Deconfined quantum criticality on a triangular Rydberg array Lisa Bombieri, Torsten V. Zache, Gabriele Calliari, Mikhail D. Lukin, Hannes Pichler, Daniel González-Cuadra Abstract Fluctuations can drive continuous phase transitions between two distinct ordered phases—so-called Deconfined Quantum Critical points (DQCPs)—which lie beyond the Landau-Ginzburg-Wilson paradigm. Despite several theoretical predictions over the past decades, experimental evidence of DQCPs remains elusive. We show that a DQCP can be explored in a system of Rydberg atoms arranged on a triangular lattice and coupled through van der Waals interactions. Specifically, we investigate the nature of the phase transition between two ordered phases at 1 / 3 1/3 and at 2 / 3 2/3 Rydberg excitation density, which were recently probed experimentally in [P. Scholl et al., Nature 595, 233 (2021)]. Using a field-theoretical analysis, we predict both the critical exponents for infinitely long cylinders of increasing circumference and the emergence of a conformal field theory near criticality showing an enlarged U( 1 1 ) symmetry—a signature of DQCPs—and confirm these predictions numerically. Finally, we extend these results to ladder geometries and show how the emergent U( 1 1 ) symmetry could be probed experimentally using finite tweezer arrays. Introduction.— Rydberg atoms trapped in optical tweezers Browaeys and Lahaye 2020 ; Kaufman and Ni 2021 have emerged in recent years as a powerful platform for quantum simulation Altman et al. 2021 , enabling the exploration of quantum many-body physics in a highly controllable setting Labuhn et al. 2016 ; Bernien et al. 2017 ; de Léséleuc et al. 2019 ; Keesling et al. 2019 ; Ebadi et al. 2021 ; Scholl et al. 2021 ; Bluvstein et al. 2021 ; Semeghini et al. 2021 ; Choi et al. 2023 ; Chen et al. 2023 ; Shaw et al. 2024 ; Manovitz et al. 2025 ; González-Cuadra et al. 2025 ; Fang et al. 2024 . In particular, the ability to arrange atoms in arbitrary geometries, combined with long-range Rydberg interactions, makes these systems especially well-suited for investigating frustrated spin models Sachdev 2023 . These models are characterized by many competing classical configurations, a setting where quantum fluctuations can give rise to long-range entangled states Wen 2007 , as exemplified by the recent experiments probing a topological spin-liquid phase Semeghini et al. 2021 . The interplay between competing orders and quantum fluctuations can also lead to deconfined quantum critical points (DQCPs) Senthil et al. 2004a ; Senthil et al. 2004b ; Shao et al. 2016 , which correspond to continuous phase transitions between two distinct ordered phases, and that lie beyond the conventional Landau-Ginzburg-Wilson (LGW) paradigm Landau and Lifshitz 1980 ; Wilson and Kogut 1974 . DQCPs are characterized by emergent fractionalized excitations and deconfined gauge fields and have been predicted in a variety of models Senthil 2024 . Despite significant theoretical and numerical efforts over the past decades, experimental evidence for DQCPs remains scarce Zayed et al. 2017 ; Guo et al. 2020 ; Hong et al. 2022 ; Cui et al. 2023 . Identifying experimental platforms capable of probing this phenomenon directly is, therefore, of great interest. In this work, we show that deconfined quantum criticality can be explored in a system of Rydberg atoms arranged on a triangular lattice and coupled via van der Waals interactions. Such systems have already been realized experimentally Scholl et al. 2021 , where the ordered phases at 1/3 and at 2/3 filling were adiabatically prepared. Although the existence of an intermediate phase driven by an order-by-disorder mechanism was initially proposed based on analogous results for the Ising model Moessner et al. 2000 ; Humeniuk 2016 ; Saadatmand et al. 2018 ; Koziol et al. 2019 ; Fey et al. 2019 , the nature of the ground state in this region remains an open question. For a point in parameter space lying in between the two ordered phases, Ref. Guo et al. 2023 found a finite-temperature Kosterlitz-Thouless (KT) phase—characterized by an emergent U( 1 ) symmetry. At zero temperature, however, it is unclear whether the system hosts an intermediate phase induced by the order-by-disorder mechanism or instead features a first- or continuous phase transition. In this work, we address this question and connect the emergent U( 1 ) symmetry to a DQCP—a continuous phase transition—separating the two distinct ordered phases observed in Ref. Scholl et al. 2021 . To this end, we first study the model on infinitely long cylinders of increasing circumference. Combining field-theoretical analysis with large-scale density-matrix renormalization group (DMRG) simulations Schollwöck 2011 , we demonstrate the presence of a DQCP in the ground-state phase diagram of these quasi-1D systems. As in previous studies of DQCPs in 1D Jiang and Motrunich 2019 ; Roberts et al. 2019 ; Mudry et al. 2019 ; Huang et al. 2019 ; Zhang and Levin 2023 ; Lee et al. 2023 ; Romen et al. 2024 ; Baldelli et al. 2024 , the critical point is described by a conformal field theory (CFT) with central charge c=1 and an emergent U( 1 ) symmetry. We interpret the cylinder circumference as an inverse temperature of the corresponding 2D model and, using a dimensional reduction argument, show that our results are consistent with a DQCP at low temperatures. Finally, we extend our results to quasi-1D ladder geometries and describe how the emergent U( 1 ) symmetry could be directly probed in more complex 2D systems experimentally via measurements in the occupation basis. Figure 1: (a) Neutral atoms trapped on a triangular tweezer array with lattice vectors {\bf a}_{x} and {\bf a}_{y} , where we depict the laser-coupled electronic levels. (b) Ground-state phase diagram for a cylinder with N_{x}=15 atoms along the x -axis and N_{y}=9 atoms along the periodic y -axis. (Top) Average occupation number \langle n_{\bf j}\rangle in the 1/3 phase ( \Delta/U=1.5 ), at the DQCP [ (\Delta/U)_{\rm c}=3.158 ], and in the 2/3 phase ( \Delta/U=5 ). (Bottom) The 1/3 ( 2/3 ) phase spontaneously breaks the \mathbb{Z}_{3} symmetry: m=|m|e^{i\phi} can point along \phi=0,2\pi/3,4\pi/3 ( \phi=\pi/3,\pi,5\pi/3 ), corresponding to a real order parameter \langle m^{3}+m^{3\dagger}\rangle=+1(-1) . At the transition, the \mathbb{Z}_{6} symmetry is enlarged to U( 1 ). The histograms show the distribution of m [Eq. ( 2 )] over 10^{4} snapshots in the occupation basis when restricting the sites to the ‘bulk’, i.e., excluding the first and the last three rows of atoms along {\bf a}_{x} . Results obtained using a finite MPS with bond dimension \chi=100 for \Delta/U=1.5 and \Delta/U=5 , and \chi=500 for \Delta/U=3.158 . Model and results.— We consider neutral atoms trapped in optical tweezers arranged on a triangular lattice with basis vectors {\bf a}_{x}=(\sqrt{3}/2,1/2)a and {\bf a}_{y}=(0,1)a , where a is the lattice constant [Fig. 1 (a)]. The atoms are driven by a coherent laser field with Rabi frequency \Omega and detuning \Delta , which couples two internal electronic states—the ground state \ket{g} and a highly excited Rydberg level \ket{r} . This system is described by the following Hamiltonian Scholl et al. 2021 : H_{\rm Ryd}=\frac{\Omega}{2}\sum_{{\bf j}}\sigma^{x}_{{\bf j}}-\Delta\sum_{{\bf j}}n_{{\bf j}}+\sum_{{\bf i}>{\bf j}}U_{{\bf i}{\bf j}}\,n_{{\bf i}}n_{{\bf j}}, (1) where n_{{\bf j}}=\ket{r}_{{\bf j}}\bra{r} , \sigma^{x}_{{\bf j}}=\ket{r}_{{\bf j}}\bra{g}+\ket{g}_{{\bf j}}\bra{r} , and U_{{\bf i}{\bf j}}=C_{6}/|x_{\bf i}-x_{\bf j}|^{6} is the van der Waals interaction strength between excited atoms at positions x_{\bf i} and x_{\bf j} . In the following, we indicate the number of atoms along {\bf a}_{x} ( {\bf a}_{y} ) with N_{x} ( N_{y} ), and the interaction strength between nearest neighbors with U . In this work, we study the phase transition between the ordered phases at 1/3 and 2/3 filling Scholl et al. 2021 —hereafter referred to as the 1/3 and 2/3 phases for brevity—characterized by a non-zero staggered magnetization m=\frac{1}{N/3}\sum_{{\bf j}}n_{{\bf j}}\,e^{i{\bf Q}\cdot x_{{\bf j}}}, (2) with {\bf Q}=\left(2\sqrt{3}/3,2/3\right)\pi/a , and x_{\bf j} the position of the \bf{j} -atom. In the 1/3 phase, the complex order parameter m=|m|e^{i\phi} can point along \phi=0,2\pi/3, or 4\pi/3 , while in the 2/3 phase, it can point along \phi=\pi/3,\pi, or 5\pi/3 [Fig. 1 (b)]. The two ordered phases thus correspond to a spontaneous symmetry breaking of two distinct \mathbb{Z}_{3} symmetries. In our work, to distinguish between the two, we use the real-valued order parameter \langle m^{3}+m^{3\dagger}\rangle=|m|^{3}\cos{(3\phi)} , which takes positive and negative values in the 1/3 and 2/3 phases. Since these phases are related by a \mathbb{Z}_{2} inversion symmetry ( \ket{g}\to\ket{r} ), a transition is expected at the point of degeneracy (\Delta/U)_{\rm c} , where the system exhibits a \mathbb{Z}_{3}\times\mathbb{Z}_{2}\simeq\mathbb{Z}_{6} symmetry. In the recent study of Ref. Guo et al. 2023 , the authors focus on the critical point (\Delta/U)_{\rm c} at fixed \Omega/U\approx 0.33 , analyzing the phase transition as a function of temperature. They show that thermal fluctuations render the \mathbb{Z}_{6} anisotropy irrelevant, resulting in an extended KT phase with quasi-long-range order and emergent continuous U( 1 ) symmetry, separated from the high-temperature disordered phase by a KT transition. In our work, we focus instead on the zero-temperature quantum phase transition as a function of \Delta/U at fixed \Omega/U=0.33 . In particular, we first consider periodic boundary conditions along {\bf a}_{y} , for cylinders with increasing circumference N_{y}=3,6 and 9 —corresponding to l_{y}=1,2 and 3 unit cells (of size 3\times 3 ) along {\bf a}_{y} . We expect the analysis of such geometries to be effectively equivalent to isotropic systems at finite temperatures, set by the inverse of the cylinder circumference, i.e., k_{\rm B}T=1/l_{y} SM . For every value of 1/l_{y} , we find that the two phases are separated by a DQCP with emergent U( 1 ) symmetry [see Fig. 1 (b)]. Below we describe this continuous transition using an effective field theory and confirm our predictions with extensive numerical simulations based on matrix product states. In the following, we truncate the interactions to the third-nearest neighbors [Fig. 1 (a)]. This approximation is justified for small Rydberg radius—the characteristic distance where the interaction strength is dominant and satisfies U_{{\bf i}{\bf j}}=\Omega . For \Omega/U=0.33 , this yields R_{\rm b}=1.203a , supporting the neglect of interactions beyond a distance of 2a . In the Supplementary Material (SM) SM , we show how our results, in particular the presence of a DQCP with an emergent U( 1 ) symmetry, remain unaffected by the presence of longer-range interaction terms and persist for a broad range of parameters up to \Omega/U\approx 0.4 . Effective field theory.— Figure 2: (a) Mapping from a 2D cylindrical lattice to a 1D field theory model. The phase of the magnetization m [Eq. ( 2 )] in each unit cell (of size 3\times 3 ) is mapped to a 2D field \phi(x,y) , and by dimensional reduction to a 1D field \phi(x) . (b)–(e) Analysis of an infinitely long cylinder with circumference N_{y}=6 ( l_{y}=2 unit cells) for increasing bond dimension \chi [see legend in panel (b)]. (b) The order parameter changes continuously from positive to negative as a function of \Delta/U , passing through zero at the critical point at (\Delta/U)_{\rm c}\approx 3.158 (dashed vertical line). (c) The correlation length \xi/a as a function of \Delta/U shows a peak at (\Delta/U)_{\rm c} , which diverges with \chi . Inset: extraction of the central charge c from the scaling S_{\rm tr}=c\log{(\xi_{\rm tr}/a)}/6 at (\Delta/U)_{\rm c} . (d) Power-law decay of the two-point correlation function \langle m(r)m(0)\rangle\approx C_{1}(r)=\rho^{2}_{0}\langle e^{i\phi(r)}e^{-i\phi(0)}\rangle at (\Delta/U)_{\rm c} , where m is evaluated on unit cells of size 3\times 3\,l_{y} . Inset: Luttinger parameter K^{\prime} as a function of the cylinder circumferences l_{y} , extracted from the power-law decay of C_{1}(r) [Tab. 1 ]. (e) Linear fit of the order parameter and the correlation length (inset) in log-log scale as a function of \bar{\Delta}=\Delta/U-(\Delta/U)_{\rm c} , used to extract the critical exponents \beta and \nu , respectively. (f) Dependence of the critical exponents on l_{y} . We compare the scaling dimension x_{3} estimated from (red) 2-1/\nu and (blue) 2\beta/(1+\beta) to the theoretical prediction (black dots) x_{3}=9K^{\prime}/4 and (black line) x_{3}=9K/(4\rho^{2}_{0}l_{y}) , with coefficient K/\rho^{2}_{0} obtained from the fit in the inset of (d). We aim to characterize the transition on infinitely long cylinders with increasing circumference and to develop a field theory that both captures the nature of the transition and connects the results across different cylinders. We begin by presenting the results for the infinitely long cylinder with circumference N_{y}=6 ( l_{y}=2 ). Further details on other geometries can be found in the SM SM . Here, the order parameter changes continuously from positive ( 1/3 phase) to negative ( 2/3 phase), vanishing at the transition point (\Delta/U)_{\rm c}\approx 3.158 [Fig. 2 (b)]. This smooth behavior signals a continuum quantum phase transition beyond the conventional LGW paradigm. This is further corroborated by the growth of the correlation length \xi as the bond dimension \chi increases in our DMRG simulations [Fig. 2 (c)]. In particular, at the critical point, the scaling of the von Neumann entanglement entropy at half cylinder with the correlation length, S_{\rm tr}=c\log{(\xi_{\rm tr}/a)}/6 , is consistent with the emergence of a CFT with central charge c=1 Calabrese and Cardy 2004 ; Tagliacozzo et al. 2008 ; Pollmann et al. 2009 . Further evidence for critical behavior is provided by the power-law decay of the two-point correlation function of the magnetization between unit cells separated by distance r , \langle m(r)m(0)\rangle [Fig. 2 (d)]. To capture these numerical findings, we formulate an effective field theory by mapping the discrete microscopic model onto a continuous field description [Fig. 2 (a)]. In the limit of quasi-1D systems ( l_{y}\ll l_{x} ), we associate a complex field \psi(x)=\rho(x)e^{i\phi(x)} to the staggered magnetization m evaluated in a unit-cell of size 3\times N_{y}=3\times 3\,l_{y} . Neglecting amplitude fluctuations, and setting \psi(x)=\rho_{0}e^{i\phi(x)} , we can effectively describe the system via the 1D theory SM : S=\int dt\,dx\left[\frac{\left(\partial_{\mu}\phi\right)^{2}}{2\pi K^{\prime}}+g^{\prime}_{3}\cos{\left(3\phi\right)}+g^{\prime}_{6}\cos{(6\phi)}\right], (3) where the parameters K^{\prime} , g^{\prime}_{3} , and g^{\prime}_{6} depend on l_{y} and \rho_{0} . In particular, K^{\prime}=K/(\rho_{0}^{2}l_{y}) , where K is a constant. This theory captures the phase diagram of the Rydberg model in Eq. ( 1 ) as a function of \Delta/U (which we assume controls g^{\prime}_{3} ) and temperature T (proportional to 1/l_{y} , and thus to K^{\prime} ) SM . In particular, it allows for a quantitative characterization of the continuous transition between the two \mathbb{Z}_{3} -ordered phases, including predictions for the critical exponents, as detailed below. Here, we focus on the regime 2/90 , corresponding to the 1/3 and 2/3 phases, respectively. At the critical point, i.e., g^{\prime}_{3}=0 , the model reduces to a Luttinger Liquid (LL) with an emergent U( 1 ) symmetry. Therefore, as g^{\prime}_{3} changes sign, the system undergoes an unconventional continuous phase transition between two distinct ordered phases, which are separated by a DQCP with U( 1 ) symmetry. We further predict the dependence of the critical exponents on the cylinder width l_{y} , as summarized in Table 1 SM . Observable Scaling Exponent C_{n}(r) r^{-2x_{n}} x_{n}=\frac{n^{2}K^{\prime}}{4}=\frac{n^{2}K}{4\rho^{2}_{0}l_{y}} \xi {|g_{3}^{\prime}|}^{-\nu} \nu=\frac{1}{2-x_{3}} \rho^{3}_{0}|\langle\cos(3\phi)\rangle| {|g_{3}^{\prime}|}^{\beta} \beta=\frac{x_{3}}{2-x_{3}} Table 1: Correlation function and critical exponents for the field theory in Eq. ( 3 ) in the regime 2/92000 ) would be necessary. The fate of the transition in the infinite-circumference limit is of particular interest for quantum simulations on experimental platforms, as we discuss now. Finite ladders and experimental proposal.— Figure 3: (a) Correlation length \xi/a as a function of \Delta/U for an infinite ladder ( N_{x}=\infty ), with N_{y}=3l_{y}=12 atoms along {\bf a}_{y} , and increasing bond dimension \chi . Inset: extraction of the central charge c from the scaling of S_{\rm tr}=c\log{(\xi_{\rm tr}/a)}/6 at (\Delta/U)_{\rm c}\approx 3.1586 , corresponding to the maximum of the correlation length. (b)-(e) Experimentally feasible system. (b) Lattice and occupation number \langle n_{\bf j}\rangle for \chi=200 and \Delta/U=2.92 . (c) Distribution of the staggered magnetization m [Eq. ( 2 )] of the state considered in (b) over 10^{4} snapshots in the occupation basis when restricting the sites to the ‘bulk’ (i.e., excluding the last two ‘rings’ of atoms). (d) Angular distribution of m . (e) Radial distribution of m fitted according to Eq. ( 4 ). Apart from probing the fate of an isotropic 2D system at zero temperature, we now argue that current Rydberg experiments could be readily used to observe a DQCP in an accessible quasi-1D geometry. This also represents an interesting direction, as various experimental proposals have been advanced for observing DQCPs in 1D systems Lee et al. 2023 ; Romen et al. 2024 ; Baldelli et al. 2024 , yet none have been realized so far. Previously, we have shown that the transition on quasi-1D cylinders is unconventional. However, realizing cylindrical geometries in experiments is challenging. To address this, we now consider ‘infinite ladders’—systems that have open boundary conditions along {\bf a}_{y} and are infinite along {\bf a}_{x} . Although open boundaries play an important role for small ladder widths, for l_{y}=4 we recover the continuous transition between the 1/3 and 2/3 phases. In particular, at criticality, both the entanglement entropy and the correlation length increase with bond dimension, consistent with a CFT with c=1 [Fig. 3 (a)]. As an experimentally feasible setup to observe this DQCP, we consider the finite array shown in Fig. 3 (b), where the six classical configurations at 1/3 and 2/3 filling are exactly degenerate in energy in the limit \Omega=0 . We note that similar geometries were realized in Ref. Scholl et al. 2021 . Remarkably, even for such a small system, we find that an approximately U( 1 )-symmetric angular distribution emerges between the 1/3 and 2/3 phases [Fig. 3 (c)], signaling the presence of deconfined quantum criticality. In particular, the angular probability distribution P(\phi) and the radial probability distribution P(\rho) agree with the field-theory predictions [Fig. 3 (d)-(e)] and remain unaffected by the presence of longer-range interaction terms SM . The distribution P(\phi) is consistent with a weak potential of the form V(\phi)=g^{\prime}_{6}\cos{(6\phi)} , while the distribution P(\rho) is compatible with a simple local effective potential V_{\rm eff}(\rho)=\mu\rho^{2}-\lambda\rho^{4} (with \rho_{0}^{2}=\mu/2\lambda ), leading to the probability distribution P(\rho)=\frac{\rho\,e^{-{\rm Vol.}\,V_{\rm eff}(\rho)}}{\int_{0}^{\infty}d\rho\,\rho e^{-{\rm Vol.}\,V_{\rm eff}(\rho)}}\,, (4) where {\rm Vol.} is an effective volume factor SM . Scaling up this geometry in one direction would allow for an experimental probe of the critical properties of the quasi-1D ladder depicted in Fig. 3 (a). Whether or not the emergent U( 1 ) symmetry is spontaneously broken in the 2D thermodynamic limit remains an open and interesting question that cannot be definitely answered from our numerics. This question could also be addressed experimentally by scaling up this geometry in both directions. Conclusion and outlook.— In this work, we have shown that quantum fluctuations render the phase transition between the 1/3 and 2/3 ordered phases of a triangular Rydberg array continuous in quasi-1D geometries, for both periodic and open boundary conditions. The transition point corresponds to a DQCP, which cannot be captured by the standard LGW formalism. Furthermore, our results demonstrated that the emergence of a continuous U( 1 ) symmetry at criticality is captured by a 1D conformal field theory and can be experimentally probed in accessible finite geometries. This opens several exciting directions for future research. An important open question is whether the DQCP persists down to zero temperature in isotropic 2D systems. This question could also be addressed experimentally—as Rydberg atom arrays enable access to system sizes and geometries beyond current numerical capabilities—by adiabatically preparing the U(1)-symmetric ground state at the transition in systems of increasing size. However, adiabatically preparing a critical state requires a preparation time that grows (polynomially) with system size, imposing limits due to finite coherence times and experimental imperfections. Possible future directions include performing numerical analyses of adiabatic state preparation to identify optimal laser sweep protocols to prepare the U(1)-symmetric angular distribution demonstrated in Fig. 3 (b)–(e), as well as dynamically probing the DQCP via the quantum Kibble-Zurek mechanism Keesling et al. 2019 ; Huang and Yin 2020 . Acknowledgments.— D.G.-C. thanks Daniel Barredo and Vincent Lienhard for helpful discussions on the experiments performed in Ref. Scholl et al. 2021 . This work is supported by the European Union’s Horizon Europe research and innovation program under Grant Agreement No. 101113690 (PASQuanS2.1), the ERC Starting grant QARA (Grant No. 101041435), the EU-QUANTERA project TNiSQ (N-6001), and by the Austrian Science Fund (FWF) (Grant No. DOI 10.55776/COE1). D.G.-C. acknowledges financial support through the Ramón y Cajal Program (RYC2023-044201-I), financed by MICIU/AEI/10.13039/501100011033 and by the FSE+. Tensor network calculations were performed using the TeNPy Library Hauschild et al. 2024 . The results in Fig. 3 have been obtained using the LEO HPC infrastructure of the University of Innsbruck. Supplemental Material to “Deconfined quantum criticality on a triangular Rydberg array” In this Supplementary Material, we provide additional details supporting the analytical and numerical results presented in the main text. In App. I , we expand on the field-theoretical analysis. In App. II , we detail the numerical methods, including the data used to extract the critical exponents shown in Fig. 2 (f), a finite-size scaling analysis confirming the dependence of the critical exponents on the cylinder circumference, an analysis of the statistics of the U( 1 )-symmetric state for cylinders of different circumferences, and a discussion on the truncation of the interaction tails. In App. III , we consider the open system in Fig. 3 (b)-(e) with interactions up to the fifth-nearest neighbor and demonstrate how the staggered magnetization serves as an experimentally accessible observable to distinguish between the ordered phases and the critical region. Finally, in App. IV , we examine the dependence on \Omega/U for the cylinder with the smallest circumference considered in the main text and show that the field-theory description—and thus the DQCP—remains valid for \Omega/U\lesssim 0.43 , with a Luttinger parameter K that depends on \Omega/U . I Details on the field-theoretical description Here, we provide further details on the field theory discussed in the main text. Our analysis begins with the formulation of a low-energy 2D effective field theory for the complex staggered magnetization m=\rho e^{i\phi} [Eq. ( 2 )]. From this, we then derive an effective field theory governing its phase \phi . We then discuss it in the 1D limit, corresponding to small circumferences. Finally, we present analytical predictions for the marginal probability distributions P(\phi) and P(\rho) . I.1 Mapping To formulate the 2D field theory, we begin by defining how the microscopic lattice model is mapped onto a continuum description. We consider blocks of size 3\times 3 , corresponding to the unit cells of the considered Rydberg model [Eq. ( 1 )]. We label each unit cell by {\bf u}=(u_{x},u_{y})\in\mathbb{Z}^{2} , e.g., indicating the position of its ‘center’ {\bf r}_{\bf u}=\left(u_{x}+\tfrac{1}{2}\right){\bf A}_{x}+\left(u_{y}+\tfrac{1}{2}\right){\bf A}_{y} , where {\bf A}_{x}=3\,{\bf a}_{x} and {\bf A}_{y}=3\,{\bf a}_{y} are the basis vectors of the effective lattice with lattice spacing 3a . Here, {\bf a}_{x} and {\bf a}_{y} are the primitive vectors, and a is the lattice spacing of the underlying triangular lattice. On this coarse-grained lattice, we define a local observable on each unit cell given by the staggered magnetization m({\bf u})=\sum_{{\bf j}\in{\bf u}}n_{\bf j}\,e^{i\,{\bf Q}\cdot x_{\bf j}} , where the sum runs over all sites {\bf j} within the 3\times 3 unit cell centered at {\bf r}_{\bf u} . In the ‘continuum limit’, where 3a\to 0 , we associate with each discrete site {\bf u}\in\mathbb{Z}^{2} a continuous position {\bf r}=(x,y)\in\mathbb{R}^{2} and to the observable defined on the discrete lattice m({\bf u}) a smooth field \psi({\bf r}) such that \psi({\bf r})=\lim_{a\to 0}m({\bf u}) . Throughout the numerics, we set a=1 and consider the limit of many unit cells to approximate the continuum behavior. I.2 Field theory in 2D Here, we formulate a continuous 2D field-theoretic description of the system in terms of the complex scalar field \psi({\bf r}) . Guided by the symmetries of the two ordered phases and the emergence of a U( 1 ) symmetry at criticality of the Rydberg model [Fig. 1 ], we propose the following 2D effective theory, described by the Euclidean-space path integral Z=\int D[\psi,\psi^{*}]\,e^{-S[\psi,\psi^{*}]} , with associated action S[\psi,\psi^{*}]=\int d\tau\,dx\int_{0}^{l_{y}}dy\left[\frac{|\partial_{\mu}\psi|^{2}}{2\pi K}-\mu|\psi|^{2}+\lambda|\psi|^{4}+\frac{g_{3}}{2}\left(\psi^{3}+\psi^{*3}\right)+\frac{g_{6}}{2}\left(\psi^{6}+\psi^{*6}\right)\right]. (S1) Here, |\partial_{\mu}\psi|^{2}=|\partial_{\tau}\psi|^{2}+|\nabla\psi|^{2} , \tau parametrizes the imaginary time, and the y -direction is compact: \psi(\tau,x,l_{y})=\psi(\tau,x,0) . Moreover, g_{3} and g_{6} are the coupling constants of the \mathbb{Z}_{3} and the \mathbb{Z}_{6} symmetry-breaking perturbations, respectively. To describe the transition between the ordered phases at 1/3 - and the 2/3 -filling, we require g_{3} to change sign across the transition and vanish at criticality. Due to the symmetry of the action under the exchange of imaginary time \tau and the compact spatial coordinate y , this theory on a spatial cylinder can be interpreted as describing an isotropic system at finite temperature k_{\rm B}T=1/\beta=1/l_{y} . Finally, we note that higher-order interaction terms in the action become irrelevant under renormalization group (RG) flow at low energies due to their negative mass dimension. Indeed, assuming K is dimensionless, i.e., [K]=0 , the field \psi then has mass dimension [\psi]=1/2 , and the coupling constant g_{n} associated with the perturbation \psi^{n}+\psi^{*n} has mass dimension [g_{n}]=3-n/2 . We now switch to polar coordinates by introducing real fields \rho and \phi via the parametrization \psi=\rho e^{i\phi} . The propagator becomes Z=\int D[\rho,\phi]\,\,e^{-S[\rho,\phi]} , where the measure is D[\rho,\phi]=\rho\,d\rho d\phi , and the action is given by S[\rho,\phi]=\int d\tau\,dx\int_{0}^{l_{y}}dy\left[\frac{(\partial_{\mu}\rho)^{2}}{2\pi K}-\mu\rho^{2}+\lambda\rho^{4}+\frac{\rho^{2}}{2\pi K}(\partial_{\mu}\phi)^{2}+g_{3}\rho^{3}\cos{(3\phi)}+g_{6}\rho^{6}\cos{(6\phi)}\right], (S2) where we have used the identity |\partial_{\mu}\psi|^{2}=(\partial_{\mu}\rho)^{2}+\rho^{2}(\partial_{\mu}\phi)^{2} . To try to solve this model, we aim to factorize the propagator as Z=Z_{\rho}Z_{\phi} , where Z_{\rho}=\int D[\rho]\,e^{-S[\rho]} and Z_{\phi}=\int D[\phi]\,e^{-S[\phi]} . This requires that the action decomposes into two independent parts: S[\rho,\phi]=S[\rho]+S[\phi] . To achieve this, we need to make the approximations and assumptions detailed in the following. I.3 Effective field theory for the phase field: from 2D to 1D First, we neglect fluctuations of the amplitude field \rho and assume it is fixed at the constant value \rho_{0} that minimizes the action. This corresponds to a saddle-point approximation, where \rho_{0} satisfies \left.\frac{\partial S}{\partial\rho}\right|_{\rho=\rho_{0}}=2\rho_{0}(2\lambda\rho_{0}^{2}-\mu)=0, (S3) yielding to \rho_{0}=0\>\>\ \text{for}\>\>\ \mu<0\>\>\ \text{and}\>\>\ \rho_{0}=\sqrt{\frac{\mu}{2\lambda}}\>\>\ \text{for}\>\>\ \mu>0. (S4) Second, we assume the system is in the phase with \mu>0 . Substituting \rho=\rho_{0} into Eq. ( S2 ), we obtain the following effective action for the phase field \phi : S[\phi]=\int d\tau dx\int_{0}^{l_{y}}dy\left[\frac{\rho_{0}^{2}}{2\pi K}(\partial_{\mu}\phi)^{2}+g_{3}\rho_{0}^{3}\cos{(3\phi)}+g_{6}\rho_{0}^{6}\cos{(6\phi)}\right]. (S5) I.3.1 Effective 1D model We now perform a dimensional reduction of the 2D field theory in Eq. ( S5 ) to a effective 1D theory. Specifically, for a small cylinder circumference l_{y} , we assume the system is translationally invariant along the y -direction [ \phi(\tau,x,y)\to\phi(\tau,x) ], and we obtain the effective 1D action S[\phi]=\int d\tau\,dx\left[\frac{(\partial_{\mu}\phi)^{2}}{2\pi K^{\prime}}+g^{\prime}_{3}\cos{(3\phi)}+g^{\prime}_{6}\cos{(6\phi)}\right], (S6) with parameters K^{\prime}=K/(\rho_{0}^{2}l_{y}),\>\>\ g^{\prime}_{3}=g_{3}l_{y}\rho_{0}^{3},\>\>\ \text{and}\>\>\ g^{\prime}_{6}=g_{6}l_{y}\rho_{0}^{6}. (S7) In particular, we find that K^{\prime} decreases as the cylinder circumference l_{y} increases, and therefore increases with temperature, i.e., K^{\prime}\sim 1/l_{y}\sim T . Thus, analyzing the phase diagram of this model for varying K^{\prime} corresponds to analyzing the phase diagram of the original Rydberg model as a function of temperature. While Eq. ( S6 ) captures the low-temperature physics, to describe the high-temperature behavior we include in the following the term g^{\prime}_{H}\cos{(2\pi\Theta)} , in analogy to the \mathbb{Z}_{q} deformed sine-Gordon model Matsuo and Nomura 2006 : \displaystyle S[\phi]=\int d\tau\,dx\Bigg[\frac{(\partial_{\mu}\phi)^{2}}{2\pi K^{\prime}}+g^{\prime}_{3}\cos(3\phi)+g^{\prime}_{6}\cos(6\phi) \displaystyle+g^{\prime}_{H}\cos\left(2\pi\Theta\right)\Bigg]. (S8) Here, the field \Theta is mutually dual to \phi , satisfying the relation \partial_{x}\phi=-\pi K^{\prime}\partial_{y}\Theta and \partial_{y}\phi=\pi K^{\prime}\partial_{x}\Theta . We note that this model has also been recently discussed in Ref. Prakash and Jones 2025 . I.3.2 Phase diagram Figure S1: Phase diagram and phase transitions for the model in Eq. ( S8 ). The phase diagram of the effective 1D model in Eq. ( S8 ) can be understood by analyzing the renormalization group (RG) flow of the perturbative terms Di Francesco et al. 1997 . The scaling dimension of a general operator O_{m,n}=e^{in\phi}e^{im\Theta} is given by Matsuo and Nomura 2006 x_{m,n}=\frac{1}{4}\left(\frac{m^{2}}{K^{\prime}}+n^{2}K^{\prime}\right). (S9) In particular, the scaling dimensions of \cos{(3\phi)} , \cos{(6\phi)} , and \cos{(2\pi\Theta)} are x_{3}=\frac{9}{4}K^{\prime},\>\>\ x_{6}=9K^{\prime},\>\>\ \text{and}\>\>\ x_{\Theta}=\frac{1}{K^{\prime}}. (S10) The relevance or irrelevance of each term depends on whether its scaling dimension is less than or greater than 2 , respectively. This allows us to distinguish different regimes as a function of K^{\prime} (interpreted as temperature), as illustrated in the phase diagram in Fig. S1 . Let us first consider the case g^{\prime}_{3}\neq 0 and the region K^{\prime}>2/9 . Here, the \mathbb{Z}_{6} symmetry breaking term \cos(6\phi) becomes irrelevant, effectively reducing the theory in Eq. ( S8 ) to a 3-state clock model Li et al. 2020 . Upon varying K^{\prime} (equivalent to T ), we encounter a phase transition between a low-temperature ordered phase, in which the \mathbb{Z}_{3} symmetry is spontaneously broken, and a high-temperature disordered phase, characterized by exponentially decaying correlations. This transition belongs to the universality class of the Potts model. The critical point is determined by the so-called ‘self-dual’ condition x_{3}=x_{\Theta} , which is satisfied for K^{\prime}=2/3 . These findings are consistent with the transition between the ordered phase at 1/3 (or at 2/3 )-filling and the high-temperature disordered phase of the Rydberg model described in Ref. Guo et al. 2023 . Next, we fix K^{\prime} within the \mathbb{Z}_{3} -ordered phase, i.e., 2/9K^{\prime}_{2}=1/2 (where \cos(2\pi\Theta) is a relevant perturbation), and an intermediate phase for K_{1}^{\prime}0 or g^{\prime}_{3}<0 ) depends on the sign of g^{\prime}_{6} . It is a first-order transition for g^{\prime}_{6}<0 , while an intermediate phase with \mathbb{Z}_{6} symmetry arises for g^{\prime}_{6}>0 . The phase transitions at the boundaries of this intermediate phase belong to the Ising universality class Prakash and Jones 2025 . We note that in this regime of small K^{\prime} , the dimensional reduction may break down, and the effective 1D field theory [Eq. ( S6 )] may no longer capture the full 2D physics [Eq. ( S2 )]. Indeed, our numerical simulations do not show clear evidence for either a first-order transition or an intermediate phase, as further discussed in App. II . I.3.3 Critical exponents Let us focus on the region 2/9\>\ \text{with}\>\>\ \nu=\frac{1}{2-x_{3}}, (S12) \displaystyle|\langle\cos(3\phi)\rangle|\sim\left|g^{\prime}_{3}\right|^{\beta},\>\>\ \text{with}\>\>\ \beta=\frac{\eta\nu}{2}=\frac{x_{3}}{2-x_{3}}. (S13) Notably, these exponents depend on the cylinder circumference via Eq. ( S7 ) and Eq. ( S9 ). Specifically, for \beta/\nu and 1/\nu , we have: \frac{\beta}{\nu}=x_{3}=\frac{9K}{4\rho^{2}_{0}}\,\frac{1}{l_{y}},\>\>\ \frac{1}{\nu}=2-x_{3}=2-\frac{9K}{4\rho^{2}_{0}}\,\frac{1}{l_{y}}. (S14) I.4 Prediction of the marginal probabilities distributions To predict the marginal probability distributions for the phase and amplitude fields, we return to the 2D field theory in Eq. ( S2 ). For simplicity, to derive a nontrivial probability distribution for \rho , we assume that the amplitude field \rho decouples from the phase field \phi and is described independently by the following action: S[\rho]=\int d\tau\,dx\int_{0}^{l_{y}}dy\left[\frac{(\partial_{\mu}\rho)^{2}}{2\pi K}-\mu\rho^{2}+\lambda\rho^{4}\right]. (S15) Together with the phase action S[\phi] given in Eq. ( S5 ), these decoupled models allow us to define marginal probability distributions for the respective fields. Specifically, we define: \displaystyle\langle\phi\rangle=\frac{1}{Z_{\phi}}\int d\phi\,\phi\,e^{-S[\phi]}=\int d\phi\,\phi\,P(\phi), \displaystyle\langle\rho\rangle=\frac{1}{Z_{\rho}}\int d\rho\,\rho\,\rho e^{-S[\rho]}=\int d\rho\,\rho\,P(\rho), (S16) where P(\phi) and P(\rho) are the marginal probability distributions for the phase and amplitude fields, respectively. In our numerical analysis, we sample the ‘total’ staggered magnetization m=\sum_{\bf u}m(\bf u) . We can relate this quantity to the zero-mode component of the field \psi(x,y) , given by \tilde{\psi}(0,0)=\sum_{x,y}\psi(x,y) , whose marginal probability distribution we can predict under certain assumptions. Assuming that the fields \rho and \phi are constant in space-time (in different unit cells), the marginal probability distributions of \psi coincide with those of \tilde{\psi} . Furthermore, the actions simplify to: S[\phi]=\mathrm{Vol.}\,V[\phi] and S[\rho]=\mathrm{Vol.}\,V[\rho] , where \mathrm{Vol.} indicates the volume factor coming from the integrals, and the effective potentials are \displaystyle V[\phi]=g_{3}\rho^{3}_{0}\cos{(3\phi)}+g_{6}\rho^{6}_{0}\cos{(6\phi)}, (S17) \displaystyle V[\rho]=-\mu\rho^{2}+\lambda\rho^{4}. (S18) From these potentials, using Eq. ( S16 ), we obtain \displaystyle P(\phi)=\frac{e^{-{\rm Vol.}\,V[\phi]}}{\int D[\phi]\,e^{-{\rm Vol.}\,V[\phi]}}, (S19) \displaystyle P(\rho)=\frac{\rho e^{-{\rm Vol.}\,V[\rho]}}{\int D[\rho]\,\rho e^{-{\rm Vol.}\,V[\rho]}}. (S20) These expressions provide analytic predictions for the form of the marginal distributions that we compare to the numerical data in our work. II Further details on the numerics on cylinders Figure S2: Analysis of infinitely long cylinders with (first column) l_{y}=1 , (second column) l_{y}=2 and (third column) l_{y}=3 unit cells along {\bf a}_{y} . For each of them, we show for different bond dimensions \chi : (First row) Power-law decay of the two-point correlation \langle m({\bf r})m(0)\rangle\approx C_{1}(r)=\rho^{2}_{0}\langle e^{i\phi(r)}e^{-i\phi(0)}\rangle at the transition point (\Delta/U)_{\rm c} , where the correlation length is maximal. The Luttinger parameter K^{\prime} , shown in the inset of Fig. 2 (d), is extracted from a linear log-log fit (red dashed line). (Second row) Evaluation of the critical exponent \beta (main panels) and \nu (insets), displayed in Fig. 2 (f), from linear log-log fits of the order parameter and the correlation length as function of \bar{\Delta}=\Delta/U-(\Delta/U)_{\rm c} . Here, we provide further details on the numerical analysis. First, we discuss the numerical method employed. Second, we present the numerical data and the log-log fits used to determine the critical exponents shown in Fig. 2 (f). Third, we perform a finite-size scaling analysis which confirms the dependence of the critical exponents on the cylinder circumference predicted by the effective 1D field theory [Eq. ( S14 )]. Fourth, we analyze the statistics of the staggered magnetization on three finite cylinders with increasing circumference, by fitting the data to the analytical predictions for the marginal distributions given in Eqs. ( S20 ) and ( S19 ). This fitting procedure enables us to estimate how the ratio |g_{6}/g_{3}| scales with the cylinder circumference. Finally, we include the interactions up to the fifth-nearest neighbor and show that the emergence of the U( 1 ) symmetry remains unaffected by the truncation at the third-nearest neighbors. II.1 Computational methods Numerically, we represent the wavefunction using a matrix product state (MPS) ansatz and compute the ground state of the Hamiltonian via the density-matrix renormalization group (DMRG) algorithm, as implemented in the Python package TeNPy Hauschild et al. 2024 . We map the lattice onto a one-dimensional MPS by wrapping it along the {\bf a}_{y} direction. Our numerical results are converged in energy below error 10^{-6} . Our sweep strategy begins from a simple product state, progressively increasing the bond dimension. At each step, the converged MPS from the previous bond dimension is used as the initial state for the next. To avoid the algorithm becoming trapped in local minima, we include a mixer term during the early stages of the sweeps. II.2 Critical exponents derived from the numerics In Fig. S2 , we show the linear fits used to extract the Luttinger parameter K^{\prime} , the critical exponent \beta , and the critical exponent \nu shown in Fig. 2 , for infinitely long cylinders with l_{y}=1,2, and 3 unit cells along the y -axis. In particular, K^{\prime} follows from the decay of the two-point correlation function \langle m(r)m(0)\rangle\approx C_{1}(r)=\rho_{0}^{2}\langle e^{i\phi(r)}e^{-i\phi(0)}\rangle , \beta from the scaling of the order parameter |\langle m^{3}+m^{\dagger 3}\rangle|\sim|\bar{\Delta}|^{\beta} , and \nu from the scaling of the correlation length \xi\sim|\bar{\Delta}|^{-\nu} . In the first case, we consider the decay of C_{1}(r) for the largest considered bond dimension \chi (as the results have converged). For the last two cases, we first extrapolate the observables as a function of \chi by performing a quadratic fit in 1/\chi , and then apply a linear fit to the extrapolated values to extract the critical exponents. To estimate the error bars on these exponents, we repeat the fitting procedure over two different ranges: one over the full range shown in red, and another excluding one or two of the smallest |\bar{\Delta}| values (for which the results are less converged in \chi ). The difference between the resulting estimates is then taken as the uncertainty reported in Fig. 2 (f). Finally, for the largest circumference l_{y}=3 , we find that the correlation length grows as |\bar{\Delta}| is reduced, i.e., as the system approaches criticality, but drops at the very closest point we consider [see inset of Fig. S2 (f)]. At first sight, this non-monotonic behavior might hint at an intermediate symmetry‑broken phase. However, up to \chi=2000 , we find that the correlation length and entanglement entropy continue to scale as expected for a CFT with central charge c\approx 0.5 ; they do not saturate as they would in a fully ordered phase. Moreover, at the second‑closest point to criticality we extract c\approx 1 in disagreement with the Ising value c=0.5 that should appear at the boundaries of the intermediate ordered phase (Fig. S1 ). These observations suggest that the dip in correlation length is most likely a finite‑ \chi artifact rather than evidence of a genuine intermediate phase. However, further numerical investigation at larger bond dimensions is required to clarify the nature of this apparent dip in the correlation length. II.3 Finite size scaling Figure S3: Finite size scaling analysis on cylinders. (a) Order parameter \langle m^{3}+m^{3\dagger}\rangle as a function of \Delta/U for cylinders with circumference (yellow) l_{y}=1 and (blue) l_{y}=2 unit cells along {\bf a}_{y} , and increasing lengths l_{x} (unit cells along {\bf a}_{x} ). The dashed vertical line is located at \Delta/U=3.158 . (b) Finite-size scaling using the ansatz in Eq. ( S21 ), with \bar{\Delta}=\Delta/U-(\Delta/U)_{\rm c} and (\Delta/U)_{\rm c}\approx 3.158 . The exponents depend on l_{y} according to Eq. ( S14 ), with K/\rho_{0}^{2}\approx 0.1 . Results obtained using a finite MPS with bond dimension \chi=200 for l_{y}=1 , and \chi=300 for l_{y}=2 . Here, we further numerically confirm the dependence of the critical exponents on the cylinder circumference predicted by the effective field theory [Eq. ( S14 )] by performing a finite-size scaling (FSS) analysis in the cylinder length. In particular, we consider cylinders of increasing lengths l_{x}=N_{x}/3 ( l_{x} is the number of unit cells along {\bf a}_{x} ) in the limit of l_{x}\gg l_{y} , i.e., l_{x}>4l_{y} . For finite cylinders, the correlation length \xi is bounded by the number of unit cells along the x -direction l_{x} , assuming that the data is converged in bond dimension \chi . Accordingly, we replace \xi\to l_{x} and identify the perturbation g^{\prime}_{3} in Eq. ( S14 ) with \bar{\Delta}=\Delta/U-(\Delta/U)_{\rm c} , arriving at the following scaling hypothesis: |\langle m^{3}+m^{\dagger 3}\rangle|\sim l^{-\beta/\nu}_{x}\mathcal{F}\left(|\bar{\Delta}|l^{1/\nu}_{x}\right). (S21) where \mathcal{F} is a universal scaling function, with \mathcal{F}\to constant at the critical point \bar{\Delta}=0 , and the exponents \beta/\nu and 1/\nu depend on the cylinder circumference l_{y} as described in Eq. ( S14 ). We first apply this scaling analysis to the smallest cylinder with l_{y}=1 [Fig. S3 ], extracting the constant K/\rho^{2}_{0} that ensures data collapse: K/\rho^{2}_{0}\approx 0.1 . (Note that this constant differs from the one extracted for infinite cylinders.) Using this value, we then verify that a consistent scaling collapse is obtained for the largest cylinder with l_{y}=2 , confirming the dependence of the critical exponents on cylinder circumference. We note that cylinders with l_{y}=3 and large l_{x} are beyond our current numerical reach. II.4 Sampling of the wavefunction on open cylinders and marginal probability distributions Figure S4: Statistics of the staggered magnetization m [Eq. ( 2 )] for three open cylinders with increasing circumference l_{y} at \Delta/U\approx 3.1615 , 3.15789 , and 3.1584 , respectively. We consider: (First column) l_{y}=1 , l_{x}=7 unit cells along {\bf a}_{y} and {\bf a}_{x} , respectively; (Second column) l_{y}=2 , l_{x}=10 ; and (Third column) l_{y}=3 , l_{x}=5 . (First row) Distribution of m obtained from 10^{5} snapshots in the occupation basis, restricted to the bulk (excluding the first and last three rows of atoms along \bf{a}_{x} ). (Second row) Corresponding angular distribution, fitted according to Eq. ( S19 ) (red dashed line). (Third row) Corresponding radial distribution, fitted according to Eq. ( S20 ) (red dashed line). Blue lines in (d)-(i) indicate a Gaussian kernel density estimate. Results obtained using a finite MPS with bond dimension \chi=200 for l_{y}=1 , \chi=250 for l_{y}=2 , and \chi=500 for l_{y}=3 . Figure S5: (a) Estimated values of \rho_{0}^{3}|g_{6}/g_{3}| obtained by fitting the angular distribution P(\phi) shown in Fig. S4 , for cylinders with l_{y}=1,2, and 3 , using Eq. ( S19 ), plotted as a function of the number of measurement shots. Solid lines indicate the fitted values, and shaded regions represent the corresponding \pm\sigma confidence intervals. (b) Log-log plot of the values in (a) at 10^{5} shots [marked by the vertical dashed line in (a)] versus cylinder circumference l_{y} . Figure S6: Estimated values of \rho_{0}^{2} from fitting the radial distribution P(\rho) using Eq. ( S20 ) for the three cylinders shown in Fig. S4 , with circumferences l_{y}=1 , 2 , and 3 , as a function of the number of measurement shots. Solid lines indicate the fitted parameter values, and shaded regions represent the \pm 1 standard error of the fit parameter obtained from the covariance matrix. Here, we analyze the statistics of the staggered magnetization m [Eq. ( 2 )] for open cylinders of increasing circumference l_{y} , near the transition point. We observe the emergence of a U(1) symmetry and explain the marginal angular and radial probability distributions using the effective field theory discussed above. We note that the emergence of a U(1) symmetry has also been observed in other Rydberg models Zhou et al. 2025 , but these are not straightforward to implement experimentally. For circumferences l_{y}=1 and l_{y}=2 , we focus on the two longest open cylinders considered for the FSS analysis presented above, with l_{x}=7 and l_{x}=10 , respectively. For l_{y}=3 , we consider the longest accessible cylinder with l_{x}=5 ( N_{x}=15 atoms). We choose values of \Delta/U close to the transition point \Delta_{{\rm c},(l_{x},l_{y})}/U , which we estimate by the point where the order parameter \langle m^{3}+m^{3\dagger}\rangle vanishes. We find \Delta_{{\rm c},(l_{x},l_{y})}\approx 3.15876 , 3.158097, and 3.15879 for the three respective geometries. The chosen values of \Delta/U correspond to the following distances from criticality: \bar{\Delta}=\Delta/U-\Delta_{{\rm c},(l_{x},l_{y})}/U\approx 0.0027 , -0.0002 , and -0.0004 , for l_{y}=1,2, and 3 , respectively. In Fig. S4 , we present the distribution of m=\rho e^{i\phi} in the complex plane, as well as the marginal distributions for its amplitude and phase, P(\rho) and P(\phi) . These marginal distributions are compared with the analytical predictions of Eq. ( S19 ) and Eq. ( S20 ) (red dashed line), respectively, evaluated using histogram binning with a chosen bin width. In all cases, the analytical fit agrees well with a Gaussian kernel density estimate of the data, which provides a smoother and more reliable characterization of the underlying distribution than binning alone. These results confirm that the radial distribution compatible with a simple local effective potential V(\rho)=-\mu\rho^{2}+\lambda\rho^{4} , with \mu,\lambda>0 , while the angular distribution is well described by a potential of the form V(\phi)=g_{3}\rho_{0}^{3}\cos(3\phi)+g_{6}\rho_{0}^{6}\cos(6\phi) . Here, the term g_{3} arises due to deviations from the critical point (i.e., g_{3}\propto\bar{\Delta}\neq 0 ), as confirmed by the fitted values of g_{3} , whose sign matches that of \bar{\Delta} . In Fig. S5 (a), we show the ratio \rho_{0}^{3}|g_{6}/g_{3}| [estimated from the radial distribution P(\rho) ] as a function of the number of shots for each l_{y} . In panel (b), we fix the number of shots to 10^{5} , where the estimates saturate, and show that \rho_{0}^{3}|g_{6}/g_{3}| scales as l_{y}^{2.7} . Analytically, this scaling is expected to arise from the dependence of |g_{6}/g_{3}| on l_{y} , while \rho_{0}^{2} is expected to be independent on l_{y} . This is supported by Fig. S6 , which shows that \rho_{0}^{2} , estimated from the radial distribution P(\rho) , exhibits only a weak dependence on l_{y} . II.5 Including the interaction up to the fifth-nearest neighbors Figure S7: Sampling of the staggered magnetization m [Eq. ( 2 )] for a cylinder with l_{y}=2 and l_{x}=10 at \Delta/U=3.1796 , including interactions up to the fifth-nearest neighbors (cf. second column in Fig. S4 for interactions up to the third-nearest neighbors). (a) Distribution of m over 10^{5} snapshots in the occupation basis, restricted to the bulk. (b) Corresponding angular distribution, fitted according to Eq. ( S19 ) (red dashed line). (c) Corresponding radial distribution, fitted according to Eq. ( S20 ) (red dashed line). Blue lines in (b)-(c) indicate a Gaussian kernel density estimate. Results obtained using a finite MPS with bond dimension \chi=250 for l_{y}=2 . Here, we consider the open cylinder with l_{y}=2 and l_{x}=5 , and we show that extending the range of interactions beyond the third-nearest neighbor—up to the fifth-nearest neighbor—does not affect the emergence of U( 1 ) symmetry at criticality. In the classical limit ( \Omega=0 ), the extended interaction tails induce a shift in the critical point, from \Delta/U\approx 3.158 to \Delta/U\approx 3.1796 . Nevertheless, as demonstrated in Fig. S7 , the two-dimensional and marginal distributions of the staggered magnetization near criticality remain consistent with those obtained in the shorter-range interaction case discussed previously [cf. second column in Fig. S4 ]. III Open systems including the interaction up to the fifth-nearest neighbors Here, we investigate the experimentally feasible setup shown in Fig. 3 (c)–(f), incorporating interactions up to the fifth-nearest neighbor. We demonstrate that the staggered magnetization m , computed from 10^{4} snapshots in the occupation basis, serves as an experimentally accessible observable to distinguish between the two ordered phases with \mathbb{Z}_{3} symmetry and the critical regime featuring an emergent U(1) symmetry. In Fig. S8 , we present the angular probability distribution P(\phi) of m as the system undergoes the phase transition between the 1/3 and 2/3 phases. In the ordered phases, P(\phi) exhibits three distinct maxima: at \phi=0,2\pi/3, and 4\pi/3 for \Delta/U=2.9 ( 1/3 phase), and at \phi=\pi/3,\pi, and 5\pi/3 for \Delta/U=3.2 ( 2/3 phase). At the transition ( \Delta/U\approx 2.92 ), an approximately U( 1 )-symmetric angular distribution emerges, which is thus robust to extension of the interaction tails [cf. Fig. 3 (c)–(e)]. In particular, from the fit of these angular distributions using Eq. ( S19 ), we can estimate the coefficients of the effective field theory presented in App. I . We consistently find that |g_{3}/g_{6}|\gg 1 in the \mathbb{Z}_{3} -ordered phases and g_{3} changes sign across the transition—being negative in the 1/3 phase and positive in the 2/3 phase. Figure S8: Statistics of the staggered magnetization for the experimentally feasible setup shown in Fig. 3 (c)–(f), incorporating interactions up to the fifth-nearest neighbor. (First column) \Delta/U=2.9 , in the 1/3 phase; (Second column) \Delta/U=2.94 , near the transition; and (Third column) \Delta/U=3.2 , in the 2/3 phase. For each cases, we show: (First row) the distribution of m obtained from 10^{5} snapshots in the occupation basis, restricted to the bulk; and (Second row) the corresponding angular distribution, fitted according to Eq. ( S19 ) (red dashed line). Results obtained using a finite MPS with bond dimension \chi=300 . IV Phase transition for varying \Omega/U Figure S9: Analysis of the transition between the 1/3 and the 2/3 phases as a function of \Omega/U , for an infinitely long cylinder with l_{y}=1 . (a) Correlation length \xi/a as a function of \Delta/U and \Omega/U . (b) Power-law decay of the two-point correlation function \langle m(r)m(0)\rangle\approx C_{1}(r)=\rho^{2}_{0}\langle e^{i\phi(r)}e^{-i\phi(0)}\rangle at criticality for \Omega/U\leq 0.43 . The colors refer to the \Omega/U values shown in (c). (c) Luttinger parameter K^{\prime} , extracted from linear fits shown (red lines) in (b), as a function of \Omega/U . Results obtained using a finite MPS with bond dimension \chi=150 . Here, we consider the infinitely long cylinder with the smallest circumference ( l_{y}=1 ) and analyze the phase diagram near the transition between the 1/3 and 2/3 phases as a function of \Omega/U . We find that the two ordered phases are separated by a line of DQCPs for small values of \Omega/U ( \Omega/U\lesssim 0.43 ) [see dashed yellow line in Fig. S9 (a)], where the two-point correlation function C_{1}(r) exhibits power-law decay [Fig. S9 (b)]. The location of the transition, i.e., the position of the maximum in the correlation length \xi , remains unchanged with varying \Omega/U . However, \Omega/U affects the Luttinger parameter K^{\prime} . Specifically, the exponent governing the power-law decay of the two-point correlation function \langle m(r)m(0)\rangle at criticality, which is given by K^{\prime}/2 , increases with increasing \Omega/U [Fig. S9 (c)].