Arxiv:2005.01614V3 [Cond-Mat.Quant-Gas] 1 Sep 2020 the Underlying Many-Body Phases

Home , Supersolid

Probing the supersolid order via high-energy scattering: analytical relations among response, density modulation, and superﬂuid fraction.

L. Chomaz1 1Institut fürExperimentalphysik, UniversitätInnsbruck, Technikerstraße 25, 6020 Innsbruck, Austria (Dated: September 2, 2020) High-energy scattering spectroscopy is a widely-established technique for probing the characteristic properties of complex physical systems. Motivated by the recent observation of long-sought supersolid states in dipolar quantum Bose gases, I investigate the general relationships existing between the density contrast, the superfluid fraction, and the response to a high-energy scattering probe of density-modulated states within a classical-field approach. I focus on the two extreme regimes of ”shallow” and ”deep” supersolids, which are of particular interest in describing the phase transitions of the supersolid to a uniform superfluid and an incoherent crystal state, respectively. Using relevant Anstze for the fields of dipolar supersolid states in these regimes, I specify and illustrate the scaling laws relating the three observables. This work was first prompted to develop an intuitive understanding of a concomitant study based on experiments and mean-field numerical simulations. Beyond this specific application, this works provides a simple and general framework to describe density-modulated states, and in particular the intriguing case of supersolids. It describes key properties characterizing the supersolid order and highlights new possibilities for probing such properties based on high-energy scattering response.

I. INTRODUCTION mental features of the SSS, which are its (partial) superfluid fraction, the contrast of the density modulation that characterizes the underlying crystal order, and the Supersolidity is a counterintuitive state of matter, first possible occurrence of phase variations and fluctuations, conceived more than half a century ago [1,2]. The which connect to the dynamics, finite temperature, and very possibility of a supersolid phase and its surprising global coherence of the state. Investigating these fea- properties have raised great debates and triggered many tures requires developing complementary experimental studies [3–11]. A particularly intriguing feature of a su- approaches, which also await further theoretical under- persolid is that it could be fully phase-coherent (Bose- standing. condensed) while only partially superfluid, its superfluid Scattering spectroscopy, i.e. recording the response of fraction, f , satisfying 0 < f < 1 [3]. The experimental sf sf a target system to a scattering probe as a function of search for supersolid states (SSS) had for long remained its energy and momentum, has long been established as inconclusive [8–11]. Complementing many efforts focus- an exquisite probe of the structures and properties of ing on helium, the past decade has seen quantum gases quantum many-body states [32, 45–48]. Scattering pro- appearing as a promising platform for this search [11– tocols have been developed and applied on a wide vari- 22]. ety of systems ranging from nuclear to condensed matter In 2019, breakthrough experiments reported on the physics [45–53]. In the case of the quantum-gas plat- observation of SSS in cigar-shaped dipolar quantum form relevant for the recent experimental realizations of gases [23–25]. In these systems, a spontaneous density supersolids, two-photon Bragg spectroscopy is a highly modulation arises in the axial direction (x), via the in- versatile technique for probing the response at momenta trinsic effect of the momentum-dependent inter-particle and energies ranging from low to high [54–65]. interactions [11, 14, 19–28]. A global phase-coherence In the weak-perturbation regime, scattering spec- was additionally shown to be preserved. Such a coex- troscopy connects to the central concept of the spectrum istence of solid and superfluid orders arises thanks to a of elementary excitations [32, 66–68], dictating the sys- subtle competition of interactions, at the mean-field and tem’s dynamical response and thermodynamical proper- (many-body) beyond-mean-field levels. Calculations per- ties. In the low-energy regime, the excitation spectrum formed within an extended mean-field (eMF) approxima- is sensitive to collective effects and its changes reflects tion show remarkable agreement with experiments. Here, arXiv:2005.01614v3 [cond-mat.quant-gas] 1 Sep 2020 the underlying many-body phases. In the supersolid the SSS is described by a classical field, Ψ(r), with am- case, various experimental and theoretical works under- plitude modulation along x, and its evolution is dictated took to characterize the specificity of its low-lying excited by an extended Gross-Pitaevskii equation including first modes [23, 24, 38, 40, 69–73]. A major signature is the order corrections from quantum fluctuations [29–37]. occurrence of two phononic branches, connected to the Several studies on the peculiar dynamical behaviors spontaneous breaking of two continuous symmetries. The arising with supersolidity in dipolar gases soon followed, high energy scattering regime, achieved by using a probe in experiments and using eMF theory [38–44]. Yet, whose energy is much higher than the characteristic en- many questions still remain open. Of prime interest ergy of the interaction system and thus well described by are those dealing with the interplay between the funda- free-particle excitations, constitutes a distinct paradigm 2 in which one probes the system’s microscopic properties. noted L with periodic boundary conditions at the interval Whereas it has proven powerful to probe key features [0,L] bounds. I focus on the system’s properties along the such as short-range correlations or beyond-mean-field ef- modulation direction and disregard the system’s behav- fects [32, 58–65, 74–78], high-energy scattering has not ior in the yz-plane by assuming the spatial dependencies been previously applied to SSSs. The present work and to be separable. Then, the wave function decomposes as Ref. [79] are a first step toward bridging this gap. Ψ(r) = ψ(x)ξ(y, z) with R |ξ(y, z)|2dy dz = 1 [32, 73, 80], In the present work, I use simple models relying on and integrating out the yz dependence of Ψ(r) yields an a classical-field description to figure out which informa- effective one-dimensional model on ψ(x). tion is contained in the high-energy scattering response The SSS’s crystal order implies that the amplitude of a SSS. Initiated in relation to the investigations of |ψ(x)| is modulated. A SSS is standardly described con- Ref. [79] based on experiments and numerical simulation, sidering a fully-coherent many-body ground state, which this work provides us with an intuitive understanding of corresponds to a single classical field with amplitude our intricate findings. Beyond this first application, the modulation and uniform phase [2,3,6, 22, 24, 25, 38– present work also provides us with a general framework to 42, 70, 81, 82]. We here extend the state’s description describe SSSs, sketches general scaling relationships be- beyond this case in order to encompass several physical tween its most fundamental properties, characterizing its effects that are relevant for dipolar SSSs. We here allow supersolid order, and shows the relevance of high-energy for spatial modulations of ψ(x) not only in amplitude scattering to probe supersolidity. but also in phase, which enable to capture the possible The paper is organized as follows. SectionII first in- existence of excitations in the state [32]. troduces our representation of a SSS and decomposes Excitations can occur in SSSs for two main reasons, its field in presence of density and phase modulations. namely dynamics and statistical fluctuations. Dynamics The SSS’s density contrast, superfluid fraction, and is relevant for dipolar SSSs since, in order to reach such high-energy scattering response are then described, and states in experiments, dynamical tuning of parameters their expressions within the present representation de- and crossing of phase transitions are performed [19, 23– tailed. SectionsIII andIV explore the behaviors and 25, 39, 40, 44]. The resulting coherent dynamics yields relations of these three quantities, in the two limiting deterministic phase patterns, i. e. which are reproducible regimes of ”shallow” (i.e. weakly modulated) and ”deep” from one realization of ψ to the other. Furthermore, (i.e. strongly modulated) supersolids, respectively. In statistical fluctuations are important due to both the Sec.III, scaling laws are derived from Taylor expansions enhanced role of quantum fluctuations in the dipolar in powers of the modulated component and exemplified SSS [19, 23–25, 33–37] and the finite temperature of ex- using a sine-modulation Ansatz. In Sec.IV, the super- perimental systems. The present classical-field model solid is described as an array of (overlapping) drops and can be used to approximately account for statistical- the scaling laws extracted from the drop’s profile. Con- fluctuation effects [83]. This is achieved by performing clusions are drawn in Sec.V. ensemble averages over sets of ψ including phase and amplitude modulations of random character, and appropri- ately sampling the phase space. For instance, one can II. GENERAL CONSIDERATIONS add thermal and quantum noise in the weakly correlated regime by superimposing excited modes on this initial The aim of this section is to introduce a general repre- state, with the complex amplitudes of these modes drawn sentation of a SSS (Sec.IIA), and to specify within this from the appropriate (Wigner) distribution [83]. I note framework the expressions of the state’s main charac- that the effect of such fluctuations will be particularly teristics (Sec.IIB), namely its density contrast [Eq. (10)] relevant in the ”deep” supersolid regime of Sec.IV, dic- and superfluid fraction [Eq. (11)], as well as of its tating the transition to an incoherent crystal state. In the response to a high-energy scattering probe [Sec.IIC, following, I will give the expressions of the state’s prop- Eq.(14)]. These general expressions are the basis to the erties for a single modulated field. A simple extension derivations of Secs.III,IV. to fluctuating configurations is provided via the above- discussed statistical sampling. To analyze the SSS’s properties, I use a convenient de- A. Supersolid state’s decomposition composition of the axial field, ψ(x), in two components: (i) a spatially uniform (u) component, ψu, corresponding to the spatial mean of the field, and (ii) a modulated (m) In the present model, a SSS is depicted by a classical component, ψm(x), accounting for spatial modulations of field, Ψ(r). This is particularly justified in the context the field’s phase and amplitude, of dipolar supersolids, given their successful description within an eMF theory [22–25, 38–43]. The density mod- ψ(x) ≡ ψu + ψm(x), (1) ulation is assumed to arise along one direction of space, Z L dx x, as relevant for the cigar-shaped configurations of the ψ ≡ ψ(x) . (2) u L above-cited experiments. For simplicity, I consider a sys- 0 tem confined in a uniform box whose size along x is de- It follows from this definition that ψm spatially averages 3 out, The superfluid order is standardly quantified via the superfluid fraction, f [32, 66]. In a seminal work [3], Z L sf Leggett explicitly related the superfluid character of a ψm(x)dx = 0. (3) 0 fully coherent state to the spatial dependence of its wave function. Leggett’s derivation is based on the concept The global phase of ψ(x) is chosen such that ψu is of non-classical rotational inertia: when a system is put real [84]. The above decomposition is particularly con- under rotation, its energy changes; the ratio between venient for our purposes, since, as we will see in Sec.IIC the quantum and the classical expectations for this en- [Eq.(14)], the amplitude of the high-energy scattering re- ergy change is given by the superfluid fraction. Thus, sponse on resonance, which is our observable of interest, by evaluating the kinetic energy of a fully coherent SSS is uniquely determined by ψu, see also ApendixA. under the assumption of a stationary flow and twisted In the remainder of this section, I detail how the SSS’s boundary conditions, Leggett found an upper bound for properties read within the decomposition of Eqs.(1)-(3). the supersolid’s superfluid fraction. This upper bound In this aim, a first step is to write the spatial density was further shown to be saturated in the case of a one- distribution. This is dimensional (axial) density modulation [82], as consid- 2 2 ered in the present work and relevant for the dipolar-SSS n(x) = |ψ(x)| = nu + |ψm(x)| + 2ψuRe[ψm(x)], (4) experiments. Therefore, I define fsf using Leggett’s for- with nu being the density of the spatially uniform com- mula, ponent, n = ψ2. The cross term ψ ψ (x) appearing in u u u m −1 Eq. (4) implies that the spatial density cannot be gener-  L L  Z Z dx ally decomposed into independent contributions from the f ≡ L2 dx|ψ(x)|2 . (9) sf  |ψ(x)|2  uniform and modulated fields. In contrast, Eq. (3) yields 0 0 the simple decomposition of the spatial mean density, Equation (9) evaluates the state’s superfluid fraction for Z dx n¯ ≡ n(x) = nu +n ¯m, (5) ψ(x) real (i.e. for a steady coherent state) [3, 81, 82]. L When ψ(x) presents phase modulations, as can be con- wheren ¯ is the mean density in the modulated field, sidered in this work (see Sec.IIA), we note that Eq. (9) m is still mathematically defined, yet without direct phys- Z 2 dx ical meaning. Equation (9) implies that, if ψ(x) has a n¯m = |ψm(x)| . (6) L uniform amplitude (cancels somewhere), then fsf = 1 (fsf = 0). When uniform, a fully coherent state is en- We also introduce the effective profile tirely superfluid, whereas it loses its superfluid character 2 if its density vanishes somewhere in space. nm(x) = n(x) − nu = |ψm(x)| + 2ψuRe[ψm(x)], (7) Using the decomposition of the density profile of which can take negative values. Despite that nm(x) dif- Secs.IIA [Eqs.(4)-(7)], C and fsf rewrite 2 fers from |ψm(x)| , these two profiles have the same av- max(nm) − min(nm) erage value,n ¯m. C = , (10) 2nu + max(nm) + min(nm)  L −1 B. Supersolid’s characteristics: density contrast Z n¯ dx and superfluid fraction fsf =   . (11) nu + nm(x) L 0 Supersolids are states of matter where both crystal and Without making additional assumptions, no general for- superfluid orders coexist. To quantify these two distinct mula relates C and fsf to the averages nu andn ¯m. In characters, two physical observables are needed. Here, the following sections, such relations will be extracted in I specify such observables and express them within the limiting cases or by relying on wave-function Anstze. model (1). The crystal order of a SSS relates to the amplitude modulations in its field, ψ(x). Their strength is charac- C. High-energy scattering response terized via the contrast of the modulations induced in the density profile [23–25, 44, 73], C, defined as As introduced in Sec.I, the focus of the present work is max(n) − min(n) to understand which properties of many-body quantum C = , (8) max(n) + min(n) supersolid states are probed via scattering spectroscopy, especially focusing on the high-energy regime. Consider- where max (min) denotes the maximum (minimum) over ing a scattering probe of momentum ¯hk and energy ¯hω, space (i.e. x ∈ [0,L]). By definition, C = 0 (C = 1) is the regime of high-energy scattering is defined, within a equivalent to ψ(x) having a uniform amplitude (vanishing classical-field picture, by comparing ¯hω and ¯hk2/2m (m somewhere). is the particle’s mass) to the characteristic energy scales 4 of the interactions of particles within the state ψ(x) and Here ψˆ(x) is the field operator, annihilating a particle at scattered to a state of momentum ¯hk [32, 74, 77, 78], position x, h·i is the state average, and n(x) = G1(x, x). 0 see also AppendixB. In the regime where the probe en- The matrix G1(x, x ) characterizes the correlations in the ergies are large compared to the interactions, the sys- system and its long-distance behavior informs on the system’s excitations are well approximated by free-particle tem’s coherence (i.e. off-diagonal long-range order). We states, mostly unaffected by interaction effects, justify- note that, because the SSS of interest here is intrinsically ing an ”impulse approximation”. For ultracold quantum density modulated, thus nonuniform,n ˜(0) provides infor- gases, this regime is readily achievable based on two pho- mation not only on the coherence properties of the state ton Bragg scattering [58–65, 79]. but also on the modulations of the order parameter. This section expresses the resonant response of a SSS Using the general decomposition of Sec.IIA, which ˜ to such a high-energy high-momentum scattering probe, yields by definition ψ(0) = Lψu [Eq. (2)], the resonant and relates it to the decomposition (1-2). Exploiting high-energy DSF of a modulated state is simply given by the geometry described in Sec.IIA, the discussion is re- stricted to axial excitations with momenta along x and S = nuL = (¯n − n¯m)L, (14) |k| = kx ≡ k. I additionally disregard the effect of trans- i.e. by the density of the uniform component. It is inter- verse excitations and assume that the excited modes’ esting to note that this response is reduced compared to wave functions decompose into ϕj(x)ξ(y, z) for all modes that of a uniform state of identical mean densityn ¯. The j responding to the probe. In the high-energy regime, reduction factor, given by the fraction of atoms encom- the excited state j of momentum√ ¯hkj = hj/L has a wave passed in the modulated field [Eq. (6)], arises from both ikj x function ϕj(x) ≈ e / L well approximated by a plane density and phase modulations, see also AppendixA. wave, and an energyhω ¯ j dominated by the kinetic term Finally, we note that the DSF on resonance is a conve- 2 2 ¯h kj /2m. nient observable that one can readily characterize from In the linear regime, the response of a physical sys- scattering spectroscopy measurements. In such a mea- tem to a scattering probe of momentum ¯hk and en- surement, after applying a scattering probe of momen- ergy ¯hω is given by the dynamic structure factor (DSF), tum ¯hk and energy ¯hω for a time τ, one typically ex- S(k, ω)[32, 85, 86]. It connects to the expectation tracts the fraction of particle excited to the momentum of the squared (equal-time) density fluctuations at mo- ¯hk or equivalently the total momentum transferred to the mentumhk ¯ . Focusing on the high-k regime where in- system. In cold-atom experiments, this is achieved by teraction effects are negligible, this is simply given by imaging the cloud’s momentum distribution after releas- P R ikx ∗ 2 j dx e ϕj (x)ψ(x) [32, 74, 77, 78]. The DSF in ing it from the trap [54–65, 79]. Such observables are ex- this so-called impulse approximation then reads pected to be directly proportional to the DSF amplitude Sτ (k, ω) Fourier-broadened by the finite probe time, ne- X n˜(k − kj) S(k, ω) = δ(¯hω − ¯hω ), (12) glecting the initial population of the high-k state [32, 87]. L j j By probing the system’s response at a given momentum and varying ω, one typically records a resonant curve of where δ(·) is the Dirac delta function, andn ˜(k) is the mo- Sτ (k, ω) with ω and the value on resonance finally gives mentum distribution corresponding to ψ(x), i.e.n ˜(k) = access to S. This scheme has been applied to super- |ψ˜(k)|2 with ψ˜(k) = R dx eikxψ(x). solids in the measurements of ref. [79] concomitant to Probing the system on resonance with a state j, i.e. at the present work. k = kj and ω = ωj, the response is given by

mn˜(0) mL III. SHALLOW-SUPERSOLID REGIME S(kj, ωj) = 2 ≡ S 2 , (13) 2π¯h kj 2π¯h kj In Sec.II, I gave general formulae expressing two char- mL The factor 2 appearing in Eq. (13) is the density of 2πh¯ kj acteristic properties of a SSS, namely its density contrast excited states at the energy ¯hωj. In general, the density C [Eq. (10)] and its superfluid fraction fsf [Eq. (11)], as of excited states could depend not only on the excita- well as our observable of interest, the high-energy reso- tion parameters but also on the state’s properties. Yet, nant scattering response S [Eq. (14)], as a function of the at large kj, this factor is purely given by the excitation state’s macroscopic field and its modulations. I remind momentum and does not depend on ψ, see Appendix that, whereas the above equations give the formulae for B. In the following, I focus on the dimensionless factor one classical field, fluctuations effects could be accounted S that accounts for all information on the state’s wave for by sampling the fields’ phase space and performing function ψ contained in the high-momentum DSF. This ensemble averages of these expressions (see discussion in factor being simply S =n ˜(0)/L implies that the res- Sec.IIA). Based on these formulae, the aim of Secs.III onant scattering response directly probes the system’s andIV, is to specify the scaling laws relating these three momentum distribution at k = 0. The componentn ˜(0) quantities, while focusing on limiting regimes, where the gives the spatial integral of the one-body density ma- SSS presents extreme characters in its solid and super- 0 ˆ† ˆ 0 R 0 0 trix, G1(x, x ) = hψ (x)ψ(x )i,n ˜(0) = dxdx G1(x, x ). fluid properties. 5

The present section focuses on the case of a weakly A. Scaling laws of the supersolid’s properties modulated SSS, i.e. with a modulated field of small amplitude compared to the uniform one. In this weak- To describe the shallow-supersolid regime, I introduce modulation regime, here referred to as a ”shallow” su- ψm(x) the normalized modulated field ψ¯ (x) = √ . I then persolid, one expects C 1 and f ≈ 1 [Eqs. (10)- m n¯ sf define the small parameter Λ 1 via (11)]. Such a regime is of particular interest to describe the transition region between a uniform superfluid and a Z n¯ Λ2 = |ψ¯ (x)|2dx = m , (15) supersolid state, provided that the discontinuity at the m n¯ transition is not too large [73, 82]. In experiments, the ¯ dynamical crossing of this transition is one of the usual such that |ψm| is of order Λ. To characterize the SSS’s paths towards SSSs [23–25], making the present regime properties, I first derive the spatial density distribution of practical relevance. (4). The modulated contribution (7) rewrites as a func- ¯ tion of ψm and Λ as ¯ ¯ 2 3 nm(x) ≈ n¯ 2Re[ψm(x)] + |ψm(x)| + O(|Λ| ) . (16) Starting from Eq. (16), I perturbatively expand Eqs (10), (11) in order of Λ to derive approximate scaling laws of C and fsf . ¯ Using the fact that the real part of ψm(x) takes posi- tive as well as negative values over space (following from Eq. (3)), the small density contrast reads ¯ ¯ 2 C = max(Re[ψm]) + max(−Re[ψm]) + O(|Λ| ). (17) It is thus of order Λ and one can write (a) (c) C ∼ Λ 1. (18)

Leggett’s formula (11) for the superﬂuid fraction rewrites ¯ as a function of ψm and Λ as

 L −1 Z 1 dx f = (19). sf  ¯ 2 ¯ 2 3  1 + 2Re[ψm] − Λ + |ψm| + O(|Λ| ) L 0 Expanding the fraction up to second order in Λ (which is the leading non-zero order) and simplifying the integral (b) (d) gives

L Z ¯ 2 dx 2 fsf ≈ 1 − 4 Re[ψm(x)] ≥ 1 − 4Λ . (20) FIG. 1. Examples of SSSs with sine-modulated fields, de- L fined by Eqs. (1) and (24), using C = 0.1 (a,b) and C = 0.7 0 (c,d). Here LK/2π = 5 is chosen. (a,c) show the axial field Keeping in mind that Eq. (11) correctly estimates the ψ(x) (solid black line), which is here real, and its general superfluid fraction only in the case of a steady coher- decomposition [Eq. (1)] in ψu [dashed red line, Eq. (2)] and ent state (see discussion in Sec.IIB), the physically ψm [dotted blue line]. (b,d) show the corresponding density profiles and average values. The total density, n(x), [thick relevant case for the inequality of Eq. (20) is thus the 2 ¯ black line, Eq. (4)] has for average ,n ¯, [thin grey line, Eq. (5)], equality, fsf ≈ 1 − 4Λ = 1 − 4¯nm/n¯, with ψm(x) be- which decomposes in nu [dashed red line] andn ¯m [thin blue ing real. Generally speaking, the value 1 − 4¯nm/n¯ ex- line] the contributions of ψu and ψm, respectively. The dot- tracted from the mean densities is a lower bound for ted blue line shows the effective profile nm(x) [Eq. (7)]. For the value of fsf , meaning that phase modulations in the (a,b) we numerically find a density contrast C = 0.1, a su- field yield an effective increase of the quantity fsf . This perfluid fraction fsf = 0.995 and a high-energy scattering re- statement should however be pondered by reminding sponse S/nL¯ = 0.99875, matching the perturbative results that the occurrence of phase modulations also changes of Eqs. (25)-(26). For (c) and (d), we find C = 0.62 (0.7), the magnitude of the modulated field. In particular, as fsf = 0.774 (0.755) and S/nL¯ = 0.942 (0.939) from numerics R |ψ(x)|dx ≥ | R ψ(x)dx|, the value ofn ¯ /n¯ is larger for (perturbative formulae using C = 0.7). m ψ(x) than for a state whose field would be |ψ(x)|, and 1−4¯nm/n¯ is not the superfluid fraction associated to the mere amplitude of ψ, see also AppendixA. 6

Combining Eqs.(20), and (18) with (14), one ﬁnds the following approximate relations between S, fsf , and C: (a) S f + 3 < sf , (21) nL¯ ∼ 4 2 1 − fsf ∼ C , (22) S 1 − ∼ C2. (23) nL¯ Those approximate relations are valid at leading order for small Λ or equivalently small C. In Eq. (21), the approximate inequality becomes an approximate equality (c) in the case of ψ(x) real, i.e. of a fully coherent stationary state. The Equations (21)-(23) constitute the central result of Sec.III, specifying how the observable S from high-energy scattering scales with the density contrast and superﬂuid fraction of a SSS in the weakly modulated regime. The scaling is mostly linear with fsf and quadratic with C. In the next subsection, we will exemplify these relations.

B. Sine-modulated Ansatz (b) (d)

To give a more concrete example of shallow SSS, one can use a special form of modulated field. Following the considerations of Refs. [14, 73, 82], I apply a sine Ansatz FIG. 2. Illustration of the scaling laws of the SSS’s properties within the sine-modulated Ansatz (24) and the shallow- to the modulation in the state’s wave function, writing supersolid approximation. (a) density contrast numerically √ ψ¯ (x) = ψ (x)/ n¯ = C sin(Kx)/2. (24) extracted using Eq. (8) as a function of the amplitude mod- m m ulation parameter C in (24) (red solid line). In the shallow Here, C is the amplitude modulation parameter, K is supersolid regime the two quantities match (dashed line). (b) the wave number associated with the density modulation. Scattering response S/nL¯ as a function of the superfluid fraction fsf . The solid red line shows the numerically extracted In order to satisfy periodic boundary conditions within values using Eqs. (14) and (9). The dashed line shows the ex- the Ansatz (24), LK/2π should be an integer. Figure pected approximate linear scaling law of Eq. (21). The panels 1 shows two examples of SSSs built on the Ansatz (24). (c,d) detail the scaling law as a function of C of fsf and S/nL¯ , The panels (a,b) exemplify a SSS with a weak amplitude respectively. The red lines show the numerically extracted modulation of C = 0.1 whereas the panels (c,d) show values, as a function of the Ansatz’s amplitude modulation a more strongly modulated case with C = 0.7. In the [Eq. (24)] (solid line) or as a function of the numerically ex- latter case, nu visibly deviates fromn ¯ (see Fig.1(d)), thus tracted density contrast (dash-dotted line), see also (a). The black dashed lines show the shallow-supersolid approximate departing from the assumptions of the present Section√ III. A simple integration of Eq. (24) yields Λ = C/ 8. Fol- scaling laws from Eqs. (25) and (26). lowing Eq. (17), the density contrast indeed matches the Ansatz’s modulation parameter, C, at leading order in Λ, i.e. for small C values. Satisfying this condition, the case and (14) yields 2 of Fig.1(a,b) has a density contrast and a modulation fsf ≈ 1 − C /2, (25) parameter that match. In contrast, in Fig.1(c,d), the S ≈ (1 − C2/8)nL, ¯ (26) actual density contrast equals 0.62. Figure2(a) further details how the actual density contrast scales with the using that ψ(x) is here real. Figure2(b-d) illustrates Ansatz’s amplitude modulation parameter over the full the relative scaling laws of fsf , S and C within this sine- range 0 ≤ C ≤ 1. The identity of the two quantities ex- modulation Ansatz and over the full range of modulation pected in the shallow supersolid regime is observed for a parameter values 0 ≤ C ≤ 1. As for the contrast’s scaling < wide range of C ∼ 0.5. For larger C, the density contrast law discussed above, see Fig.2(a), the shallow-supersolid is reduced compared to the amplitude parameter. predictions are found to describe well the numerically ex- The sine-modulated Ansatz (24) additionally enables tracted values over a relatively broad parameter range. to confirm and specify the approximate quadratic scaling Figure2(b) reports on the relative behavior of S and fsf . laws derived in Eqs. (22)-(23) of the superfluid fraction The approximate linear scaling law of Eq. (21), generally and scattering response versus density contrast. Insert- expected in the shallow-supersolid regime, appears here > ing the expression of Λ as a function of C in Eqs. (20) relevant for fsf ∼ 0.6. Figure2(c,d) show that the ex- 7 pected quadratic scaling laws of fsf and S as a function Finally, the total field for the drop array writes of C, given by Eqs. (25)-(26), are well satisfied for density N contrasts, C < 0.5. Interestingly, the shallow-supersolid P D iθj ∼ ψ(x) = j=0 χ(x − jd)e . (27) scaling laws still describe well the properties’ variations at larger C, yet when expressed in terms of the modula- In order to satisfy periodic boundary conditions within tion parameter (solid line) and not of the actual density the Ansatz (27), L should be L = d(ND − 1) and θ1 = contrast (dash-dotted line). θND . There is then effectively only ND − 1 independent drops in the array. For the Ansatz (27) to be relevant (i.e. to be able to IV. DEEP-SUPERSOLID REGIME indeed isolate drops), the wave-function overlap between neighboring drops is required to be small. This require- This section explores the opposite limiting case of a ment imposes a steep-enough functional dependence of deep supersolid, i.e. a state with a strong amplitude mod- |χ(x)| over the range set by the drop distance d. Relying ulation such that C ≈ 1. In this case, as the spatial on the monotonic character of |χ(x > 0)|, we generally dependence of the field is in the dominant term, per- define the steep-|χ| condition as turbative treatments from the general decomposition of Eqs. (1)-(3) do not easily help to simplify the expression |χ (d) | |χ (d/2) | 1, (28) of fsf and C, and to ultimately derive their relationships |χ (d/2) | |χ (0) | with S. To go further, I therefore rely on a specific Ansatz for the deep SSS’s field. see also Sec.IVB and AppendixC. The steep- |χ| condition (28) has strong connections to a localization condition, constraining the localization parameter (i.e. drops’ A. The drop-array Ansatz extent-over-distance ratio) to small values, λ 1. In particular, |χ(d/2)| ∼ |χ(0)| generally implies λ ∼ 1, In order to describe a strongly modulated SSS, a stan- and, for a fixed functional form for |χ(x)|, the ratios ap- dard model, previously used in the context of dipo- pearing in Eq. (28) decrease with λ. lar gases, assumes that the field ψ(x) consists of a set In the present model, one should note that, if the of ND widely spaced and localized density peaks (or drops’ phases θj are not equal, then the wave function’s drops) [19, 25, 73]. The present work considers such a phase would jump in between the drops. In a more real- model, additionally assuming the drops to be regularly istic model, the phase gradients should be finite. Due to spaced by a distance d along x and to have identical the continuity equation, it is however physically expected density profiles. This implies that the drops can be de- that, in a steady state, phase gradients are concentrated scribed by a unique wave function, χ(x) here taken to be in the region of lower densities [3, 32]. In a array of steep centered at x = 0. Further, |χ(x)| is assumed to be sym- drops (see also Sec.IVB), this is exactly in between the metric and monotonically decreasing towards zero for in- drops, as assumed by the present model. creasing |x| (c.f. ”localized” peak). A localization param- In the following, most conclusions will be drawn with- eter, λ, is defined by the ratio of the root-mean-square out specifying a functional form for χ(x) and simply as- (rms) width of the drop density profile to the distance d, suming it to be steep [Eq. (28)]. In the context of dipo- q R 2 2 R 2 lar gases, drop Anstze were found to be appropriate in i.e. λ = x |χ(x)| dx/N1d d where N1d = |χ(x)| dx regimes similar to that of the present work, describing is the number of atoms in one drop. either single-drop states or density-modulated ones, see In our model, the drops can additionally have indepen- Refs. [19, 20, 25, 35, 73]. These works typically rely on a dent phase profiles. As introduced in Sec.IIA, the inclu- specific Ansatz for χ(x), assuming it of Gaussian profile. sion of phase patterns enables us to encompass various In the following, I will use such a Gaussian Ansatz to physical phenomena, ranging from dynamics to fluctu- exemplify the relationships generally derived, i.e. posing ations. Of particular relevance for the deep-supersolid regime, we note that the strength of statistical phase r 2 fluctuations connects to the degree of global phase coher- N1d x χ(x) = χg(x) ≡ exp − . (29) ence of the underlying state, larger fluctuations meaning π1/2λd 2λ2d2 weaker coherence. In the present modeling, the phases are assumed to be uniform over each drop, with value Examples of phase-coherent stationary Gaussian-drop θj for the drop j. This description of the phase pro- arrays are shown in Fig.3 for two values of λ. The pan- file over an array is reminiscent of an array of Josephson els (a,b) show an array made of relatively broad drops junctions [88]. Josephson-junction models were also pre- with λ = 0.2, and (c,d) exemplify an array of much viously fruitfully applied to the case of density modulated steeper drops with λ = 0.1. The upper row shows the states of dipolar gases, see Refs. [19, 25, 44]. The phase- states’ wave function following from Eqs. (27) and (29). coherent stationary case where ψ(x) is real corresponds The lower panels show the corresponding density profiles, to θj ≡ 0. which will be the focus of the next SectionIVB. 8

2 2.5 drops’ spacing, s = l − j. Equation (30) then rewrites 2 1.5 ND −1 ND −j−1 X X i(θj −θj+s) 1.5 n(x) = e ηs(x). (31) 1 j=0 s=−j 1 0.5 where ηs is the general s-overlap function, not accounting 0.5 for the drop’s individual phases but for their identical 0 0 shapes, and deﬁned via

-0.5 ∗ -0.5 ηs(x) = χ(x)χ (x − sd) . (32) (a) (c) -1 -1 The η0-function matches the individual drop density 3 6 profile, maximum at x = 0. The η1-function is the 2.5 5 nearest-neighbor overlap. Under the steep-|χ| assump- 2 tion (28), η1 is maximum at x ≈ d/2, with a value 4 η1(d/2) ≈ η0(d/2). Under this same approximation, the 1.5 3 ηs-functions are steeply decreasing in magnitude with s. 1 Therefore, at any x, in the sum of Eq. (31), one can ne- 2 glect the s >1-contributions, giving 0.5 1 0 n(x) ≈ n0(x) + n1(x). (33) 0 -0.5 The s= 0-contribution, n (x), matches the density of all (b) (d) 0 -1 -1 the individual drops, 0 1 2 3 4 5 0 1 2 3 4 5 N XD FIG. 3. Examples of SSSs based on the drop-array Ansatz, n0(x) = η0(x − jd). (34) defined by Eqs. (27), and (29), using λ = 0.2 (a,b) and λ = 0.1 j=1 (c,d), and a phase-coherent stationary configuration θj = 0. Here L = 5d is chosen. The panels and the color code are It dominates overall. As the steep-|χ| condition (28) ad- similar to those of Fig.1. (a,c) show ψ(x) (solid black line), ditionally implies that, at the drop j’s center, the con- and its general decomposition in ψu (dashed red line) and ψm tributions of all other drops can be neglected, n0(x) has (dotted blue line). (b,d) show the corresponding density pro- maxima at each x ≈ jd, approximately equaling η0(0), files and average values: the total density n(x), (thick black and minima at x ≈ (j + 1/2)d, approximately equaling line), its averagen ¯ (thin grey line), the uniform contribution 2η0(d/2), from the contributions of the two neighboring n (dashed red line), the average modulated contributionn ¯ u m drops. The s= 1-overlap contribution, n1(x), writes (thin blue line), and the effective profile nm (dotted blue line). −10 For a,b (c,d) we find 1 − C = 0.015 (1.1 · 10 ), fsf = 0.11 ND −1 −9 (7.6 · 10 ) and S/nL¯ = 0.71 (0.35) both from numerics and X i∆θj n1(x) = 2 Re e η1(x − jd) . (35) from the perturbative formulae, Eqs. (41),(48)-(50)). j=1

with ∆θj = θj − θj+1. The contribution n1(x) competes with n (x) only at x ≈ (j + 1/2)d. In absence of phase B. Spatial density in the drop array 0 profiles (phase-coherent stationary case), the two contributions have the same magnitude at these intermediate To express our physical observables C, fsf , and S for positions, n0((j + 1/2)d) ≈ n1((j + 1/2)d) ≈ 2η0(d/2). the drop-array model (27), we start, as in the previous The complex numbers ei∆θj coming into play in Eq.(35) sections, by writing the density profile. It generally is tend to make the overlap contribution reduce, and even counteract (by changing sign) the presence of n0(x), in N −1 between the neighboring drops (j,j +1) when their phase XD n(x) = ei(θj −θl)χ(x − jd)χ∗(x − ld). (30) difference increases. When the two drops are π-shifted, the total density n(x) between them approximately can- j,l=0 cels out as n1((j+1/2)d) ≈ −n0((j+1/2)d) ≈ −2η0(d/2). In presence of statistical phase fluctuations, this means The steep-|χ| condition (28) allows to simplify Eq. (30) by that the total density n(x) in between the drops reduces sorting the terms by magnitude orders for each position with the degree of phase coherence. of space, see also AppendixC for further details. By integrating Eq. (33) over x ∈ [0,L], one gets the The terms of the sum of Eq. (30) correspond to the mean densityn ¯. At leading perturbation order, it is wave-function overlaps for the drop couple (j, l) . These dominated by the s=0 contribution and, using peri- terms can be relevantly differentiated as a function of the odic boundary conditions (see discussion below Eq. (27), 9

L = (ND − 1)d and the array effectively contains ND − 1 It is reduced compared to the full-contrast value by two drops), it writes features of the array, (i) the relative value of the individual drop’s density remaining at half-distance from its (N − 1)N N n¯ ≈ D 1d = 1d . (36) neighboring drop, compared to its value at center, (ii) L d the degree of phase locking in the array, as encompassed by φ. We note that the degree of phase locking coming The Gaussian-drop model (29) enables to exemplify into play in the contrast value is a rather local one, as φ the above general derivations. In this case, is sensitive to phase differences in between neighboring −(x/d)2 − (x/d − s)2 drops only. η (x) = η (0) exp (37) s 0 2λ2 In the Gaussian-model case (29), the contrast reads √ with η0(0)=n/ ¯ πλ. (38) 1 C ≈ 1 − 4φ exp − . (41) 4λ2 The steep-|χ| condition (28) here simply writes 2 exp(−1/4λ ) 1. In the Gaussian-drop Ansatz, the and its difference to unity steeply decreases with decreas- drop’s extent-over-distance ratio λ is thus the parameter ing λ. This is exemplified in Fig.3 where 1 − C is re- controlling the drop’s steepness and the steep-drop con- duced by eight orders of magnitude from λ = 0.2 to < dition appears to be satisfied once λ ∼ 0.3. Examples λ = 0.1. The steep scaling law of the contrast with λ of Gaussian-drop-array density profiles are given in the for arrays of Gaussian drops is further illustrated in the lower panels of Fig.3. Matching the above description, semi-logarithmic plot of Fig.4(a). The approximate scal- the density profiles show maxima at the drop positions < ing law of Eq. (41) is observed up to λ ∼ 0.3. and minima in between. This is true both for the relatively broad drops (b) and for steeper ones (d). The density in between the drops sharply decreases from λ = 0.2 D. Superfluid fraction within the drop-array 2 to λ = 0.1, as expected from the exp(−1/4λ )-scaling Ansatz law of η0(d/2)/η0(0). To evaluate the superfluid density using Leggett’s formula [Eq. (9)], one has to consider the integral of the C. Density contrast in the drop array inverse density profile. Starting from Eqs. (33)-(35), the terms of the sums that effectively contribute to n(x) at Summarizing the above discussion, under the steep- a given x, depend on x. As under the steep-|χ| con- |χ| assumption (28), the total density, given by Eq. (33), dition (28), a drop does not contribute significantly to has maxima at x ≈ jd, approximately equaling n(jd) ≈ the density profile at distance larger than d of its center, η0(0), and local minima at x ≈ (j +1/2)d, approximately one can decompose the integral of Eq. (9) by segment equaling n((j + 1/2)d) ≈ 2(1 + cos ∆θj)η0(d/2) n(jd). x ∈ [jd, (j + 1/2)d]. Also using L = (ND − 1)d, Eq. (9) A first consequence of the above discussion is that, at writes leading order, the density profile is nearly fully con-  N −1 0 trasted, C ≈ 1; see also Fig.3(b,d). This qualifies the n¯ XD X f ≈ drop-array state as a deep supersolid under the steep-|χ| sf d(N − 1) condition (28), justifying the use of Eqs.(29)-(39) in the D j=0 s=−1 −1 remainder of Sec.IV. d/2  At next order, we should specify the overall minimum. Z dx  This is 2φη (d/2) where i∆θ . (42) 0 η0(x) + η0(x − d) + 2Re [e j+s η1(x)] 0 φ = 1 + min [cos ∆θj] (39) j Equation (42) includes the phase factors ei∆θj . Yet is a numerical factor, ranging between 0 and 2, and char- it should be kept in mind that, as discussed above acterizing the phase variations on the drop array. For a (Sec.IIB), fsf is physically meaningful only for a phase- i∆θj phase-locked array, θj ≡ 0, as relevant for a fully coher- coherent stationary array, i.e. e ≡ 1. ent steady state, φ = 2. For a strongly phase-modulated The x-range dominating the integral of Eq. (42) is array such that the phase of a drop can be π shifted where the density is the smallest, i.e. at x ≈ d/2. In compared to its neighbor, φ = 0. For what concerns this range, the denominator approximately equals 2(1 + fluctuations, as highlighted above, the strength of the cos ∆θj)η0(d/2). The spatial extent of the low-density fluctuations in the drops’ phases relates to the degree of region depends on the steepness of |χ|2 in this x-range. coherence that thus connects to a reduced φ. A rough estimate, setting this extent to the rms width The density contrast then reads λd andn ¯ ∼ η0(0)λ [see Eq. (38)], gives

η0(d/2) η0(d/2) 1 0 C ≈ 1 − 4φ . (40) fsf ∼ 2 φ , (43) η0(0) η0(0) λ 10

0 1 10 (a) (c) Using a Gaussian-drop model (29) enables us to go beyond this rough estimate. Here Eq. (42) yields -2 10 0.8 0 1 φ exp − 4λ2 −1 fsf ≈ 2 I , (46) 10-4 0.6 λ 1 √ Z 2λ π exp(u2)du -6 I = u . (47) 10 0.4 1 (φ − 1) + cosh − 2λ λ

-8 In the fraction of Eq. (46), one recognizes Eq. (43). The 10 0.2 inverse of the integral I gives the numerical prefac- -10 tor. For λ 1, I is well approximated by I ≈ 10 0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4 2π exp − 1 Erfi 1 and, at leading order in λ, I ≈ √ 4λ2 2λ 1 100 4 πλ. Here Erfi is the imaginary error function. There- (b) (d) fore, fsf rewrites -2 0.8 10 0 1 0 φ exp − 2 1 − C φ f ≈ √ 4λ ≈ √ . (48) sf 4 πλ3 λ3 16 πφ 0.6 10-4 In this Ansatz, |χ| scales very steeply and fsf vanishes 0.4 10-6 for decreasing λ, as shown in Fig.4(a), blue line. The scaling law of Eq. (48) describes well the numerical eval- < 0.2 10-8 uation of fsf from Leggett’s formula (9) up to λ ∼ 0.3. Furthermore, the scaling laws of both 1 − C and fsf are 0 -10 dominated by the exponential function, suggesting that 0 0.5 1 10 0.4 0.6 0.8 1 the two quantities scale together. Their relative scaling is illustrated in Fig.4(b). Here the approximation of FIG. 4. Illustration of the scaling laws of the SSS’s properties Eq. (48) holds quantitatively only for C → 1, evidencing within the Gaussian drop-array Ansatz [Eqs. (29),(27)] and the sensitivity to the exact prefactors. deep-supersolid approximation. The array is here assumed fully phase coherent. (a) remainder of the density contrast 1 − C and superfluid fraction as a function of the parameter E. High-energy scattering response within the controlling the drop steepness, λ. Numerically extracted val- drop-array Ansatz ues from the full density profiles are represented by the solid lines, red and blue, respectively. Approximate scaling laws from the steep-|χ| condition (Eqs.(41) and (48)) are given by We finally focus on high-energy scattering as it is the the black lines, dashed and dotted, respectively. (b) fsf versus observable of focus for the present paper. As shown in 1−C from numerical extractions (solid red line) and approxi- Sec.IIC [Eq. (14)], the resonant response factor S sim- mate scaling laws (dashed line). (c) scattering response S/nL¯ ply scales with the density of the uniform contribution as a function of λ from a numerical extraction (solid red line) from the decomposition (1). For a drop array of wave and approximate scaling law of Eq. (50) (dashed line). (d) function (27), the uniform field (2) writes scaling law of fsf versus 1 − C as a function of S/nL¯ , same N −1  colorcode as (a). XD Z dx ψ = eiθj χ(x) (49) u   L j=1 with using periodic boundary conditions, which imply that the 0 array effectively counts only N − 1 independent drops φ = 1 + hcos ∆θjij ≥ φ, (44) D and L = (ND − 1)d (see discussion below Eq. (27)). This being a numerical phase factor similar to φ [Eq.(39)], but yields the response factor here using the average over the array, h.ij, instead of the 2 0 R 2 ND −1 minimum. Similar to φ, the factor φ probes the local χ(uλd)du 1 X S ≈ nLλ¯ eiθj . (50) phase modulations on the array, i.e. at the level of neigh- R 2 |χ(uλd)| du ND − 1 boring drops. In the physically relevant phase-coherent j=1 0 stationary case, φ ≡ 2 and this factor can be discarded From Eq. (50), S can be decomposed into two contri- from the scaling law (43). butions, (i) a phase-locked contribution S0 and (ii) a re- The smallness of fsf depends on the steepness of the duction factor due to the phase modulations in the array, drop’s wave function: fsf is small only if |χ(x)| decreases ν . Precisely, > θ faster than 1/x for large x ∼ d/2. Using Eq. (40), one also finds an approximate scaling law with the density S ≈ S0νθ, with (51) R 2 contrast, S = λ | χ(uλd)du| nL,¯ (52) 0 R |χ(uλd)|2du 0 φ 1 − C 2 1 PND −1 iθj fsf ∼ 2 . (45) νθ = e . (53) φ λ ND −1 j=1 11

The phase-locked contribution is the only relevant one tions and deterministic modulations on the full extent of for the case of a phase coherent stationary array, where a strongly density-modulated state. as νθ encompasses additional effects from dynamics and fluctuations. On the one hand, the phase-locked contribution, S0, V. CONCLUSION is reduced compared to the unform valuenL ¯ by a factor which is sensitive to the arrangement of the drops in By describing supersolid states using a classical-field the array. Interestingly, this dependence is mostly via an approach, I derived relationships between two of their approximate linear scaling law with the localization pa- fundamental properties, namely the density contrast and rameter λ. This scaling law fundamentally differs from the superfluid fraction, and their resonant response to a those of the superfluid fraction and of the density con- high-energy scattering probe. I specified these relation- trast, see Eqs. (40)-(41). The latter scaling laws generally ships in two opposite regimes. involve η0(d/2)/η0(0), which depends√ on the steepness of In the case of a weakly modulated (”shallow”) su- |χ|. If |χ(x)| is steeper than 1/ x at large distances, persolid, the high-energy scattering response generally the superfluid fraction and of the density contrast scales scales approximately linearly with the superfluid frac- faster than S0 with λ. tion [Eq. (21)] and quadratically with the density con- To specify these statements, we consider the Gaussian- trast [Eq. (23)] . This establishes high-energy scattering drop model (29). Here the scaling laws for the superfluid spectroscopy as a sensitive probe of the spontaneous ap- fraction and density contrast are dictated by the factor pearance of density modulation in a superfluid. For a exp(−1/4λ2) [see Eqs. (45)-(48)], which is much steeper fully coherent stationary state and as long as fsf remains than the linear scaling of S0 in λ. The contribution S0 close to unity, the scattering response is additionally ex- can also be specified in this Gaussian case, pected to provide an estimate of the superfluid fraction. √ S = 2 πλ nL.¯ (54) In the case of a strongly modulated (”deep”) super- 0 solid, a drop-array Ansatz is used. In this case, the confirming the linear scaling of Eq. (52). As illustrated scaling laws of the physical properties depend on the in Fig.4(c), this scaling for S in the phase-locked case steepness of the drop’s wave function over the inter-drop < appears to hold up to λ ∼ 0.25. The much weaker scaling distance. Assuming Gaussian-shaped drops, the density law of S0 with λ compared to those of 1 − C and 1 − contrast and superfluid fraction vary very steeply with fsf is further illustrated by their relative plots in semi- the ratio λ of the drops’ extent to their distance, mostly logarithmic scale given in Fig.4(d). This is also visible as exp(−1/4λ2) [Eqs. (41), (48)]. In contrast, the scatter- in Fig.3, where nu, matching S0/L, remains significant ing response varies much more smoothly with this ratio, in highly contrasted drop arrays. roughly linearly [Eqs. (52), (54)]. Hence, in contrast to On the other hand, the phase-dependent contribution, the shallow-supersolid case, the sensitivity of the scat- νθ, brings an additional reduction factor which is sen- tering probe to density modulation or superfluid fraction sitive to the phase modulation in the array. Generally appears to be lost in the deep-supersolid regime. Yet, a speaking νθ ≤ 1, and νθ = 1 if and only if the ar- distinct feature of the scattering response is that it has an ray is in a fully coherent stationary configuration (phase acute sensitivity to the global modulations of the phase locked). The comparison of Eq. (53) with Eqs. (39)-(40) of the deep supersolid [Eq. (53)]. This makes high-energy shows that the phase information contained in the scat- scattering spectroscopy a promising path to investigate tering response is more global than that contained in the both coherent dynamics and fluctuation effects in super- other quantities considered up to now. In φ (φ0), which solid systems. sets the sensitivity of C (fsf ) to the phase modulation in In our complementary work of Ref. [79], we report on the array, only the phase difference between neighboring a first study of dipolar gases across the transitions from drops are considered. In contrast, νθ measures the ac- regular superfluid to supersolid and to incoherent crys- tual length of the full phase vector over the drop array tal based on a high-energy two-photon Bragg-scattering in comparison to the fully stretched configuration. By technique. The simple models developed in the present expanding the norm, Eq. (53) rewrites work are there applied and compared to both eMF-theory and experimental results. They are found to provide an 1 PND −1 intuitive understanding of the physics at play, and to νθ = (N −1)2 j,l=1 cos(θj − θl), (55) D show a quantitative predictive power. They are crucial where one sees that the phase differences between all drop to not only understand but also demonstrate the central couples come into play regardless of their distances. As role played by the phase effects induced by the dynamical highlighted above (see e.g. Secs.IIA,IVA), the phase dif- crossing of the superfluid-to-supersolid phase transition ferences between drops can be deterministic, e.g. when in experiments. arising from coherent dynamics, or of statistical charac- By figuring out and disentangling the sensitivity of ter when connecting to thermal or quantum fluctuations the high-energy scattering response to both density and of the state. The high-energy scattering response thus phase modulations occurring in self-modulated quan- appears as a highly sensitive probe of the phase fluctua- tum states, the present work ultimately establishes high- 12 energy scattering spectroscopy as an exquisite probe of easily exemplified using a real field with a pure amplitude the characteristic properties of supersolid states. High- modulation such as the one used in our sine-modulation energy scattering is readily achievable in a variety of sys- Ansatz of Sec.IIIB. Using a fully contrasted modula- tems [45–53] and, in particular, in ultracold quantum tion with a wavelength matching the system length, i.e. √ gases, via two photon Bragg scattering [58–65, 79]. This R L ψ(x) = 2¯n cos(2πx/L), we find 0 ψ(x)dx = 0, yield- makes our study relevant for the dipolar supersolids and ing ψu = 0, nu = 0 and ψm = ψ,n ¯m =n ¯. Therefore, beyond. Furthermore, by deriving general and simple a pure amplitude modulation may yield a state with no relations, the present work provides a hopefully useful uniform component. framework for future studies of such intriguing states, in For the effect of phase modulations, one can con- any system, as long as they can be pictured within a sider yet another Cauchy-Schwarz inequality, which reads classical-field theory. R L R L | 0 ψ(x)dx| ≤ 0 |ψ(x)|dx. As the terms of the inequality match ψuL either for the full field (left) or for its mere amplitude (right), one sees that the presence ACKNOWLEDGMENTS of phase modulations in ψ yields a further decrease of nu compared to the pure amplitude modulated case. I thank Francesca Ferlaino, Daniel Petter, Danny To exemplify the striking effect of phase modulations, Bailly, Blair Blakie, Mikhail Baranov, Alessio Recati, let’s consider an exemplary wave function with a pure Gabriele Natale, Alexander Patscheider, and Manfred J. phase modulation making a full rotation√ over the sys- Mark for stimulating discussions and comments on the tem’s length. This writes ψ(x) = n¯ exp(i2πx/L). We manuscript. I thank especially Blair Blakie and Mikhail R L then find ψ(x)dx = 0 such that ψu = 0, nu = 0 and Baranov for many insightful explanations. I thank Danny 0 ψm = ψ,n ¯m =n ¯. Therefore, a pure phase modulation Baillie for a careful check of my derivations. I thank can also transfer a state completely into its modulating Philippe Chomaz for enriching comments. This work is field. financially supported through a DFG/FWF grant (FOR 2247/PI2790), a joint RSF/FWF grant (I 4426), and the FWF Elise Richter Fellowship number V792. Appendix B: Full expression of the resonant * Correspondence and requests for materials should be high-energy scattering response addressed to [email protected]. SectionIIC describes the SSS’s response to a high- energy scattering probe in the linear and weakly inter- Appendix A: Respective roles of amplitude and acting regime. If the elementary excitations dictating phase modulations in the field decomposition. the response can be well approximated to plane waves, Eq. (12) gives the DSF values, S(k, ω), and Eq. (13) its The present work describes a SSS via a single macro- resonant value. In the main text, details on the derivation scopic field with both amplitude and phase modulations, of the prefactor were skipped to focus on the information see discussion in Sec.IIA. Equations. (1)-(7) specify a de- related to the state’s wave function. In this appendix, composition of the field into uniform and modulated com- I provide an additional description to derive the full ex- ponents, ψ(x) = ψu + ψm(x), which also yield a simple pression of the resonant response, Eq. (13). decomposition of the mean densityn ¯ = nu +n ¯m. In this Here, I consider again a system well described by an appendix, I illustrate the respective roles of the ampli- effective 1D model and confined along x in a uniform tude and phase modulations in these decompositions. box of size L, see also Sec.IIA. In this case, the plane- First, we show that any source of modulation yields a wave basis is discrete, with the allowed momentum values 2 2 non-zero ψm. Indeed, using the Cauchy-Schwarz equal- beinghk ¯ j = hj/L, j ∈ Z. I note j = ¯h kj /2m the R L 2 R L free-particle (kinetic) energy associated to the plane wave ity condition for the inequality | 0 ψ(x)dx| ≤ 0 dx · R L 2 of index j, with m being the particle’s mass. In the 0 |ψ(x)| dx, one finds that nu =n ¯ only if ψ is colin- ear with a constant function over space, i.e. if ψ is uni- regime of validity of Eqs. (12)-(13), the response of the form. This means that once a modulation, of any kind, is system to a scattering probe is dictated by elementary present, the density in the uniform component is reduced excitations. If the scattering probe is of large-enough compared ton ¯. Below, I detail the respective effects of momentum ¯hk, then the elementary excitation yielding amplitude and phase modulations in this reduction. a response is well approximated by the plane wave of index j ≈ kL/(2π)[89]. The energy of this elementary The effects of amplitude modulations can be seen excitation readshω ¯ ≈ +nV ¯ (k ) wherenV ¯ (k ) is a from the Cauchy-Schwarz inequality | R L |ψ(x)|dx|2 ≤ j j j j 0 mean-field interaction shift. In this case, the regime of R L 2 2 L 0 |ψ(x)| dx. As the right-hand side matchesnL ¯ , high-energy scattering where the impulse approximation 2 and the left-hand side nuL while only considering the is valid is defined bynV ¯ (kj) ¯hωj, j. In the case of amplitude of ψ, it implies that the presence of ampli- dipolar gases, the interaction term, V (k), can in general tude modulations indeed results in a decrease of nu with depend on k. For the dipolar system considered here, respect ton ¯. The role of amplitude modulation can be effective expression of V (k) has been derived assuming an 13 effective 1D model with transverse wave functions ξ(y, z) symmetry of |χ(x)| and its monotonic character for x > 0, of Gaussian shapes, see refs. [73, 90, 91]. Remarkably, for we deduce two properties: (i), |ηs(x)| ≤ |χ(0)χ(sd/2)| for large-enough k, V (k) was found to recover a momentum- any x; (ii), if |χ(0)χ∗(sd)| |χ(sd/2)|2, the maximum independent character with V (k) ∼ V∞. In this regime amplitude of ηs is found at x ≈ sd/2 and it approxi- ∗ and assumingnV ¯ ∞ ¯hωj, j, the Dirac factor appearing mately equals η0(sd/2). By assuming (B), |χ(0)χ (d)| in Eq. (12) can then be simply written as: |χ(d/2)|2, one deduces from the properties (i) and (ii) that h2j2 δ(¯hω − ¯hω ) = δ ¯hω − nV¯ − (B1) j ∞ 2mL2 1. |ηs(x)| η1(x), η0(x), for all x ∈ [−d/2, d/2] and 2πmL s ≥ 2; = 2 δ (jω − j) . (B2) h |kω|

L p 2. η1(x) ≈ η0(x) for x ≈ (j + 1/2)d, which is where η1 with jω = h 2m(¯hω − nV¯ ∞) and kω = 2πjω/L. The summation of the factors (B2) over j gives the density is approximately maximum, and η1(x) η0(x) for P 2πmL x ≈ jd. of excited states at energy ¯hω, δ(¯hω − ¯hωj) = 2 . j h kω It matches the density of free-particle states in 1D taken at the mean-field-shifted energyhω ¯ − nV¯ ∞. Combining The result (1), combined with the above consideration this expression with Eq. (12) yields Eq. (13) for the DSF on n0, implies that all terms with s ≥ 2 can be neglected value on resonance with the state j (k = kj, ω = ωj). in Eq. (31), yielding Eq. (33). The result (2) implies the In conclusion, the dependency of the resonant scattering dominant character of the s=0-contribution to the den- response on the probed state at high momentum is con- sity, and the deep-supersolid character of the drop array, tained in the factor S of Eq. (13) whereas the density-of- C ≈ 1. Result (2) also determines the exact strength state factor yields an additional momentum dependence. of the density minima of the SSS, and thus the values of the density contrast and of the superfluid fraction, Eqs. (40), (43). Appendix C: Density contribution in the drop-array In conclusion, the steep-|χ| assumption needed to draw Ansatz the conclusion given in the main text combines the conditions (A) and (B). The condition (B) rewrites In Sec.IVB, an approximate expression of the density as |χ(d)|/|χ(d/2)| |χ(d/2)|/|χ(0)|, matching the profile in the drop array is derived under the assumption first inequality of Eq. (28), which defines the steep-|χ| that the drops have a wave function that is steep enough condition. The last inequality appearing in Eq. (28), over the range set by their half-distance. In the present |χ(d/2)|/|χ(0)| 1, is not exactly needed but it is suf- Appendix, I detail the considerations that enable to sim- ficient to imply (A) to be satisfied. We thus choose the plify the density profile n(x) as Eq. (33). We see how condition Eq. (28) as the relevant one to have both con- these considerations yield the formulation of the steep- ditions (A) and (B) simultaneously satisfied. |χ| condition as Eq. (28). In the last paragraph of this appendix, we finally As stated in the main text, a primary step is to dif- illustrate the general considerations above, using the ferentiate the terms of the sum appearing in the general Gaussian-drop model (29). Using this model, the ηs- definition of the density, Eq. (30), as a function of the functions take simple forms, given by Eq. (37). By per- spacing, s = l − j, between the two drops identified by forming a simple spatial derivation, the couple of sum indexes (j, l). This yields the expression Eq. (31) where the s-overlap functions, η , defined dη 2x − sd s s (x) = − · η (x), (C1) in Eq.(32), come into play. dx λ2d2 s To simplify Eq. (31) into Eq. (33), let us first focus on the s = 0 contribution. Its overall value, n0(x), is given and one finds that the maximum amplitudes of ηs(x) are by Eq. (34), matching the sum of the individual drop indeed found at x = sd/2 and equaling η0(sd/2). The density profiles. As |χ(x)| is monotonically decreasing n0(x) minima are similarly found at x = d(j + 1/2) 2 2 for x > 0, if one assumes (A) |χ(d/2)| |χ(0)| , one and equaling 2η0(d/2). Combining these results with can neglect the contributions of all other drops at one Eq. (37) implies that the exponential factor exp(−1/4λ2) drop j’s center. In this way, one finds that n0(x) is fixes the ratio of the maximum amplitudes of |ηs| and structured as described in the main text. It has max- |ηs−1|. It also gives the ratio between the minima and ima at each drop center, x ≈ jd, approximately equal- maxima of n0(x) up to a factor 2. Finally, the steep- ing η0(0). Additionally, it has minima in between the |χ| condition (28) also writes as a function of this same drop, at x ≈ d(j + 1/2), with an approximate value of factor, simply as exp(−1/4λ2) 1. Therefore, under 2η0(d/2), from the contributions of the two neighboring the steep-|χ| condition, in the Gaussian-drop Ansatz, all drops, which is furthermore negligible compared to the s¿1 contributions are indeed found to be negligible and maximum value. n1 only compete with n0 at the intermediate positions Differently, in the case s 6= 0, the maximum amplitude x ≈ d(j + 1/2). This example corroborates the general of ηs(x) is found at an intermediate x in [0, sd]. Using the considerations made above and given in the main text. 14

[1] O. Penrose and L. Onsager, Phys. Rev. 104, 576 (1956). (1961). [2] E. P. Gross, Phys. Rev. 106, 161 (1957). [32] L. Pitaevskii and S. Stringari, Bose-Einstein condensa- [3] A. J. Leggett, Phys. Rev. Lett. 25, 1543 (1970). tion and superfluidity, Vol. 164 (Oxford University Press, [4] A. Andreev and I. Lifshits, Zhur Eksper Teoret Fiziki 56, Oxford, 2016). 2057 (1969), [Sov. Phys. JETP 29, 1107 (1969)]. [33] F. Wächtler and L. Santos, Phys. Rev. A 93, 061603(R) [5] G. V. Chester, Phys. Rev. A 2, 256 (1970). (2016). [6] D. Kirzhnits and Y. A. Nepomnyashchii, Zhur Eksper [34] R. N. Bisset, R. M. Wilson, D. Baillie, and P. B. Blakie, Teoret Fiziki 59, 2203 (1970), [Sov. Phys. JETP 32, 1191 Phys. Rev. A 94, 033619 (2016). (1971)]. [35] F. Wächtler and L. Santos, Phys. Rev. A 94, 043618 [7] T. Schneider and C. P. Enz, Phys. Rev. Lett. 27, 1186 (2016). (1971). [36] L. Chomaz, S. Baier, D. Petter, M. J. Mark, F. Wächtler, [8] S. Balibar, Nature (London) 464, 176 (2010). L. Santos, and F. Ferlaino, Phys. Rev. X 6, 041039 [9] A. B. Kuklov, N. V. Prokofev, and B. V. Svistunov, (2016). Physics 4, 109 (2011). [37] I. Ferrier-Barbut, H. Kadau, M. Schmitt, M. Wenzel, [10] M. Boninsegni and N. V. Prokof’ev, Rev. Mod. Phys. 84, and T. Pfau, Phys. Rev. Lett. 116, 215301 (2016). 759 (2012). [38] G. Natale, R. M. W. van Bijnen, A. Patscheider, D. Pet- [11] V. I. Yukalov, Physics 2, 49 (2020). ter, M. J. Mark, L. Chomaz, and F. Ferlaino, Phys. Rev. [12] N. Henkel, R. Nath, and T. Pohl, Phys. Rev. Lett. 104, Lett. 123, 050402 (2019). 195302 (2010). [39] L. Tanzi, S. Roccuzzo, E. Lucioni, F. Famà,A. Fioretti, [13] F. Cinti, P. Jain, M. Boninsegni, A. Micheli, P. Zoller, C. Gabbanini, G. Modugno, A. Recati, and S. Stringari, and G. Pupillo, Phys. Rev. Lett. 105, 135301 (2010). Nature 574, 382 (2019). [14] Z.-K. Lu, Y. Li, D. S. Petrov, and G. V. Shlyapnikov, [40] M. Guo, F. Böttcher, J. Hertkorn, J.-N. Schmidt, Phys. Rev. Lett. 115, 075303 (2015). M. Wenzel, H. P. Büchler, T. Langen, and T. Pfau, [15] J.-R. Li, J. Lee, W. Huang, S. Burchesky, B. Shteynas, Nature (London) 574, 386389 (2019). F. Ç. Top, A. O. Jamison, and W. Ketterle, Nature [41] J. Hertkorn, F. Böttcher, M. Guo, J. N. Schmidt, T. Lan- (London) 543, 91 (2017). gen, H. P. Büchler, and T. Pfau, Phys. Rev. Lett. 123, [16] J. Léonard,A. Morales, P. Zupancic, T. Esslinger, and 193002 (2019). T. Donner, Nature (London) 543, 87 (2017). [42] L. Tanzi, J. G. Maloberti, G. Biagioni, A. Fioretti, [17] A. Macia, J. Sánchez-Baena, J. Boronat, and F. Maz- C. Gabbanini, and G. Modugno, (2019), zanti, Phys. Rev. Lett. 117, 205301 (2016). arXiv:1912.01910 [cond-mat.quant-gas]. [18] F. Cinti and M. Boninsegni, Phys. Rev. A 96, 013627 [43] S. M. Roccuzzo, A. Gallem´ı,A. Recati, and S. Stringari, (2017). Phys. Rev. Lett. 124, 045702 (2020). [19] M. Wenzel, F. Böttcher, T. Langen, I. Ferrier-Barbut, [44] P. Ilzhfer, M. Sohmen, G. Durastante, C. Politi, and T. Pfau, Phys. Rev. A 96, 053630 (2017). A. Trautmann, G. Morpurgo, T. Giamarchi, L. Chomaz, [20] D. Baillie and P. B. Blakie, Phys. Rev. Lett. 121, 195301 M. J. Mark, and F. Ferlaino, (2019), arXiv:1912.10892 (2018). [cond-mat.quant-gas]. [21] L. Chomaz, R. M. W. van Bijnen, D. Petter, G. Faraoni, [45] H. Palevsky, K. Otnes, K. E. Larsson, R. Pauli, and S. Baier, J. H. Becher, M. J. Mark, F. Wächtler, L. San- R. Stedman, Phys. Rev. 108, 1346 (1957). tos, and F. Ferlaino, Nat. Phys. 14, 442 (2018). [46] D. G. Henshaw, Phys. Rev. Lett. 1, 127 (1958). [22] S. M. Roccuzzo and F. Ancilotto, Phys. Rev. A 99, [47] J. L. Yarnell, G. P. Arnold, P. J. Bendt, and E. C. Kerr, 041601(R) (2019). Phys. Rev. 113, 1379 (1959). [23] L. Tanzi, E. Lucioni, F. Famà,J. Catani, A. Fioretti, [48] M. Breidenbach, J. I. Friedman, H. W. Kendall, E. D. C. Gabbanini, R. N. Bisset, L. Santos, and G. Modugno, Bloom, D. H. Coward, H. DeStaebler, J. Drees, L. W. Phys. Rev. Lett. 122, 130405 (2019). Mo, and R. E. Taylor, Phys. Rev. Lett. 23, 935 (1969). [24] F. Böttcher, J.-N. Schmidt, M. Wenzel, J. Hertkorn, [49] R. E. Taylor, Rev. Mod. Phys. 63, 573 (1991). M. Guo, T. Langen, and T. Pfau, Phys. Rev. X 9, 011051 [50] H. W. Kendall, Rev. Mod. Phys. 63, 597 (1991). (2019). [51] J. I. Friedman, Rev. Mod. Phys. 63, 615 (1991). [25] L. Chomaz, D. Petter, P. Ilzhöfer,G. Natale, A. Traut- [52] A. Griffin, Excitations in a Bose-condensed Liquid, Cam- mann, C. Politi, G. Durastante, R. M. W. van Bijnen, bridge Classical Studies (Cambridge University Press, A. Patscheider, M. Sohmen, M. J. Mark, and F. Fer- New York, 1993). laino, Phys. Rev. X 9, 021012 (2019). [53] C. GIACOVAZZO, H. L. MONACO, D. VITERBO, [26] L. Santos, G. V. Shlyapnikov, and M. Lewenstein, Phys. F. SCORDARI, G. GILLI, G. ZANOTTI, and Rev. Lett. 90, 250403 (2003). M. CATTI, Fundamentals of crystallography (OXFORD [27] D. H. J. O’Dell, S. Giovanazzi, and G. Kurizki, Phys. UNIVERSITY PRESS, 1992). Rev. Lett. 90, 110402 (2003). [54] J. Stenger, S. Inouye, A. P. Chikkatur, D. M. Stamper- [28] Y. Kora and M. Boninsegni, Journal of Low Temperature Kurn, D. E. Pritchard, and W. Ketterle, Phys. Rev. Physics 197, 337347 (2019). Lett. 82, 4569 (1999). [29] N. N. Bogoliubov, J. Phys. USSR 11, 23 (1947). [55] D. M. Stamper-Kurn, A. P. Chikkatur, A. Görlitz,S. In- [30] T. D. Lee, K. Huang, and C. N. Yang, Phys. Rev. 106, ouye, S. Gupta, D. E. Pritchard, and W. Ketterle, Phys. 1135 (1957). Rev. Lett. 83, 2876 (1999). [31] E. P. Gross, Il Nuovo Cimento (1955-1965) 20, 454 [56] J. Steinhauer, R. Ozeri, N. Katz, and N. Davidson, Phys. 15

Rev. Lett. 88, 120407 (2002). [74] P. C. Hohenberg and P. M. Platzman, Phys. Rev. 152, [57] R. Ozeri, N. Katz, J. Steinhauer, and N. Davidson, Rev. 198 (1966). Mod. Phys. 77, 187 (2005). [75] V. F. Sears, E. C. Svensson, P. Martel, and A. D. B. [58] S. Richard, F. Gerbier, J. H. Thywissen, M. Hugbart, Woods, Phys. Rev. Lett. 49, 279 (1982). P. Bouyer, and A. Aspect, Phys. Rev. Lett. 91, 010405 [76] T. R. Sosnick, W. M. Snow, P. E. Sokol, and R. N. Silver, (2003). Europhysics Letters (EPL) 9, 707 (1989). [59] T. Stöferle, H. Moritz, C. Schori, M. Köhl, and [77] F. Zambelli, L. Pitaevskii, D. M. Stamper-Kurn, and T. Esslinger, Phys. Rev. Lett. 92, 130403 (2004). S. Stringari, Phys. Rev. A 61, 063608 (2000). [60] S. B. Papp, J. M. Pino, R. J. Wild, S. Ronen, C. E. [78] J. Hofmann and W. Zwerger, Phys. Rev. X 7, 011022 Wieman, D. S. Jin, and E. A. Cornell, Phys. Rev. Lett. (2017). 101, 135301 (2008). [79] D. Petter, A. Patscheider, G. Natale, M. J. Mark, M. A. [61] E. D. Kuhnle, H. Hu, X.-J. Liu, P. Dyke, M. Mark, P. D. Baranov, R. v. Bijnen, S. M. Roccuzzo, A. Recati, Drummond, P. Hannaford, and C. J. Vale, Phys. Rev. B. Blakie, D. Baillie, L. Chomaz, and F. Ferlaino, (2020), Lett. 105, 070402 (2010). arXiv:2005.02213 [cond-mat.quant-gas]. [62] N. Fabbri, D. Clément, L. Fallani, C. Fort, and M. In- [80] L. Salasnich, A. Parola, and L. Reatto, Phys. Rev. A guscio, Phys. Rev. A 83, 031604(R) (2011). 65, 043614 (2002). [63] S. Hoinka, M. Lingham, K. Fenech, H. Hu, C. J. Vale, [81] C. Josserand, Y. Pomeau, and S. Rica, The European J. E. Drut, and S. Gandolfi, Phys. Rev. Lett. 110, 055305 Physical Journal Special Topics 146, 47 (2007). (2013). [82] N. Sepúlveda, C. Josserand, and S. Rica, Phys. Rev. B [64] C. Carcy, S. Hoinka, M. G. Lingham, P. Dyke, C. C. N. 77, 054513 (2008). Kuhn, H. Hu, and C. J. Vale, Phys. Rev. Lett. 122, [83] P. Blakie, A. Bradley, M. Davis, R. Ballagh, and C. Gar- 203401 (2019). diner, Advances in Physics 57, 363 (2008). [65] R. Lopes, C. Eigen, A. Barker, K. G. H. Viebahn, [84] Starting from an arbitrary phase choice, the appropriate M. Robert-de Saint-Vincent, N. Navon, Z. Hadzibabic, correction would be made by subtracting a uniform phase and R. P. Smith, Phys. Rev. Lett. 118, 210401 (2017). matching the argument of R ψ(x)dx. [66] L. D. Landau, J. Phys. (Moscow) 5, 71 (1941). [85] D. Pines and P. Nozières, The Theory of Quantum Liq- [67] L. D. Landau, J. Phys. (Moscow) 11, 91 (1947). uids V. 1: Normal Fermi Liquids (Benjamin, 1966). [68] R. P. Feynman, Rev. Mod. Phys. 29, 205 (1957). [86] L. Van Hove, Phys. Rev. 95, 249 (1954). [69] A. F. Andreev and I. M. Lifshitz, Sov. Phys. JETP 29, [87] P. B. Blakie, R. J. Ballagh, and C. W. Gardiner, Phys. 1107 (1969). Rev. A 65, 033602 (2002). [70] Y. Pomeau and S. Rica, Phys. Rev. Lett. 72, 2426 (1994). [88] R. Fazio and H. [van der Zant], Physics Reports 355, 235 [71] S. Saccani, S. Moroni, and M. Boninsegni, Phys. Rev. (2001). Lett. 108, 175301 (2012). [89] I note that the resonant character of the response is also [72] T. Macr`ı,F. Maucher, F. Cinti, and T. Pohl, Phys. Rev. discretized and possible only if kL/(2π) ∈ Z. A 87, 061602(R) (2013). [90] F. Deuretzbacher, J. C. Cremon, and S. M. Reimann, [73] P. B. Blakie, D. Baille, L. Chomaz, and F. Ferlaino, Phys. Rev. A 81, 063616 (2010). arXiv:2004.12577 (2020). [91] S. Sinha and L. Santos, Phys. Rev. Lett. 99, 140406 (2007).