Bose Glass Phase Overview

• Bose glass is a gapless, compressible phase where disorder localizes bosonic excitations, preventing global superfluid coherence.
• It is characterized by finite compressibility and vanishing superfluid stiffness, with local superfluid puddles existing in an insulating background.
• Realizations span disordered Bose-Hubbard models, quasiperiodic lattices, and cavity-mediated setups, supported by local diagnostics and rare-region effects.

Bose glass is a compressible, gapless, non-superfluid phase of interacting bosons in which disorder, quasidisorder, or self-generated inhomogeneity destroys global phase coherence without producing a Mott gap. In the standard disordered Bose-Hubbard setting it appears between Mott-insulating and superfluid regions, while in broader realizations it is controlled by rare low-energy regions, local superfluid puddles, or non-percolating superfluid clusters. The same phenomenology has been established or proposed in quenched-disorder models, quasiperiodic lattices, cavity-mediated self-organization, quantum magnets mapped to dirty bosons, and several cold-atom and correlated-oxide experiments (Pal et al., 2018, Niederle et al., 2013, Habibian et al., 2012, Yu et al., 2023).

1. Defining characteristics

The operational definition of the Bose glass is remarkably stable across models: it is insulating or non-superfluid, yet compressible and gapless. In lattice formulations this is expressed through finite compressibility,

$$\kappa=\frac{\partial \langle \hat n\rangle}{\partial \mu},$$

together with vanishing superfluid stiffness,

$$\rho_s=\frac{L}{8\pi^2}\left.\frac{\partial^2 E_0}{\partial\varphi^2}\right|_{\varphi=0},$$

or an equivalent current-correlation diagnostic, and the absence of global phase coherence (Pal et al., 2018). In one-dimensional and continuum formulations, the same distinction is often made through exponentially decaying one-body correlations and zero superfluid fraction at finite $\kappa$ (Yao et al., 2020).

Within the disordered Bose-Hubbard model, the standard picture is that Mott insulator (MI), Bose glass (BG), and superfluid (SF) phases coexist, with BG intervening between MI and SF in the presence of disorder (Pal et al., 2018). A geometric refinement identifies BG as the Griffiths region of the disordered Bose-Hubbard model: finite quasi-superfluid clusters exist in a Mott-insulating background, but these clusters do not percolate. The MI is then the regime with no such clusters, while SF begins when they percolate and support macroscopic phase coherence (Niederle et al., 2013).

This distinction is especially important on finite disordered systems. Disorder-averaged gaps can remain misleadingly finite even when rare low-energy regions render the thermodynamic phase compressible. For that reason, the full gap distribution $P(E_g)$ and the minimum-gap estimator $E_g^{\rm min}$ can be more informative than averaged gaps when identifying BG on finite clusters (Carrasquilla et al., 2010). At finite temperature, several works further distinguish BG from a normal fluid: both are compressible and non-superfluid, but the BG retains disorder-dominated short-range coherence or local superfluid islands, whereas the normal fluid is the finite-temperature continuation of a melted glass (Pal et al., 2018, Yao et al., 2020).

2. Microscopic routes to Bose-glass formation

The most familiar route to Bose-glass physics is diagonal quenched disorder. In the two-dimensional disordered Bose-Hubbard model, random site-energy offsets $\epsilon_{p,q}\in[-D,D]$ create spatially inhomogeneous regions where some sites or clusters are locally superfluid while others remain insulating; this Griffiths structure yields a phase that is compressible but not globally phase coherent (Pal et al., 2018). Disorder can also enter through site dilution in magnetic systems, where it localizes bosonic quasiparticles generated by a spin-to-boson mapping (Yu et al., 2010).

Quasiperiodicity provides a second route. In a blue-detuned optical Penrose quasicrystal, increasing the potential depth causes a two-dimensional condensate to approach a gapless insulating phase consistent with BG behavior; because the potential consists of peaks rather than wells, the usual Mott-insulating mechanism is absent, so insulating behavior is attributed to localization in the quasiperiodic structure (Cetoli et al., 2011). In a one-dimensional continuum Lieb-Liniger gas in a shallow bichromatic potential, a genuine single-particle localization transition occurs at

$V_c \simeq 1.375\pm 0.008,$

and for $V>V_c$ a Bose glass emerges between superfluid and Mott regions in an intermediate interaction window (Yao et al., 2020). An eight-fold optical quasicrystal realizes the same logic experimentally in two dimensions: weak repulsive interactions connect localized puddles, and the BG-to-SF transition occurs when those puddles merge into a globally coherent state (Yu et al., 2023).

A third route dispenses with external disorder altogether. In a high-finesse cavity with an optical lattice incommensurate with the cavity wavelength, cavity backaction produces a density-dependent, nonlocal, self-consistent potential. Quantum fluctuations enable coherent photon scattering into the cavity, the resulting intracavity field amplifies density modulations, and the atoms reorganize into clusters with checkerboard density order; in the resulting regime the ground state has finite compressibility but no superfluidity, which is identified as a Bose glass (Habibian et al., 2012). A related self-organized mechanism appears in quantum cluster quasicrystals with a Lifshitz-Petrich-Gaussian pair interaction: increasing interaction strength drives a sequence consistent with super quasicrystal $\to$ Bose glass $\to$ quasicrystal insulator, with global superfluidity essentially zero but local ring superfluidity still finite in selected coronas (Grossklags et al., 2023).

An artificial gauge field does not by itself create BG in the cited work, but it enlarges its domain once quenched disorder is present. In the two-dimensional disordered Bose-Hubbard model, a Peierls phase

$$\rho_s=\frac{L}{8\pi^2}\left.\frac{\partial^2 E_0}{\partial\varphi^2}\right|_{\varphi=0},$$0

suppresses itinerancy and enhances the Bose-glass region; at $$\rho_s=\frac{L}{8\pi^2}\left.\frac{\partial^2 E_0}{\partial\varphi^2}\right|_{\varphi=0},$$1 and $$\rho_s=\frac{L}{8\pi^2}\left.\frac{\partial^2 E_0}{\partial\varphi^2}\right|_{\varphi=0},$$2, for example, the BG-like region extends to larger $$\rho_s=\frac{L}{8\pi^2}\left.\frac{\partial^2 E_0}{\partial\varphi^2}\right|_{\varphi=0},$$3 for $$\rho_s=\frac{L}{8\pi^2}\left.\frac{\partial^2 E_0}{\partial\varphi^2}\right|_{\varphi=0},$$4 than for $$\rho_s=\frac{L}{8\pi^2}\left.\frac{\partial^2 E_0}{\partial\varphi^2}\right|_{\varphi=0},$$5 (Pal et al., 2018).

Route	Representative realization	Key feature
Quenched disorder	Disordered Bose-Hubbard model	Local SF islands in insulating background
Quasiperiodicity	Penrose or bichromatic optical lattices	Gapless localization without random disorder
Cavity backaction	Incommensurate lattice in a resonator	Self-generated nonlocal quasidisorder
Self-organized quasicrystal	Lifshitz-Petrich-type interactions	Global insulating response with local ring superfluidity
Gauge-field-enhanced glassiness	Disordered Bose-Hubbard model with flux	Synthetic flux enlarges BG domain

3. Field-theoretic structure and glassy phenomenology

In one-dimensional disordered Bose fluids, bosonization and nonperturbative FRG describe the BG as a fully attractive strong-disorder fixed point. The clean low-energy theory is the Tomonaga-Luttinger Hamiltonian

$$\rho_s=\frac{L}{8\pi^2}\left.\frac{\partial^2 E_0}{\partial\varphi^2}\right|_{\varphi=0},$$6

while disorder generates replica-coupled backscattering terms (Dupuis et al., 2019). The FRG flow drives the renormalized Luttinger parameter to zero, $$\rho_s=\frac{L}{8\pi^2}\left.\frac{\partial^2 E_0}{\partial\varphi^2}\right|_{\varphi=0},$$7, and yields a singular disorder correlator with a cusp. This cusp is interpreted as the field-theoretic signature of metastable states, pinning, and static avalanches, placing the Bose glass within a droplet picture of glassy matter (Dupuis, 2019, Dupuis et al., 2019).

At any finite RG scale, quantum tunneling rounds the cusp into a quantum boundary layer (QBL). Its width is set by the running Luttinger parameter,

$$\rho_s=\frac{L}{8\pi^2}\left.\frac{\partial^2 E_0}{\partial\varphi^2}\right|_{\varphi=0},$$8

and the rounded QBL encodes rare superfluid regions that dominate low-energy dynamics (Dupuis et al., 2019). The same FRG construction yields a low-frequency dissipative conductivity with

$$\rho_s=\frac{L}{8\pi^2}\left.\frac{\partial^2 E_0}{\partial\varphi^2}\right|_{\varphi=0},$$9

a hallmark traced to tunneling between metastable states and to rare superfluid droplets embedded in an otherwise localized background (Dupuis, 2019).

The same fixed-point structure implies chaos in the glassy sense. For two copies of a one-dimensional disordered Bose fluid with slightly different disorder realizations or slightly different Luttinger parameters, inter-copy correlations are lost beyond an overlap length

$\kappa$0

and the linear stability analysis yields $\kappa$1 (Daviet et al., 2021). In this formulation, the instability of the BG ground state to infinitesimal perturbations is tied directly to the cuspy fixed-point correlator and to a chaos boundary layer analogous to the QBL.

In higher-dimensional lattice theories, strong-coupling expansion plus replica RG leads to a complementary glassy picture. For the disordered Bose-Hubbard model in $\kappa$2, the BG carries a replica-symmetry-breaking order parameter in particle-density fluctuations rather than in the superfluid order parameter itself. The MI-to-BG transition is governed by a one-step RSB fixed point, and the susceptibility to infinitesimal RSB perturbations diverges with exponent

$\kappa$3

which is interpreted as evidence of non-self-averaging and ergodicity breaking (Thomson et al., 2013). In disordered dimerized quantum antiferromagnets, where the bosonic description applies to triplons, the inclusion of RSB is likewise essential to obtain a finite compressibility in the compressible glass phase; away from the lobe tips the glass is a standard compressible BG, whereas near the tips compressibility is strongly suppressed and the results indicate a rare Mott glass (Thomson et al., 2015).

4. Diagnostics and observables

Phase identification in Bose-glass problems relies on combining local and global observables. Standard lattice diagnostics include the local condensate amplitude $\kappa$4, local number fluctuations

$\kappa$5

compressibility $\kappa$6, and superfluid stiffness $\kappa$7 (Pal et al., 2018). In the disordered Bose-Hubbard model, the Edwards-Anderson order parameter

$\kappa$8

serves as a localization measure: it is zero in MI, small in weakly disordered SF, and substantially larger in BG. The cited work emphasizes, however, that near the tip of the Mott lobes number fluctuations are strongly suppressed, so $\kappa$9 alone becomes unreliable and must be combined with other observables (Pal et al., 2018).

Local-density statistics provide a more microscopic view. In the canonical disordered Bose-Hubbard model, the BG at incommensurate filling $P(E_g)$0 and $P(E_g)$1 emerges at a superfluid-to-BG transition near $P(E_g)$2, and the local-density distribution becomes strongly skewed, with a peak near $P(E_g)$3 and a long tail toward higher occupancies. By contrast, near commensurate filling $P(E_g)$4 and $P(E_g)$5, the MI-to-BG transition occurs near $P(E_g)$6, but the local-density distribution remains much more Gaussian-like (Hettiarachchilage et al., 2017). The same work reports multifractal behavior near the incommensurate transition and a percolation probability of non-integer-filled sites reaching about $P(E_g)$7 near $P(E_g)$8, close to but below the superfluid critical disorder (Hettiarachchilage et al., 2017).

Correlation functions and collective modes are equally diagnostic in quasiperiodic continua. In the Penrose quasicrystal study, the spatially averaged one-body correlator

$P(E_g)$9

shrinks rapidly with increasing lattice depth, low-lying dipole and quadrupole modes soften and split, and the normal fraction inferred from rotational response grows, all consistent with a crossover toward a gapless insulator (Cetoli et al., 2011). In shallow bichromatic continua, BG and normal fluid are distinguished not by $E_g^{\rm min}$0 or superfluid fraction alone, since both are compressible and non-superfluid, but by the temperature dependence of the correlation length extracted from

$E_g^{\rm min}$1

BG exhibits a low-temperature plateau in $E_g^{\rm min}$2, whereas the normal fluid shows a strong thermal decrease (Yao et al., 2020).

Site-resolved experiments have added genuinely local glass diagnostics. In a two-dimensional square lattice with reproducible disorder, the local Edwards-Anderson parameter

$E_g^{\rm min}$3

rises in the BG regime, while Talbot interferometry measures the short-range coherence length. The cited quantum-gas-microscope study reports a maximum possible $E_g^{\rm min}$4 but experimental values around $E_g^{\rm min}$5 at the strongest disorder, which the authors attribute to finite temperature, and a coherence length decreasing from $E_g^{\rm min}$6 at $E_g^{\rm min}$7 to $E_g^{\rm min}$8 at the largest disorder (Koehn et al., 17 Apr 2025).

5. Variants and related glass phases

Magnetic realizations make explicit the dirty-boson mapping that underlies much of BG theory. In a two-dimensional site-diluted array of coupled antiferromagnetic dimers, the external magnetic field acts as a chemical potential for bosonic quasiparticles. Site dilution liberates local moments at low field and localizes triplet quasiparticles on intact dimers at higher field, producing two continuously connected Bose-glass regimes and, for

$E_g^{\rm min}$9

an overlap region interpreted as a two-species Bose glass (Yu et al., 2010). Finite-size scaling of the BG-to-ordered transition in this magnetic system yields

$\epsilon_{p,q}\in[-D,D]$0

consistent with the expected dirty-boson relation $\epsilon_{p,q}\in[-D,D]$1 in two dimensions (Yu et al., 2010).

Spin degrees of freedom can qualitatively reshape BG formation. In the spin-1 Bose-Hubbard model with diagonal disorder, disorder mixes neighboring Mott states of different filling and generates a compressible but insulating glass. Antiferromagnetic spin correlations can stabilize singlet Mott lobes and produce a singlet Bose glass, while in some disorder channels the BG disappears because fluctuations are aligned with a Mott boundary and therefore cannot mix distinct MI sectors (Paganelli et al., 2011). This is a sharp departure from scalar Bose-Hubbard intuition.

Mixtures support still more structure. In a one-dimensional disordered Bose-Fermi mixture, weak disorder and density-density coupling can produce not only the standard bosonic BG but also a Bose-Fermi glass in which both bosonic and fermionic density fluctuations are localized and remain coupled. The RG criterion

$\epsilon_{p,q}\in[-D,D]$2

governs the relevance of disorder, and once fermions localize their finite localization length feeds back onto the bosonic flow, creating coupled glass phases beyond the single-component BG scenario (Crépin et al., 2010). This suggests that Bose-glass phenomenology is not restricted to purely bosonic Hilbert spaces so long as the low-energy sector retains localized, compressible bosonic modes.

The relation between BG and Mott glass remains model-dependent. In dimerized quantum antiferromagnets, the rare incompressible glass inferred near particle-hole-symmetric lobe tips contrasts with the standard compressible BG found away from the tips (Thomson et al., 2015). The cited results indicate that glassy insulating phases in disordered bosonic systems are not exhausted by weak-localization language alone.

6. Experimental realizations and open problems

Cold-atom experiments now probe BG directly in two dimensions. In an eight-fold optical quasicrystal loaded with about $\epsilon_{p,q}\in[-D,D]$3 $\epsilon_{p,q}\in[-D,D]$4 atoms, time-of-flight coherence measurements map the BG-to-SF boundary in the weakly interacting regime. The critical depth in the noninteracting limit approaches

$\epsilon_{p,q}\in[-D,D]$5

and increasing interactions shifts this boundary to larger lattice depths, consistent with interaction-driven restoration of coherence (Yu et al., 2023). The same experiment demonstrates non-ergodicity by showing that coherence is not fully restored after the system has evolved in the Bose-glass regime and is then transformed into a periodic lattice where the ground state should be superfluid (Yu et al., 2023).

Quantum-gas microscopy has made the local structure of BG experimentally accessible. In a two-dimensional square optical lattice with DMD-generated site-resolved disorder, in-situ density distributions, local particle fluctuations, the local Edwards-Anderson parameter, and Talbot interferometry are combined to identify a finite-temperature Bose-glass regime and to track its short-range coherence (Koehn et al., 17 Apr 2025). By driving the system into and out of the glass, the same work finds incomplete recovery of coherence and enhanced EA response, which it interprets as evidence of non-ergodic dynamics (Koehn et al., 17 Apr 2025).

Beyond cold atoms, a correlated-oxide realization has been proposed in Ca$\epsilon_{p,q}\in[-D,D]$6RuO$\epsilon_{p,q}\in[-D,D]$7 nanofilms. The authors report a Bose-glass regime in the weak-localization region, characterized by

$\epsilon_{p,q}\in[-D,D]$8

a distinctive “vertical flow” in $\epsilon_{p,q}\in[-D,D]$9-function scaling, and two quantum critical points separating BG from superconducting and Mott-insulating regimes. They further interpret the data as evidence that localized Cooper pairs persist up to about $V_c \simeq 1.375\pm 0.008,$0 and that vortex mobility produces finite resistance despite bosonic order (Nobukane et al., 29 Sep 2025). This interpretation remains specific to that material platform, but it extends BG phenomenology into strongly correlated electronic matter.

Several open issues remain active across the cited literature. Finite temperature is one. Some works present the BG as historically a ground-state phase but also argue that it survives at finite temperature in experimentally relevant windows; others introduce a normal fluid that replaces the BG as local superfluid puddles melt (Pal et al., 2018, Koehn et al., 17 Apr 2025). Universality is another. In shallow bichromatic continua, the emergence of BG only after renormalization-group generation of a dense set of harmonics raises the possibility that the transition may not fall into the same universality class as standard random-disorder problems (Yao et al., 2020). A further unresolved theme is the relation between Bose glass, many-body localization, and broader glassy dynamics. Optical-quasicrystal experiments and site-resolved disordered lattices both emphasize non-ergodicity and incomplete adiabatic response, suggesting that BG is not only a thermodynamic phase label but also a dynamical regime with slow relaxation, memory, and strong sensitivity to microscopic structure (Yu et al., 2023, Koehn et al., 17 Apr 2025).

In that sense, the Bose glass is best understood not as a single microscopic mechanism but as a phase class: a gapless, compressible, globally insulating state whose detailed realization may involve rare superfluid regions, non-percolating clusters, quasiperiodic localization, self-generated nonlocal feedback, or replica-symmetry-broken glassiness. The common content is the coexistence of bosonic localization with low-energy density response, and the persistent failure of local coherence to organize into macroscopic superfluid order.