Nonlocal magic of fermions is bounded by entanglement fluctuations and is logarithmic for free fermions

Nonlocal Magic in Free‑Fermion Systems: A Logarithmic Bound

Free‑fermion states—such as those describing non‑interacting electrons, Hartree‑Fock approximations, or BCS mean‑field wavefunctions—are famously easy to simulate classically. Yet when expressed in the qubit language of quantum information, they often exhibit a substantial amount of magic, the resource that distinguishes generic quantum states from stabilizer states and enables universal quantum computation. This article explains why most of that magic is actually local and why the truly nonlocal component grows only logarithmically with system size.

---

1. Magic, Stabilizer Rényi Entropy, and Gaussian Restrictions

In the stabilizer formalism, magic quantifies how far a state lies outside the convex hull of stabilizer states. One convenient measure is the stabilizer Rényi entropy, which captures the amount of non‑stabilizer content. For fermionic systems, a natural class of local operations are Gaussian unitaries—those that preserve the quadratic structure of the fermionic Hamiltonian. Earlier work showed that if one restricts local unitaries to Gaussian ones, the nonlocal magic of a system consisting of (m) partially entangled fermion pairs grows linearly with (m).

However, this linear scaling is an artifact of the Gaussian restriction. Gaussian unitaries minimize the stabilizer Rényi entropy only for up to three entangled pairs; beyond that, more general parity‑preserving unitaries can achieve a lower nonlocal magic.

---

2. From Linear to Logarithmic Growth: The Role of Entanglement Fluctuations

When unrestricted local unitaries (still respecting fermion parity) are allowed, the picture changes dramatically. These operations can convert a collection of free‑fermion pairs into fermionic Bell pairs with a fidelity that decays at most as (1/\sqrt{m}\). By contrast, Gaussian unitaries lead to an exponential decay of fidelity. The key insight is that Bell pairs possess a flat entanglement spectrum: all Schmidt coefficients are equal. A state can remain far from such a spectrum only if its Schmidt weights span many orders of magnitude.

The spread of Schmidt weights is captured by the capacity of entanglement (CE), defined as the variance of the entanglement Hamiltonian. Importantly, the nonlocal magic of any fermionic state is bounded above by a logarithmic function of (CE). Since each free‑fermion pair contributes independently to (CE), the total capacity grows linearly with (m). Consequently, the nonlocal magic grows at most as \(\log CE \sim \log m\), and the stabilizer fidelity declines no faster than \(1/\sqrt{m}\).

---

3. Implications for Common Fermionic States

The logarithmic bound has concrete consequences for widely used fermionic wavefunctions:

• Hartree‑Fock states: Although they can be expressed as Slater determinants with extensive magic in the qubit basis, the nonlocal component is limited by the entanglement capacity and therefore scales only logarithmically.
• BCS mean‑field states: Even though they involve pairing correlations that look highly non‑local, the same capacity argument applies, keeping their nonlocal magic in check.
• General free‑fermion ensembles: Any state built from independent fermionic modes inherits the linear growth of (C_E) and thus the logarithmic bound on nonlocal magic.

In practice, this means that the bulk of the magic present in such states can be removed by local (possibly non‑Gaussian) operations, leaving only a modest, slowly growing nonlocal residue.

---

4. Why This Matters for Quantum Simulation

Understanding the distribution of magic is crucial for assessing the classical simulability of quantum systems. If most magic is local, one can apply local Clifford‑plus‑non‑Gaussian circuits to reduce the state to a form that is easier to sample classically. The logarithmic scaling of the remaining nonlocal magic suggests that even large free‑fermion systems may be amenable to efficient approximate simulation, provided the simulation algorithm can handle the modest residual nonlocal resource.

Moreover, the capacity‑of‑entanglement framework offers a quantitative tool for estimating magic in other many‑body contexts, linking entanglement fluctuations directly to computational hardness.

---

Closing Thoughts

The study of nonlocal magic in free‑fermion systems reveals a reassuring message: despite the apparent abundance of magic when fermionic states are mapped to qubits, the truly hard‑to‑remove, nonlocal component is tightly constrained. By relating magic to the variance of the entanglement Hamiltonian, the authors demonstrate that the nonlocal magic grows only logarithmically with the number of entangled pairs, a far slower pace than the linear growth seen under Gaussian‑only operations. This insight not only deepens our theoretical understanding of quantum resources in fermionic models but also points toward more efficient classical simulation strategies for a broad class of quantum many‑body states.