---
title: Spatial Superposition of Quantum Channels
url: https://www.emergentmind.com/topics/spatial-superposition-of-quantum-channels
type: topic
---

# Spatial Superposition of Quantum Channels

Searching arXiv for recent papers on spatial superposition of quantum channels and related coherent-control/channel-superposition work.
Spatial superposition of quantum channels is a coherent-control paradigm in which a target quantum system is routed through two or more spatially distinct communication links while an additional control degree of freedom—typically a path qubit or qudit—remains in superposition. In this setting, the effective transformation is not an ordinary classical mixture of channels but a channel whose off-diagonal interference terms depend on the coherent implementation of the underlying noisy processes, including their vacuum extensions and relative phases. Recent work has shown that such superpositions can enhance communication figures of merit, generate bipartite and multipartite entanglement during transmission, cancel certain noise processes by destructive interference, and in some formulations reproduce the effective action of the quantum switch [2605.02564], [2302.14820], [2606.10744], [2508.18077].

## 1. Formal definition and channel model

A conventional noisy channel $\hat E$ acting on a target Hilbert space $H_t$ admits a Kraus decomposition
$$
\hat E(\rho)=\sum_k E_k\,\rho\,E_k^\dagger,\qquad \sum_k E_k^\dagger E_k=I_t.
$$
To describe spatial superposition, the channel description is enlarged so that one can represent the possibility that the particle does not traverse a given link. One formulation introduces an extended space $H_t\oplus|vac\rangle$ together with “vacuum amplitudes” $\{\alpha_k\}$ satisfying $\sum|\alpha_k|^2=1$, yielding a dilation in which physically different choices of $\{\alpha_k\}$ correspond to different coherent implementations of the same reduced CPTP map $\hat E$ [2605.02564]. Closely related work formulates the same necessity as an expansion from qubit channels to qutrit channels with basis $\{|0\rangle,|H\rangle,|V\rangle\}$, where $|0\rangle$ denotes vacuum and each arm is properly described as a qutrit channel rather than a qubit channel [2302.14820].

For two channels $\hat E_1$ and $\hat E_2$ with Kraus data $\{F_i,\alpha_i\}$ and $\{N_j,\beta_j\}$, one convenient superposition operator is
$$
S_{ij}=\beta_j F_i\otimes|0_c\rangle\langle0_c|+\alpha_i N_j\otimes|1_c\rangle\langle1_c|,
$$
with joint map
$$
S(\hat E_1,\hat E_2)(\rho_t\otimes\rho_c)=\sum_{i,j}S_{ij}(\rho_t\otimes\rho_c)S_{ij}^\dagger.
$$
If the control is initialized in $|+_c\rangle=(|0\rangle+|1\rangle)/\sqrt2$, then tracing or measuring the control induces an effective channel on the target [2605.02564]. In a random-unitary formulation, the post-selected effective Kraus operators take the form
$$
K_{(i,j)}^{(p)}=\sqrt{p_i^{(a)}p_j^{(b)}}\,\frac{U_i^{(a)}+U_j^{(b)}}{2},
$$
which makes the coherent sum explicit [2302.14820].

A distinct but related formulation superposes Stinespring dilation unitaries $U_0$ and $U_1$ under a control qubit prepared in $\ket{+}$, postselects the control, and traces out the ancilla. The resulting map contains cross-terms of the form $U_0(\cdot)U_1^\dagger$, so it is not simply a convex mixture. In that framework, validity as a CPTP channel requires a state-independent normalization, equivalently
$$
\sum_i K_{1,i}^\dagger K_{0,i}\propto \mathbb I_S,
$$
a condition identified as the relevant Kraus-operator constraint for the superposed map [2606.10744].

## 2. Interference structure, vacuum coherence, and effective dynamics

The operational distinction between spatial superposition and classical path mixing lies in the preservation of off-diagonal control-space coherences. In the channel-block description for $d$ alternative paths, the joint input can be written as
$$
|\Psi^{\rm in}\rangle=\sum_{i=0}^{d-1}a_i\,|i\rangle_c\otimes|\phi^{\rm in}\rangle_t,
$$
and the output state decomposes into diagonal blocks propagated by the ordinary path channels and off-diagonal blocks
$$
\rho^{\rm out}(ij)=F^{(i)}_{\rm vio}\,\rho^{\rm in}(ij)\,F^{(j)\dagger}_{\rm vio},\qquad i\neq j,
$$
where
$$
F^{(i)}_{\rm vio}=\sum_{s_i}\alpha^{(i)*}_{s_i}K^{(i)}_{s_i}.
$$
These terms quantify the residual interference enabled by the vacuum amplitudes of the untraversed paths [2510.19092].

This same point appears in the qutrit-channel description of optical spatial superposition. For full depolarization $\alpha=1$, the phase-coherent implementation yields non-zero coherences in the Choi state between $\ket{00}$ and $\ket{VH}$ or $\ket{HV}$, whereas the phase-incoherent implementation produces a diagonal Choi state with no coherences [2302.14820]. The contrast isolates the role of coherent phase control: when relative phases are randomized, the effective map reduces to an ordinary mixture; when phases are fixed, a partially coherent map remains.

Vacuum coherence is therefore central. One source states explicitly that if $\alpha^{(i)}_s\equiv 0$ for all but one $s$, there is no path interference, whereas stronger vacuum coherence gives larger interference and better noise cancellation [2510.19092]. A microscopic dephasing example yields
$$
F_{\rm vio}=(1-2p_0)^{1/4}\,\mathbb1,
$$
providing a quantitative measure of residual vacuum coherence after noise [2510.19092]. This suggests that the physically relevant resource is not merely the existence of multiple paths, but the controlled retention of coherent amplitudes associated with path occupancy and non-occupancy.

## 3. Entanglement generation during distribution

A central recent development is the demonstration that spatial superposition of noisy links can generate entanglement inherently during transmission, including in regimes where each channel alone destroys entanglement [2605.02564]. For two-qubit outputs, the analysis uses concurrence
$$
C(\rho')=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},
$$
where the $\lambda_i$ are the square roots of the eigenvalues of $\rho'(\sigma_y\otimes\sigma_y)\rho'^*(\sigma_y\otimes\sigma_y)$ in decreasing order; negativity $N(\rho')=||\rho'^{T_A}||_1-1>0$ provides an equivalent entanglement witness [2605.02564].

The simplest deterministic example superposes the unitaries $U_1=X\otimes X$ and $U_2=Z\otimes Z$ acting on $\rho_t=|00\rangle\langle00|$ under a control prepared in $|+\rangle_c$. Writing
$$
S=U_1\otimes|0\rangle\langle0|+U_2\otimes|1\rangle\langle1|,
$$
one obtains
$$
S(|00\rangle\otimes|+\rangle)=\frac{|11\rangle\otimes|0\rangle+|00\rangle\otimes|1\rangle}{\sqrt2}
=\frac{|\Phi^+\rangle\otimes|+\rangle+|\Phi^-\rangle\otimes|-\rangle}{\sqrt2}.
$$
Measurement of the control in the $\{|+\rangle,|-\rangle\}$ basis therefore projects the target deterministically onto Bell states $|\Phi^\pm\rangle$ [2605.02564].

The same work states a multipartite extension: superposing $X^{\otimes n}$ and $Z^{\otimes n}$ on $|0\rangle^{\otimes n}$ under $|+\rangle_c$ deterministically generates the $n$-qubit GHZ state $(|0\ldots0\rangle+|1\ldots1\rangle)/\sqrt2$, up to a phase; using an $n$-dimensional control qudit prepared in $(1/\sqrt n)\sum|j\rangle$ and superposing $n$ single-bit flips $X_j$ on $|0\ldots0\rangle$ yields the $W$ state $(|10\ldots0\rangle+|01\ldots0\rangle+\ldots+|00\ldots1\rangle)/\sqrt n$ [2605.02564].

The noisy-channel version is more striking. For two identical depolarizing channels on two qubits,
$$
\hat E_p(\rho)=(1-p)\rho+\frac p3(X\otimes X\,\rho\,X\otimes X+Y\otimes Y\,\rho\,Y\otimes Y+Z\otimes Z\,\rho\,Z\otimes Z),
$$
appropriate choices of vacuum amplitudes can give output fidelity $F=1$ and concurrence $C=1$ even in zero-capacity regimes. Specifically, for $p=q=1$,
$$
\alpha_0=\beta_0=0,\quad \alpha_1=-\alpha_2=-\alpha_3=1/\sqrt3,\quad \beta_1=-\beta_2=-\beta_3=1/\sqrt3
$$
implies $F=1$ and $C=1$; for $p=q=1/2$,
$$
\alpha_0=-\beta_0=1/\sqrt2,\quad \alpha_1=-\alpha_2=-\alpha_3=1/\sqrt6,\quad \beta_1=-\beta_2=-\beta_3=1/\sqrt6
$$
again gives $F=1$ and $C=1$ [2605.02564]. In the formulation of that result, even though each channel alone breaks all entanglement, their coherent superposition reconstructs a perfect Bell pair.

## 4. Noise cancellation and capacity activation

Spatial superposition has also been used to show that some noise processes can interfere destructively at the level of effective channels. In the Stinespring-superposition framework, superposing two dephasing channels
$$
\mathcal D_p(\rho)=(1-p)\rho+p\,\sigma_z\rho\sigma_z
$$
yields another dephasing channel $\mathcal D_{\mathbb p}$ with
$$
\mathbb p=\frac1{2\gamma}\bigl(\sqrt{p_a}\cos\alpha+\sqrt{p_b}\sin\alpha\bigr)^2,
$$
where
$$
\gamma=\tfrac12\bigl[1+\sin2\alpha\,(\sqrt{p_a p_b}+\sqrt{(1-p_a)(1-p_b)})\bigr].
$$
Perfect cancellation occurs when $\sqrt{p_a}\cos\alpha+\sqrt{p_b}\sin\alpha=0$, equivalently $\tan\alpha=-\sqrt{p_a/p_b}$; for $p_a=p_b$, choosing $\alpha=3\pi/4$ makes $\mathbb p=0$ and restores the off-diagonal coherence [2606.10744].

The same framework treats depolarizing channels
$$
\mathcal P_q(\rho)=(1-q)\rho+\frac q3\sum_{i=1}^3 \sigma_i\rho\sigma_i.
$$
For $q\in[1/2,1]$, the channel is entanglement-breaking and has zero quantum capacity. Yet superposing two such channels produces an effective depolarizing channel $\mathcal P_{\mathbb q}$ with
$$
\mathbb q=\frac{3}{2\gamma}\Bigl(\sqrt{\frac q3}\cos\alpha+\sqrt{\frac{\tilde q}3}\sin\alpha\Bigr)^2,
$$
and this parameter can be pushed below the entanglement-breaking threshold even when both original channels are above it [2606.10744]. The same source states that for $\mathbb q\lesssim 0.189$, the channel acquires positive quantum capacity $Q>0$, establishing superactivation of quantum capacity by spatial superposition [2606.10744].

A separate line of work analyzes capacity through coherent information in optical spatial superposition experiments. For a channel $\mathcal N$ and input $\rho$,
$$
I_c(\rho,\mathcal N)=S[\mathcal N(\rho)]-S[(\mathcal N\otimes I)(|\Psi\rangle\langle\Psi|)],
$$
and numerical optimization shows a region $0<\alpha\lesssim0.4$ in which the post-selected superposed map $\Phi^{(p)}(\alpha)$ has larger maximal coherent information than the simple qubit depolarizing channel $N_\alpha$ [2302.14820]. However, that same analysis reports a strict hierarchy
$$
I_c(N_\alpha)\le I_c(\Phi^{(p)})\le I_c(qutrit)\le I_c(two\text{-}qutrit),
$$
and concludes that the apparent advantage disappears once the enlarged qutrit-channel description is used as the correct baseline [2302.14820].

These two strands are not identical. One concerns performance relative to a reduced qubit description versus a full vacuum-extended description [2302.14820]; the other concerns constructive interference between Stinespring dilations that produces effective Pauli channels with improved parameters, including positive capacity from zero-capacity constituents [2606.10744]. A plausible implication is that “capacity enhancement by superposition” is model-dependent: it can either denote an advantage over a restricted effective description or a genuinely improved effective channel relative to the original component channels, depending on what is held fixed.

## 5. Relation to the quantum switch and pure superchannels

Spatial superposition of channels is related to, but distinct from, coherent superposition of causal orders. The quantum switch is the prototypical process in which two input operations are applied in a coherent superposition of the orders $A\to B$ and $B\to A$. In the theory of pure two-slot superchannels, every reversibility-preserving bipartite superchannel is unitarily equivalent either to a fixed-order pure comb or to a coherent superposition of the two orders [2003.05682]. In the control-subspace form,
$$
U=|0\rangle\langle0|_C\otimes(U_BU_A)+|1\rangle\langle1|_C\otimes(U_AU_B),
$$
which is the quantum-switch structure [2003.05682].

Recent work based on quantum walks proposes that a two-hop walk in a spatial superposition of channels can reproduce the action of the quantum switch [2508.18077]. With system Hilbert space $\mathcal H_S$ and walker/path space $\mathcal H_W\cong\mathbb C^2$, one prepares
$$
|\psi_W\rangle=\frac{|0\rangle+|1\rangle}{\sqrt2},
$$
uses conditional-channel Kraus operators
$$
S_{ij}=E_i\otimes|0\rangle\langle0|+D_j\otimes|1\rangle\langle1|,
$$
and interleaves two such steps with a walker Pauli-$X$. After two hops, tracing out the walker gives
$$
\Phi_{\rm eff}(\rho_S)=\sum_{i,j}\bigl(E_iD_j\,\rho_S\,D_j^\dagger E_i^\dagger+D_jE_i\,\rho_S\,E_i^\dagger D_j^\dagger\bigr),
$$
which that work identifies with the trace-over-control form of the switch, hence $\Phi_{\rm eff}=\Phi_{\rm switch}$ [2508.18077].

On that basis, the same source states that all known capacity results of the quantum switch carry over. In particular, for two entanglement-breaking channels $\mathcal E,\mathcal D$, each with $Q=0$, it gives
$$
Q(\Phi_{\rm eff})=1
$$
for a noiseless qubit transmission, and for qubit depolarizing channels it states that the switch capacity exceeds $\max\{Q(\mathcal D_p),Q(\mathcal D_q)\}$ [2508.18077]. This suggests a possible bridge between experimentally accessible spatial superposition and indefinite-order advantages, though the proposal is explicitly presented as preliminary theoretical results [2508.18077].

## 6. Experimental realizations and network-level optimization

Optical experiments implement spatial superposition by encoding the control in interferometric path and the target in polarization. One representative setup uses a heralded single-photon source, a Sagnac interferometer to prepare the path qubit in $|+\rangle$, a Mach–Zehnder geometry with one channel per arm, and a balanced beam splitter for recombination and post-selection [2302.14820]. The channels in each arm are programmable random unitaries using liquid-crystal waveplates that choose among $\{I,X,Y,Z\}$, while a glass plate controls the relative path phase [2302.14820]. Another interferometric proposal for entanglement generation specifies beam-splitter transmissivities and phase shifters to tune the vacuum amplitudes, with final polarization-path measurement in the $\{|+\rangle,|-\rangle\}$ basis [2605.02564]. The same proposal lists feasibility parameters including path-length difference $\Delta \ell\ll$ coherence length of the photon, waveplates or decohering elements to realize arbitrary Pauli or depolarizing noise, and output tomography to verify fidelity and concurrence [2605.02564].

NMR has provided an alternative platform. A three-qubit NMR register was used to realize cancellation of two dephasing channels by superposing controlled Stinespring unitaries; by varying the control angle $\alpha$ and postselecting in $\ket{+}$, the experiment observed restoration of coherence for dephasing strengths $p=0.1$ and $\tilde p=0.45$ near $\alpha\approx2.70$, with postselection probability as low as $\gamma\sim0.06$ [2606.10744]. A five-qubit NMR register implemented superposition of two entanglement-breaking depolarizing channels with strengths $q=0.55$ and $\tilde q=0.70$, reporting a region $\alpha\in[2.37,2.51]$ where $\mathbb q<0.189$, indicating positive quantum capacity, and a near-perfect point $\mathbb q\approx0$, with the smallest postselection probability in that region falling to $\sim0.006$ [2606.10744].

At the network level, variational optimization has been proposed to choose the amplitudes and phases of path superpositions without channel tomography. In that framework, Alice applies $U_{c1}(\boldsymbol\theta_{c1})\in SU(d)$ to prepare $\sum_i a_i|i\rangle_c$, Bob applies $U_{c2}(\boldsymbol\theta_{c2})\in SU(d)$ before measuring the control, and then applies an outcome-dependent correction $V_b^{(j)}(\boldsymbol\theta_{\rm corr})\in SU(2)$ [2510.19092]. Performance is measured by the Choi–Jamiołkowski fidelity
$$
F_{\rm CJ}=\langle\Phi^+|\rho^{\rm out}_{ab}|\Phi^+\rangle,
$$
with deterministic and probabilistic objective functions defined in terms of measurement outcomes and success probabilities [2510.19092]. The protocol is explicitly described as a black-box loop using end-to-end fidelity evaluation and classical optimizers such as COBYLA or Nelder–Mead, with no process tomography of individual paths [2510.19092].

Reported performance figures include, for $d$ identical dephasing channels with perfect vacuum coherence, relative infidelity improvement up to factor $d$ in the limit $p_0\to0$, and for $d=2$ dephasing, $\mathcal R_{\rm det}=\mathcal R_{\rm prob}\approx2$ [2510.19092]. For non-identical dephasing and depolarizing channels with equal CJ fidelities, deterministic $\mathcal R_{\rm det}=3/2$ and probabilistic $\mathcal R_{\rm prob}=2$ are stated [2510.19092]. A 12-stage random network with approximately 40 channels is reported to achieve probabilistic $\mathcal R\approx2.25$ with success $\sim0.97$, converging within $O(10^2$–$10^3)$ outer-loop iterations even with 60 variational parameters [2510.19092].

## 7. Conceptual issues, limitations, and outlook

A recurring conceptual issue is whether the operational gains attributed to spatial superposition arise from genuinely new resources or from a higher-dimensional channel description already implicit in the implementation. One position states that the apparent gain in coherent information under superposition of depolarizing qubit channels is fully explained by the expanded qutrit channel description and its physical phase structure, so “no enhancement remains once the proper qutrit description is used from the outset” [2302.14820]. Another body of work treats the coherent implementation itself—through vacuum amplitudes, Stinespring superposition, or path interference—as the mechanism that converts noise into a constructive resource for entanglement generation and error cancellation [2605.02564], [2606.10744]. These views are not strictly contradictory, but they impose different comparison baselines.

Several limitations are stated explicitly in the source material. Postselection can be costly: in the NMR experiments, the success probability becomes very small near optimal interference points, reaching $\sim0.06$ for dephasing cancellation and $\sim0.006$ in the depolarizing-capacity experiment [2606.10744]. General non-Pauli channels require more careful enforcement of the CPTP constraint [2606.10744]. In network optimization, deterministic advantage can vanish when noise in the control degree of freedom exceeds the channel-cancellation scale, though the probabilistic method is described as more robust [2510.19092]. For quantum-walk emulations of the switch, generalization beyond two channels requires more hops and higher-dimensional coins, and full finite-size analysis of coin and routing noise remains open [2508.18077].

Within those constraints, the current literature converges on a common picture. Spatial superposition of quantum channels is a framework in which path coherence, vacuum extension, and controlled interference modify the effective channel beyond ordinary stochastic routing. Depending on the architecture and comparison class, it can deterministically generate Bell, GHZ, and W states during distribution [2605.02564], cancel dephasing and depolarizing noise by destructive interference [2606.10744], enhance coherent information relative to reduced descriptions [2302.14820], emulate the effective map of the quantum switch via a quantum-walk construction [2508.18077], and support black-box optimization of high-fidelity transmission across noisy quantum networks [2510.19092].

Source: https://www.emergentmind.com/topics/spatial-superposition-of-quantum-channels