, whose commutator is [ϕ, ˆ χˆ] = i. These can be expanded we obtain: in terms of the creation and annihilation operators: 2 i 1 i ϕˆ = ϕ (â +a ˆ†) , eiχˆ(x)/~peˆ −iχˆ(x)/~ =p ˆ + [ˆχ(x), pˆ] + [χ(ˆx), [χ(ˆx), pˆ]] + 0 ~ 2 ~ † ... =p ˆ − ∂xχˆ(x) , (26) χˆ = −iχ0(â − aˆ ) , √ √ where we have used the result [ˆχ(x), pˆ] = i ∂xχˆ(x). In 2 2 ~ where ϕ0 = 2e L~Ω = 2~e Z, and χ0 = conclusion, using Eq. (26), Eq. (25) becomes p p C~Ω/(8e2) = ~/8e2Z. The Hamiltonian of the LC oscillator can be expressed eiχˆ(x)/~O(x, pˆ)e−iχˆ(x)/~ = O(x, pˆ − ∂ χˆ(x)) . (27) x as Considering now the special function ϕˆ2 Hˆ = 4E ζˆ2 + E , (33) osc C L 2 χˆ(x) = xAˆ0 , (28) 2 2 where EC = e /(2C) and EL = (φ0/2π) /L. where Aˆ0 ≡ Aˆ(x0) is the field potential calculated at the The capacitive coupling (charge gauge) between the atom position x0, we obtain fluxonium and the LC oscillator can be described by making the substitution Nˆ → Nˆ +χ ˆ in the fluxonium ˆ ∂xχˆ(x) = A0 . (29) Hamiltonian in Eq. (31). The total Hamiltonian then becomes: If we plug this result into Eq. (27), we get ˜ 2 ˆ ˜ ˆ 2 ˆ EL ˆ2 2 ϕˆ iχˆ(x)/~ −iχˆ(x)/~ ˆ HC = 4EC (N +χ ˆ) −EJ cos(φ)+ φ +4EC χˆ +EL . e O(x, pˆ)e = O(x, pˆ − A0) , (30) 2 2 (34) demonstrating that the unitary transformation in This replacement is different from the standard minimal Eq. (27) corresponds to the application of the minimal coupling replacement, it involves two conjugate momenta, coupling replacement in the dipole approximation. instead of a conjugate momentum and a field coordinate. This minimal coupling replacement can also be obtained by applying a unitary transformation to the fluxonium SUPPLEMENTARY NOTE 2: FLUXONIUM LC Hamiltonian Hˆflux: OSCILLATOR HAMILTONIAN IN THE CHARGE GAUGE † HˆC = Hôsc + Rˆ HˆfluxRˆ , (35)

Here we derive the total Hamiltonian describing a flux- where Rˆ = exp[iφˆχˆ]. This can be proved by following the onium qubit (TLS) interacting with a superconducting procedure described in the next section. ˆ LC oscillator [7, 60]. Considering for simplicity the case Diagonalizing Hflux and then projecting Eq. (34) in a of a zero-flux offset, the fluxonium Hamiltonian is: two-level space, spanned by the eigenstates |0i (ground state) and |1i (excited state), we obtain E˜ ˆ ˜ ˆ 2 L ˆ2 ˆ 0 † † 2 † 2 Hflux = 4EC N + φ − EJ cos φ . (31) Hˆ = Hˆ + ω aˆ aˆ+i g σˆ (â−aˆ )−4E˜ χ (â−aˆ ) , 2 C flux ~ c ~ C y C 0 (36) ˆ ˆ ˆ ˆ where in the two-level approximation the reduced flux where φ = Φ/Φ0 = 2e/~ Φ is the reduced flux opera- ˆ ˆ tor with conjugate momentum Nˆ = Q/ˆ 2e (the reduced operator becomes φ = Φ10σˆx (Φ10 ≡ he|φ|gi is assumed ˆ real), and g = ω φ χ . We also have: charge), such that [φ, Nˆ] = i. In Eq. (31), E˜C , E˜L, C 10 10 0 and EJ are, respectively, the capacitive, inductive, and ˆ ~ω10 Josephson energies. Hflux = σˆz , (37) 2 The LC oscillator is characterized by a capacitance C with σˆ = |1ih1| − |0ih0|. and√ an inductance L. Its resonance frequency is ωc = z 1/ LC and its characteristic impedance is Z = pL/C. It can be shown that the total Hamiltonian in Eq. (36) The Hamiltonian of the LC oscillator is is not gauge invariant because its derivation does not take into account the presence of an effective nonlocal poten- Qˆ2 Φˆ2 tial which originates from the two-level truncation. In Hˆ = + , (32) osc 2C 2L analogy with the procedure employed in the main text, the correct gauge invariant total Hamiltonian can be ob- ˆ ˆ where Q and Φ are the charge and the flux operator, tained applying the following unitary transformation to respectively. Hˆflux: We now introduce the reduced flux operator ϕˆ = ˆ ˆ † ˆ ˆ ˆ† 2πΦ/Φ0 and the reduced charge operator χˆ = Q/(2e), HC = ~ωc aˆ aˆ + R HfluxR , (38) 9 where [15] P. Nataf and C. Ciuti, “No-go theorem for superradiant quantum phase transitions in cavity QED and counter- gC † Rˆ = exp σˆx(â − aˆ ) (39) example in circuit QED,” Nat. Commun. 1, 72 (2010). ω10 [16] A. Vukics, T. Grießer, and P. Domokos, “Elimination is the unitary operator resulting from the truncation in of the A-square problem from cavity QED,” Phys. Rev. the two-level space of the unitary operator Rˆ. The mini- Lett. 112, 073601 (2014), arXiv:1310.0694. [17] T. Grießer, A. Vukics, and P. Domokos, “Depolarization mal coupling replacement in the truncated space Eq. (38) shift of the superradiant phase transition,” Phys. Rev. A gives 94, 033815 (2016), arXiv:1604.03531. [18] M. Bamba, K. Inomata, and Y. Nakamura, “Super- radiant Phase Transition in a Superconducting Circuit ˆ † ~ω10 2gC † HC = ~ωcaˆ aˆ + cosh (â − aˆ ) σˆz+ in Thermal Equilibrium,” Phys. Rev. Lett. 117, 173601 2 ω10 (2016), arXiv:1605.01124. 2g [19] F. Yoshihara, T. Fuse, S. Ashhab, K. Kakuyanagi, + i sinh C (â − aˆ†)ˆσ . ω y S. Saito, and K. Semba, “Superconducting qubit- 10 oscillator circuit beyond the ultrastrong-coupling This equation describes the correct (gauge invariant) to- regime,” Nat. Phys. 13, 44 (2017), arXiv:1602.00415. [20] F. Yoshihara, T. Fuse, Z. Ao, S. Ashhab, K. Kakuyanagi, tal Hamiltonian for a fluxonium qubit interacting with S. Saito, T. Aoki, K. Koshino, and K. Semba, “Inver- an LC circuit in the charge gauge. sion of qubit energy levels in qubit-oscillator circuits in the deep-strong-coupling regime,” Phys. Rev. Lett. 120, 183601 (2018), arXiv:1602.00415. [21] S. Ashhab and F. Nori, “Qubit-oscillator systems in the ultrastrong-coupling regime and their potential for ∗ corresponding author: [email protected] preparing nonclassical states,” Phys. Rev. A 81, 042311 [1] I.J. R. Aitchison and A. J. G. Hey, “Gauge theories in (2010), arXiv:0912.4888. particle physics,” Adam Hilger , 327 (1989). [22] J. Casanova, G. Romero, I. Lizuain, J. J. Garc´ıa-Ripoll, [2] F. Wilczek, “In search of symmetry lost,” Nature 433, and E. Solano, “Deep Strong Coupling Regime of the 239–247 (2005). Jaynes-Cummings Model,” Phys. Rev. Lett. 105, 263603 [3] M. Maggiore, A modern introduction to quantum field (2010), arXiv:1008.1240. theory, Vol. 12 (Oxford university press, 2005). [23] T. Niemczyk, F. Deppe, H. Huebl, E. P. Menzel, [4] F. Wilczek, “Particle physics: Minimalism triumphant,” F. Hocke, M. J. Schwarz, J. J. Garcia-Ripoll, D. Zueco, Nature 496, 439–441 (2013). T. Hummer,¨ E. Solano, A. Marx, and R. Gross, “Circuit [5] D. De Bernardis, T. Jaako, and P. Rabl, “Cavity quan- quantum electrodynamics in the ultrastrong-coupling tum electrodynamics in the nonperturbative regime,” regime,” Nat. Phys. 6, 772 (2010), arXiv:1003.2376. Phys. Rev. A 97, 043820 (2018), arXiv:1712.00015. [24] A. Ridolfo, M. Leib, S. Savasta, and M. J. Hartmann, [6] D. De Bernardis, P. Pilar, T. Jaako, S. De Liberato, and “Photon Blockade in the Ultrastrong Coupling Regime,” P. Rabl, “Breakdown of gauge invariance in ultrastrong- Phys. Rev. Lett. 109, 193602 (2012), arXiv:1206.0944. coupling cavity QED,” arXiv:1805.05339 (2018). [25] A. Ridolfo, S. Savasta, and M. J. Hartmann, “Nonclassi- [7] A. Stokes and A. Nazir, “Gauge ambiguities in cal Radiation from Thermal Cavities in the Ultrastrong qed: Jaynes-Cummings physics remains valid in the Coupling Regime,” Phys. Rev. Lett. 110, 163601 (2013), ultrastrong-coupling regime,” arXiv:1805.06356 (2018). arXiv:1212.1280. [8] A. F. Kockum, A. Miranowicz, S. De Liberato, [26] R. Stassi, A. Ridolfo, O. Di Stefano, M. J. Hartmann, S. Savasta, and F. Nori, “Ultrastrong coupling between and S. Savasta, “Spontaneous Conversion from Virtual light and matter,” arXiv:1807.11636 (2018). to Real Photons in the Ultrastrong-Coupling Regime,” [9] P. Forn-D´ıaz, L. Lamata, E. Rico, J. Kono, and Phys. Rev. Lett. 110, 243601 (2013), arXiv:1210.2367. E. Solano, “Ultrastrong coupling regimes of light-matter [27] M. Cirio, S. De Liberato, N. Lambert, and F. Nori, interaction,” arXiv:1804.09275 (2018). “Ground State Electroluminescence,” Phys. Rev. Lett. [10] K. Hepp and E. H. Lieb, “On the superradiant phase 116, 113601 (2016), arXiv:1508.05849. transition for molecules in a quantized radiation field: [28] S. De Liberato, D. Gerace, I. Carusotto, and C. Ciuti, the dicke maser model,” Annals of Physics 76, 360–404 “Extracavity quantum vacuum radiation from a single (1973). qubit,” Phys. Rev. A 80, 053810 (2009), arXiv:0906.2706. [11] Y. K. Wang and F. T. Hioe, “Phase transition in the [29] S. De Liberato, “Light-Matter Decoupling in the Deep dicke model of superradiance,” Phys. Rev. A 7, 831–836 Strong Coupling Regime: The Breakdown of the Pur- (1973). cell Effect,” Phys. Rev. Lett. 112, 016401 (2014), [12] K. Rza˙zewski, K. Wódkiewicz, and W. Zakowicz,˙ “Phase arXiv:1308.2812. transitions, two-level atoms, and the A2 term,” Phys. [30] L. Garziano, R. Stassi, V. Macr`ı, A. F. Kockum, Rev. Lett. 35, 432–434 (1975). S. Savasta, and F. Nori, “Multiphoton quantum Rabi os- [13] Neill Lambert, Clive Emary, and Tobias Brandes, “En- cillations in ultrastrong cavity QED,” Phys. Rev. A 92, tanglement and the phase transition in single-mode su- 063830 (2015), arXiv:1509.06102. perradiance,” Phys. Rev. Lett. 92, 073602 (2004). [31] L. Garziano, V. Macr`ı,R. Stassi, O. Di Stefano, F. Nori, [14] J. Keeling, “Coulomb interactions, gauge invariance, and and S. Savasta, “One Photon Can Simultaneously Ex- phase transitions of the Dicke model,” J. Phys.: Condens. cite Two or More Atoms,” Phys. Rev. Lett. 117, 043601 Matter 19, 295213 (2007). (2016), arXiv:1601.00886. 10

[32] L. Garziano, A. Ridolfo, S. De Liberato, and 439–460 (1983). S. Savasta, “Cavity QED in the Ultrastrong Coupling [46] W. E. Lamb, “Fine structure of the hydrogen atom. iii,” Regime: Photon Bunching from the Emission of Indi- Phys. Rev. 85, 259–276 (1952). vidual Dressed Qubits,” ACS Photonics 4, 2345 (2017), [47] W. E. Lamb, R. R. Schlicher, and M. O. Scully,“Matter- arXiv:1702.00532. field interaction in atomic physics and quantum optics,” [33] O. Di Stefano, R. Stassi, L. Garziano, A. F. Kockum, Phys. Rev. A 36, 2763–2772 (1987). S. Savasta, and F. Nori, “Feynman-diagrams approach [48] A. F. Starace, “Length and velocity formulas in approx- to the quantum Rabi model for ultrastrong cavity QED: imate oscillator-strength calculations,” Phys. Rev. A 3, stimulated emission and reabsorption of virtual particles 1242–1245 (1971). dressing a physical excitation,” New J. Phys. 19, 053010 [49] R. Girlanda, A. Quattropani, and P. Schwendimann, (2017), arXiv:1603.04984. “Two-photon transitions to exciton states in semiconduc- [34] A. F. Kockum, A. Miranowicz, V. Macr`ı, S. Savasta, tors. application to CuCl,” Phys. Rev. B 24, 2009–2017 and F. Nori, “Deterministic quantum nonlinear optics (1981). with single atoms and virtual photons,” Phys. Rev. A [50] S. Ismail-Beigi, E. K. Chang, and S. G. Louie, “Coupling 95, 063849 (2017), arXiv:1701.05038. of Nonlocal Potentials to Electromagnetic Fields,” Phys. [35] S. Felicetti, D. Z. Rossatto, E. Rico, E. Solano, and Rev. Lett. 87, 087402 (2001), arXiv:cond-mat/0101383. P. Forn-D´ıaz,“Two-photon quantum Rabi model with su- [51] F. Bassani, J. J. Forney, and A. Quattropani, “Choice perconducting circuits,” Phys. Rev. A 97, 013851 (2018), of Gauge in Two-Photon Transitions: 1s-2s Transition arXiv:1711.10032. in Atomic Hydrogen,” Phys. Rev. Lett. 39, 1070–1073 [36] P. Nataf and C. Ciuti, “Vacuum Degeneracy of a Circuit (1977). QED System in the Ultrastrong Coupling Regime,” Phys. [52] R. H. Dicke, “Coherence in spontaneous radiation pro- Rev. Lett. 104, 023601 (2010), arXiv:0909.3505. cesses,” Phys. Rev. 93, 99–110 (1954). [37] P. Nataf and C. Ciuti, “Protected Quantum Computa- [53] M. Combescot and O. Betbeder-Matibet, “The exciton- tion with Multiple Resonators in Ultrastrong Coupling polariton: Microscopic derivation of the phenomenologi- Circuit QED,” Phys. Rev. Lett. 107, 190402 (2011), cal theory,” Solid State Commun. 80, 1011–1015 (1991). arXiv:1106.1159. [54] S. Savasta and R. Girlanda, “The particle-photon inter- [38] G. Romero, D. Ballester, Y. M. Wang, V. Scarani, and action in systems descrided by model Hamiltonians in E. Solano, “Ultrafast quantum gates in circuit QED,” second quantization,” Solid State Commun. 96, 517–522 Phys. Rev. Lett. 108, 120501 (2012), arXiv:1110.0223. (1995). [39] T. H. Kyaw, S. Felicetti, G. Romero, E. Solano, and L.- [55] S. Savasta and R. Girlanda, “Quantum description of the C. Kwek, “Scalable quantum memory in the ultrastrong input and output electromagnetic fields in a polarizable coupling regime,” Sci. Rep. 5, 8621 (2015). confined system,” Phys. Rev. A 53, 2716–2726 (1996). [40] Y. Wang, J. Zhang, C. Wu, J. Q. You, and G. Romero, [56] W. H. Louisell, Quantum statistical properties of radia- “Holonomic quantum computation in the ultrastrong- tion, Vol. 7 (Wiley New York, 1973). coupling regime of circuit QED,”Phys. Rev. A 94, 012328 [57] M. Malekakhlagh, A. Petrescu, and H. E. Tureci,¨ (2016), arXiv:1605.05449. “Cutoff-Free Circuit Quantum Electrodynamics,” Phys. [41] R. Stassi, V. Macr`ı, A. F. Kockum, O. Di Stefano, A. Mi- Rev. Lett. 119, 073601 (2017), arXiv:1701.07935. ranowicz, S. Savasta, and F. Nori, “Quantum nonlinear [58] M. F. Gely, A. Parra-Rodriguez, D. Bothner, Ya. M. optics without photons,”Phys. Rev. A 96, 023818 (2017), Blanter, S. J. Bosman, E. Solano, and G. A. Steele,“Con- arXiv:1702.00660. vergence of the multimode quantum Rabi model of cir- [42] A. F. Kockum, V. Macr`ı,L. Garziano, S. Savasta, and cuit quantum electrodynamics,” Phys. Rev. B 95, 245115 F. Nori, “Frequency conversion in ultrastrong cavity (2017), arXiv:1701.05095. QED,” Sci. Rep. 7, 5313 (2017), arXiv:1701.07973. [59] C. S. Muñoz, F. Nori, and S. De Liberato, “Resolution of [43] F. Armata, G. Calajo, T. Jaako, M. S. Kim, and P. Rabl, superluminal signalling in non-perturbative cavity quan- “Harvesting multiqubit entanglement from ultrastrong tum electrodynamics,” Nat. Commun. 9, 1924 (2018). interactions in circuit quantum electrodynamics,” Phys. [60] V. E. Manucharyan, J. Koch, L. I. Glazman, and Rev. Lett. 119, 183602 (2017), arXiv:1707.08969. M. H. Devoret, “Fluxonium: Single Cooper-pair cir- [44] R. Stassi and F. Nori, “Long-lasting quantum memories: cuit free of charge offsets,” Science 326, 113–116 (2009), Extending the coherence time of superconducting arti- arXiv:0906.0831. ficial atoms in the ultrastrong-coupling regime,” Phys. [61] N. Shammah, N. Lambert, F. Nori, and S. De Liberato, Rev. A 97, 033823 (2018), arXiv:1703.08951. “Superradiance with local phase-breaking effects,” Phys. [45] M. Babiker and R. Loudon, “Derivation of the Power- Rev. A 96, 023863 (2017), arXiv:1704.07066. Zienau-Woolley Hamiltonian in quantum electrodynam- [62] E. Cortese, L. Garziano, and S. De Liberato, “Polariton ics by gauge transformation,” Proc. R. Soc. Lond. A 385, spectrum of the Dicke-Ising model,” Phys. Rev. A 96, 053861 (2017), arXiv:1707.03681.