Two edge classes from one bulk state
The paper asks whether the edge of a U(1) Dirac spin liquid can behave differently depending on how the emergent gauge field is terminated, without changing the bulk. The bulk is described at low energy by massless quantum electrodynamics in (2+1) dimensions (QED3), with Dirac spinons coupled to an emergent U(1) gauge field. The authors treat the system as a flat half-space and impose a reflecting boundary condition on the fermions, (1+γ_z)ψ=0, which forces the normal current to vanish at the surface. The gauge field is then given either a Neumann condition, F_za=0, or a Dirichlet condition, A_a=0, and the two choices are shown to define distinct boundary universality classes.
The distinction matters because a boundary universality class is not fixed by the bulk theory alone: it is set by which boundary conditions are compatible with the bulk and which boundary operators survive. The paper's answer is that the gauge field's boundary condition is one such selector. In the image-charge picture used for the propagators, Neumann corresponds to a same-sign image charge and Dirichlet to an opposite-sign image charge, which changes the one-loop corrections that boundary fields receive.
How the half-space RG isolates edge scaling
The calculation uses a perturbative renormalization group near four spacetime dimensions, with ε measuring the deviation from d=4. Bulk renormalization constants are the standard QED ones, with the infrared fixed point e*²=6π²ε/N for N fermion flavors. The new ingredient is a set of boundary renormalization constants, Z₂^∂ and Z₃^∂, which absorb ultraviolet divergences that appear only when external points are placed on the boundary at z=0.
The boundary fermion and gauge fields are written as ψ̂=√(Z₂Z₂^∂)ψ̂_R and Â_μ=√(Z₃Z₃^∂)Â_μ,R. The boundary scaling dimensions follow as Δ₂^∂=(d−1)/2+η₂^∂ and Δ₃^∂=(d−2)/2+η₃^∂, with the anomalous dimensions read off from the logarithmic scale derivative of the boundary renormalization factors. This construction is what allows the paper to assign separate edge scaling dimensions to the two gauge boundary conditions while keeping the same bulk fixed point.
The boundary fermion splits by gauge condition
The one-loop boundary fermion renormalization factor is Z₂^∂=1−(4+w)e²/(8π²ε), where w=+1 for Neumann and w=−1 for Dirichlet. The resulting boundary fermion scaling dimensions in Table I are Δ₂^∂=3/2+(15−4N)ε/(8N) for the Neumann case, labeled BC-I, and Δ₂^∂=3/2+(9−4N)ε/(8N) for the Dirichlet case, labeled BC-II. The difference comes from the fermion self-energy diagram, which contains the gauge propagator and therefore inherits its boundary condition.
The gauge field itself behaves differently. Its boundary renormalization factor is Z₃^∂=1−Ne²/(4π²ε) in both cases, so the boundary gauge field dimension is Δ₃^∂=1+ε/4 for Neumann and Dirichlet alike. The reason given is that the gauge-field renormalization is generated by the fermion polarization bubble, and the fermions obey the same reflecting boundary condition in both setups. At one loop, then, the gauge sector does not distinguish the two classes; the fermion sector does.
A pseudoscalar bilinear carries the observable signature
The paper identifies gauge-invariant fermion bilinears as the quantities that would make the boundary class visible. The scalar channel O_S=ψ̄ψ is forbidden at the boundary: the fermion boundary condition and its Hermitian conjugate imply ψ̂=−γ_zψ̂, which forces the boundary scalar bilinear to vanish. The pseudoscalar channel O_P=ψ̄γ⁵ψ survives, and its boundary anomalous dimension is computed from the three-point function with the operator inserted on the boundary.
The result is η_O^∂=−3/(4N)ε for Neumann and +3/(4N)ε for Dirichlet, so the two boundary classes differ in sign as well as in the fermion dimension. This is the paper's most decision-relevant comparison: the same bulk spin liquid, terminated with two gauge boundary conditions, produces opposite-sign anomalous dimensions for the same boundary operator. The authors connect this operator to a (π,π) charge-density-wave observable O=Σ_i(−1)^{i_x+i_y}c_i†c_i, which maps in the continuum onto the pseudoscalar bilinear ψ†τ_xσ_zψ, i.e. the boundary operator whose dimension is listed in Table I.
A lattice model with tunable edge coupling
To connect the continuum result to a microscopic setting, the paper proposes a square-lattice fermion-gauge model. The bulk Hamiltonian combines fermion hopping −tΣ c_i†e^{iθ_ij}c_j with a compact U(1) gauge field and a magnetic term KΣ_□cos(curl θ). For K>0 the ground state favors a π-flux configuration, realizing a U(1) deconfined spin liquid, and the π-flux pattern doubles the unit cell into A and B sublattices. In the absence of boundaries, the π-flux phase hosts two gapless Dirac cones at K=(π/2,π/2) and K′=(−π/2,π/2), whose low-energy expansion gives the Dirac Hamiltonian used in the continuum analysis.
Open boundaries at x=0 and x=L mix the two valleys, and requiring the wavefunction to vanish outside the system yields boundary conditions that support gapless boundary modes for L=2Z, matching the conformal fermion boundary condition −γ_xψ=ψ on the edge. The gauge boundary condition is then made tunable through a boundary interaction H_bdy=−K_BΣcos(θ_ij−φ_i+φ_j), where φ_i is a charged field on the boundary link. When K_B=0 the boundary gauge field is unconstrained and the Neumann condition follows; for K_B≫K the interaction locks θ_ij=φ_i−φ_j and pins the transverse boundary field, realizing the Dirichlet condition. This gives a concrete route to realizing and probing the two boundary classes, for example with quantum Monte Carlo or cold-atom platforms as the paper suggests.
Where the two boundary classes diverge
The paper's central comparison is not between bulk phases but between two edge terminations of the same bulk state. Neumann and Dirichlet gauge boundary conditions produce boundary fermion dimensions that differ at order ε, and pseudoscalar bilinear anomalous dimensions that differ in sign, while the one-loop boundary gauge dimension is shared. A reader evaluating an edge measurement should therefore track which boundary condition the gauge field obeys, because that choice, not the bulk spin-liquid identity, sets the boundary fermion and bilinear scaling.
The scope is set by the method: the results are one-loop perturbative values in an ε expansion near four spacetime dimensions, with the fermion boundary condition fixed to the reflecting choice and the gauge field restricted to homogeneous Neumann or Dirichlet conditions. The lattice model is proposed as a microscopic setting rather than a solved realization, and the paper frames nonperturbative study of the resulting boundary phase structure as the next step. Within those limits, the paper's priority is clear: the gauge field's boundary condition is the variable that separates the two boundary universality classes, and the pseudoscalar bilinear's sign change is the sharpest edge signature to look for.
Sources used in this answer
Boundary Criticality in (2+1)-dimensional U(1) Dirac Quantum Spin Liquid
A one-loop half-space RG analysis shows that Neumann and Dirichlet boundary conditions for the emergent gauge field produce distinct boundary universality classes in a (2+1)D U(1) Dirac spin liquid, with different boundary fermion and pseudoscalar bilinear scaling dimensions but the same one-loop boundary gauge dimension.