---
title: Peaked Circuits in Quantum Computing
url: https://www.emergentmind.com/topics/peaked-circuits
type: topic
---

# Peaked Circuits in Quantum Computing

A peaked circuit is a quantum circuit whose output probability distribution, when acting on a fixed input basis state (often $\lvert 0^n\rangle$), exhibits an anomalously high probability—far above the Haar-random background—on a specific computational basis string, known as the heavy or peaked output. Peaked circuits are of central importance in quantum computing as they simultaneously enable classically efficient verification (by checking the frequency of the heavy-output) and promise hardness of classical simulation, making them strong candidates for verifiable quantum advantage experiments [2404.14493].

## 1. Formal Definition and Quantitative Measures

Given an $n$-qubit unitary $C \in U(2^n)$, define the output probability for basis string $s \in \{0,1\}^n$:
\[
p_C(s) = |\langle s | C | 0^n \rangle|^2.
\]
$C$ is called $\delta$-peaked if
\[
\max_{s \in \{0,1\}^n} p_C(s) \geq \delta.
\]
The unique $x^* = \operatorname{argmax}_s p_C(s)$ is the heavy string, and $\delta_{x^*} = p_C(x^*)$ is the peak weight [2404.14493, 2510.00132, 2510.25838]. In practice, $\delta$ is usually set to be inverse polynomial in $n$ ($\delta \sim 1/\mathrm{poly}(n)$). Peakedness is contrasted to the Haar-typical weight $2^{-n}$ for random states, where all amplitudes are highly delocalized. Classical simulation algorithms exploit the fact that such peakedness can sometimes be correlated with the circuit structure, but for sufficiently deep random circuits peakedness is believed to be hard to engineer classically [2510.00132].

## 2. Circuit Constructions and Models

A generic peaked circuit is constructed as follows:
\[
C = P_{\tau_p} \cdot R_{\tau_r} \cdots R_1
\]
where
- $R_j$ are layers of uniformly random two-qubit gates (e.g., from the Haar measure), composing the scrambling or random phase.
- $P_i$ are parameterized (optimizable) "peaking" layers, engineered via classical optimization to maximize $\delta$ [2404.14493].
- $\tau_r$ denotes the number of random layers, $\tau_p$ the number of peaking layers.

For $\tau_p \ge \tau_r$, one can invert the randomization and attain $\delta = 1$; the regime $\tau_p \ll \tau_r$ is of maximal practical interest.

Alternative explicit protocols (e.g., [2508.07491]) construct a brick-wall circuit $Q$ and append (approximately) its inverse, with or without deliberate hidden modifications and obfuscations. Peaked multi-output and "double peaked" circuits can be realized by embedding entangler blocks at the circuit's front, followed by randomized and mirrored layers.

Random-peaked circuits can also be realized via postselection: two $k$-designs $C, C'$ are combined, and $C'$ is resampled until the overlap $|\langle r | C' C | 0^n \rangle|^2 \geq \delta$ for a fixed $r$; the resulting $C'^{\dagger}C$ is then $\delta$-peaked at $r$, with typical circuit complexity $\widetilde\Omega(nk)$ [2510.00132].

## 3. Analytical Complexity and Lower Bounds

Using approximate $t$-design properties and state complexity lower bounds, [2404.14493] showed that, for a polynomially small target peak weight $\delta = 1/\mathrm{poly}(n)$, achieving peakedness necessitates
\[
\tau_p = \Omega\left( \left(\frac{\tau_r}{n}\right)^{0.19} \right)
\]
with overwhelming probability, once the scrambling random circuit is deep enough ($\tau_r \gg n$). The proof applies design-theoretic arguments: random circuits of depth $\gtrsim n t^{5+o(1)}$ approach unitary $t$-designs; distinguishing such a state from the maximally mixed one with bias $\gtrsim 1/\mathrm{poly}(n)$ using shallow circuits would violate "strong state complexity" [2404.14493]. This constrains the overhead required for embedding a verifiable quantum advantage.

Typical random peaked circuits, constructed via postselected $k$-designs or equivalent protocols, also obey circuit complexity lower bounds: the expected gate count is $\Omega(nk/\log(nk))$, and with high probability cannot be implemented by circuits of substantially lower depth [2510.00132]. This rules out trivial compressibility or classical simulatability via circuit minimization or known tensor network contraction unless the protocol is structurally "mirror reducible" [2604.21908].

## 4. Numerical and Empirical Properties

Extensive simulations (e.g., [2404.14493], [2508.07491]) reveal that achievable peak weights for brick-wall architectures scale as
\[
\overline{\delta}_{0^n} \approx a^{-n}
\]
with $a \approx 1.19$, a significant enhancement over the Haar value $2^{-n}$ for moderate $n$ and depth, and with attainable $\overline\delta \sim 0.15$ for $n=10-12$, $\tau_r=50$, $\tau_p\approx8$. Even for $\tau_p \ll \tau_r$, a nontrivial exponential base is retained, confirming that circuit engineering outperforms Haar-typical random sampling [2404.14493].

Classical simulation of shallow peaked circuits is tractable via Matrix Product States (MPS) and related tensor network algorithms, with threshold bond dimension $\chi_{th}$ saturating for small circuit depth [2508.07491, 2309.08405]. For deeper instances, $\chi_{th} \rightarrow 2^{n/2}$, causing simulation cost to grow exponentially. Nonetheless, certain "mirror symmetric" constructions (including some HQAP circuits) are vulnerable to efficient classical contraction if their structure allows iterative reduction, as established in [2604.21908]. There, full extraction of the peak was achieved in $\approx 1$ hour on a single GPU, directly challenging claims of quantum advantage for these specific mirror-based families.

## 5. Verifiability, Quantum Advantage, and Hardness

Peaked circuits address the critical challenge of quantum advantage verifiability: given $(x^*, \delta)$, a verifier can ask for $m = O(1/\delta^2)$ samples from the circuit, and check whether $x^*$ appears with the expected frequency. This protocol requires only polynomial time in $n$ and $1/\delta$ for both the quantum device and the classical verifier, as opposed to the exponential post-processing for generic random circuit sampling [2404.14493, 2510.00132]. Crucially, without prior knowledge of $x^*$, a classical spoofer can do no better than random guessing.

Quantum advantage rooted in peaked circuits is supported by worst-case and average-case classical hardness results:
- Estimating the weight of the peak to exponential accuracy is #P-hard on average [2510.00132].
- For inverse-polynomial additive error, the problem becomes BQP-complete in the worst case.
- Under anti-concentration and random circuit simulation hardness assumptions, no polynomial-time sequential simulator can achieve inverse-polynomial precision for the peak on a non-negligible fraction of instances [2510.00132].
- In certain constructions, the underlying hardness is equivalent to weight enumerator estimation of random codes, known to be #P-hard [2510.05262].

The Hidden Code Sampling (HCS) protocol further generalizes this approach using CSS codes and transversal gates. There, output distributions are conditionally peaked: when conditioned on an observed syndrome string $s$, the logical register is almost deterministic. Verification requires two classical stages: checking the unique logical decoding per syndrome ("peak verification") and confirming the expected marginal distribution on syndromes ("syndrome verification") [2510.05262]. This procedure is exponentially more efficient than classical full simulation and is believed to resist spoofing on complexity-theoretic grounds.

## 6. Practical Implementations, Simulation, and Benchmarking

The PRC (Peaked Random Circuit) paradigm has been established as a powerful system-level fidelity benchmark for quantum hardware [2605.25983]. The fidelity metric
\[
F(N, d) = \hat p_{peak}(N, d) - 1/2^N
\]
(where $\hat p_{peak}$ is the observed frequency of the known peak string) is both easy to measure experimentally and sensitive to system decoherence, gate errors, and architectural constraints. Compared with Quantum Volume, PRC-based benchmarking decouples circuit width and depth, avoids the exponential post-processing hurdle, and exhibits higher signal-to-noise due to the unique peak structure. Experimental studies reveal clear fidelity "frontiers" as a function of $(N, d)$ across architectures, empirically mapping the device's capacity for coherent interference [2605.25983].

For practical quantum advantage demonstrations, modern protocols (e.g., HQAP) employ obfuscation layers, permutation masking, and circuit patch optimizations to make the hidden peak inaccessible to classical simulation. However, structural vulnerabilities such as U(θ)U(θ)† mirror symmetry can allow polynomial-time classical extraction of the peak (via MPO-based iterative contraction and "unswapping") unless high mid-circuit entanglement or genuinely hard code structures are used [2604.21908, 2510.25838]. Proposals leveraging code-based or syndrome-conditional peakedness offer fundamentally stronger guarantees of classical hardness [2510.05262].

## 7. Open Problems and Future Directions

Several main research directions remain for peaked circuits [2404.14493, 2510.00132, 2508.07491]:
- Efficient classical generation: Can one design provably efficient algorithms to construct physically realizable peaked circuits (given $n, \tau_r, \delta$) without suffering from barren plateau issues in the optimization landscape?
- Classical distinguishers: Determining whether a black-box circuit is $\delta$-peaked (without knowledge of the heavy string) is a candidate for post-quantum pseudorandomness; the computational complexity of this problem is open.
- Generalization to other ensembles: The extension to IQP, bosonSampling, and identity-obfuscated circuit classes is a rich area, potentially opening new quantum cryptographic protocols.
- Tightening analytic bounds: The established scaling $\tau_p = \Omega((\tau_r/n)^{0.19})$ may be improvable. Clarifying the optimal overhead for robust peakedness under circuit depth or peaking layer constraints is ongoing.
- Embedding computationally hard problems: Integration of intrinsically hard problem instances (e.g., codeword weights, NP-hard optimization) into the predicate determining the peak offers a robust path to cryptographically secure quantum advantage protocols [2510.00132, 2510.05262].
- Benchmarks of average-case vs. worst-case simulatability in the context of emerging classical algorithms.

The field is actively advancing toward quantum protocols that are simultaneously feasible, verifiable in polynomial time, and strongly resistant to classical spoofing, cementing peaked circuits as a foundational tool for quantum advantage, benchmarking, and post-quantum cryptography.

Source: https://www.emergentmind.com/topics/peaked-circuits