State-Dependent Quantum Copying

• The paper establishes that a family of quantum states can be exactly copied if and only if their density matrices commute, thereby linking cloning to simultaneous diagonalizability.
• It introduces conditional copying mechanisms where state-specific ancillas enable exact replication without violating the no-cloning theorem, as seen in stimulated emission models.
• The study also details approximate cloning strategies, highlighting the influence of ensemble geometry and asymmetry on optimal fidelity and implications for quantum cryptography.

State-dependent quantum copying denotes a class of copying tasks in which the copying map is defined only on a specified family of inputs, or in which exact copying becomes possible only because the task is weakened or supplemented by side information. In finite-dimensional quantum theory, its sharp exact boundary is given by the no-broadcasting theorem: a family of states admits a common broadcasting operation if and only if its density matrices commute pairwise, and for pure states this reduces to the condition that the corresponding vectors are pairwise orthogonal (Maassen et al., 2 Jul 2026). Around that exact boundary, the subject includes restricted-ensemble approximate cloners, local copying of bipartite resources whose performance depends on the input state, virtual copying with non-CPTP maps, and cryptographic settings in which the number of copies of a specific state family is itself the central resource or threat model (Bi et al., 23 Jul 2025).

1. Exact copying, cloning, and broadcasting

In the finite-dimensional $C^*$-algebraic formulation, a state on an algebra $A\subset M_d$ is a positive normalized linear functional, and copying devices are unital completely positive maps $C:A\otimes A\to A$ in the Heisenberg picture, with dual $C^*:S(A)\to S(A\otimes A)$ in the Schrödinger picture. For a pure state $\varphi$, cloning means

$C^*\varphi=\varphi\otimes\varphi,$

whereas broadcasting means that both output marginals equal the input state: $$\varphi\circ C(a\otimes \mathbf 1)=\varphi(a),\qquad \varphi\circ C(\mathbf 1\otimes a)=\varphi(a).$$ For pure states these notions coincide; for mixed states broadcasting is weaker because the two outputs need not be independent (Maassen et al., 2 Jul 2026).

The central exact theorem is state-dependent rather than universal. For a specified set $S$ of states on a matrix algebra $A\subset M_d$, the following are equivalent: the states in $S$ have a common broadcaster, every state in $A\subset M_d$0 is a convex combination of states with pairwise orthogonal support projections, and the density matrices of the states in $A\subset M_d$1 commute pairwise. In finite dimensions this is equivalent to simultaneous diagonalizability and containment in a common abelian subalgebra. For pure states, the same statement reduces to the usual no-cloning criterion: a family of pure states has a common cloner if and only if the corresponding vectors are orthogonal (Maassen et al., 2 Jul 2026).

This exact criterion separates state-dependent copying from universal copying. A universal copier exists only when the observable algebra is commutative, in which case copying reduces to the classical diagonal-copy map. By contrast, a singleton family is trivially broadcastable, because one may always output a fixed bipartite state with the required marginals. The theory is therefore fundamentally about a single operation that must work on an entire specified family, not about custom operations tailored separately to each input.

2. Conditional exact copying and side-information models

A distinct exact notion appears when the ancillary system is allowed to depend on the input state. In that setting one may define a unitary $A\subset M_d$2 such that

$A\subset M_d$3

where the ancilla $A\subset M_d$4 is linearly correlated with $A\subset M_d$5. The construction is explicit in a basis $A\subset M_d$6, with

$A\subset M_d$7

so that linearity gives $A\subset M_d$8. This is exact copying, but it is not universal cloning, because the ancilla is not a fixed blank register. The same paper interprets stimulated emission as a physical model of this state-dependent mechanism: an excited atom functions as an “adaptive ancilla,” and the effective cloning domain is restricted to those photon polarizations for which the dipole matrix element

$A\subset M_d$9

is nonzero, equivalently to states in

$C:A\otimes A\to A$0

The paper also formulates the corresponding symmetry condition as $C:A\otimes A\to A$1 (Kadam, 21 Jul 2025).

Another exact but narrower model appears when the state is already known. A state-parameterized $C:A\otimes A\to A$2 unitary depending on amplitudes $C:A\otimes A\to A$3 can map

$C:A\otimes A\to A$4

for a given qubit $C:A\otimes A\to A$5. A single unitary can also clone the two orthogonal Hadamard states

$C:A\otimes A\to A$6

and only those two by that particular transformation. The same work’s “nesting doubled qubits” circuit is not an orthodox cloning machine: after entangling gates and measurement, it yields post-measurement branches of the form

$C:A\otimes A\to A$7

rather than $C:A\otimes A\to A$8. This clarifies a recurrent terminological issue: exact copying by a state-specific unitary or state-matched ancilla does not contradict the no-cloning theorem, because the operation is neither universal nor blank-state based (Grigoryan et al., 2022).

3. Approximate copying on restricted state families

The standard approximate theory of state-dependent copying is organized by the input ensemble. For $C:A\otimes A\to A$9 qubit cloning under the assumptions

$C^*:S(A)\to S(A\otimes A)$0

the optimization can be written in terms of the operator

$C^*:S(A)\to S(A\otimes A)$1

and reduced to the spectral problem of an anisotropic Heisenberg star Hamiltonian. The ensemble enters through

$C^*:S(A)\to S(A\otimes A)$2

while asymmetry enters through the weights $C^*:S(A)\to S(A\otimes A)$3. Universal qubit cloning corresponds to $C^*:S(A)\to S(A\otimes A)$4, equatorial cloning to $C^*:S(A)\to S(A\otimes A)$5. In the symmetric case the paper gives closed formulas for the optimal fidelity; for the universal ensemble it recovers

$C^*:S(A)\to S(A\otimes A)$6

and for equatorial cloning it obtains

$C^*:S(A)\to S(A\otimes A)$7

It also proves that odd $C^*:S(A)\to S(A\otimes A)$8 admits economical implementation, even $C^*:S(A)\to S(A\otimes A)$9 needs at most one ancilla qubit, and arbitrary asymmetric equatorial cloning is economical. In the $\varphi$0 equatorial case the trade-off connects directly to CHSH monogamy through

$\varphi$1

(Kay et al., 2012).

A different state-dependent phenomenon appears for finite prescribed sets of coplanar qubit states. In a deterministic ancilla-free $\varphi$2 cloner optimized for equally spaced states within a fixed angular spread $\varphi$3, the relevant quality measure is the average global fidelity $\varphi$4. The paper tests the intuition that more a priori information, quantified by the Shannon entropy $\varphi$5 of an equiprobable $\varphi$6-state ensemble, should always imply better cloning. It finds instead that the decisive geometric quantity is the “denseness”

$\varphi$7

and that for fixed $\varphi$8 the optimal global fidelity can increase with $\varphi$9 when $C^*\varphi=\varphi\otimes\varphi,$0. The authors’ conclusion is explicit: Shannon entropy of the prior does not determine clone quality; the no-cloning obstruction is dominant for $C^*\varphi=\varphi\otimes\varphi,$1, while denseness is more crucial for $C^*\varphi=\varphi\otimes\varphi,$2 (Hwang et al., 2010).

These two strands capture the main approximate use of the term. State dependence may arise because a machine is deliberately optimized for a restricted ensemble, as in phase-covariant or equatorial cloning, or because the geometry of the allowed set changes the attainable fidelity even when the task remains deterministic.

4. Bipartite resources, entanglement copying, and local broadcasting

State dependence becomes especially sharp for composite systems. In the experiment on “Cloning of Quantum Entanglement,” the elementary devices are two local universal $C^*\varphi=\varphi\otimes\varphi,$3 single-qubit cloners, but the induced two-qubit entanglement-cloning performance depends on the initial bipartite state. The protocol takes an input

$C^*\varphi=\varphi\otimes\varphi,$4

and two singlet ancilla pairs,

$C^*\varphi=\varphi\otimes\varphi,$5

applies local partial-teleportation cloners on $C^*\varphi=\varphi\otimes\varphi,$6 and $C^*\varphi=\varphi\otimes\varphi,$7, and traces out the unused modes. In the symmetric setting $C^*\varphi=\varphi\otimes\varphi,$8, the Bell-state input $C^*\varphi=\varphi\otimes\varphi,$9 leads theoretically to

$$\varphi\circ C(a\otimes \mathbf 1)=\varphi(a),\qquad \varphi\circ C(\mathbf 1\otimes a)=\varphi(a).$$0

with Bell fidelity

$$\varphi\circ C(a\otimes \mathbf 1)=\varphi(a),\qquad \varphi\circ C(\mathbf 1\otimes a)=\varphi(a).$$1

The experiment reports

$$\varphi\circ C(a\otimes \mathbf 1)=\varphi(a),\qquad \varphi\circ C(\mathbf 1\otimes a)=\varphi(a).$$2

with negative witness values

$$\varphi\circ C(a\otimes \mathbf 1)=\varphi(a),\qquad \varphi\circ C(\mathbf 1\otimes a)=\varphi(a).$$3

so both clone pairs remain entangled. The paper states explicitly that, despite using universal single-state quantum cloning machines locally, the bipartite entanglement cloning is state-dependent and that the maximally entangled Bell state is the worst case for cloning fidelity (Peng et al., 2020).

Related local-broadcasting behavior is analyzed in a theoretical study of two-qubit universal and state-dependent cloners. There the naive local use of the Buzek–Hillery single-qubit machine on

$$\varphi\circ C(a\otimes \mathbf 1)=\varphi(a),\qquad \varphi\circ C(\mathbf 1\otimes a)=\varphi(a).$$4

produces the reduced clone fidelity

$$\varphi\circ C(a\otimes \mathbf 1)=\varphi(a),\qquad \varphi\circ C(\mathbf 1\otimes a)=\varphi(a).$$5

so local copying of a bipartite state is already state-dependent. The same paper proposes an optimal universal local cloner with state-independent fidelity

$$\varphi\circ C(a\otimes \mathbf 1)=\varphi(a),\qquad \varphi\circ C(\mathbf 1\otimes a)=\varphi(a).$$6

and two state-dependent local machines with fidelities

$$\varphi\circ C(a\otimes \mathbf 1)=\varphi(a),\qquad \varphi\circ C(\mathbf 1\otimes a)=\varphi(a).$$7

and

$$\varphi\circ C(a\otimes \mathbf 1)=\varphi(a),\qquad \varphi\circ C(\mathbf 1\otimes a)=\varphi(a).$$8

In the subsequent broadcasting analysis, quantum discord survives over wider parameter intervals than entanglement for both pure and Werner inputs, leading to the paper’s conclusion that discord is broadcast more robustly than entanglement (Kheirollahi et al., 2016).

A more specialized exact construction appears in entangled cloning of stabilizer codes and free fermions. There the target state is not arbitrary but the ground state of a known Hamiltonian $$\varphi\circ C(a\otimes \mathbf 1)=\varphi(a),\qquad \varphi\circ C(\mathbf 1\otimes a)=\varphi(a).$$9, and the coupled auxiliary system $S$0 is arranged so that the ground state of

$S$1

coincides with the ground state of

$S$2

For stabilizer codes, $S$3 is a time-reversed copy induced by singlet-based operator duality; for free fermions, tunneling produces a particle-hole transformed clone, while pairing yields an identical clone. The copy is exact at the ground-state level but state-dependent in a strong sense, since it requires knowledge of a Hamiltonian class with the required dual-operator structure (Hsieh, 2016).

5. Beyond CPTP cloning: virtual, observable, and copy-assisted variants

A major recent generalization replaces physical channels by Hermitian-preserving trace-preserving maps. In this virtual-cloning framework, a linear map $S$4 need only satisfy

$S$5

on a prescribed finite set $S$6, without being completely positive. The exact criterion is that such a virtual-cloning operation exists if and only if the density operators in the set are linearly independent. This is state-dependent in a different sense from no-broadcasting: nonorthogonal pure states and noncommuting mixed states can be cloned exactly, but only virtually, and only when the target set is linearly independent. The simulation overhead is quantified by a quasiprobability cost

$S$7

and the optimal protocol can be found by semidefinite programming. For two pure states, the optimal $S$8 cloning cost is

$S$9

and more generally

$A\subset M_d$0

The paper’s explicit example clones $A\subset M_d$1 and $A\subset M_d$2 exactly, something impossible for a CPTP cloner (Bi et al., 23 Jul 2025).

Another variant clones not states but observable statistics. The “biomimetic cloning” protocol defines exact cloning of an observable $A\subset M_d$3 by requiring

$A\subset M_d$4

The copying is exact for commuting observables, while information about a generally noncommuting operator $A\subset M_d$5 is transferred into a correlation: $A\subset M_d$6 For $A\subset M_d$7, the controlled-shift unitary reduces to CNOT. This is state-dependent only indirectly: arbitrary inputs are allowed, but only selected commuting statistics are cloned locally, with noncommuting coherence redistributed globally (Alvarez-Rodriguez et al., 2013).

A still broader use of copies appears in state-based quantum simulation. There, the paper does not propose a cloning protocol at all; instead it assumes that auxiliary systems can be identically prepared in an arbitrary number of copies of known states, or that copies of the simulator’s current or past state are explicitly supplied. These copies drive effective Hamiltonians of the form

$A\subset M_d$8

and, in the nonlinear setting,

$A\subset M_d$9

The relation to state-dependent quantum copying is therefore indirect but substantive: the paper treats copies of states as consumable resources for state-dependent dynamics, while explicitly insisting that no unknown-state cloning is involved (Alipour et al., 20 May 2025).

6. Asymptotic and cryptographic regimes

In asymptotic settings, state dependence reappears as a property of the source family. Quantum superreplication studies whether $S$0 inputs from a structured family can be transformed into $S$1 approximate outputs with vanishing global error. For continuous families of states, deterministic cloning obeys an asymptotic no-cloning theorem: no deterministic process can reliably clone at rate $S$2, where

$S$3

For clock states and related families, however, probabilistic cloning achieves every rate $S$4. In the equatorial-qubit case the fidelity satisfies

$S$5

and more generally finite-dimensional families containing clock-state structure obey a quadratic ceiling: rates above $S$6 are impossible, and fidelity vanishes in the strong converse. The same paper shows that whenever superreplication is possible it can also be achieved by estimate-and-prepare strategies, though with slower power-law error $S$7 rather than the faster quantum decay in $S$8 (Chiribella et al., 2015).

In quantum cryptography, the issue is often not whether a state can be physically cloned, but how the security of a primitive changes when an adversary receives one copy, $S$9 copies, or $A\subset M_d$00 identical copies of a specific keyed state. The paper on copy complexity makes this dependence explicit in definitions for pseudorandom state generators, pseudorandom unitaries, public-key quantum money, and copy-protection. It proves a generic simulator theorem that approximates many-copy pure-state access from $A\subset M_d$01 i.i.d. samples of an underlying mixed-state family with trace-distance error bounded by

$A\subset M_d$02

From this it derives that one-copy stretch pseudorandom state generators imply $A\subset M_d$03-copy stretch pseudorandom state generators for any fixed polynomial $A\subset M_d$04, one-query pure pseudorandom unitaries imply $A\subset M_d$05-query non-adaptive pseudorandom unitaries, and, under the stated assumptions, there exist identical-copy secure unclonable primitives such as public-key quantum money and quantum copy-protection (Ananth et al., 6 Oct 2025).

Taken together, these developments show that “state-dependent quantum copying” is not a single protocol class but a family of logically distinct regimes. Exact common copying of a family is controlled by commutativity; exact copying with matched ancillas is possible but conditional; approximate copying depends on ensemble geometry and symmetry; composite-state copying can inherit state dependence even from universal local components; virtual copying replaces positivity by simulation cost; and cryptographic copy complexity turns the number of copies of a particular state family into a first-class security parameter. The common thread is that copying becomes possible, meaningful, or optimizable only after the universal unknown-state problem has been restricted, reparameterized, or operationally reinterpreted.