Trotterization

qc/algorithms qc/math

The standard technique for turning $e^{-iHt}$ into an actual circuit when $H={\sum }_{j=1}^N{h}_j{P}_j$ is a sum of non-commuting Pauli terms — you can’t just exponentiate each term independently and multiply, because $e^{-iHt}\ne {\prod }_j{e}^{-i{h}_j{P}_{j}t}$ unless the terms commute. Trotterization approximates the product by breaking the evolution into $r$ small time slices.

First-order Trotter

$e^{-iHt}\approx {\left({\prod }_{j=1}^N{e}^{-i{h}_j{P}_{j}t/r}\right)}^r,\text{error }O(t^2/r)$

More slices ($r$) means a better approximation, at the cost of a deeper circuit — one layer of gates per slice.

Second-order (symmetric) Trotter

Splitting $H$ into two non-commuting pieces (e.g. $H_X$ and $H_{ZZ}$ for a transverse-field model) and symmetrizing:

$e^{-iHt}\approx {\left(e^{-i{H}_{X}t/2r}{e}^{-i{H}_{ZZ}t/r}{e}^{-i{H}_{X}t/2r}\right)}^r,\text{error }O(t^3/r^2)$

Key insight: the leading first-order error term cancels in the symmetric ordering, so second-order Trotter reaches the same accuracy with far fewer slices — better error scaling in $r$, at the cost of a more complex (three-part) gate sequence per slice.

The exact circuit family you already know

This is the same transverse-field Ising Hamiltonian and Trotter-step structure as 1D Ising Chain and the Mirror Trick — $R_x(\theta )$ layers for the transverse-field term, $CZ$-based layers for the $ZZ$ interaction term, repeated per Trotter step. The two notes use it for opposite purposes: that note treats the circuit as a fixed-structure noise benchmark (append $U^{†}$, check you get back to $|0{⟩}^{\otimes n}$) and never varies $t$ or $r$ for physics reasons. This note is about Trotterization as a simulation technique in its own right — choosing $r$ and the order to trade circuit depth against approximation error for an actual physics question, independent of any noise-benchmarking use.

Beyond Trotter (named, not derived here)

• QDRIFT — randomized Trotterization; sample which term to apply next, weighted by $|h_j|$, rather than a fixed deterministic order.
• Taylor expansion / quantum walk / Qubitization methods — fault-tolerant-era techniques with better asymptotic query complexity (Qubitization: $O(k\Vert H\Vert t+\log (1/\epsilon ))$ for sparsity $k$ and error $\epsilon$), not used in near-term algorithms in this vault.

Related

• Hamiltonians and Encoding for Quantum Circuits
• 1D Ising Chain and the Mirror Trick — the same circuit family, used as a mitigation benchmark instead
• Quantum Phase Estimation (QPE) — the next step once you can implement $e^{-iHt}$

Self-Check

• Why doesn’t $e^{-iHt}$ just equal ${\prod }_j{e}^{-i{h}_j{P}_{j}t}$ when the $P_j$ don’t commute?
• Why does second-order Trotter need fewer slices than first-order for the same accuracy?
• What’s the actual difference in purpose between this note’s use of the Ising Hamiltonian and 1D Ising Chain and the Mirror Trick’s use of it, given the circuit is nearly identical?