Arxiv:1911.08427V2 [Quant-Ph] 5 Feb 2020 Tense Discussions [15, 40, 62, 74–82]

Relevance of the quadratic diamagnetic and self-polarization terms in cavity quantum electrodynamics

Christian Schäfer,1, ∗ Michael Ruggenthaler,1, † Vasil Rokaj,1, ‡ and Angel Rubio1, 2, § 1Max Planck Institute for the Structure and Dynamics of Matter and Center for Free-Electron Laser Science & Department of Physics, Luruper Chaussee 149, 22761 Hamburg, Germany 2Nano-Bio Spectroscopy Group, Dpto. Fisica de Materiales, Universidad del Pa´ıs Vasco, 20018 San Sebastián,Spain (Dated: February 6, 2020) Experiments at the interface of quantum-optics and chemistry have revealed that strong coupling between light and matter can substantially modify chemical and physical properties of molecules and solids. While the theoretical description of such situations is usually based on non-relativistic quantum electrodynamics, which contains quadratic light-matter coupling terms, it is common- place to disregard these terms and restrict to purely bilinear couplings. In this work we clarify the physical origin and the substantial impact of the most common quadratic terms, the diamagnetic and self-polarization terms, and highlight why neglecting them can lead to rather unphysical results. Specifically we demonstrate its relevance by showing that neglecting it leads to the loss of gauge invariance, basis-set dependence, disintegration (loss of bound states) of any system in the basis set-limit, unphysical radiation of the ground state and an artificial dependence on the static dipole. Besides providing important guidance for modeling strongly coupled light-matter systems, the presented results do also indicate under which conditions those effects might become accessible.

Driven by substantial experimental progress in the field gard these quadratic terms and at the same time indi- of cavity-modified chemistry [1–11], theoretical methods cates under which conditions substantial influence can at the border between quantum-chemical ab initio meth- be expected [12], accessible with ab initio techniques ods and optics have become the focus of many recent such as quantum-electrodynamic density-functional the- investigations [12–58]. The high complexity of a molec- ory (QEDFT) [12, 16, 22, 25, 83]. Before we do so, let ular system, which can undergo, e.g., chemical reactions us briefly outline the theory we consider and collect a or quantum phase-transitions, coupled strongly to pho- set of fundamental conditions we deem important for a tons makes the use of some sort of approximation strat- reasonable theoretical description. egy necessary. A common approach is to use approxi- Any theory we employ to model coupled light-matter mation strategies designed for atomic two-level-like sys- systems should obey certain fundamental constraints. tems in high-quality optical cavities [59–61] and to ap- Which ones these are often depends on the specific sit- ply them to the quite different situation of molecular uation we consider. For instance, in the case of high- systems. However, under the generalized conditions of energy physics an adherence to special relativity (physi- cavity-modified chemistry usually disregarded contribu- cal laws should be Lorentz invariant) is paramount and tions in the theoretical description, e.g., quadratic cou- hence the use of Dirac’s equation becomes necessary to pling terms between light and matter, can become impor- capture the behavior of electrons. If we further want tant [15, 62, 63] and might even dominate the physical to ensure that all interactions among the electrons are properties [13–16, 64, 65]. While the existence of these local and our theory should stay invariant under local quadratic terms is well-known [66–73] their origin, in- phase transformations we find the Maxwell field coupled terpretation and consequences are less clear, and when to Dirac’s momentum operator in a linear (minimal) fash- to include them has become the subject of recent in- ion. However, the resulting theory - which, if quantized, arXiv:1911.08427v2 [quant-ph] 5 Feb 2020 tense discussions [15, 40, 62, 74–82]. In this work we is called quantum electrodynamics (QED) and perfectly will elucidate these terms for the most relevant setting describes high-energy scattering events - has many sub- of cavity-modified chemistry, i.e., in Coulomb gauge and tle issues [84]. For low-energy physics a simplified ver- in the long-wavelength limit, clarify their origins, phys- sion, where instead of the relativistically invariant mo- ical interpretations and consequences as well as show mentum the non-relativistic momentum is employed, has under which conditions and for which observables they been shown to be able to resolve many of these issues [73]. become relevant. This will also highlight a domain The resulting theory of non-relativistic QED (also some- of applicability of common approximations that disre- times called molecular QED [69, 72, 85, 86]) is ideally suited to describe atoms, molecules or solids interacting with the quantized light field [87–89]. The coupling be- ∗ Electronic address: [email protected] tween light and matter is, however, only defined up to † Electronic address: [email protected] a phase and we need to make a specific choice for this ‡ Electronic address: [email protected] phase, i.e., we need to fix a gauge. Changing the gauge § Electronic address: [email protected] or performing a local unitary transformation should not 2 modify physical observables but merely affect their repre- Maxwell equations [85] sentation in terms of canonical coordinates. While gauge E(r, t) = ∂tB(r, t) independence is respected by non-relativistic QED, this ∇ × − constraint is specifically challenging for dimensionally- 1 2 B(r, t) = ∂tE(r, t) + µ0c j(r, t) . reduced, simplified models [15, 77, 80, 82]. Beside gauge ∇ × c2 independence, non-relativistic QED guarantees a set of This representation of light-matter coupling is by no further intuitive and essential conditions. For instance, means unique and many different formulations, such the physical observables are independent of the chosen as the Riemann-Silberstein [90, 91] or the macroscopic coordinate system (or in more quantum-chemical terms, Maxwell’s equations, have been developed over the years. where a specific spatial basis is just one of many basis-set To arrive at the latter, let us first rewrite j = jb + jf , choices, basis-set independent) and it also guarantees the where jb is a bound and jf a free current, and define a stability of matter, i.e., atoms and molecules are stable if bound charge current coupled to the vacuum of the electromagnetic field [73]. b j (r, t) = M(r, t) + ∂tP(r, t), A direct consequence of this fundamental condition is ∇ × that the combined ground state of light and matter has with M the magnetization and P the polarization of the a zero transversal electric-field expectation value. If this matter system. If we then define the displacement field 1 would not be the case the system could emit photons and D = 0E + P and the magnetization field H = B M µ0 − lower its energy. To summarize, a few basic constraints we end up with we want a theory of light-matter interactions to adhere f to are: All physical observables should be independent of ∂tD(r, t) + H(r, t) = j (r, t) − ∇ × the gauge choice and of the choice of coordinate system which takes the back-reaction of a given medium on the (for instance, it would be unphysical that the properties electromagnetic field into account by the constitutive of atoms and molecules would depend on the choice of equations. Clearly, the classical description of electro- the origin of the laboratory reference frame), the theory magnetic interaction can take different forms for which, should support stable ground states (else we could not without further simplifications, none is superior over the define equilibrium properties and identify specific atoms other. These forms deviate merely in their choice of and molecules) and the coupled light-matter ground state canonical variables, the very same variables that will be should have a zero transversal electric field (else the sys- quantized to reach QED. The electromagnetic field en- tem would radiate and cascade into lower-energy states). ergy In the following we will introduce non-relativistic QED Z and some of its unitarily equivalent realizations, highlight 0 3 2 2 2 em(t) = d r E (r, t) + c B (r, t) (1) the physical implications of the associated transforma- H 2 tions and further approximations that lead to the non- is of quadratic form and therefore substituting relativistic QED in the long-wavelength limit in Sec. I, D = 0E + P into Eq. (1) will naturally lead to and illustrate that the aforementioned fundamental phys- quadratic self-polarization P2 and self-magnetization ical conditions will not be retained when disregarding M2 terms. quadratic components in Sec. II. Finally we discuss implications and perspectives in Sec. III. We provide fur- While many equivalent ways of formulating the ther details in the Appendix. In Appendix A, the ba- Maxwell’s equations exist, there will be accordingly also sic approximations leading to non-relativistic QED are several (unitarily equivalent) forms of the resulting non- presented, App. B and C provide additional details that relativistic QED Hamiltonian. Let us in the follow- complement our discussion regarding the Power-Zienau- ing see how this equivalence in QED manifests. A rel- Wooley transformation and transversal basis functions, ativistic quantization procedure with subsequent non- and in App. D we discuss some implications that go hand relativistic limit, as illustrated in App. A, is indeed in hand with approximating operators. Our discussions equivalent to introducing the covariant derivative for will be presented first in a field-theoretical convention the electronic system and then quantizing the resulting µ where the four vector potential A is given in Volts and gauge field [12, 25, 73]. This minimal-coupling procedure µ the four charge-current density j is given in Coulomb makes the invariance under local phase transformations per meter squared per second and later in atomic units. Ψ0 = eiθ(r)Ψ explicit and with the Coulomb gauge con- µ By multiplying A by 1/c we find the standard conven- dition Aˆ (r) = 0 fixes the local phase θ(r) uniquely. tion in terms of Volt second per meter. The momentum∇ · of each particle is shifted according to q ˆ i~ ( i~ c A(r)), where q is the charge of the particle− ∇ → and− the∇ −quantized vector potential is I. FROM MICROSCOPIC TO MACROSCOPIC s 2 MAXWELL’S EQUATIONS ˆ ~c X 1 h † ∗ i A(r) = aˆn,λSn,λ(r) +a ˆn,λSn,λ(r) 0 √2ω n,λ n Classical electrodynamics is at the heart of QED. ikn·r √ 3 Consider for instance the inhomogeneous microscopic Sn,λ(r) = n,λe / L . (2) 3

Here we defined the transversal polarization vectors n,λ tion for mode and polarization (n, λ) [92], and the creation Ne Z 1 and annihilation operators can be expressed in terms of ˆ X 3 q † P(r) = rj δ (r srj)ds displacement coordinates qn,λ = ~ (â +a ˆn,λ) and − 0 − 2ωn n,λ j=1 q their conjugate momenta i∂ = i ~ωn (â† aˆ ) Nn Z 1 qn,λ 2 n,λ n,λ X 3 − − + ZjRj δ (r sRj)ds of harmonic oscillators with the allowed frequencies ωn = − j=1 0 c kn . | | The non-relativistic minimally coupled Hamiltonian where the jth nucleus has the effective positive charge Zj. (including Ne electrons and Nn nuclei) in Coulomb gauge This implies that all physical charges contribute to the 0 PNe+Nn f then reads with ˆj (r) = cnˆ(r) = c qiδ(r ri) bound current such that j = 0, D = ε0 E+ P = 0 i=1 − ∇· ∇· ∇· and therefore D = D⊥. However, since the vector- potential operator is purely transversal it also only cou- Ne+Nn ˆ X 1 qi ˆ 2 ˆ ⊥ ˆ ples to the transversal part of the polarization operator, H = i~ i A(ri) + Hem + Hint,k 2m − ∇ − c which can be expressed in terms of the transversal delta- i=1 i distribution [92] or we use the mode expansion of the ⊥ 1 X 2 2 2 Hˆ = ( i∂q ) + ω q (3) vector potential directly. To do so we first, for notational em 2 − n,λ n n,λ n,λ simplicity, introduce the abbreviation α (n, λ), then ≡ Ne+Nn use that the vector-valued functions Sn,λ(r) in (2) form ˆ 1 X qiqj Hint,k = . a basis for the transversal square-integrable vector fields 8πε0 ri rj (see App. C), and find with Sα(r) = α Sα(r) i6=j | − | · Ne Z 1 Each charged particle then evolves under the influence X X ∗ Pˆ ⊥(r) = α (rj α) S (r)Sα(srj)ds (4) q ˆ 2 − · α of the kinetic-energy operator ( i~ c A(r)) and j=1 α 0 at the same time experiences the− instantaneous∇ − longi- N tudinal field (Coulomb potential Hˆ ) created by all Xn X Z 1 int,k + Z (R ) S∗ (r)S (sR )ds. the other charged particles. The nonrelativistic limit j α j α α α j α · 0 of the minimal coupling procedure leads therefore nat- j=1 urally to the appearance of a quadratic term (see also The resulting Hamiltonian [29, 69, 72] has the advantage App. A). This quadratic term provides the diamagnetic that it can be conveniently expanded in multipoles of the shift [93] of the bare modes and introduces a lowest al- interaction. We note, however, that the validity of such lowed frequency [94] which then removes the infrared di- an expansion depends critically on whether it is consid- vergence [73]. It is therefore not a drawback to have ered as a perturbation of the wavefunction or affecting this term [74]. In contrast, it affects for instance op- the operator itself and subsequently its self-consistent tical spectroscopy [95, 96], is responsible for diamag- solution (see App. D). The mode-expansion provides a netism [97] and hence implies very important physical consistent regularization such that terms like Pˆ (r)2 are processes, such as the famous Meissner effect. Recent ⊥ well-defined, a necessity when multiplying delta distri- theoretical [14, 94, 98] and experimental [76] studies fo- butions (see App. C). This avoids the usual auxiliary as- cused on the ultra-strongly coupled light-matter dynam- sumption that some of these terms, which contribute to ics as well as the prediction of enhanced electron-phonon the polarization self-energy, are only taken into account coupling [64, 65, 99] highlight the non-negligible influence perturbatively [79]. It furthermore highlights how the of the collective diamagnetic shift. condition of transversality of the Maxwell field also af- Let us next, for convenience assume linear polarization fects matter-only operators like the polarization. So far, ∗ n,λ = n,λ and to connect to the more common formula- the only restriction we employed was that we considered tion, switch to atomic units, such that 0 = 1/(4π), c = non-relativistic particles. This simplification is, however, 1/α0 and the elementary charge e = 1. There is a usually not yet enough to allow for practical calculations. well-known procedure to connect to a Hamiltonian corre- In the following we do not want to consider this more sponding to the macroscopic Maxwell’s equation by em- general case but assume only dipole interactions. This ploying the unitary Power-Zienau-Woolley (PZW) trans- approximation is very accurate provided the dominating formation (see also App. B) modes of the photon field have wavelengths that are large compared to the extend of the matter subsystem. In the multipole form of the non-relativistic QED Hamiltonian Z this means that we discard the integration over s in our ˆ 3 ˆ ˆ U = exp iα0 d rP(r) A(r) transformation and the polarization operator [69]. The − · Hamiltonian we then find is the same as the one that we get if we approximate Aˆ (r) Aˆ (rMatter) for the bilin- for all charged particles contributing to the polariza- ear and quadratic coupling terms.' This does not restrict 4 the form of the cavity modes itself but merely its spatial energy of the electrons (an extended discussion can be extension in relation to the matter subsystem. In prac- found in App. C). The operator Eq. (6) contains the bi- tice where, e.g., an ensemble of molecules interacts with linear matter-photon coupling and the quadratic dipole 2 the cavity mode, this simplification can become question- 1 R 1 PMp 2 self-energy term drPˆ (r) (λα R) . 2ε0 ⊥ 2 α=1 able. Such an ensemble might extend over macroscopic The fundamental light-matter→ coupling to mode· α is scales such that individual molecules will experience dif- then denoted by ferent couplings. In the following we take rMatter = 0 for simplicity such that Sα(0) is real and we can λα = √4πSα(0)α, (7) straightforwardly perform the unitary PZW, also referred to as length gauge, transformation Hˆ = UˆHˆ Uˆ † with which depends on the form of the mode functions and the √ P PNe PNn −i ( 4πSα(0)α· − rj + Zj Rj )qα chosen reference point for our matter subsystem [25, 100]. Uˆ = e α j=1 j=1 . We accompany this by a canonical transformation which This can lead to an increase of the fundamental coupling to a specific mode and is an inherent feature of the phys- swaps the photon coordinates and momenta i∂qα −1 − → ical set-up, e.g., the form and nature of the cavity. In the ωαpα, qα iωα ∂pα while preserving the commutation− relations→ [22].− The non-relativistic QED Hamilto- following we will treat λα and the corresponding frequen- nian then reads cies ωα as parameters that we can adopt freely to match different physical situations, motivated by the recent ex-

Hˆ = Hˆn + Hê + Hˆne + Hˆp + Hêp + Hˆnp, (5) perimental progress to sub-wavelength effective cavity volumes [101, 102]. This also highlights that the self- where the nuclear Hamiltonian is energy term depending on λα is influenced directly by the properties of the cavity, i.e, obtains increasing relevance Nn Nn X 1 1 X ZiZj √ 3 Hˆ = Tˆ + Wˆ = 2 + with decreasing effective mode volume Sα(0) = 1/ L n n nn Rj − 2Mj ∇ 2 Ri Rj and increasing number of participating modes M . j=1 i6=j p | − | Importantly, since the PZW gauge is equivalent to the with the bare nuclear masses Mj. The electronic Hamil- Maxwell’s equation in matter as introduced earlier, we tonian is now work in terms of the, purely transversal, displacement field [62, 72, 79] Ne Ne 1 X 2 1 X 1 Hˆ = Tˆ + Wˆ = + ωα e e ee rj ˆ X −2me ∇ 2 ri rj D⊥ = λαpα j=1 i6=j 4π | − | α with the bare electron mass me and the nuclear-electron and the transversal polarization operator interaction is given by X 1 Nn Ne Pˆ ⊥ = λα (λα R) . X X Zj 4π · Hˆne = . α − r R j=1 i=1 i j | − | By construction, the electric-field operator in PZW Further, the photonic contribution for Mp modes is then gauge, no longer representing the conjugate momentum, given by becomes Mp " 2# Eˆ = 4π Dˆ Pˆ . (8) 1 X λα ⊥ ⊥ ⊥ Hˆ + Hˆ + Hˆ = ∂2 + ω2 p R , − p ep np pα α α 2 − − ωα · α=1 The combination of PZW and canonical-momentum (6) transformation changed our canonical operators Bˆ i∂ , Dˆ p and consequentially the representation∝ PNe pα ⊥ α which incorporates the total dipole R = rj + − ∝ − j=1 of our original creation and annihilation operators to PNn j=1 ZjRj of electrons and nuclei. The resulting bilinear coupling or R itself might be occasionally defined 1 aˆα = ( i∂pα iωαpα + iλα R) , with the opposite sign as a consequence of the inversion √2ωα − − · (9) symmetry of Eq. (5). Even if we break the inversion † 1 aˆα = ( i∂pα + iωαpα iλα R) . symmetry of the matter Hamiltonian as e.g. in Sec. II C, √2ωα − − · the photonic symmetry pα pα will retain the trivial connection between both↔ Hamiltonians − and their re- We might yet again define a new harmonic oscilla- spective observables. Here Mp is a finite but arbitrar- tor algebra solely in terms of our new canonical op- p ωα † ily large amount of photon modes which are the most erators i∂p = i (â aˆα,P ZW ), pα = − α 2 α,P ZW − relevant modes but in principle run from the fundamen- q 1 † (â +âα,P ZW ) to reach a potentially more fa- tal mode of our arbitrarily large but for simplicity fi- 2ωα α,P ZW † nite quantization volume up to a maximum sensible fre- miliar representations in terms of differenta ˆPZW , aˆPZW . quency, for example, an ultra-violet cut-off at rest-mass However, we have to consider then that the expression of 5 physical observables in terms of creation and annihila- to be employed to arrive at the above Hamiltonian which tion operators is not invariant under the PZW transfor- represents the usual starting point of most considera- mation, i.e.,a ˆPZW =a ˆ. Special care has to be taken tions in cavity QED and cavity-modified chemistry. Each on how we interpret6 observables and design possible ap- approximation restricts its applicability but the basic proximations as otherwise unphysical consequences arise physical constraints, i.e., gauge and coordinate-system as highlighted explicitly in Sec. II B. We also see that in (basis-set) independence, existence of a ground state, and accordance to the Maxwell’s equation in matter, by work- radiation-less eigenstates, are as of yet conserved. It is ing with Dˆ ⊥ we implicitly take into account the back- now subject of the following sections to emphasize that reaction (polarization) of matter on the electromagnetic 1 PMp 2 ignoring the transversal self-polarization 2 α=1(λα R) field. The physical field is found with the constitutive 2 2 · or diamagnetic q Aˆ (0) terms will inevitably break some relation of Eq. (8). Finally we note that the PZW trans- c2 of those fundamental constraints. This implies by no formation has removed the explicit diamagnetic contri- means that perturbative treatments that ignore these bution of the current and the physical current is now terms, either by restricting the Hilbert space of the mat- equivalent to the paramagnetic current. The diamag- ter subsystem or by perturbation theory on top of free netic term has, however, not vanished but is contained matter observables, might not provide accurate predic- in the introduced phase of the coupled light-matter wave tions [15, 77, 107]. It shows however that care has to be function [62]. taken when the quadratic terms are disregarded. Let us clearly state a warning at this point. Unitary equivalence or gauge invariance, which implies that we obtain the same predictions in Coulomb and PZW gauge, II. NECESSITY AND IMPLICATIONS OF is only fulfilled when the full Hilbert space (full basis- QUADRATIC COUPLINGS IN THE DIPOLE set) is considered. Any approximation in the molecular APPROXIMATION or photonic space will violate this equivalence and therefore result in deviating predictions [72, 77, 80, 82, 103]. Let us now consider concretely what happens if we We remain with three different strategies. We can ac- discard the quadratic terms and which further physical knowledge this failure and focus on reasonable domains in constraints we violate. The example will be a simple which predictions remain in reasonably close agreement, molecular system, a slightly asymmetric one-dimensional e.g. focus on resonant interactions [15, 72]. Alterna- Shin-Metiu model, coupled strongly to a single cavity tively, we adjust the PZW transformation to compensate mode as illustrated in Fig. 1. We subsume the rest of the reduced space accordingly [80], or consider as much of the space as possible which leads us into the realm of first-principles cavity QED. This break-down of gauge invariance can have very fundamental consequences, as the long-standing debate of the (non-)existence of a Dicke superradiant phase shows [63, 104, 105]. Disregarding ˆ 2 ˆ 2 quadratic contributions (A /P⊥) will consequentially merely allow to obtain perturbatively similar results. Finally, there is one subtle question left. If we consider many photon modes they give rise to the radiative losses, that is, they constitute the photon bath of the matter subsystem into which the excited states can dis- sipate their energy [50]. Vacuum fluctuations give rise to effects like spontaneous emission, that is, turning the discrete eigenstates of the closed system described by a Schödingerequation into resonances with finite line width [50]. Selecting furthermore only one or a very limited set of modes α will restrict retardation effects and can lead to unphysical superluminal transfer appearing in the (deep) ultra-strong coupling regime [106]. In this work we are, however, not interested in lifetimes but in equilibrium states of the coupled light-matter system. In Figure 1. Illustration of the Shin-Metiu model insight the this case we can instead of keeping many modes sub- cavity. The two outer nuclei with the distance L and charges sume the vacuum photon bath by renormalizing the bare Z± are fixed in position while central nucleus and electron masses me and Mj of the charged particles, i.e., we use can move freely within one dimension. the usual physical masses such as me = 1 in atomic units, and only keep a few important modes that are enhanced with respect to the free-space vacuum. the photon bath in our description approximately into We have seen that already several approximations have the physical mass of electron and nuclei. The Shin-Metiu 6 model features one nucleus moving in between two pinned basis functions (grid-points), keeping other parameters nuclei with in total one electron and is a paradigmatic fixed, and present first the light-matter correlated ener- model for non-adiabatic electron-nucleus coupling that gies as well as the total dipole in Fig. 2. We find that gives rise to many interesting chemical processes. The Hamiltonian of this model with moving nuclear charge Nuclear box size (A)˚ ω 1 2 3 4 5 6 7 8 Z = +1 and removed vacuum shift 2 is given by 0 0.0 9 1 1 − 2 2 0 Hˆ = ∂ ∂ + V (x X) (10) 2.5 10 ˚ A) 2M X 2m x enx − − − − e − 10 5.0 − 2 L L L − 62.5 65.0 67.5 R + Z−Vnn(X ) + Z+Vnn(X + ) + Z−Ven(x ) − 2 2 − 2 7.5 2 − " 2 # R L 1 λ 2 2 2 10.0 1 + Z+Ven(x + ) + ∂ + ω p R ω , − no R 10 2 2 − p − ω − − GS total dipole R ( Correlated eigenvalues (eV) 12.5 2 − no R with the total dipole R = x + ZX and electron-nuclear 10 20 30 40 50 60 70 80 and nuclear-nuclear potentials− Electronic box size (A)˚

L Z Figure 2. First four light-matter correlated eigenvalues with Vnn(X ) = ± 2 X L (blue, solid) and without (blue, dashed) self-polarization con- | ± 2 | L tribution as well as total dipole R = −x + ZX with (red, L erf ( x 2 )/Rc solid) and without (red, dashed) self-polarization. While ob- Ven(x ) = | ± | 2 L servables start to be converged with a box size around 50 A,˚ ± x 2 | ± | without self-polarization the system starts to disintegrate al- Z erf [( x X )/Rf ] Ven (x X) = | − | ready for slightly larger box-values as highlighted by the inset. x − x X | − | where erf represents the error-function. For the follow- by increasing the space of allowed wave functions, i.e., ing calculations we consider parameters Z+ = 1,Z− = approaching the basis-set limit, the minimal-energy solu- 1.05, M = 1836 me,L = 18.8973,Rc = 2.8346, Rf = tion without the self-polarization term does not converge 3.7795 with an electronic and nuclear spacing of ∆x = and minimizes the energy (dashed blue line) by increasing 0.4, ∆X = 0.04 between the equidistant grid-points and 40 photon number-states. Furthermore, we couple elec- the total dipole (dashed red line). To put it differently, tron and nucleus to a single cavity-mode with the fre- the system is torn apart and electrons occupy one side of quency ω = 0.00231, resonant to the vibrational excita- the simulation box, while nuclei the other. tion. We achieve ultra-strong vibrational coupling with On the other hand, with the self-polarization term we g/ω = λ/√2ω = 0.40748 in atomic units, where by no see how we approach quickly the basis-set limit such that means the following results qualitatively depend on cou- we have a basis-set independent result (red and blue solid pling or frequency. The strength of the light-matter in- line, respectively). The complete disintegration of the teraction solely determines how quickly given effects will system without the self-polarization happens at a criti- be visible and the selected values are close to those of cal box size, which is just marginally larger than a box previous publications in this field of research. It is im- leading to converged results when the self-polarization portant to realize that the associated wavelength to this is included. With further increasing box size, the en- frequency is 1.9724 105 A˚ = 19.724 µm and thus differs ergies resemble more and more those of an inverted by about four orders· of magnitude from the computa- shifted harmonic oscillator, which only supports scatter- ˚ ing states [108]. This illustrates that by a small ( 20%) tional box ( 60 A) that is considered for the matter ∼ system. Our≈ example is thus safely within the validity variation of the simulation box we lose the physical char- of the long wavelength approximation when considering acter of the model and enter a non-physical regime. How e.g. one-dimensional cavities. drastic this effect will appear is given by the ratio of 1 2 quadratically divergent potential energy 2 (λ R) and the energy that is demanded to ionize the− system· from a A. No bound eigenstates without self-polarization given eigenstate εi (assuming non-interacting electrons for simplicity) such− that a pure bilinear treatment would 2 be perturbatively only reasonable for (λ R) /2εi 1. Let us start to investigate the most fundamental − · problem of discarding the quadratic term in the non- In this sense the common ratio between coupling and excitation energy g/ω, assuming resonance ε2 ε1 = ω, can perturbative regime: The instability of the coupled sys- − tem, i.e., that electrons and nuclei fly apart if coupled with slight adjustments to even in the slightest to the photon field unless we restrict 2 the Hilbert space [15, 62]. To illustrate this we increase λ Extension criterion = 2 (in atomic units) (11) the simulation box stepwise by increasing the number of 4εi 7

0.6 0.6 be seen as an estimate how quick the given eigenstate 0.00 2 ”i” will become unstable without self-polarization com- 0.4 0.05 0.4 ponent. This extension criterion (11) can be motivated − 2 50 60 70 80 90 100 110 by the single-particle Schrödinger equation [ −~ ∆ + 0.2 0.2 2me v(r)]Ψi(r) = εiΨi(r) in the limit r , v(r) 0, ∆ 2 → ∞ → → 0.0 0.0 ∂r such that the long-range exponential decay of the −√2me(−εi)r/ −r/ai 0.2 0.2 state Ψi e ~ = e is defined by its ∝ p − − characteristic extension ai = ~/ 2me( εi) (e.g. for the Correlated eigenvalues (eV) 0.4 0.4 hydrogen atom the Bohr radius). The− simulation box − − Correlated eigenvalues (eV) no R has to be large enough to at least fit the state ”i” to 0.6 0.6 an amount that we resolve an exponential decay e−1 − 50 100 150 200 250 − ˚ (which is far from numerical convergence in fact).∼ This Electronic box size (A) provides an estimate of the extension of the eigenstate Figure 3. First four light-matter correlated eigenvalues with of interest and its associated self-polarization energy 2 2 2 (blue, solid) and without (red, dashed) self-polarization con- (λ R) (λai) = λ /2εi. While this might provide · ≈ − tribution for the Rydberg-type weakly bound hydrogen model an orientation for theoretical calculations when instabil- with grid spacing ∆x = 0.8 and 120 photon number-states. ities are to be expected without self-polarization, even Until the box reaches large extends ≈ 200 A˚ the ground state before the system is torn apart we see that the eigenval- without self-polarization is merely slightly deviating from the ues and the total dipole differ noticeably when increasing correct one. The excited states however, i.e., also the spec- the basis set. Also other observables are changing with- tra and all observables involving excited states, are relatively out the self-polarization term, e.g., the non-perturbative weakly bound and experience unphysical behavior even before Rabi splitting. The observables with self-polarization re- entering a converged regime. The inset magnifies the unphys- main completely size-independent once reaching a suffi- ical cross-over from physically bound into scattering states. cient basis-resolution. The disintegration effect is qualitatively independent of the frequency. Let us illustrate how weakly bound states are affected with the help of a second numerical example. We select a simple one-dimensional soft-Coulomb Hydrogen atom but screen the nuclear charge Z that binds the electron is typically weaker affecting the ground state, features with v(x) = Z/√x2 + 1 to Z = 1/20. We couple − the contrary tendency and their competition will deter- this system rather weakly g/ω = 0.006 with frequency mine the qualitative distribution of charges inside the ω = 0.01368 in resonance to its first excitation (when cavity [15, 109]. The resulting consequences can e.g. in- converged) and as before increase step-wise the simula- clude a reduced equilibrium bond length [13, 14] with tion box. Figure 3 illustrates that although the ground an earlier onset of static correlation [14] that could be state is merely perturbatively affected for up to 200 A,˚ steered on demand by controlling the polarization of the the excited states immediately turn into, for this case, field and therefore implies interesting opportunities for unphysical scattering states. As before adding the self- chemical considerations and electronic devices. polarization term will result in the expected spectrum, Let us briefly inspect how the same system behaves very much in contrast to the spectrum without the self- in the Coulomb gauge instead. Fig. 4 illustrates the polarization component. The extension criterion λ2/4ε2 i same correlated eigenvalues with increasing boxsize as leads for the ground state to 0.0011 and for the first ex- the previous Fig. 3. Coulomb and PZW gauge lead to cited state to 0.1664. While the ratio g/ω = 0.006 gives accurate agreement, a numerical difference in energy of the impression of rather weak coupling, the extension cri- less than 10−7 eV for a box size of more than 40 A,˚ terion provides a first indication that the excited states when both quadratic components are included. Omit- will be substantially affected by the self-polarization com- ting the diamagnetic term leads to a slight negative shift ponent. in the correlated eigenvalues and illustrates the rele- While the bilinear interaction reduces the ground state vance of the diamagnetic component, i.e., shifting the energy with increasing coupling, the self-polarization photonic excitations (see also App. A). In the Coulomb contribution counteracts by an increase in energy and gauge, couplings between higher excited states rescale dominates for typical couplings the bilinear contribution, lower matrix elements, demanding a well converged set of i.e., even the sign, thus the qualitative behavior, of ener- electronic eigenstates, and therefore the bilinear compo- getic shift within the cavity can alter depending on the nent accounts for the major effect of the self-polarization presence/absence of the self-polarization [13, 15]. This [15, 77, 82]. We should recall however, that the dia- qualitative change is also represented in spatial observ- magnetic contribution scales with the amount of polar- ables, i.e., the self-polarization term favors a reduced po- izable material and it attains increasing importance the larizability and thus focuses charge density in domains smaller the frequency of the field. In this sense, per- where charge is already present [14–16]. The bilinear forming the same investigation with a ten times smaller coupling, which furthermore scales with the frequency, ω = 0.00137 = 0.0372 eV , the lowest excited states are 8

0.6 0.6 0.00 Nuclear box size (A)˚

2 1 2 3 4 5 6 7 8 0.05 0.4 − 0.4 0.0000 2 0.04 50 60 70 80 90 100 110 no R 0.2 0.2 0.0002 − 2 0.03 0.0 0.0 R 0.0004 − 0.2 0.2 0.02 − − 0.0006 − Correlated eigenvalues (eV) 0.4 0.4 0.01 Electric field E (a.u.) − − Correlated eigenvalues (eV) no A Displacement field D (a.u.) 0.0008 2 2 0.6 0.6 − R no R − 50 100 150 200 250 − 0.00 10 20 30 40 50 60 70 80 Electronic box size (A)˚ Electronic box size (A)˚ Figure 4. First four light-matter correlated eigenvalues in Figure 5. Displacement field with (blue, solid) and without Coulomb gauge with (blue, solid) and without (red, dashed) (blue, dashed) self-polarization contribution as well as the diamagnetic Aˆ (r)2 contribution for the Rydberg-type weakly electric field with (red, solid) and without (red, dashed) self- bound hydrogen model (same parameters as Fig. 3). Disre- polarization. While the electric field is independent of the garding the quadratic (diamagnetic) term omits this time the molecular convergence, therefore even in a restricted subspace diamagnetic shift that leads to accurate agreement between we do not radiate and have a well defined equilibrium solu- both gauges. The inset magnifies the same region as in Fig. 3. tion, the displacement field depends on the convergence of the Recall that the diamagnetic contribution attains increasing molecular system. impact the smaller the frequency and the more polarizable matter is available.

45 50 A˚ (blue and red solid lines). Only extending photon replica with an energetic spacing between ground ≈the box− slightly to 60 A,˚ the system without the self- and first excited state of 0.0372 eV with diamagnetic polarization desintegrates≈ (blue and red dashed lines). contribution and 0.0254 eV without, highlighting a sub- Recall here that this is well within the validity of the long- stantial deviation. wavelength approximation as the matter and photon-field As a side remark, although the validity of the dipole scales are separated by four orders of magnitude. By approximation for high frequencies is questionable, the construction the system with the self-polarization term quadratic self-polarization term guarantees that the high- always obeys the basic constraint of zero electric field, frequency photons essentially decouple from the matter while the system with only bilinear-coupling leads for subsystem. If only a purely linear-coupling is assumed, large extensions to an eigenstate with non-zero electric then the ultra-violet behavior is completely wrong as field. This cannot be a physical ground state. photons with arbitrarily high energies still interact with Realizing the connection between observable field and the matter subsystem [100]. canonical momentum, let us turn our attention to the number of photons in the ground state. The photon number operator (the electromagnetic field occupation) B. Radiating eigenstates without self-polarization is defined in Coulomb gauge as

X Let us look at another unphysical feature that appears Nˆ = aˆ† aˆ . when the self-polarization is neglected. In the case of α α α the simple Rabi or Dicke model, where the particle is assumed perfectly localized (assuming effectively classical In the PZW gauge these annihilation and creation op- particles), the polarization is zero and we can associate erators are given by Eq. (9), and as a consequence the the expectation values of the modes in PZW gauge with ˆ the electric field. However, if we consider an ab initio number operator N in this gauge is treatment, this is no-longer the case and we need to use the correct definition of the electric field of Eq. (8). In Mp " 2 # ˆ X 1 2 ωα λα 1 Fig. 5 we then show the displacement as well as the elec- N = ∂ + pα R . −2ω pα 2 − ω · − 2 tric field expectation values as we increase the number α=1 α α of basis functions. This is again the same as allowing the electrons and nuclei to extend over an ever increas- Originating from the change of conjugate momentum ing spatial region which is equivalent to exploring the from electric to displacement field, we see that the self- full Hilbert space. Also in these observables we see that polarization enters the definition of the photon number the system with the self-polarization term included leads operator when we work in the PZW gauge. Without sur- to simulation-box size independent results after roughly prise, this leads to different occupations as if we would 9 naively use such that also the total dipole operator R is shifted. Note that this also changes the origin of the cavity as the long- Mp wavelength approximation enforces that all molecules see X 1 ωα 1 Nˆ 0 = ∂2 + p2 , the same field (of the now also shifted reference point) −2ω pα 2 α − 2 α=1 α of the cavity. However, due to the zero-electric-field condition of a physical ground state we explicitly know the and we illustrate this difference in Fig. 6. The alleged 0 relation between the (shifted) dipole expectation value occupation N (blue) is higher than the physical occupa- R of the matter subsystem and the expectation values tion N (red) caused by the permanent dipole. Only for h i λα of the displacement fields as pα = R [15, 25, 62]. ωα If we then further re-expressh thei light-matter· h i coupling Nuclear box size (A)˚ 1 2 3 4 5 6 7 8 with fluctuation quantities ∆R = R R such that Eq. (6) becomes − h i no R2 no R2 2 Mp " # 1 10 2 10 1 X 2 2 λα ∂p + ωα ∆pα ∆R , 2 2 − α − ω · R α=1 α 101 we find that in equilibrium the shifts cancel and the only remaining contribution in the Hamiltonian is given by 100 the fluctuations around the mean values. Indeed, the

Naive mode occupation 2 0 R 10 Physical mode occupation equilibrium wave function in the new coordinate system 0 0 is just the original wave function translated in space and 10 20 30 40 50 60 70 80 the photon subsystem coherently shifted. Electronic box size (A)˚ As a consequence, the light-matter coupled system is Figure 6. Naive Nˆ 0 (blue) and physical photon occupations invariant under shifts of the origin in equilibrium, no Nˆ (red) in PZW gauge with (solid) and without (dashed) self- physical observable has a dependence on the permanent polarization contribution during the self-consistent solution. dipole. That the equilibrium properties of light-matter coupled systems do not depend on a possible permanent dipole, is merely a consequence of how particles couple to two-level models such as the Rabi model both definitions the transversal electromagnetic field: only currents can agree [15]. Comparing Fig. 5 and 6, it is instructive to interact with photons. A permanent dipole does only shift the photonic displacement field, which is not a phys- observe that displacement field D⊥ and naive mode occupation N 0 behave qualitatively very similar, obtaining ical observable, and the permanent dipoles of molecules relevant non-zero values only after a sufficiently large nu- will only contribute when the combined system is moved merical box size is reached. In contrast the electric field E out of equilibrium. remains system-size independent and the physical mode Only upon neglecting the self-polarization term can an occupation N adjusts merely quantitatively to the simu- unphysical dependence on the permanent dipole in equi- lation box. Not surprisingly, ignoring the self-interaction librium arise. To illustrate that even for small systems contributions in general leads to different results for dif- and shifts this can have a large influence, we consider ferent gauge choices. the Shin-Metiu model from before, however, we slightly charge the complete system by using Z = +1.05 e . We then perform a small shift x x + µ in the coordinate| | system, solve the corresponding→ Shin-Metiu model and C. Coordinate system and dipole dependence µ without self-polarization determine the electronic ground state density ne (x) and µ then translate back, i.e., ne (x µ), to compare with the original (unshifted) solution n−(x). 1 As expected, when The Hamiltonian of Eq. (5) (and its variants) guaran- e the self-polarization is included, we just recover the old tees that all physical observables in equilibrium are inde- density and nµ(x µ) n (x) 0. In contrast, in Fig. 7 pendent of the chosen coordinate system. This is obvious e e we show the differences− − without≡ the self-polarization and if we have a charge neutral system, where the Hamilto- find an ever increasing difference for larger µ (increas- nian of Eq. (5) is completely translationally invariant. ing permanent dipole). The behavior of the Shin-Metiu This constraint is physically very reasonable, because without a spatial dependence, i.e., the manifestation of the long-wavelength approximation, the electromagnetic field cannot break the translational symmetry of the bare 1 Note that in this case this is not a physical translation but merely molecular system. If the system is not charge neutral, a shift of origin for the coordinate system. However, within the long-wavelength approximation, physical (or active) and coordi- e.g., when we only consider electrons in an external bind- nate system (or passive) translations are synonymous and only ing potential, we do no longer have trivial translational when extending beyond this prototypical approximation observ- invariance. To see this, consider a shift of the origin of able differences between active and passive translations will ap- the coordinate system along the polarization of the field pear. 10

0.002 to the self-polarization contribution [85]. We can further approximately distinguish between situations where the wavefunctions of the diﬀerent constituents overlap 0.001 strongly and situations where there is no strong overlap. The previous situation, often referred to as intra- 0.000 molecular, demands to carefully consider Coulomb and self-polarization contributions simultaneously, where the 0.001 − Translation by 1.06 A˚ Coulomb contribution is dominating in most situations. Translation by 3.18 A˚ The latter situation of contact-free interactions, referred 0.002 − Translation by 5.29 A˚ to as inter-molecular, between matter sub-systems leads

Electronic density difference (a.u.) to a perturbative (dipolar approximation) cancellation 4 2 0 2 4 6 8 10 R − − of instantaneous interactions such that drPA PB 0 Electronic position x (A)˚ and we are left approximately with purely bilinear· and≈ retarded interactions between those separated matter sub- Figure 7. Electronic density difference between the trans- µ systems [112]. This situation, however, does not allow to lated (by a shift µ) ne (x) and original system ne(x), i.e., µ neglect longitudinal or transversal interactions when per- ne (x − µ) − ne(x), without the self-polarization term. If the self-polarization is included the difference is always zero, i.e., forming calculations locally for one of the sub-systems. the equilibrium-physics remains independent of the coordi- A consistent calculation considering, for example, molec- nate system and the permanent dipole. In the Shin-Metiu ular rearrangements during chemical reactions due to the model the moving nucleus was slightly charged by Z = +1.05 influence of cavity-mediated strong light-matter coupling and an electronic and nuclear box size of 59.27 and 5.93 A˚ was will thus demand also a consistent treatment of Coulomb chosen, i.e., before any scattering states appear. All other pa- and self-polarization contributions. rameters remain as before. If we now enter into the realm where instantaneous contributions to the inter-molecular interaction cancel, it is often instructive to assume that indeed the local, model without the self-polarization is clearly unphysi- i.e., sub-system, eigenstates are not affected by inter- cal, since observables should not depend on the coordi- molecular interactions and do not need to be updated nate system. Any approximate method tailored to per- during the process. In this case we can perform the form self-consistent calculations should respect the above pinned-dipole approximation which implies that each coordinate system independence by retaining the bal- subsystem is localized at a specific position and distin- ance between bilinear and quadratic contributions. Con- guishable. Starting from Eq. (5), we can then recover sider for example the performance of the non-variational the Dicke model by absorbing the self-polarization con- Krieger-Li-Iafrate approximation for ab initio quantum tribution perturbatively by renormalizing the mass of electrodynamical density-functional theory presented in the particles (similar to the perturbative treatment of Ref. [16] which is breaking this balance. the Lamb shift) such that the effective interaction re- Notice that quadratic components also necessarily ap- duces to the common bilinear coupling [72]. In the case pear in other situations, i.e., they are indeed a quite of the pinned-dipole approximation the bilinear coupling general feature of non-relativistic Hamiltonians. If for to the displacement field becomes equivalent to a cou- example nuclear vibrations are approximated as phonon pling to the electric fields since the local polarization in ˆ E⊥ = 4π(D⊥ P⊥) is zero by construction. To assume modes, the non-linear Debye-Waller-term k k0 H has − to be added to the bilinear interaction [110,∝ ∇ 111].∇ This that the quantum sub-systems are perfectly localized term originates from the quadratic elements in a Born- and distinguishable is in stark contrast to a quantum- Huang expansion [15] with the very same physical effects mechanical ab initio description of molecules. Thus ap- plying the Dicke model to deduce the influence of strong ˆ 2 ˆ 2 as the quadratic components A or P⊥, e.g., enforcing coupling on the local molecular states calls for a very translational invariance and renormalizing the excitation careful analysis of all the applied approximations and energies. their consequences. It furthermore permits physical features such as when charge-distributions start to overlap, as often the case in quantum chemical calculations, lead- D. Collectivity, the limit of the Dicke model and ing to a dependence of local observables on the surround- plasmonic systems ing (collective) ensemble [16]. The occurrence of quadratic, i.e., Debye-Waller, terms When considering a system of several molecules we in the electron-nuclei coupling highlights how general can separate their instantaneous interactions mediated quadratic components are in a non-relativistic theory. by longitudinal and transversal polarization fields (in More closely connected to our current situation is the R ˆ 2 R ˆ 2 ˆ 2 the PZW gauge) by drP = drPk + P⊥. Here coupling to modes of a plasmonic environment. In prin- the first term on the right-hand side corresponds to ciple, if we describe the plasmonic environment as part the Coulomb interaction, the second term corresponds of the full system [113, 114], the density oscillations of 11 the plasmonic environment are captured in an ab initio of-the-art models might provide insightful perturbative description by the Coulomb interaction and the induced results, the development of corrected models should ob- transversal photon field and hence Eq. (5) is directly ap- tain additional interest and ab initio calculations could plicable. Let us assume, however, that we are not in- prove beneficial to foster this effort. terested in a self-consistent calculation, can safely dis- R regard contact-terms drPA PB = 0, and rather care · about perturbative corrections. This consideration will III. SUMMARY lead to van-der-Waals (dipole-dipole) type interactions with different scalings in terms of their inter-molecular It is the very nature of physics that our descriptions distance R , independent of the choice of Coulomb or AB are necessarily approximate and that every theory PZW gauge [72]. The consideration of large distances has its limitations and drawbacks. And even if we subject to significant retardation are described by at- have seemingly very accurate theories like QED, we tractive Casimir-Polder interactions [115, 116] scaling as need to reduce their complexity by employing further (R−7 ). For smaller R retardation might be omit- AB AB approximations and assumptions to render them practi- Oted and we enter the realm of instantaneous attractive cal. For QED this was historically done by employing interactions captured by the London dispersion potential perturbation theory and/or restrictions to a minimal set (R−6 ). While those considerations are well tested and AB of dynamical variables. The resulting simplified versions Oallow for excellent perturbative results, they would not of QED are due to their clarity and elegance a very good allow a self-consistent calculation as these forces would starting point for further investigations, provide for good merely result in a collapse of the wave function onto reasons a common language for a variety of subjects a singular point due to its unbalanced attraction. As- and have provided tremendous insight over decades suming for instance a set of harmonic oscillators describ- of research. Nevertheless, we need to be aware under ing the plasmonic excitations coupled merely bilinear to which conditions these simplifications are valid and what the system of interest would introduce divergent forces their consequences are. With the recent experimental P λ (λ R) [15]. Coulomb potential, wavefunc- α α α advances in combining quantum-optical, chemical and tion∝ − overlap and· the Pauli principle give rise to repulsive material science aspects [121] and the subsequent components for small R , modeled for example by the AB merging of ab initio approaches with quantum-optical empiric (R−12) of the Lennard-Jones potential or the AB methods, it has become important to scrutinize these (e−RABO) of the Buckingham potential. It is therefore common assumptions [12]. theO higher-order components that are ensuring the stability of matter. A self-consistent treatment of molecules In this work we have elucidated and illustrated in a polaritonic cavity, which itself is modeled as, e.g., a the consequences of discarding quadratic terms that simple harmonic oscillator [40, 117], thus needs to include arise naturally in non-relativistic QED. Omitting them higher-order couplings to describe a stable and physical breaks gauge invariance, introduces a dependence on system. Self-consistent calculations would therefore de- the coordinate system (or basis set), leads to radiating mand extending the quasistatic approximation [118–120] ground states, introduces an artificial dependence on for plasmonic systems such that the plasmonic cluster the total dipole and in the basis-set limit leads to a responds to the coupled molecule. This is precisely the disintegration of the complete system. However many of physical origin of the quadratic terms in QED, they al- these effects can be mitigated if one works perturbatively low the photonic or vice-versa the matter system to re- or restricts the space. This is in accordance with many spond to the coupling by adjusting their excitation en- years of successful application of such approximations, ergies. For instance, the A2 part can be subsumed into but also highlights their limits of applicability. However, adjusted mode frequencies and further defines a mini- estimates of their applicability, such as the extension cri- mal frequency, i.e., cures the infrared divergence, while terion Eq. (11) discussed in Sec. II A, become nowadays the P2 term renormalizes the energies of the material, ⊥ relevant for practical calculations. Certainly when strong all within the long-wavelength approximation. The very coupling between light and matter modifies the local same effects should be present for a plasmonic cavity matter subsystem, as is suggested by recent experimental when consistently quantized. Such effects are already ob- results [3, 11, 99, 122–124], the quadratic terms can served when performing ab initio calculations with solely become important and determine the physical properties. the longitudinal Coulomb interaction [114]. For small clusters, therefore small effective volume and high coupling strength, the modification of the response and vol- When looking beyond the simple Rabi splitting of spec- ume due to the presence of the coupled molecule is non- tral lines, which is the accepted indicator of the onset negligible and modifies the plasmonic modes of the clus- of strong coupling, other observables that contain fur- ter. A purely bilinear coupling dictates entirely different ther information about the matter subsystem should be physics (see App. D for a detailed discussion), violates able to highlight the necessity to modify the common all the aforementioned basic constraints and leaves such Dicke or Rabi models, e.g., as demonstrated in Ref. [76]. a simplification as inherently perturbative. While state- By considering photonic as well as matter observables at the same time, the dipole-approximated bilinear coupling 12 can be further scrutinized, the influence of quadratic cou- (ERC-2015-AdG694097), the Cluster of Excellence ’Ad- pling terms revealed and effects that are due to spatially vanced Imaging of Matter’ (AIM), Grupos Consolidados inhomogeneous fields (beyond the long-wavelength ap- (IT1249-19), partially by the Federal Ministry of Edu- proximation) observed. Furthermore, when the light- cation and Research Grant RouTe-13N14839, and the matter coupling renders bilinear, self-polarization and SFB925 ”Light induced dynamics and control of corre- Coulomb interaction act on comparable energy scales, lated quantum systems”. non-perturbative effects can be expected. At this point it is important to realize that this statement also holds spatially, i.e., that while a coupling might be considered Appendix A: Fundamental coupling of light with small for certain bondlengths/extensions of the molecu- matter and the emergence of diamagnetism lar system, at other parts or on other scales it might become substantial. The extension criterion Eq. (11), that Let us here briefly show how QED and its non- weights the with spacial extension increasing divergent relativistic limit can be set up starting from classical elec- forces against the ionization energy, is motivated pre- trodynamics. In vector-potential form the microscopic cisely with this spirit in mind, providing the to g/ω com- description of the electromagnetic fields are given by plementary parameter estimate. Consider e.g. the binding curve of a molecule, probing the dissociative regime 1 0 E(r, t) = ∂tA(r, t) A (r, t), with large distances will change the ratio of the afore- − c − ∇ 1 mentioned contributions until van-der-Waals type of in- B(r, t) = c A(r, t). teractions containing retardation effects have to be con- ∇ × sidered, a problem of also chemical interest [125]. Not For later reference we use the vacuum permeability µ0 just the equilibrium distance of molecules will change but and vacuum permittivity 0, which are connected to the especially their behavior in the stretched configuration speed of light by c = 1/√µ00. If we then choose the, will be effected [13, 14], a feature essential to describe in the non-relativistic limit most convenient, Coulomb chemical reactions. For relatively large systems, which gauge condition A(r, t) = 0, the energy expression of are yet still small compared to the relevant wavelengths the classical electromagnetic∇ · field is given by [92] of the photon field, stronger effects would be expected. In the simple models we presented here we could have Z ⊥ (t) = 0 d3r E2 (r, t) + c2B2(r, t) (A1) used a smaller coupling strength yet a spatially more- Hem 2 ⊥ extended system and we would have found similar effects. Dynamics that probe the long-range part of potential- and the interaction among charged particles emerges via energy surfaces should be affected more strongly, espe- Z cially true when compared to dynamics due to classi- 1 3 int(t) = d r j(r, t) A(r, t) (A2) cal external laser fields in dipole approximation ignoring H − c · the self-polarization term. While our focus remained on | {z } =H (t) the single molecular limit an additional essential scale int,⊥ 1 Z of the system is represented by the amount of charged + d3r j0(r, t)A0(r, t) . carriers, amplifying the dipole, polarizability, and there- 2c | {z } fore the self-polarization contribution. Extended systems =Hint,k(t) (e.g. solids and liquids) and molecular ensembles with charge contact (e.g. biomolecules) are therefore expected Here we have used the decomposition of the electric field to experience quite sizable influences by quadratic com- in a purely transversal part (polarized perpendicular to ponents, i.e., perturbatively seen renormalizing photonic 1 the propagation direction) E⊥(r, t) = c ∂tA(r, t) and a or matter excitations due to the collective light-matter purely longitudinal part (polarized along− the propagation 0 interaction [15, 62, 75]. Recent investigations on cavity direction) Ek(r, t) = A (r, t). The electromagnetic enhanced electron-phonon coupling [64, 65] and its role field is coupled to a charge−∇ current j(r, t) that obeys the for superconductivity [99] might indicate the substantial 1 0 continuity equation c ∂tj (r, t) = j(r, t). We there- scientific impact of this realization. Exploring these sit- fore see that it is the moving charges−∇ via · their combined uations where our theoretical descriptions begin to differ charge current that induce and modify the electromag- strongly thus hold promise of revealing further yet undis- netic fields. covered effects [12]. The above decomposition is furthermore very convenient to single out electrostatic contributions, which are 0 given exclusively in terms of Ek and j . With the Poisson ACKNOWLEDGMENTS equation in full space, which determines the zero component of the electromagnetic vector potential

We would like to thank S. Buhmann, J. Feist, A. Z 0 0 Salam and I. Tokatly for insightful discussions. This 0 1 3 0 j (r , t) A (r, t) = d r 0 , work was supported by the European Research Council c 4π0 r r | − | 13

q ˆ the term int,k(t) in Eq. (A2) can be brought into the mc nˆ(r)A(r) [25, 72, 92] that in the Dirac current arises form H −only implicitly as can be seen by the Gordon decomposition [25, 126]. This quadratic coupling term captures 1 ZZ j0(r0, t)j0(r, t) 3 3 0 the effective photon-photon interaction due to the dis- int,k(t) = 2 d r d r 0 (A3) H 2c 4π0 r r | − | carded positronic degrees of freedom. A direct beneficial and thus corresponds to the longitudinal Coulomb inter- consequence of this explicit diamagnetic term is that it action, typically dominating the electronic structure of removes the infrared divergence of relativistic QED [73]. condensed matter. If we are not in vacuum on all of R3 This can be best understood by considering the Heisen- but instead have, e.g., boundaries with certain boundary berg equation of motion, analogue to the inhomogeneous conditions, the Coulomb kernel 2 1 = δ3(r r0) microscopic Maxwell’s equation, −∇ 4π|r−r0| − changes accordingly. This also changes the Coulomb in- 1 2 2 ˆ ˆ 2 ∂ A(r, t) = µ0cj⊥(r, t) . (A5) teraction among charged particles, e.g., for cavity sit- c t − ∇ ˆ uations it can lead to the inclusion of mirror-charges, Here j⊥(r) is the transversal part of the physical current depending on the selected gauge [112]. Further, due to ˆ ˆ ˆ operator j(r) = jp(r) + jd(r). Grouping the diamagnetic the Coulomb-gauge condition, the first term on the right- current with the vector potential on the left-hand side hand side of Eq. (A2) is merely coupling to the transver- shows that the mere existence of charged particles will sal part of the charge current and can thus be rewritten modify the frequency of the bare fields (see also Sec. II as for the dipole case and recall Sec. I). Z Only when longitudinal (A3) and transversal (A4) cou- 1 3 int,⊥(t) = d r j⊥(r, t) A(r, t). (A4) H − c · pling are treated consistently they provide a local interaction. However, in practice often only one of the two in- We have therefore divided the interaction due to cou- teractions is treated explicitly depending on which prop- pling with a charge current into a purely longitudinal erties of the combined light-matter system one is inter- (electrostatic) and a purely transversal one. To quantize ested in [12]. Focusing on quantum mechanics, e.g., the the theory we need to promote the classical vector poten- electronic structure as essential to describe chemical reac- tial to a quantum field Aˆ (r), which is basically a sum of tions, the transversal interaction is often omitted and one quantum harmonic oscillators [25, 92] (see also Eq. (2)). merely implicitly considers the fluctuations in form of the The quantum fields include the transversal character via physical mass [15, 73]. Quantum optical considerations ij the transversal delta-function δ⊥ in the commutation re- on the other hand focus typically on the description of lations between conjugate fields. We furthermore need the transversal fields and thus strongly simplify the elec- to promote the classical charge current to the conserved tronic structure. The resulting quantum-optical models charge-current operators ˆj(r) and ˆj0(r) = cnˆ(r) of the are designed to predict specific photonic observables and non-interacting matter subsystem [25, 92]. In this way are consequentially limited in their predictability for the the total charge current of the quantized particles gener- matter subsystem [14, 15]. Recent interest in strong-light ates the quantized electromagnetic field, and at the same matter interaction is calling now for a consistent treat- time the photon field modifies the movement of the quan- ment of those historically as complementary perceived tized particles. Hence QED becomes a self-consistent limits. theory of light and matter, and equilibrium is reached Under certain assumptions the diamagnetic term can when a force balance among the constituents is reached. indeed be absorbed by a redefinition of the frequencies This clear procedure holds true if we consider QED and polarizations of the field modes [15, 75, 93, 94]. with Dirac particles and thus the Dirac current. If we, These redefinitions depend on the matter subsystem however, take the non-relativistic limit for the particles (more specifically the number of charged particles) and and thus also for the conserved charge current, a sub- lead to the diamagnetic shift of the photon field which can tlety arises with important consequences. By express- be observed experimentally [98, 127, 128]. Since the dif- ing the positronic degrees of freedom to first order in ference between the bare and the diamagnetically-dressed 1/mc2 in terms of the electronic components, a term photonic quantities go as pN/V , where N is the total quadratic in the vector potential appears [25, 126]. This number of particles and V the quantization volume, it means that in Eq. (A4), if we use the conserved cur- is often argued [93] that one can use the bare quanti- rent j(r, t) = jp(r, t) + jd(r, t) that consists of the para- ties for finite systems. This is not entirely correct. The magnetic current jp plus the diamagnetic current jd [97], same argument would predict that the coupling between 1 R 0 q 2 a correction term of the form c j (r, t) 2mc2 A (r, t) light and matter (see Eq. (7)) would be zero. The rea- has to be added [25]. This leads− to the appearance of a son for non-zero coupling lies in the fact that by mak- quadratic coupling term. This quadratic term renders ing the quantization volume larger (and approaching free the coupling defined by Eq. (A4) consistent with the space) the amount of modes in any frequency interval minimal coupling prescription also in the non-relativistic approaches infinity as well. So while indeed the coupling limit (see the usual minimal-coupling form of Eq. (3)). to an individual mode becomes zero, the coupling to the Indeed, this extra term is due to the explicit appear- continuum of modes is non-zero. Thus when we keep indi- ˆ ance of the diamagnetic current contribution jd(r) = vidual modes in our theoretical description we effectively 14 treat a small but finite frequency interval of modes. This necessarily complicated, as the above Hamiltonian con- frequency interval can be related to the effective mode verges in the norm resolvent sense to the infinite-space volume, since it gives the spacing between the effective Hamiltonian for Mp [132]. Hence the above Hamil- modes. Consequently we also need to dress the photon tonian can be made→ equivalent ∞ to the full infinite-space operators diamagnetically. Hamiltonian if we increase the quantization volume. Appendix D: Spectral features of operators

Appendix B: The Power-Zienau-Wooley gauge transformation Most arguments for performing a multipole expansion for a Hamiltonian are based on perturbation theory, While we here perform the unitary PZW transforma- where local properties derived from a fixed wave function tion after having chosen the Coulomb-gauge quantiza- do only slightly depend on higher-order contributions of tion, which leaves the vector-potential operator invariant a perturbing operator [72]. This does, however, not mean but leads to an adjusted conjugate photon-field momen- that such arguments still apply for non-perturbative con- tum and coupling, one can equivalently use the PZW siderations, i.e., when the operator itself is changed and (multipolar) gauge of the field to perform the quanti- we solve the resulting equation self-consistently. Indeed, zation procedure [68]. This gauge is connected to the if we consider the influence of such expansions on the Coulomb gauge by adjusting the phase of each particle Hamiltonian directly, the opposite is usually true: The R 1 by θ(r) = qα0 r A(sr)ds [68]. This extra phase highest order determines all basic properties. An in- − 0 · removes the explicit diamagnetic component from the structive example is a one-dimensional model atom with ˆ 2 physical current but also assigns a longitudinal compo- H = ∂x/2 + v(x) and v(x) some binding potential cen- nent to the vector potential that can be associated with tered− at x = 0 with v(x) 0 for x . Its spectrum the Coulomb interaction. While the PZW gauge fea- as a self-adjoint operator→ in L2(R) contains→ ∞ both, bound tures, similarly to the Coulomb gauge, a purely transver- (eigenfunctions exponentially localized around x = 0) sal light-matter coupling, it mixes, in accordance with and scattering (distributional eigenfunctions correspond- the macroscopic Maxwell’s equations, light and matter ing to the continuous spectrum) states. In such cases a degrees of freedom [66–68, 79, 85]. harmonic approximation for certain ground-state properties where v(x) v(0) + v0(0)x + v00(0)x2/2 is reasonable (assuming 'v00(0) > 0) and the perturbative in- Appendix C: Transversal basis and distributions fluence of higher-order terms proportional to xn with n > 2 is minor. However, if we consider the actual For an arbitrary cavity geometry, the vector-valued Hamiltonian and treat higher-order terms proportional eigenfunctions can be very complicated, deviating from to xn non-perturbatively in L2(R), we see that either we simple plane waves. Still this basis Sα(r) is assumed to have an operator that is unbounded from below (hav- obey the condition Sα(r) = 0. In this case we need to ing no ground state) with purely continuous spectrum perform the mode expansion∇ · of the vector-potential oper- (no eigenfunctions but only scattering states) for n odd ator and the polarization operator with the correspond- [62], or bounded from below with only bound states for ing modes. Selecting a basis will define the representation n even (again assuming that all v(n)(0) > 0). So all ba- of the delta distribution and the according polarization. sic properties are only determined by the highest order It is important to note that while there are many equiva- of n. We therefore find that an expansion of an opera- lent representations of the delta distribution, e.g., by us- tor only becomes meaningful if we also indicate whether ing different basis sets Sα(r), multiplications of distribu- we consider it perturbatively for a fixed wave function or tions are not uniquely defined [129, 130]. Indeed, the ori- non-perturbatively. In this work we focus on the non- gin of the divergence in quantum field theories stems from perturbative situation. Let us also mention that alter- the fact that (operator-valued) distributions are multi- natively to perturbation theory a non-perturbative con- plied [131] and a regularization and renormalization pro- sideration but on a different Hilbert space of a restricted cedure needs to be employed to give a finite answer. The domain ϑ R3, i.e., we consider the operators on L2(ϑ) usual way of regularization is equivalent to introducing a with appropriate⊂ boundary conditions, or on a restricted 2 cutoff in the mode expansion α and hence by keeping this state space Ψ1, ..., Ψr L (R) becomes possible (see explicit instead of working with an unspecified represen- also Sec. II).{ As illustrated} ⊂ in Sec. II, this will render the tation of the delta distribution we avoid non-uniqueness restricted domain or subset a relevant parameter for the problems [79]. We could straightaway also use the full in- theoretical prediction since the physical properties can finite space [73], but this will just make the notation un- then crucially depend on this parameter.

[1] James A. Hutchison, Tal Schwartz, Cyriaque Genet, ing chemical landscapes by coupling to vacuum ﬁelds,” Elo¨ıse Devaux, and Thomas W. Ebbesen, “Modify- 15

Angewandte Chemie International Edition 51, 1592– [13] Johannes Flick, Michael Ruggenthaler, Heiko Ap- 1596 (2012). pel, and Angel Rubio, “Atoms and molecules in [2] David M. Coles, Niccolo Somaschi, Paolo Michetti, Cas- cavities, from weak to strong coupling in quantum- par Clark, Pavlos G. Lagoudakis, Pavlos G. Savvidis, electrodynamics (qed) chemistry,” Proceedings of the and David G. Lidzey, “Polariton-mediated energy trans- National Academy of Sciences 114, 3026–3034 (2017), fer between organic dyes in a strongly coupled optical http://www.pnas.org/content/114/12/3026.full.pdf. microcavity,” Nature Materials 13, 712–719 (2014). [14] Christian Schäfer, Michael Ruggenthaler, Heiko Ap- [3] Anoop Thomas, Jino George, Atef Shalabney, Mar- pel, and Angel Rubio, “Modification of excitation ian Dryzhakov, Sreejith J. Varma, Joseph Moran, and charge transfer in cavity quantum-electrodynamical Thibault Chervy, Xiaolan Zhong, Elo¨ıseDevaux, Cyr- chemistry,” Proceedings of the National Academy of Sci- iaque Genet, James A. Hutchison, and Thomas W. ences 116, 4883–4892 (2019). Ebbesen, “Ground-state chemical reactivity under vi- [15] Christian Schäfer,Michael Ruggenthaler, and Angel brational coupling to the vacuum electromagnetic field,” Rubio, “Ab initio nonrelativistic quantum electrody- Angewandte Chemie International Edition 55, 11462– namics: Bridging quantum chemistry and quantum op- 11466 (2016). tics from weak to strong coupling,” Phys. Rev. A 98, [4] Xiaolan Zhong, Thibault Chervy, Lei Zhang, Anoop 043801 (2018). Thomas, Jino George, Cyriaque Genet, James A. [16] Johannes Flick, Christian Schäfer, Michael Ruggen- Hutchison, and Thomas W. Ebbesen, “Energy trans- thaler, Heiko Appel, and Angel Rubio, “Ab initio op- fer between spatially separated entangled molecules,” timized effective potentials for real molecules in op- Angewandte Chemie International Edition 56, 9034– tical cavities: Photon contributions to the molecu- 9038 (2017). lar ground state,” ACS Photonics 5, 992–1005 (2018), [5] Battulga Munkhbat, Martin Wersäll,Denis G. Bara- https://doi.org/10.1021/acsphotonics.7b01279. nov, Tomasz J. Antosiewicz, and Timur Shegai, “Sup- [17] Johannes Flick, Heiko Appel, Michael Ruggen- pression of photo-oxidation of organic chromophores by thaler, and Angel Rubio, “Cavity born-oppenheimer strong coupling to plasmonic nanoantennas,” Science approximation for correlated electron-nuclear-photon Advances 4, eaas9552 (2018). systems,” Journal of Chemical Theory and Com- [6] Bo Xiang, Raphael F. Ribeiro, Adam D. Dunkelberger, putation 13, 1616–1625 (2017), pMID: 28277664, Jiaxi Wang, Yingmin Li, Blake S. Simpkins, Jeffrey C. http://dx.doi.org/10.1021/acs.jctc.6b01126. Owrutsky, Joel Yuen-Zhou, and Wei Xiong, “Two- [18] V. M. Agranovich, M. Litinskaia, and D. G. Lidzey, dimensional infrared spectroscopy of vibrational polari- “Cavity polaritons in microcavities containing disor- tons,” Proceedings of the National Academy of Sciences dered organic semiconductors,” Physical Review B 67 115, 4845–4850 (2018). (2003), 10.1103/physrevb.67.085311. [7] Kati Stranius, Manuel Hertzog, and Karl Börjesson, [19] P. Michetti and G. C. La Rocca, “Polariton states in “Selective manipulation of electronically excited states disordered organic microcavities,” Physical Review B through strong light–matter interactions,” Nature Com- 71 (2005), 10.1103/physrevb.71.115320. munications 9 (2018), 10.1038/s41467-018-04736-1. [20] M. Ruggenthaler, F. Mackenroth, and D. Bauer, [8] Daqing Wang, Hrishikesh Kelkar, Diego Martin-Cano, “Time-dependent kohn-sham approach to quantum Dominik Rattenbacher, Alexey Shkarin, Tobias Utikal, electrodynamics,” Phys. Rev. A 84, 042107 (2011). Stephan Götzinger, and Vahid Sandoghdar, “Turning [21] A. González-Tudela, P. A. Huidobro, L. Mart´ın-Moreno, a molecule into a coherent two-level quantum system,” C. Tejedor, and F. J. Garc´ıa-Vidal,“Theory of strong Nature Physics 15, 483–489 (2019). coupling between quantum emitters and propagating [9] Vanessa N. Peters, Md Omar Faruk, Joshua Asane, Ro- surface plasmons,” Physical Review Letters 110 (2013), han Alexander, D’angelo A. Peters, Srujana Prayakarao, 10.1103/physrevlett.110.126801. Sangeeta Rout, and M. A. Noginov, “Effect of strong [22] Ilya V. Tokatly, “Time-dependent density functional coupling on photodegradation of the semiconducting theory for many-electron systems interacting with cav- polymer p3ht,” Optica 6, 318 (2019). ity photons,” Phys. Rev. Lett. 110, 233001 (2013). [10] Oluwafemi S. Ojambati, Rohit Chikkaraddy, William D. [23] Stefan Yoshi Buhmann, Dispersion forces I: Macro- Deacon, Matthew Horton, Dean Kos, Vladimir A. scopic quantum electrodynamics and ground-state Turek, Ulrich F. Keyser, and Jeremy J. Baum- Casimir, Casimir–Polder and van der Waals Forces, berg, “Quantum electrodynamics at room tempera- Vol. 247 (Springer, 2013). ture coupling a single vibrating molecule with a plas- [24] Stefan Buhmann, Dispersion Forces II: Many-Body Ef- monic nanocavity,” Nature Communications 10 (2019), fects, Excited Atoms, Finite Temperature and Quantum 10.1038/s41467-019-08611-5. Friction, Vol. 248 (Springer, 2013). [11] Jyoti Lather, Pooja Bhatt, Anoop Thomas, Thomas W. [25] Michael Ruggenthaler, Johannes Flick, Camilla Pelle- Ebbesen, and Jino George, “Cavity catalysis by cooper- grini, Heiko Appel, Ilya V. Tokatly, and Angel Rubio, ative vibrational strong coupling of reactant and solvent “Quantum-electrodynamical density-functional theory: molecules,” Angewandte Chemie International Edition Bridging quantum optics and electronic-structure the- 58, 10635–10638 (2019). ory,” Phys. Rev. A 90, 012508 (2014). [12] Michael Ruggenthaler, Nicolas Tancogne-Dejean, Jo- [26] Javier Galego, Francisco J. Garcia-Vidal, and Johannes hannes Flick, Heiko Appel, and Angel Rubio, “From a Feist, “Cavity-induced modifications of molecular struc- quantum-electrodynamical light–matter description to ture in the strong-coupling regime,” Phys. Rev. X 5, novel spectroscopies,” Nature Reviews Chemistry 2, 041022 (2015). 0118 (2018). [27] Javier del Pino, Johannes Feist, and Francisco J Garcia- Vidal, “Quantum theory of collective strong coupling of 16

molecular vibrations with a microcavity mode,” New 91, 194 (2018). Journal of Physics 17, 053040 (2015). [44] Norah M Hoffmann, Christian Schäfer,Angel Rubio, [28] Johannes Schachenmayer, Claudiu Genes, Edoardo Aaron Kelly, and Heiko Appel, “Capturing vacuum Tignone, and Guido Pupillo, “Cavity-enhanced trans- fluctuations and photon correlations in cavity quantum port of excitons,” Phys. Rev. Lett. 114, 196403 (2015). electrodynamics with multitrajectory ehrenfest dynam- [29] Akbar Salam, Non-relativistic QED theory of the Van ics,” Physical Review A 99, 063819 (2019). der Waals dispersion interaction (Springer, 2016). [45] Nicholas Rivera, Johannes Flick, and Prineha Narang, [30] Markus Kowalewski, Kochise Bennett, and Shaul “Variational theory of nonrelativistic quantum elec- Mukamel, “Cavity femtochemistry: Manipulat- trodynamics,” Physical Review Letters 122 (2019), ing nonadiabatic dynamics at avoided cross- 10.1103/physrevlett.122.193603. ings,” J. Phys. Chem. Lett. 7, 2050–2054 (2016), [46] Johan F. Triana and JoséLuis Sanz-Vicario, “Revealing http://dx.doi.org/10.1021/acs.jpclett.6b00864. the presence of potential crossings in diatomics induced [31] Felipe Herrera and Frank C. Spano, “Cavity-controlled by quantum cavity radiation,” Physical Review Letters chemistry in molecular ensembles,” Phys. Rev. Lett. 122 (2019), 10.1103/physrevlett.122.063603. 116, 238301 (2016). [47] Michael Reitz, Christian Sommer, and Claudiu Genes, [32] M. Ahsan Zeb, Peter G. Kirton, and Jonathan Keel- “Langevin approach to quantum optics with molecules,” ing, “Exact states and spectra of vibrationally dressed Physical Review Letters 122 (2019), 10.1103/phys- polaritons,” ACS Photonics 5, 249–257 (2017). revlett.122.203602. [33] Hoi Ling Luk, Johannes Feist, J Jussi Toppari, and [48] Uliana Mordovina, Callum Bungey, Heiko Appel, Pe- Gerrit Groenhof, “Multiscale molecular dynamics sim- ter J. Knowles, Angel Rubio, and Frederick R. Manby, ulations of polaritonic chemistry,” Journal of chemical “Polaritonic coupled-cluster theory,” arXiv preprint theory and computation 13, 4324–4335 (2017). arXiv:1909.02401 (2019). [34] Simone De Liberato, “Virtual photons in the ground [49] Norah M Hoffmann, Christian Schäfer, Niko Säkki- state of a dissipative system,” Nature Communications nen, Angel Rubio, Heiko Appel, and Aaron Kelly, 8, 1465 (2017). “Benchmarking semiclassical and perturbative methods [35] Oriol Vendrell, “Collective jahn-teller interactions for real-time simulations of cavity-bound emission and through light-matter coupling in a cavity,” Phys- interference,” arXiv preprint arXiv:1909.07177 (2019). ical Review Letters 121 (2018), 10.1103/phys- [50] Johannes Flick, Davis M Welakuh, Michael Ruggen- revlett.121.253001. thaler, Heiko Appel, and Angel Rubio, “Light–matter [36] Raphael F. Ribeiro, Adam D. Dunkelberger, Bo Xi- response in nonrelativistic quantum electrodynamics,” ang, Wei Xiong, Blake S. Simpkins, Jeffrey C. Owrut- ACS photonics 6, 2757–2778 (2019). sky, and Joel Yuen-Zhou, “Theory for nonlinear spec- [51] Matthew Du, Luis A Mart´ınez-Mart´ınez, Raphael F troscopy of vibrational polaritons,” The Journal of Ribeiro, Zixuan Hu, Vinod M Menon, and Joel Yuen- Physical Chemistry Letters 9, 3766–3771 (2018). Zhou, “Theory for polariton-assisted remote energy [37] Norah M Hoffmann, Heiko Appel, Angel Rubio, and transfer,” Chemical science 9, 6659–6669 (2018). Neepa T Maitra, “Light-matter interactions via the [52] Raphael F Ribeiro, Luis A Mart´ınez-Mart´ınez,Matthew exact factorization approach,” The European Physical Du, Jorge Campos-Gonzalez-Angulo, and Joel Yuen- Journal B 91, 180 (2018). Zhou, “Polariton chemistry: controlling molecular dy- [38] David Hagenmüller,Stefan Schütz,Johannes Schachen- namics with optical cavities,” Chemical science 9, 6325– mayer, Claudiu Genes, and Guido Pupillo, “Cavity- 6339 (2018). assisted mesoscopic transport of fermions: Coherent and [53] Johannes Feist, Javier Galego, and Francisco J Garcia- dissipative dynamics,” Phys. Rev. B 97, 205303 (2018). Vidal, “Polaritonic chemistry with organic molecules,” [39] Jacopo Fregoni, Giovanni Granucci, Emanuele Coccia, ACS Photonics 5, 205–216 (2017). Maurizio Persico, and Stefano Corni, “Manipulating [54] Anton Frisk Kockum, Adam Miranowicz, Simone azobenzene photoisomerization through strong light– De Liberato, Salvatore Savasta, and Franco Nori, “Ul- molecule coupling,” Nature communications 9, 4688 trastrong coupling between light and matter,” Nature (2018). Reviews Physics 1, 19–40 (2019). [40] Javier Galego, ClàudiaCliment, Francisco J Garcia- [55] Gerrit Groenhof, Clàudia Climent, Johannes Feist, Vidal, and Johannes Feist, “Cavity casimir-polder Dmitry Morozov, and J Jussi Toppari, “Tracking po- forces and their effects in ground-state chemical reac- lariton relaxation with multiscale molecular dynamics tivity,” Physical Review X 9, 021057 (2019). simulations,” The journal of physical chemistry letters [41] Javier del Pino, Florian A. Y. N. Schröder, Alex W. 10, 5476–5483 (2019). Chin, Johannes Feist, and Francisco J. Garcia-Vidal, [56] Jorge A Campos-Gonzalez-Angulo, Raphael F Ribeiro, “Tensor network simulation of non-markovian dynam- and Joel Yuen-Zhou, “Resonant catalysis of thermally ics in organic polaritons,” Physical Review Letters 121 activated chemical reactions with vibrational polari- (2018), 10.1103/physrevlett.121.227401. tons,” Nature Communications 10, 1–8 (2019). [42] A. Strathearn, P. Kirton, D. Kilda, J. Keeling, and [57] Alexander Semenov and Abraham Nitzan, “Electron B. W. Lovett, “Efficient non-markovian quantum dy- transfer in confined electromagnetic fields,” The Journal namics using time-evolving matrix product operators,” of chemical physics 150, 174122 (2019). Nature Communications 9 (2018), 10.1038/s41467-018- [58] Denis G Baranov, Battulga Munkhbat, Elena Zhukova, 05617-3. Benjamin Rousseaux, Ankit Bisht, Adriana Canales, [43] Ali Abedi, Elham Khosravi, and Ilya V. Tokatly, “Shed- GöranJohansson, Tomasz Antosiewicz, and Timur She- ding light on correlated electron–photon states using the gai, “Ultrastrong coupling between nanoparticle plas- exact factorization,” The European Physical Journal B mons and cavity photons at ambient conditions,” arXiv 17

preprint arXiv:1912.07739 (2019). [76] Jino George, Thibault Chervy, Atef Shalabney, Eloise [59] R. H. Dicke, “Coherence in spontaneous radiation pro- Devaux, Hidefumi Hiura, Cyriaque Genet, and cesses,” Phys. Rev. 93, 99–110 (1954). Thomas W. Ebbesen, “Multiple rabi splittings under ul- [60] E.T. Jaynes and F.W. Cummings, “Comparison of trastrong vibrational coupling,” Phys. Rev. Lett. 117, quantum and semiclassical radiation theories with ap- 153601 (2016). plication to the beam maser,” Proceedings of the IEEE [77] Daniele De Bernardis, Philipp Pilar, Tuomas Jaako, Si- 51, 89–109 (1963). mone De Liberato, and Peter Rabl, “Breakdown of [61] Barry M. Garraway, “The dicke model in quantum op- gauge invariance in ultrastrong-coupling cavity qed,” tics: Dicke model revisited,” Philosophical Transactions Physical Review A 98, 053819 (2018). of the Royal Society of London A: Mathematical, Phys- [78] Daniele De Bernardis, Tuomas Jaako, and Peter ical and Engineering Sciences 369, 1137–1155 (2011). Rabl, “Cavity quantum electrodynamics in the non- [62] Vasil Rokaj, Davis M Welakuh, Michael Ruggenthaler, perturbative regime,” Physical Review A 97 (2018), and Angel Rubio, “Light–matter interaction in the long- 10.1103/physreva.97.043820. wavelength limit: no ground-state without dipole self- [79] David L Andrews, Garth A Jones, A Salam, and R Guy energy,” Journal of Physics B: Atomic, Molecular and Woolley, “Perspective: Quantum hamiltonians for opti- Optical Physics 51, 034005 (2018). cal interactions,” The Journal of chemical physics 148, [63] Oliver Viehmann, Jan von Delft, and Florian Mar- 040901 (2018). quardt, “Superradiant phase transitions and the stan- [80] Omar Di Stefano, Alessio Settineri, Vincenzo Macr`ı, dard description of circuit qed,” Physical review letters Luigi Garziano, Roberto Stassi, Salvatore Savasta, 107, 113602 (2011). and Franco Nori, “Resolution of gauge ambiguities in [64] Michael A Sentef, Michael Ruggenthaler, and An- ultrastrong-coupling cavity quantum electrodynamics,” gel Rubio, “Cavity quantum-electrodynamical polari- Nature Physics (2019). tonically enhanced electron-phonon coupling and its [81] Federico Hernández and Felipe Herrera, “Multi-level influence on superconductivity,” Science advances 4, quantum rabi model for anharmonic vibrational polari- eaau6969 (2018). tons,” arXiv preprint arXiv:1906.04374 (2019). [65] David Hagenmüller, Johannes Schachenmayer, Cyri- [82] Jiajun Li, Denis Golez, Giacomo Mazza, Andrew Mil- aque Genet, Thomas W Ebbesen, and Guido Pupillo, lis, Antoine Georges, and Martin Eckstein, “Elec- “Enhancement of the electron–phonon scattering in- tromagnetic coupling in tight-binding models for duced by intrinsic surface plasmon–phonon polaritons,” strongly correlated light and matter,” arXiv preprint ACS Photonics 6, 1073–1081 (2019). arXiv:2001.09726 (2020). [66] Edwin Albert Power and Sigurd Zienau, “Coulomb [83] Johannes Flick, Michael Ruggenthaler, Heiko Appel, gauge in non-relativistic quantum electro-dynamics and and Angel Rubio, “Kohn-sham approach to quan- the shape of spectral lines,” Philosophical Transactions tum electrodynamical density-functional theory: Ex- of the Royal Society of London. Series A, Mathematical act time-dependent effective potentials in real space,” and Physical Sciences 251, 427–454 (1959). Proc. Natl. Acad. Sci. U. S. A. 112, 15285–15290 (2015), [67] RG Woolley, “Gauge invariant wave mechanics and http://www.pnas.org/content/112/50/15285.full.pdf. the power-zienau-woolley transformation,” Journal of [84] John C Baez, Irving E Segal, and Zhengfang Zhou, Physics A: Mathematical and General 13, 2795 (1980). Introduction to algebraic and constructive quantum field [68] M Babiker and Rodney Loudon, “Derivation of the theory, Vol. 174 (Princeton University Press, 2014). power-zienau-woolley hamiltonian in quantum electro- [85] Claude Cohen-Tannoudji, Jacques Dupont-Roc, and dynamics by gauge transformation,” Proceedings of the Gilbert Grynberg, “Photons and atoms-introduction Royal Society of London. A. Mathematical and Physical to quantum electrodynamics,” Photons and Atoms- Sciences 385, 439–460 (1983). Introduction to Quantum Electrodynamics, by Claude [69] Rodney Loudon, The quantum theory of light (Oxford Cohen-Tannoudji, Jacques Dupont-Roc, Gilbert Gryn- Science Publications, 1988). berg, pp. 486. ISBN 0-471-18433-0. Wiley-VCH, Febru- [70] C Baxter, M Babiker, and R Loudon, “Canonical ap- ary 1997. , 486 (1997). proach to photon pressure,” Physical Review A 47, 1278 [86] Akbar Salam, “Molecular quantum electrodynamics,” (1993). Long-Range Intermolecular Interactions (2010). [71] VE Lembessis, M Babiker, C Baxter, and R Loudon, [87] Volker Bach, JürgFröhlich, and Israel Michael Sigal, “Theory of radiation forces and momenta for mobile “Spectral analysis for systems of atoms and molecules atoms in light fields,” Physical Review A 48, 1594 coupled to the quantized radiation field,” Communica- (1993). tions in Mathematical Physics 207, 249–290 (1999). [72] D.P. Craig and T. Thirunamachandran, Molecu- [88] Jan Derezińskiand Vojkan Jakˇsić,“Spectral theory of lar Quantum Electrodynamics: An Introduction to pauli–fierz operators,” Journal of Functional Analysis Radiation-molecule Interactions, Dover Books on 180, 243–327 (2001). Chemistry Series (Dover Publications, 1998). [89] Takeru Hidaka and Fumio Hiroshima, “Pauli–Fierz [73] Herbert Spohn, Dynamics of charged particles and their model with Kato-class potentials and exponential de- radiation field (Cambridge university press, 2004). cays,” Reviews in Mathematical Physics 22, 1181–1208 [74] AndrásVukics, Tobias Grießer, and Peter Domokos, (2010). “Elimination of the a-square problem from cavity qed,” [90] A. Gersten, “Maxwell’s Equations as the One-Photon Phys. Rev. Lett. 112, 073601 (2014). Quantum Equation,” Foundations of Physics Letters [75] Yanko Todorov, “Dipolar quantum electrodynamics 12, 291–298 (1999). theory of the three-dimensional electron gas,” Physical [91] Ole Keller, Quantum Theory of Near-Field Electrody- Review B 89, 075115 (2014). namics (Springer, 2011). 18

[92] Walter Greiner and Joachim Reinhardt, Field quantiza- (2019), 10.1038/s41467-018-08101-0. tion (Springer, 1996). [108] Dariusz Chruscinski, “Quantum mechanics of damped [93] F. H. Faisal, Theory of Multiphoton Processes (Springer, systems,” Journal of Mathematical Physics 44, 3718 Berlin, 1987). (2003). [94] Vasil Rokaj, Markus Penz, Michael A. Sentef, Michael [109] Florian Buchholz, Iris Theophilou, Soeren EB Nielsen, Ruggenthaler, and Angel Rubio, “Quantum electro- Michael Ruggenthaler, and Angel Rubio, “Reduced dynamical bloch theory with homogeneous magnetic density-matrix approach to strong matter-photon inter- fields,” Phys. Rev. Lett. 123, 047202 (2019). action,” ACS photonics 6, 2694–2711 (2019). [95] Kayn A Forbes, David S Bradshaw, and David L An- [110] Emil Antonˇc´ık,“On the theory of temperature shift of drews, “Identifying diamagnetic interactions in scatter- the absorption curve in non-polar crystals,” Cechoslo- ing and nonlinear optics,” Physical Review A 94, 033837 vackij fiziceskij zurnal 5, 449–461 (1955). (2016). [111] Xavier Gonze, Paul Boulanger, and Michel Côté,“The- [96] Kayn A Forbes, “Role of magnetic and diamagnetic in- oretical approaches to the temperature and zero-point teractions in molecular optics and scattering,” Physical motion effects on the electronic band structure,” An- Review A 97, 053832 (2018). nalen der Physik 523, 168–178 (2011). [97] Gianluca Stefanucci and Robert van Leeuwen, Nonequi- [112] EA Power and T Thirunamachandran, “Quantum elec- librium Many-Body Theory of Quantum Systems: A trodynamics in a cavity,” Physical Review A 25, 2473 Modern Introduction (Cambridge University Press, (1982). 2013). [113] René Jestädt, Michael Ruggenthaler, Micael JT [98] Y. Todorov, A. M. Andrews, R. Colombelli, S. De Lib- Oliveira, Angel Rubio, and Heiko Appel, “Real-time erato, C. Ciuti, P. Klang, G. Strasser, and C. Sirtori, solutions of coupled ehrenfest-maxwell-pauli-kohn-sham “Ultrastrong light-matter coupling regime with polari- equations: fundamentals, implementation, and nano- ton dots,” Phys. Rev. Lett. 105, 196402 (2010). optical applications,” arXiv preprint arXiv:1812.05049 [99] Anoop Thomas, Eloise Devaux, Kalaivanan Nagarajan, (2018). Thibault Chervy, Marcus Seidel, David Hagenmüller, [114] Tuomas P. Rossi, Timur Shegai, Paul Erhart, and Stefan Schütz, Johannes Schachenmayer, Cyriaque Tomasz J. Antosiewicz, “Strong plasmon-molecule Genet, Guido Pupillo, and Thomas W. Ebbesen, coupling at the nanoscale revealed by first-principles “Exploring superconductivity under strong coupling modeling,” Nature Communications 10 (2019), with the vacuum electromagnetic field,” arXiv preprint 10.1038/s41467-019-11315-5. arXiv:1911.01459 (2019). [115] Hendrik BG Casimir and Dirk Polder, “The influence of [100] Camilla Pellegrini, Johannes Flick, Ilya V. Tokatly, retardation on the london-van der waals forces,” Physi- Heiko Appel, and Angel Rubio, “Optimized effective cal Review 73, 360 (1948). potential for quantum electrodynamical time-dependent [116] Akbar Salam, “On the contribution of the diamagnetic density functional theory,” Phys. Rev. Lett. 115, 093001 coupling term to the two-body retarded dispersion in- (2015). teraction,” Journal of Physics B: Atomic, Molecular and [101] Rohit Chikkaraddy, Bart de Nijs, Felix Benz, Steven J. Optical Physics 33, 2181 (2000). Barrow, Oren A. Scherman, Edina Rosta, Angela Deme- [117] Claudia Climent, Javier Galego, Francisco J Garcia- triadou, Peter Fox, Ortwin Hess, and Jeremy J. Baum- Vidal, and Johannes Feist, “Plasmonic nanocavities en- berg, “Single-molecule strong coupling at room temper- able self-induced electrostatic catalysis,” (2019). ature in plasmonic nanocavities,” Nature 535, 127–130 [118] R Englman and R Ruppin, “Optical lattice vibrations (2016). in finite ionic crystals: Iii,” Journal of Physics C: Solid [102] Denis G Baranov, Martin Wersäll, Jorge Cuadra, State Physics 1, 1515 (1968). Tomasz J Antosiewicz, and Timur Shegai, “Novel [119] R Brako, J Hrnˇcević, and M Sunjić,“Curvatureˇ de- nanostructures and materials for strong light–matter in- pendence of surface phonon and plasmon frequencies,” teractions,” Acs Photonics 5, 24–42 (2017). Zeitschrift fürPhysik B Condensed Matter 21, 193–201 [103] Christian Schäfer,Michael Ruggenthaler, and Angel (1975). Rubio, “Ab initio nonrelativistic quantum electrody- [120] Filippo Alpeggiani and Lucio Claudio Andreani, “Quan- namics: Bridging quantum chemistry and quantum op- tum theory of surface plasmon polaritons: planar and tics from weak to strong coupling,” Physical Review A spherical geometries,” Plasmonics 9, 965–978 (2014). 98, 043801 (2018). [121] Thomas W. Ebbesen, “Hybrid light–matter states in a [104] J. M. Knight, Y. Aharonov, and G. T. C. Hsieh, “Are molecular and material science perspective,” Accounts super-radiant phase transitions possible?” Phys. Rev. A of Chemical Research 49, 2403–2412 (2016). 17, 1454–1462 (1978). [122] Jino George, Atef Shalabney, James A. Hutchi- [105] AndrásVukics, Tobias Grießer, and Peter Domokos, son, Cyriaque Genet, and Thomas W. Ebbe- “Fundamental limitation of ultrastrong coupling be- sen, “Liquid-phase vibrational strong coupling,” tween light and atoms,” Phys. Rev. A 92, 043835 J. Phys. Chem. Lett. 6, 1027–1031 (2015), (2015). http://dx.doi.org/10.1021/acs.jpclett.5b00204. [106] Carlos Sánchez Muñoz, Franco Nori, and Simone [123] A. Thomas, L. Lethuillier-Karl, K. Nagarajan, R. M. A. De Liberato, “Resolution of superluminal signalling Vergauwe, J. George, T. Chervy, A. Shalabney, E. De- in non-perturbative cavity quantum electrodynamics,” vaux, C. Genet, J. Moran, and T. W. Ebbesen, “Tilt- Nature communications 9, 1924 (2018). ing a ground-state reactivity landscape by vibrational [107] Adam Stokes and Ahsan Nazir, “Gauge ambiguities strong coupling,” Science 363, 615–619 (2019). imply jaynes-cummings physics remains valid in ul- [124] Ran Damari, Omri Weinberg, Daniel Krotkov, Na- trastrong coupling QED,” Nature Communications 10 talia Demina, Katherine Akulov, Adina Golombek, Tal 19

Schwartz, and Sharly Fleischer, “Strong coupling of Keller, Mattias Beck, Nicola Bartolo, Clemens Rössler, collective intermolecular vibrations in organic materials Thomas Ihn, Klaus Ensslin, Cristiano Ciuti, Giacomo at terahertz frequencies,” Nature Communications 10 Scalari, and JérômeFaist, “Magneto-transport con- (2019), 10.1038/s41467-019-11130-y. trolled by landau polariton states,” Nature Physics 15, [125] Ylva Andersson, Erika Hult, Henrik Rydberg, Peter 186–190 (2019). Apell, Bengt I Lundqvist, and David C Langreth, “Van [129] Laurent Schwartz, “Sur limpossibilite de la multipli- der waals interactions in density functional theory,” in cation des distributions,” Comptes Rendus Hebdo- Electronic Density Functional Theory (Springer, 1998) madaires des Seances de L Academie des Sciences 239, pp. 243–260. 847–848 (1954). [126] E. Engel and R.M. Dreizler, Density Functional The- [130] JF Colombeau, “Elementary introduction to new gen- ory: An Advanced Course, Theoretical and Mathemat- eralized functions, volume 113 of north-holland math- ical Physics (Springer, 2011). ematics studies, notes on pure mathematics, 103,” [127] Xinwei Li, Motoaki Bamba, Qi Zhang, Saeed Fallahi, (1985). Geoff C. Gardner, Weilu Gao, Minhan Lou, Katsumasa [131] Walter Thirring, Quantum mathematical physics: Yoshioka, Michael J. Manfra, and Junichiro Kono, atoms, molecules and large systems (Springer Science “Vacuum bloch–siegert shift in landau polaritons with & Business Media, 2013). ultra-high cooperativity,” Nature Photonics 12, 324– [132] Asao Arai and Masao Hirokawa, “On the existence and 329 (2018). uniqueness of ground states of a generalized spin-boson [128] Gian L. Paravicini-Bagliani, Felice Appugliese, Eli model,” journal of functional analysis 151, 455–503 Richter, Saeed Fallahi, Federico Valmorra, Janine (1997).