Interaction-resolved decomposition of multi-qubit unitaries via computational-basis phases

This paper introduces an interaction-resolved decomposition method using support-selective phase invariants to uniquely resolve and control the k-body interaction structure of multi-qubit unitaries, enabling the selective synthesis of specific many-body interactions as demonstrated in a nitrogen-vacancy spin register.

The Gist

Imagine you are trying to bake a complex cake, but instead of tasting the final product to see if it's good, you are only allowed to look at a single, blurry photo of the whole cake. You know the cake should have chocolate, vanilla, and strawberry layers, but the photo just shows a big, messy blob. You can't tell which flavor is where, or if the baker accidentally mixed them all together.

This is the problem with current quantum computers. Scientists want to build specific "quantum cakes" (operations that link multiple particles together). Usually, they check their work by comparing the final result to a perfect picture of the target. If the picture is slightly off, they know something went wrong, but they don't know what went wrong. Did the chocolate layer get too thick? Did the vanilla disappear? They have to guess.

This paper introduces a new way to look at these quantum operations. Instead of looking at the blurry "whole cake" photo, the authors give us a special pair of glasses that lets us see the individual ingredients (the interactions between particles) clearly, even while they are mixed together.

Here is how their method works, broken down into simple concepts:

1. The "Diagonalizing Frame": Rotating the Cake

Many quantum operations are like a cake that has been spun around. It's hard to see the layers when it's spinning. The authors suggest a trick: rotate the cake until the layers line up perfectly with your view. In physics terms, they apply a simple local rotation to the system. Once rotated, the complex operation becomes "diagonal."

Think of this as turning a jumbled pile of colored marbles so that all the red ones are on the left, blue in the middle, and green on the right. Suddenly, you can see exactly how much of each color you have without the colors blending into a muddy brown.

2. The "Phases": The Secret Recipe Numbers

Once the operation is "rotated" into this clear view, it leaves behind a set of numbers called phases. You can think of these phases as the "recipe numbers" for the cake.

• Some numbers tell you about the local flavor of a single particle (like just vanilla).
• Other numbers tell you about how two particles talk to each other (vanilla and chocolate mixing).
• The most important numbers tell you about three or more particles talking at once (a complex triple-flavor swirl).

3. The "Support-Selective Phase Invariants": The Magic Sieve

This is the paper's biggest innovation. The authors created a mathematical "sieve" (a filter).

• If you put the recipe numbers through this sieve, it keeps the numbers for a specific group of particles (say, particles A, B, and C) and throws away everything else.
• It's like having a sieve that only lets through the "three-way conversation" between particles A, B, and C, while ignoring any two-way chats or single-party monologues.

They call these filtered numbers "support-selective phase invariants." They are "invariants" because they stay the same even if you change the local details (like the order of the ingredients), but they change if the actual interaction between the particles changes.

4. The Result: Cooking with Precision

Using this new "sieve," the authors showed they could control quantum computers much more precisely.

• The Goal: They wanted to create a specific interaction where three particles (an electron and two nuclear spins in a diamond) talk to each other in a very specific way, without any of them accidentally talking to just one or two others.
• The Method: Instead of trying to hit a perfect "whole cake" target, they told the computer: "Make sure the 'three-way conversation' number is exactly 45 degrees, and make sure all the 'two-way conversation' numbers are zero."
• The Outcome: They successfully baked this specific "three-way interaction cake" using a single pulse of microwave energy.
  • For a "diagonal" interaction (where the particles just talk in a straight line), they achieved 99.78% accuracy.
  • For a "non-diagonal" interaction (where the particles talk in a more twisted, complex way), they achieved 99.85% accuracy.

Why This Matters (According to the Paper)

Currently, to get a three-particle interaction, scientists usually have to string together many smaller two-particle gates, like building a tower out of many small bricks. This paper shows you can build that same tower with one single, shaped brick (one control pulse).

By using these "magic sieves" (the invariants), they can tell the computer exactly which interaction they want to build and ignore the rest. This makes the process faster and cleaner, potentially reducing errors that happen when you have to stack too many small steps on top of each other.

In short: The paper gives us a new way to "see" and "tune" the specific conversations happening between quantum particles, allowing us to build complex quantum interactions in a single step rather than a long, messy chain of steps.

Technical Summary

Technical Summary: Interaction-Resolved Decomposition of Multi-Qubit Unitaries via Computational-Basis Phases

Problem Statement
In multi-qubit quantum control, target unitary operations are traditionally specified by full-unitary descriptions and assessed using global comparison measures (e.g., process fidelity). The authors argue that these conventional approaches fail to explicitly resolve the underlying many-body interaction structure of a realized operation or its deviation from a target. In multi-qubit systems, where operations comprise combinations of interaction terms across various qubit subsets, this limitation hinders the direct assessment and targeting of specific many-body terms. This is particularly relevant for stabilizer-based quantum error correction and related settings, where specific multi-qubit parity operators play a central role but are typically realized through decompositions into lower-order gates rather than direct synthesis.

Methodology
The paper introduces an interaction-resolved decomposition of $n$-qubit unitaries that provides explicit access to their many-body interaction structure. The core methodology relies on working within a diagonalizing frame where the unitary is fully characterized by the phases acquired by computational-basis states.

1. Diagonalizing Frames: For a broad class of operationally relevant operations (including gates generated by single Pauli strings or commuting sets of Pauli strings, such as stabilizer operations, controlled-phase gates, and Ising interactions), a local transformation $U_P$ exists that simultaneously diagonalizes the generator components. In this frame, the unitary $U_{\text{diag}}$ acts on basis states $|\vec{x}\rangle$ as $U_{\text{diag}} |\vec{x}\rangle = e^{-i\phi(\vec{x})} |\vec{x}\rangle$.
2. Support-Selective Phase Invariants: The authors derive a method to recover the strength of specific $k$-body interaction terms ($\theta_S$) from the computational-basis phases $\phi(\vec{x})$. By constructing parity-weighted sums of these phases, they define quantities termed support-selective phase invariants, denoted $\phi(S)$:
   $$\phi(S) = 2^{-n} \sum_{\vec{x} \in \{0,1\}^n} (-1)^{\sum_{i \in S} x_i} \phi(\vec{x})$$
   These invariants exactly and uniquely resolve the interaction strength supported on a specific subset $S$ of qubits while canceling contributions from complementary subsets. Mathematically, this corresponds to the Walsh-Hadamard transform of the computational-basis phases.
3. Equivalence Classes: The framework induces generalized equivalence classes where two unitaries are considered equivalent if they share specific interaction coordinates (phase invariants) for a chosen set of supports, while remaining free to differ on others. This recovers "local equivalence" as a special case (where all non-local interaction coordinates are fixed).
4. Optimal Control Formulation: The authors formulate quantum optimal control objectives directly in terms of these invariants. Instead of minimizing the deviation from a full target unitary, the cost function $J$ minimizes the deviation of selected phase invariants $\phi(S)$ from target values $\phi^*_S$:
   $$J = \sum_{S \in S_{\text{target}}} w_S \left[ 1 - \cos\left(2(\phi(S) - \phi^*_S)\right) \right]$$
   This allows for "interaction-selective optimization," where specific many-body terms are synthesized while local or irrelevant degrees of freedom are left unconstrained.

Key Results
The methodology was demonstrated through numerical simulations on a realistic solid-state spin register model based on a nitrogen-vacancy (NV) center in diamond coupled to two proximal $^{13}\text{C}$ nuclear spins. The control fields were optimized to synthesize specific three-qubit entangling gates within a single shaped microwave pulse.

• Diagonal Case (ZZZ): The authors synthesized a diagonal three-qubit entangler $e^{i\frac{\pi}{4} Z \otimes Z \otimes Z}$. By optimizing the phase invariants to target the tripartite interaction while suppressing pairwise terms, they achieved a final fidelity of $F_{ZZZ} = 0.9978$ with a pulse duration of 1.5 $\mu$s.
• Non-Diagonal Case (XZZ): The framework was extended to a non-diagonal target $e^{i\frac{\pi}{4} X \otimes Z \otimes Z}$. This was achieved by performing the optimization in a Hadamard-sandwiched diagonalizing frame, where the target becomes diagonal. After optimization, the diagonalizing transformations were removed. The resulting pulse achieved a fidelity of $F_{XZZ} = 0.9985$ with a duration of 1.25 $\mu$s.

In both cases, the time evolution of the phase invariants confirmed that the control pulse successfully built up the desired three-body interaction while suppressing two-body contributions. Residual local phases were corrected via virtual Z corrections (or additional single-qubit gates in the non-diagonal case).

Significance and Claims
The paper claims that this interaction-resolved decomposition provides a coordinate system that organizes unitary operations according to their multipartite interaction structure. This enables:

• Direct Access: Explicit access to local, pairwise, tripartite, and general $k$-partite interaction content underlying entangling operations without requiring full process tomography.
• Selective Synthesis: The ability to formulate control objectives that synthesize prescribed combinations of many-body interactions while suppressing unwanted terms, effectively allowing the direct synthesis of isolated tripartite interactions.
• Efficiency: The demonstration that nontrivial higher-weight multi-qubit interactions can be synthesized by a single shaped pulse, suggesting a route toward reduced circuit depth and potentially lower accumulated control errors compared to decomposing these operations into sequences of two-qubit gates.

The authors conclude that this approach is particularly relevant for stabilizer-based quantum error correction and that the underlying phase maps can be experimentally reconstructed via conditional Ramsey interferometry, enabling closed-loop interaction-resolved characterization and optimization.