Quantum Channels Are More General Than Unitary Operators

The highlighted sentence says:

A quantum operation that takes input states to output states is called a quantum channel when it is physically allowed and deterministic.

This is a compact definition, but it hides an important distinction. A unitary operator is one special kind of quantum evolution. A quantum channel is the more general mathematical object used to describe every deterministic physical transformation of quantum states, including noisy evolution, loss of information into an environment, discarding part of a system, and adding an ancillary system.

So the short answer is: a quantum channel is not usually a unitary operator. Every unitary evolution gives a quantum channel, but many quantum channels are not unitary.

From state vectors to density operators

In the simplest textbook picture, a closed quantum system in a pure state is represented by a unit vector $|\psi\rangle$ in a Hilbert space $\mathcal H$. A reversible physical evolution is represented by a unitary operator $U$, and the state changes as

\[ |\psi\rangle \mapsto U|\psi\rangle. \]

A unitary operator satisfies

\[ U^\dagger U = UU^\dagger = I, \]

which means it preserves inner products, probabilities, and the total norm of the state vector.

But the parent document has already moved to the more general language of density operators. A density operator $\rho$ is a positive semidefinite operator with trace one. Pure states are included as the special case

\[ \rho = |\psi\rangle\langle \psi|. \]

In density-operator language, the same unitary evolution is written as

\[ \rho \mapsto U\rho U^\dagger. \]

This map from input density operators to output density operators is a unitary channel. Notice the difference between $U$ itself and the channel it defines. The unitary operator $U$ acts on vectors in $\mathcal H$. The associated channel

\[ \Phi_U(\rho)=U\rho U^\dagger \]

acts on operators $\rho \in \mathcal L(\mathcal H)$.

This distinction becomes important because not every physical transformation of density operators has the form $\rho \mapsto U\rho U^\dagger$.

Why unitary operators are not enough

Unitary operators describe closed, reversible quantum evolution. But real quantum devices are rarely closed. A qubit may interact with uncontrolled surroundings, lose energy, suffer noise, or become entangled with an environment that is later ignored. When we describe only the system we keep, the effective evolution can be irreversible, even if the larger system-plus-environment evolves unitarily.

For example, consider the bit-flip noise channel from the parent document:

\[ \Phi(\rho) = (1-p)\rho + pX\rho X, \]

where $X$ is the Pauli bit-flip operator and $0 \leq p \leq 1$. This says: with probability $1-p$, nothing happens; with probability $p$, the bit is flipped.

This map is deterministic in the sense that it always produces an output state with trace one. But it is generally not unitary. If $0<p<1$, the output is a probabilistic mixture of two possible evolutions. There is no single unitary operator $U$ such that

\[ (1-p)\rho + pX\rho X = U\rho U^\dagger \]

for every input state $\rho$.

One way to see this is that unitary channels preserve purity. If $\rho=|\psi\rangle\langle\psi|$ is pure, then

\[ U\rho U^\dagger \]

is also pure. But noisy channels can turn pure states into mixed states. For instance, if $p=1/2$ and the input is $|0\rangle\langle 0|$, then

\[ \Phi(|0\rangle\langle 0|) = \frac12 |0\rangle\langle 0| + \frac12 |1\rangle\langle 1|, \]

which is a mixed state. That cannot be the result of a unitary evolution on the qubit alone.

This is one reason the word channel is needed. The theory needs a term for physically allowed state transformations that include, but are not limited to, reversible unitary motion.

What makes a map a quantum channel?

In finite-dimensional quantum information theory, a quantum channel is usually defined as a linear map

\[ \Phi:\mathcal L(\mathcal H_{\mathrm{in}}) \to \mathcal L(\mathcal H_{\mathrm{out}}) \]

that is completely positive and trace preserving [Nielsen and Chuang 2010; Watrous 2018].

Each part of this definition has physical meaning.

First, $\Phi$ is linear. If a state is a mixture

\[ \rho = q\rho_1 + (1-q)\rho_2, \]

where $0\leq q\leq 1$, then the output should be

\[ \Phi(\rho)=q\Phi(\rho_1)+(1-q)\Phi(\rho_2). \]

This reflects the idea that classical uncertainty about which state was prepared should be carried consistently through the device.

Second, $\Phi$ is trace preserving. Since $\operatorname{Tr}(\rho)=1$ represents total probability one, a deterministic operation should satisfy

\[ \operatorname{Tr}(\Phi(\rho))=1 \]

for every input density operator $\rho$. More generally,

\[ \operatorname{Tr}(\Phi(X))=\operatorname{Tr}(X) \]

for all operators $X$ in the relevant space. If a map is not trace preserving, it may describe a probabilistic event, such as one outcome of a measurement, but not a deterministic channel by itself.

Third, $\Phi$ is completely positive. Positivity alone would mean that if $\rho$ is a valid state, then $\Phi(\rho)$ is also positive semidefinite. Complete positivity is stronger. It requires that for every auxiliary system $\mathcal K$, the extended map

\[ \Phi \otimes \operatorname{id}_{\mathcal K} \]

also sends positive operators to positive operators.

This condition matters because the input system may be entangled with another system that the device does not touch. A physically allowed operation on one part of an entangled pair must still produce a valid joint state. Complete positivity is the mathematical condition that protects this requirement [Stinespring 1955; Watrous 2018].

Why say “channel” instead of just “operator”?

The word operator is too broad and can be ambiguous. In quantum theory, an operator might mean many different things:

• an observable, such as a Hamiltonian $H$;
• a unitary evolution operator $U$;
• a density operator $\rho$;
• a measurement effect $E_a$;
• a Kraus operator $A_k$;
• a linear map acting on density operators.

A quantum channel is not usually an operator on the Hilbert space $\mathcal H$ itself. It is a map on operators:

\[ \Phi:\mathcal L(\mathcal H_{\mathrm{in}}) \to \mathcal L(\mathcal H_{\mathrm{out}}). \]

Because it maps operators to operators, it is sometimes called a superoperator. But “superoperator” only describes the type of mathematical object. It does not by itself say that the map is physically allowed or deterministic. The word channel carries the physical conditions: complete positivity and trace preservation.

The terminology also comes from information theory. A classical communication channel takes input probability distributions or symbols and produces output distributions or symbols, possibly with noise. A quantum channel similarly takes quantum states as inputs and produces quantum states as outputs, possibly with quantum noise. The word emphasizes that the object describes an information-processing transformation, not merely a matrix.

For example, a classical noisy bit channel might flip a bit with probability $p$. The quantum bit-flip channel

\[ \Phi(\rho)=(1-p)\rho+pX\rho X \]

is the quantum analogue: it transmits a qubit imperfectly. Calling it a channel makes its operational meaning clearer.

Unitary channels as a special case

Although a channel is not generally a unitary operator, unitary dynamics fit perfectly inside the channel framework. Given a unitary operator

\[ U:\mathcal H_{\mathrm{in}}\to \mathcal H_{\mathrm{out}}, \]

with matching dimensions, we get the channel

\[ \Phi_U(\rho)=U\rho U^\dagger. \]

This channel is completely positive and trace preserving.

It is completely positive because, even on a larger system,

\[ (\Phi_U\otimes \operatorname{id})(R) = (U\otimes I)R(U^\dagger\otimes I), \]

which preserves positive semidefiniteness. It is trace preserving because

\[ \operatorname{Tr}(U\rho U^\dagger) = \operatorname{Tr}(U^\dagger U\rho) = \operatorname{Tr}(\rho), \]

using cyclicity of the trace and $U^\dagger U=I$.

So a unitary operator $U$ gives a channel $\Phi_U$, but the channel is the map $\rho\mapsto U\rho U^\dagger$, not simply the operator $U$ itself.

This distinction is especially important in the parent document’s topic: programmable quantum devices. A programmable processor is not only asked to implement unitary gates. It may be asked to implement arbitrary noisy channels, measurements, instruments, or higher-order transformations. For that broader theory, “unitary operator” is far too narrow.

Channels include environment interactions

A powerful reason quantum channels are the right language is the dilation picture. Stinespring’s dilation theorem says, roughly, that every finite-dimensional quantum channel can be represented by a unitary interaction with a larger environment, followed by discarding part of the system [Stinespring 1955; Nielsen and Chuang 2010; Watrous 2018].

In finite dimensions, one common form is

\[ \Phi(\rho) = \operatorname{Tr}_{\mathcal E} \left[ V\rho V^\dagger \right], \]

where $V$ is an isometry from the input space into a larger output-plus-environment space, and $\operatorname{Tr}_{\mathcal E}$ means that the environment is ignored.

Equivalently, many channels can be written in Kraus form:

\[ \Phi(\rho)=\sum_k A_k \rho A_k^\dagger, \]

where the Kraus operators satisfy

\[ \sum_k A_k^\dagger A_k = I. \]

A unitary channel is the special case with one Kraus operator $A_1=U$. Noisy channels usually require more than one Kraus operator.

This gives a useful physical picture. At the deepest level, if we include everything, quantum mechanics may use unitary evolution. But when we describe only the system available to us, the effective deterministic transformation is usually a channel.

The key distinction for the highlighted passage

The highlighted passage is therefore setting up the correct general language. A quantum operation that takes input states to output states is called a quantum channel when it is deterministic and physically allowed. In finite dimensions, that means it is a completely positive, trace-preserving linear map on density operators.

A unitary operator is enough only for closed, reversible evolution. A quantum channel is needed for the general case: open systems, noise, discarded environments, mixed states, and programmable processors that may implement more than just gates.

So the phrase “quantum channel” is not extra terminology for something already covered by “operator.” It names the physically meaningful class of deterministic state transformations. Unitary channels are inside that class, but they are only one corner of it.

References

Nielsen, M. A., and Chuang, I. L. (2010). Quantum Computation and Quantum Information: 10th Anniversary Edition. Cambridge University Press.

Stinespring, W. F. (1955). “Positive functions on C-algebras.” Proceedings of the American Mathematical Society*, 6(2), 211–216. https://doi.org/10.1090/S0002-9939-1955-0069403-4

Watrous, J. (2018). The Theory of Quantum Information. Cambridge University Press. https://doi.org/10.1017/9781316848142