We study invertible phases protected by non-onsite symmetries carrying nontrivial lattice anomaly indices. While lattice anomalies are usually viewed as obstructions to symmetric short-range-entangled (SRE) states, we show that anomalous lattice symmetries can nevertheless admit invertible phases that are not SRE, and can moreover shift the set of symmetry-compatible invertible phases. In particular, we consider a fermionic
Z4F symmetry in (2+1) dimensions. For an onsite
Z4F symmetry, symmetric invertible phases have integer chiral central charge
c−∈Z, whereas a non-onsite symmetry with nontrivial lattice anomaly index obstructs all such phases despite having trivial continuum 't Hooft anomaly. We resolve this by constructing an exact exponentially quasi-local
Z4F symmetry of a
p+ip superconductor with
c−=1/2. We compute its lattice anomaly, and find that the symmetry indeed forbids
c−∈Z invertible states. The allowed invertible phases are consequently shifted to
c−∈Z+1/2. By gauging the fermion parity of the
p+ip superconductor, one obtains an Ising topological order enriched by a
Z4 symmetry. We describe the corresponding symmetry structure in terms of a fusion 2-category. We further show that the
p+ip state admits an enlarged non-onsite
U(1)f symmetry, providing an analogue of fractional quantum Hall response in an invertible phase. This non-onsite
U(1)f symmetry again forbids Chern insulators carrying
c−∈Z, while being compatible with invertible states with
c−∈Z+1/2. This result motivates a systematic study of new classes of invertible phases and spin liquids enriched by non-onsite symmetries.
Nonlinear dynamics generate non-Gaussian states and operations, but they also make continuously monitored quantum systems difficult to track. A quantum filter performs this task by compressing a noisy measurement record into a set of evolving variables that predict future observables. For Gaussian dynamics, this compression is exact and finite. In nonlinear systems, conventional moment equations generally expand without bound, but the failure of one representation does not prove that every finite description is impossible. Here we formulate quantum filtering as a realization problem for general quantum input--output maps. We introduce the linearized observable Hankel operator
KO, which maps changes in the past measurement record to changes in a future conditional observable. Any finite-dimensional smooth filter forces
KO to have finite rank, so infinite rank provides a coordinate-independent proof that no exact finite-dimensional filter exists. Applied to single-mode polynomial bosonic systems under continuous quadrature monitoring, this yields a sharp boundary: Gaussian dynamics and the measurement-aligned Conditional Momentum Moment class admit finite realizations, whereas dynamics outside these classes produce infinitely many response directions and admit no finite-dimensional smooth exact filter. Exact impossibility, however, need not imply large practical complexity. The singular values of
KO quantify how many response directions matter at a chosen local accuracy, defining the effective memory of an observable. For the Kerr and Duffing oscillators, these singular values decay rapidly despite infinite exact rank, showing that infinite exact memory can coexist with low finite-accuracy complexity. The framework therefore separates exact realizability from effective complexity in nonlinear quantum filtering.
A qubit carries one of two non-orthogonal label states, exchanged by a Pauli-𝑍 symmetry. A channel must produce a record of the label while returning the qubit, with the returned qubit’s deviation from the label held within a fixed disturbance budget; the record’s deviation from the label is the error to be minimised. We compare three implementation classes at one operating point (label overlap 3/5, budget 1/10): environment-free isometries intertwining the symmetry; channels that may use and discard an environment on which the symmetry acts trivially (equivalently, classical mixtures of the first class, with the budget on the mixture); and all channels. The optimal errors are bracketed by exact rational certificates: a feasible channel and an exactly dual-feasible point for each of the two convex classes, and an exact one-variable reduction with a rigorous interval enclosure for the nonconvex environment-free class. The unrestricted class lowers the minimal error by between 17.2789% and 17.3105% relative to the environment-free class; the environment-free optimum and the mixture-class optimum differ by less than 1.09 ⋅ 10−6, so more than 99.968640% of the reduction lies in the step from the mixture class to the unrestricted class. Equality of the first two optima is not established. All certificates are verified by exact arithmetic from standalone files; the verification requires no optimisation solver.