Struct ProcessTomography 

pub struct ProcessTomography;

Process tomography utilities

Implementations

impl ProcessTomography

pub fn reconstruct_channel( input_states: &[Array2<Complex<f64>>], output_states: &[Array2<Complex<f64>>], ) -> QuantRS2Result<QuantumChannel>

Reconstruct a quantum channel from process tomography data via linear inversion / least-squares estimation of its Choi matrix.

Given input density matrices ρ_k and the measured outputs ρ'_k = Λ(ρ_k), the channel is determined by its Choi matrix J. Using the column-stacking Choi convention of this module (the same one used by QuantumChannel::kraus_to_choi), the action of the channel is linear in the entries of J:

Λ(ρ)[i, i'] = Σ_{j, j'} ρ[j, j'] · J[i + j·d_out, i' + j'·d_out]

Each (ρ_k, ρ'_k) pair therefore contributes d_out² scalar linear constraints A · vec(J) = b. Stacking all pairs and solving the least-squares problem (via the Hermitian normal equations AᴴA · vec(J) = Aᴴb, solved with a spectral pseudo-inverse) yields the best-fit Choi matrix, from which the channel is rebuilt with QuantumChannel::from_choi.

Errors

Returns QuantRS2Error::InvalidInput if the input/output counts differ, if any density matrix is not square / dimensions are inconsistent, or if the supplied input states are not informationally complete (fewer than d² linearly independent inputs, leaving the Choi matrix under-determined). No identity / placeholder channel is fabricated.

pub fn generate_input_states(dim: usize) -> Vec<Array2<Complex<f64>>>

Generate an informationally-complete set of input states for a dim-dimensional system.

Process tomography of a dim × dim channel needs dim² linearly independent input density matrices. This routine returns exactly such a set:

• the dim computational-basis populations |i⟩⟨i|, and
• for every pair i < j, the two superposition states |+_{ij}⟩ = (|i⟩ + |j⟩)/√2 and |+i_{ij}⟩ = (|i⟩ + i|j⟩)/√2 as density matrices |ψ⟩⟨ψ|.

The diagonal states fix the populations while each pair of off-diagonal states fixes the real and imaginary parts of the corresponding coherence, giving dim + 2 · C(dim, 2) = dim² linearly independent operators that span the full space of Hermitian matrices.