---
schema_version: 1
id: PHYS-2026-08-03-01
date: 2026-08-03
updated_at: 2026-08-03
title: A trivial Floquet endpoint does not trivialize the drive path
summary: "What the spin-chain twist algebra proves when the one-period unitary is the identity, what it does not prove, and how micromotion topology requires an independent path or defect invariant."
language: en
entry_kind: daily
status: published
level: graduate-advanced
user_difficulty: unrated
domains:
  - condensed-matter
  - quantum-information
  - mathematical-physics
estimated_minutes: 60
---

## Key claim

**The equation $U_F=I$ fixes the endpoint of a periodic drive. It does not specify the path $U(t)$, and it does not alter the symmetry representation carried by each unit cell.**

For a spin-$1/2$ chain, the algebra between a $2\pi$ twist and one-site translation survives even when the Floquet endpoint is the identity. That algebra records the microscopic Lieb–Schultz–Mattis input.

It does not, on its own, prove that the micromotion is a nontrivial Floquet loop. A zero Hamiltonian has the same endpoint and the same microscopic LSM data.

## Theme

**Flux insertion, LSM data, and the information lost by stroboscopic evolution.**

## Guiding question

Which statements follow from

$$
U_F=\mathcal T\exp\!\left[-\ii\int_0^\tau\dd t\,H(t)\right]=I,
$$

and which require the complete path $U(t)$ or its response to symmetry defects?

## Setup

Consider an even periodic chain of $L$ spin-$1/2$ sites. Let

$$
Q=\sum_{j=1}^LS_j^z
$$

generate an exact $U(1)$ spin rotation, and let $\mathsf T_x$ translate the chain by one site:

$$
\mathsf T_xS_j^z\mathsf T_x^{-1}=S_{j+1}^z,
\qquad
S_{L+1}^z=S_1^z.
$$

Work in the $Q=0$ sector. Define the large twist

$$
G=\exp\!\left(\frac{2\pi\ii}{L}
\sum_{j=1}^LjS_j^z\right).
$$

The drive is local and periodic,

$$
H(t+\tau)=H(t),
$$

and preserves $Q$ and $\mathsf T_x$ at every time. When a background $U(1)$ flux $\alpha$ threads the ring, denote the evolution by $U_F[\alpha]$.

> [!margin: Two distinct inputs]
> The twist algebra below follows from the on-site spin representation and translation. Whether a particular drive path is topologically nontrivial is a separate dynamical question.

## Analysis

### Derivation: the exact twist–translation algebra

Translate the exponent of $G$:

$$
\mathsf T_xG\mathsf T_x^{-1}
=\exp\!\left(\frac{2\pi\ii}{L}
\sum_{j=1}^LjS_{j+1}^z\right).
$$

Reindexing the sum gives

$$
\sum_{j=1}^LjS_{j+1}^z
=\sum_{j=1}^LjS_j^z+LS_1^z-Q.
$$

All terms in the exponent commute, so

$$
\mathsf T_xG\mathsf T_x^{-1}
=G\ee^{2\pi\ii S_1^z}\ee^{-2\pi\ii Q/L}.
$$

For a spin-$1/2$ site,

$$
\ee^{2\pi\ii S_1^z}=-I.
$$

Therefore, in the $Q=0$ sector,

$$
\boxed{\mathsf T_xG\mathsf T_x^{-1}=-G}.
$$

If

$$
\mathsf T_x\ket{\psi_p}=\ee^{\ii p}\ket{\psi_p},
$$

then

$$
\mathsf T_xG\ket{\psi_p}
=\ee^{\ii(p+\pi)}G\ket{\psi_p}.
$$

The large twist shifts crystal momentum by $\pi$. This is an exact finite-size operator identity; no spectrum or adiabatic approximation has entered.

The familiar static flux-insertion argument adds dynamics. If a unique gapped ground state can be followed under flux insertion, the state obtained after $2\pi$ flux has the same Hamiltonian after a large gauge transformation but a different momentum.

That conflicts with a unique symmetric gapped ground state at half-odd-integer spin per cell. Oshikawa formulated the flux-insertion version for conserved filling on a periodic lattice [Commensurability, excitation gap and topology](https://arxiv.org/abs/cond-mat/9911137).

> [!margin: What is exact]
> The momentum-shift algebra is exact. The static conclusion also uses locality, a spectral gap, a ground state, and control of flux insertion. Keep the algebra separate from those additional steps.

For the driven system, consistent coupling to a background gauge field gives endpoint covariance

$$
U_F[2\pi]=G,U_F[0],G^{-1},
$$

up to the chosen orientation of the flux and twist. If

$$
U_F[0]\ket{\psi_p}
=\ee^{-\ii\varepsilon\tau}\ket{\psi_p},
$$

then

$$
U_F[2\pi]G\ket{\psi_p}
=\ee^{-\ii\varepsilon\tau}G\ket{\psi_p}.
$$

After returning to the original gauge, the quasienergy is unchanged while the translation quantum number is shifted by $\pi$.

If $U_F[0]=I$, this covariance remains true. It also becomes spectrally weak: every state already has quasienergy $0$ modulo $2\pi/\tau$.

The twist still diagnoses the microscopic symmetry representation. It does not force a crossing inside a completely degenerate Floquet endpoint.

### Scope and assumptions: endpoint, loop, and effective Hamiltonian

The one-period operator remembers only the endpoint

$$
U(0)=I,
\qquad
U(\tau)=U_F.
$$

Two drives may have the same $U_F$ and different intermediate unitaries. When $U_F=I$, the evolution is a unitary loop:

$$
U(0)=U(\tau)=I.
$$

Some loops can be contracted to the constant loop through local symmetry-preserving evolutions. Others cannot within the phase equivalence relation being used.

The endpoint alone cannot distinguish them. Choosing

$$
H_F=\frac{\ii}{\tau}\log U_F=0
$$

also discards this distinction.

> [!margin: Classification must be named]
> “Nontrivial loop” is incomplete without a dimension, symmetry group, locality condition, and stability regime. Free-particle winding, MBL Floquet SPT pumps, and chiral many-body edge indices are related examples, not one universal invariant.

In two-dimensional noninteracting systems, Rudner and collaborators constructed a winding invariant for the full time evolution that detects anomalous chiral edge states even when bulk Floquet-band Chern numbers vanish [Anomalous edge states and the bulk-edge correspondence](https://arxiv.org/abs/1212.3324).

In interacting two-dimensional bosonic systems with a localized bulk, Po and collaborators classified chiral Floquet phases by a many-body edge index [Chiral Floquet Phases of Many-body Localized Bosons](https://arxiv.org/abs/1609.00006).

In symmetry-protected localized Floquet systems, Else and Nayak described additional dynamical phases as lower-dimensional SPT states pumped to a boundary during each period [Classification of topological phases in periodically driven interacting systems](https://arxiv.org/abs/1602.04804).

These examples establish that $U_F$ on a closed bulk can omit micromotion topology. They do not prove that every spin-$1/2$ chain with $U_F=I$ realizes one of these phases.

Indeed, take

$$
H(t)=0.
$$

Then $U(t)=I$ at all times. The spin chain still satisfies

$$
\mathsf T_xG\mathsf T_x^{-1}=-G,
$$

but its drive path is the constant loop. The microscopic LSM datum and the Floquet-loop invariant cannot be the same object.

### Physical interpretation: open boundaries and defect response

A loop that returns a closed bulk to the identity may leave a nontrivial locality-preserving unitary at an open boundary. A charge pump offers the simplest picture: motion through the periodic bulk cancels after a cycle, while an open sample accumulates opposite boundary charges.

This is why opening the system can reveal information absent from the closed-system endpoint.

Background fields provide another diagnostic. For a density matrix $\rho$, define the Schwinger–Keldysh functional

$$
Z[A_1,A_2]
=\Tr\!\left(U[A_1]\rho U^\dagger[A_2]\right).
$$

At equal fields, unitarity gives $Z[A,A]=1$. Relative fields probe response of the driven path.

Glorioso, Gromov, and Ryu used this construction to formulate many-body response invariants for selected Floquet topological systems [Effective response theory for Floquet topological systems](https://arxiv.org/abs/1908.03217).

> [!margin: Crystalline caveat]
> Their demonstrated examples do not by themselves construct the mixed translation–$U(1)$ response of this spin chain. Translation defects require crystalline background data beyond an ordinary electromagnetic gauge field.

The LSM obstruction can itself be described through symmetry defects and the obstruction to gauging internal symmetry while preserving translation. A lattice formulation is developed by [Seifnashri](https://arxiv.org/abs/2308.05151), and a broader 2026 review is given by [Zou and Cheng](https://arxiv.org/abs/2604.00347).

Combining a crystalline LSM defect theory with a Floquet Schwinger–Keldysh response is a reasonable research program. The sources above supply its ingredients in different settings.

They do not establish the blanket theorem

$$
\text{LSM datum}+U_F=I
\quad\Longrightarrow\quad
\text{nontrivial micromotion loop}.
$$

Heating supplies a further limitation. A generic clean interacting Floquet system may absorb energy and fail to support a stable eigenstate phase.

Many-body localization, integrability, coupling to a suitable environment, or a long prethermal window can make dynamical classifications useful, but each option changes the assumptions.

## False claim to diagnose

> If $U_F=I$, the drive is topologically trivial and the spin chain has evaded its LSM obstruction.

The first clause confuses an endpoint with a path. The second confuses a choice of dynamics with the projective symmetry representation in each unit cell.

The corrected statement is:

> $U_F=I$ makes all stroboscopic quasienergies trivial on the closed system. Microscopic LSM data remain, while micromotion topology must be tested independently.

## What follows — and what does not

| Statement | Status |
| --- | --- |
| $\mathsf T_xG\mathsf T_x^{-1}=-G$ in the stated $Q=0$ spin-$1/2$ sector. | Exact finite-size algebra |
| $G$ shifts momentum by $\pi$. | Exact consequence of that algebra |
| The static system cannot have a unique symmetric gapped ground state. | LSM conclusion with locality and ground-state assumptions |
| The same proof forces a Floquet quasienergy gap closing. | False; quasienergy has no ground-state ordering |
| $U_F=I$ determines the entire drive. | False; it fixes only the endpoint |
| $U_F=I$ permits nontrivial unitary loops. | True in specified Floquet classifications |
| The LSM twist algebra proves this particular loop is nontrivial. | False without a path or defect invariant |
| A trivial loop removes the microscopic LSM datum. | False |
| An ordinary $U(1)$ SK response already captures translation defects. | False in general |
| A generic clean interacting drive realizes a stable Floquet phase forever. | Not without control of heating |

## Exercise

Use the periodic spin-$1/2$ chain defined above.

1. Derive $\mathsf T_xG\mathsf T_x^{-1}$ by reindexing the twist exponent.
2. Show that $G\ket{\psi_p}$ has momentum $p+\pi$ in the $Q=0$ sector.
3. Prove the quasienergy relation using $U_F[2\pi]=GU_F[0]G^{-1}$.
4. Explain why no quasienergy crossing follows when $U_F=I$.
5. Compare the constant loop $U_0(t)=I$ with any loop $U_1(t)$ that has the same endpoint but a nontrivial open-boundary action. Which data distinguish them?

<details>
<summary>Hint 1</summary>

After translating, isolate the boundary contribution $LS_1^z$ and the total charge $Q$.

</details>

<details>
<summary>Hint 2</summary>

The endpoint spectrum cannot distinguish two loops that both satisfy $U(\tau)=I$. Inspect a homotopy invariant, an open-boundary unitary, or a background-field response.

</details>

**Oral check 1.** Which result survives every change of $H(t)$ that preserves the same local spin representation and symmetries?

**Oral check 2.** Give one reason the static variational LSM argument cannot be copied directly to quasienergy.

<details class="solution">
<summary>Solution</summary>

Translation gives

$$
\sum_{j=1}^LjS_{j+1}^z
=\sum_{j=1}^LjS_j^z+LS_1^z-Q.
$$

Therefore

$$
\mathsf T_xG\mathsf T_x^{-1}
=G\ee^{2\pi\ii S_1^z}\ee^{-2\pi\ii Q/L}.
$$

For spin $1/2$ and $Q=0$, this reduces to $-G$. Hence

$$
\mathsf T_xG\ket{\psi_p}
=-G\mathsf T_x\ket{\psi_p}
=\ee^{\ii(p+\pi)}G\ket{\psi_p}.
$$

Gauge covariance gives

$$
U_F[2\pi]G\ket{\psi_p}
=GU_F[0]\ket{\psi_p}
=\ee^{-\ii\varepsilon\tau}G\ket{\psi_p}.
$$

Thus the gauge-related state has the same quasienergy and shifted momentum.

When $U_F=I$, all states already share quasienergy zero modulo $2\pi/\tau$. The momentum shift acts inside this extensive degeneracy and does not force a distinguished crossing at quasienergy zero or $\pi/\tau$.

Both $U_0(t)$ and $U_1(t)$ have the same endpoint spectrum. They may differ in the homotopy class of the local-unitary path, their open-boundary endpoint, or their response to symmetry defects.

The twist–translation algebra is identical for both because it belongs to the microscopic spin representation, not to the chosen loop.

</details>

## Check your understanding

Suppose two translation- and $U(1)$-symmetric drives act on the same spin-$1/2$ chain and both satisfy $U_F=I$. One is the constant loop; the other pumps a protected boundary charge in a regime where that invariant is well defined.

List the data that agree and the data that differ. Which part is fixed by the LSM anomaly?

You may also reply with “deeper,” “too easy,” “too hard,” or your derivation.

## Further Reading

- [Commensurability, excitation gap and topology in quantum many-particle systems](https://arxiv.org/abs/cond-mat/9911137)
- [Lieb-Schultz-Mattis anomalies as obstructions to gauging symmetries](https://arxiv.org/abs/2308.05151)
- [Lieb-Schultz-Mattis Anomalies and Anomaly Matching](https://arxiv.org/abs/2604.00347)
- [Anomalous edge states and the bulk-edge correspondence for periodically driven systems](https://arxiv.org/abs/1212.3324)
- [Classification of topological phases in periodically driven interacting systems](https://arxiv.org/abs/1602.04804)
- [Chiral Floquet Phases of Many-body Localized Bosons](https://arxiv.org/abs/1609.00006)
- [Effective response theory for Floquet topological systems](https://arxiv.org/abs/1908.03217)

## Connections and next step

- Microscopic layer: the projective spin representation fixes the twist–translation algebra.
- Static layer: energy ordering and a ground-state gap turn that algebra into the LSM constraint.
- Floquet endpoint: $U_F$ fixes quasienergies but discards the path taken during a cycle.
- Dynamical layer: loop homotopy, boundary action, and defect response can retain micromotion topology.
- Stability layer: heating determines whether a Floquet phase is asymptotic or only prethermal.
- Next step: construct a specific symmetric loop and compute its open-chain endpoint rather than inferring it from $U_F$.
- Revisit: compare the SK response here with the counting-field deformation in the July 31 entry.