Use FillArrays.jl for handling superoperators (#589)

Author: albertomercurio

CHANGELOG.md

@@ -10,6 +10,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Change default solver detection in `eigensolve` when using `sigma` keyword argument (shift-inverse algorithm). If the operator is a `SparseMatrixCSC`, the default solver is `UMFPACKFactorization`, otherwise it is automatically chosen by LinearSolve.jl, depending on the type of the operator. ([#580]) 
 - Add keyword argument `assume_hermitian` to `liouvillian`. This allows users to disable the assumption that the Hamiltonian is Hermitian. ([#581])
 - Use LinearSolve's internal methods for preconditioners in `SteadyStateLinearSolver`. ([#588])
+- Use `FillArrays.jl` for handling superoperators. This makes the code cleaner and potentially more efficient. ([#589])
 
 ## [v0.38.1]
 Release date: 2025-10-27
@@ -364,3 +365,4 @@ Release date: 2024-11-13
 [#580]: https://github.com/qutip/QuantumToolbox.jl/issues/580
 [#581]: https://github.com/qutip/QuantumToolbox.jl/issues/581
 [#588]: https://github.com/qutip/QuantumToolbox.jl/issues/588
+[#589]: https://github.com/qutip/QuantumToolbox.jl/issues/589

Project.toml

@@ -10,6 +10,7 @@ DiffEqCallbacks = "459566f4-90b8-5000-8ac3-15dfb0a30def"
 DiffEqNoiseProcess = "77a26b50-5914-5dd7-bc55-306e6241c503"
 Distributed = "8ba89e20-285c-5b6f-9357-94700520ee1b"
 FFTW = "7a1cc6ca-52ef-59f5-83cd-3a7055c09341"
+FillArrays = "1a297f60-69ca-5386-bcde-b61e274b549b"
 Graphs = "86223c79-3864-5bf0-83f7-82e725a168b6"
 IncompleteLU = "40713840-3770-5561-ab4c-a76e7d0d7895"
 LaTeXStrings = "b964fa9f-0449-5b57-a5c2-d3ea65f4040f"
@@ -50,6 +51,7 @@ DiffEqCallbacks = "4.2.1 - 4"
 DiffEqNoiseProcess = "5"
 Distributed = "1"
 FFTW = "1.5"
+FillArrays = "1"
 GPUArrays = "10, 11"
 Graphs = "1.7"
 IncompleteLU = "0.2"

src/QuantumToolbox.jl

@@ -59,6 +59,7 @@ import DiffEqNoiseProcess: RealWienerProcess!, RealWienerProcess
 import ArrayInterface: allowed_getindex, allowed_setindex!
 import Distributed: RemoteChannel
 import FFTW: fft, ifft, fftfreq, fftshift
+import FillArrays: Eye
 import Graphs: connected_components, DiGraph
 import IncompleteLU: ilu
 import LaTeXStrings: @L_str

src/deprecated.jl

@@ -138,3 +138,44 @@ function ProgressBar(args...; kwargs...)
         "`ProgressBar` is deprecated and will be removed in next major release. Use `Progress` from `ProgressMeter.jl` instead.",
     )
 end
+
+function spre(A::AbstractQuantumObject{Operator}, Id_cache)
+    Base.depwarn(
+        "The argument `Id_cache` for `spre` is now deprecated, as we now internally use FillArrays.jl, which preserves the same efficiency without requiring this parameter.",
+        :spre,
+        force = true,
+    )
+    return spre(A)
+end
+
+function spost(A::AbstractQuantumObject{Operator}, Id_cache)
+    Base.depwarn(
+        "The argument `Id_cache` for `spost` is now deprecated, as we now internally use FillArrays.jl, which preserves the same efficiency without requiring this parameter.",
+        :spost,
+        force = true,
+    )
+    return spost(A)
+end
+
+function lindblad_dissipator(C::QuantumObject{Operator}, Id_cache)
+    Base.depwarn(
+        "The argument `Id_cache` for `lindblad_dissipator` is now deprecated, as we now internally use FillArrays.jl, which preserves the same efficiency without requiring this parameter.",
+        :lindblad_dissipator,
+        force = true,
+    )
+    return lindblad_dissipator(C)
+end
+
+function liouvillian(
+    H::QuantumObject{HOpType},
+    c_ops::AbstractVector,
+    Id_cache;
+    kwargs...,
+) where {HOpType<:Union{Operator,SuperOperator}}
+    Base.depwarn(
+        "The argument `Id_cache` for `liouvillian` is now deprecated, as we now internally use FillArrays.jl, which preserves the same efficiency without requiring this parameter.",
+        :liouvillian,
+        force = true,
+    )
+    return liouvillian(H, c_ops; kwargs...)
+end

src/qobj/boolean_functions.jl

@@ -85,7 +85,7 @@ Test whether the [`QuantumObject`](@ref) ``U`` is unitary operator. This functio
 
 Note that all the keyword arguments will be passed to `Base.isapprox`.
 """
-isunitary(U::QuantumObject; kwargs...) = isoper(U) ? isapprox(U.data * U.data', I(size(U, 1)); kwargs...) : false
+isunitary(U::QuantumObject; kwargs...) = isoper(U) ? isapprox(U.data * U.data', Eye(size(U, 1)); kwargs...) : false
 
 @doc raw"""
     SciMLOperators.iscached(A::AbstractQuantumObject)

src/qobj/states.jl

@@ -149,10 +149,10 @@ The `dimensions` can be either the following types:
     If you want to keep type stability, it is recommended to use `maximally_mixed_dm(dimensions)` with `dimensions` as `Tuple` or `SVector` from [StaticArrays.jl](https://github.com/JuliaArrays/StaticArrays.jl) to keep type stability. See the [related Section](@ref doc:Type-Stability) about type stability for more details.
 """
 maximally_mixed_dm(dimensions::Int) =
-    QuantumObject(I(dimensions) / complex(dimensions), Operator(), SVector(dimensions))
+    QuantumObject(Eye(dimensions) / complex(dimensions), Operator(), SVector(dimensions))
 function maximally_mixed_dm(dimensions::Union{Dimensions,AbstractVector{Int},Tuple})
     N = prod(dimensions)
-    return QuantumObject(I(N) / complex(N), Operator(), dimensions)
+    return QuantumObject(Eye(N) / complex(N), Operator(), dimensions)
 end
 
 @doc raw"""

src/qobj/superoperators.jl

@@ -6,76 +6,73 @@ export spre, spost, sprepost, liouvillian, lindblad_dissipator
 
 # intrinsic functions for super-operators
 ## keep these because they take AbstractMatrix as input and ensure the output is sparse matrix
-_spre(A::AbstractMatrix, Id::AbstractMatrix) = kron(Id, sparse(A))
-_spre(A::AbstractSparseMatrix, Id::AbstractMatrix) = kron(Id, A)
-_spost(B::AbstractMatrix, Id::AbstractMatrix) = kron(transpose(sparse(B)), Id)
-_spost(B::AbstractSparseMatrix, Id::AbstractMatrix) = kron(transpose(B), Id)
+_spre(A::AbstractMatrix) = kron(Eye(size(A, 1)), sparse(A))
+_spre(A::AbstractSparseMatrix) = kron(Eye(size(A, 1)), A)
+_spost(B::AbstractMatrix) = kron(transpose(sparse(B)), Eye(size(B, 1)))
+_spost(B::AbstractSparseMatrix) = kron(transpose(B), Eye(size(B, 1)))
 _sprepost(A::AbstractMatrix, B::AbstractMatrix) = kron(transpose(sparse(B)), sparse(A))
 _sprepost(A::AbstractMatrix, B::AbstractSparseMatrix) = kron(transpose(B), sparse(A))
 _sprepost(A::AbstractSparseMatrix, B::AbstractMatrix) = kron(transpose(sparse(B)), A)
 _sprepost(A::AbstractSparseMatrix, B::AbstractSparseMatrix) = kron(transpose(B), A)
-function _sprepost(A, B) # for any other input types
-    Id_cache = I(size(A, 1))
-    return _spre(A, Id_cache) * _spost(B, Id_cache)
-end
+_sprepost(A, B) = _spre(A) * _spost(B) # for any other input types
 
 ## if input is AbstractSciMLOperator 
 ## some of them are optimized to speed things up
 ## the rest of the SciMLOperators will just use lazy tensor (and prompt a warning)
-_spre(A::MatrixOperator, Id::AbstractMatrix) = MatrixOperator(_spre(A.A, Id))
-_spre(A::ScaledOperator, Id::AbstractMatrix) = ScaledOperator(A.λ, _spre(A.L, Id))
-_spre(A::AddedOperator, Id::AbstractMatrix) = AddedOperator(map(op -> _spre(op, Id), A.ops))
-function _spre(A::AbstractSciMLOperator, Id::AbstractMatrix)
+_spre(A::MatrixOperator) = MatrixOperator(_spre(A.A))
+_spre(A::ScaledOperator) = ScaledOperator(A.λ, _spre(A.L))
+_spre(A::AddedOperator) = AddedOperator(map(op -> _spre(op), A.ops))
+function _spre(A::AbstractSciMLOperator)
+    Id = Eye(size(A, 1))
     _lazy_tensor_warning(Id, A)
     return kron(Id, A)
 end
 
-_spost(B::MatrixOperator, Id::AbstractMatrix) = MatrixOperator(_spost(B.A, Id))
-_spost(B::ScaledOperator, Id::AbstractMatrix) = ScaledOperator(B.λ, _spost(B.L, Id))
-_spost(B::AddedOperator, Id::AbstractMatrix) = AddedOperator(map(op -> _spost(op, Id), B.ops))
-function _spost(B::AbstractSciMLOperator, Id::AbstractMatrix)
+_spost(B::MatrixOperator) = MatrixOperator(_spost(B.A))
+_spost(B::ScaledOperator) = ScaledOperator(B.λ, _spost(B.L))
+_spost(B::AddedOperator) = AddedOperator(map(op -> _spost(op), B.ops))
+function _spost(B::AbstractSciMLOperator)
     B_T = transpose(B)
+    Id = Eye(size(B, 1))
     _lazy_tensor_warning(B_T, Id)
     return kron(B_T, Id)
 end
 
 ## intrinsic liouvillian 
 function _liouvillian(
     H::MT,
-    Id::AbstractMatrix,
     ::Val{assume_hermitian},
 ) where {MT<:Union{AbstractMatrix,AbstractSciMLOperator},assume_hermitian}
-    H_spre = _spre(H, Id)
-    H_spost = assume_hermitian ? _spost(H, Id) : _spost(H', Id)
+    H_spre = _spre(H)
+    H_spost = assume_hermitian ? _spost(H) : _spost(H')
     return -1im * (H_spre - H_spost)
 end
-function _liouvillian(H::MatrixOperator, Id::AbstractMatrix, assume_hermitian::Val)
+function _liouvillian(H::MatrixOperator, assume_hermitian::Val)
     isconstant(H) ||
         throw(ArgumentError("The given Hamiltonian for constructing Liouvillian must be constant MatrixOperator."))
-    return MatrixOperator(_liouvillian(H.A, Id, assume_hermitian))
+    return MatrixOperator(_liouvillian(H.A, assume_hermitian))
 end
-_liouvillian(H::ScaledOperator, Id::AbstractMatrix, assume_hermitian::Val{true}) =
-    ScaledOperator(H.λ, _liouvillian(H.L, Id, assume_hermitian))
-_liouvillian(H::AddedOperator, Id::AbstractMatrix, assume_hermitian::Val) =
-    AddedOperator(map(op -> _liouvillian(op, Id, assume_hermitian), H.ops))
+_liouvillian(H::ScaledOperator, assume_hermitian::Val{true}) = ScaledOperator(H.λ, _liouvillian(H.L, assume_hermitian))
+_liouvillian(H::AddedOperator, assume_hermitian::Val) =
+    AddedOperator(map(op -> _liouvillian(op, assume_hermitian), H.ops))
 
 # intrinsic lindblad_dissipator
-function _lindblad_dissipator(O::MT, Id::AbstractMatrix) where {MT<:Union{AbstractMatrix,AbstractSciMLOperator}}
+function _lindblad_dissipator(O::MT) where {MT<:Union{AbstractMatrix,AbstractSciMLOperator}}
     Od_O = O' * O
-    return _sprepost(O, O') - (_spre(Od_O, Id) + _spost(Od_O, Id)) / 2
+    return _sprepost(O, O') - (_spre(Od_O) + _spost(Od_O)) / 2
 end
-function _lindblad_dissipator(O::MatrixOperator, Id::AbstractMatrix)
+function _lindblad_dissipator(O::MatrixOperator)
     _O = O.A
     Od_O = _O' * _O
-    return MatrixOperator(_sprepost(_O, _O') - (_spre(Od_O, Id) + _spost(Od_O, Id)) / 2)
+    return MatrixOperator(_sprepost(_O, _O') - (_spre(Od_O) + _spost(Od_O)) / 2)
 end
-function _lindblad_dissipator(O::ScaledOperator, Id::AbstractMatrix)
+function _lindblad_dissipator(O::ScaledOperator)
     λc_λ = conj(O.λ) * O.λ
-    return ScaledOperator(λc_λ, _lindblad_dissipator(O.L, Id))
+    return ScaledOperator(λc_λ, _lindblad_dissipator(O.L))
 end
 
 @doc raw"""
-    spre(A::AbstractQuantumObject, Id_cache=I(size(A,1)))
+    spre(A::AbstractQuantumObject)
 
 Returns the [`SuperOperator`](@ref) form of `A` acting on the left of the density matrix operator: ``\mathcal{O} \left(\hat{A}\right) \left[ \hat{\rho} \right] = \hat{A} \hat{\rho}``.
 
@@ -86,15 +83,12 @@ Since the density matrix is vectorized in [`OperatorKet`](@ref) form: ``|\hat{\r
 ```
 (see the section in documentation: [Superoperators and Vectorized Operators](@ref doc:Superoperators-and-Vectorized-Operators) for more details)
 
-The optional argument `Id_cache` can be used to pass a precomputed identity matrix. This can be useful when the same function is applied multiple times with a known Hilbert space dimension.
-
 See also [`spost`](@ref) and [`sprepost`](@ref).
 """
-spre(A::AbstractQuantumObject{Operator}, Id_cache = I(size(A, 1))) =
-    get_typename_wrapper(A)(_spre(A.data, Id_cache), SuperOperator(), A.dimensions)
+spre(A::AbstractQuantumObject{Operator}) = get_typename_wrapper(A)(_spre(A.data), SuperOperator(), A.dimensions)
 
 @doc raw"""
-    spost(B::AbstractQuantumObject, Id_cache=I(size(B,1)))
+    spost(B::AbstractQuantumObject)
 
 Returns the [`SuperOperator`](@ref) form of `B` acting on the right of the density matrix operator: ``\mathcal{O} \left(\hat{B}\right) \left[ \hat{\rho} \right] = \hat{\rho} \hat{B}``.
 
@@ -105,12 +99,9 @@ Since the density matrix is vectorized in [`OperatorKet`](@ref) form: ``|\hat{\r
 ```
 (see the section in documentation: [Superoperators and Vectorized Operators](@ref doc:Superoperators-and-Vectorized-Operators) for more details)
 
-The optional argument `Id_cache` can be used to pass a precomputed identity matrix. This can be useful when the same function is applied multiple times with a known Hilbert space dimension.
-
 See also [`spre`](@ref) and [`sprepost`](@ref).
 """
-spost(B::AbstractQuantumObject{Operator}, Id_cache = I(size(B, 1))) =
-    get_typename_wrapper(B)(_spost(B.data, Id_cache), SuperOperator(), B.dimensions)
+spost(B::AbstractQuantumObject{Operator}) = get_typename_wrapper(B)(_spost(B.data), SuperOperator(), B.dimensions)
 
 @doc raw"""
     sprepost(A::AbstractQuantumObject, B::AbstractQuantumObject)
@@ -132,7 +123,7 @@ function sprepost(A::AbstractQuantumObject{Operator}, B::AbstractQuantumObject{O
 end
 
 @doc raw"""
-    lindblad_dissipator(O::AbstractQuantumObject, Id_cache=I(size(O,1))
+    lindblad_dissipator(O::AbstractQuantumObject)
 
 Returns the Lindblad [`SuperOperator`](@ref) defined as
 
@@ -141,21 +132,18 @@ Returns the Lindblad [`SuperOperator`](@ref) defined as
 \hat{O}^\dagger \hat{O} \hat{\rho} - \hat{\rho} \hat{O}^\dagger \hat{O} \right)
 ```
 
-The optional argument `Id_cache` can be used to pass a precomputed identity matrix. This can be useful when the same function is applied multiple times with a known Hilbert space dimension.
-
 See also [`spre`](@ref), [`spost`](@ref), and [`sprepost`](@ref).
 """
-lindblad_dissipator(O::AbstractQuantumObject{Operator}, Id_cache = I(size(O, 1))) =
-    get_typename_wrapper(O)(_lindblad_dissipator(O.data, Id_cache), SuperOperator(), O.dimensions)
+lindblad_dissipator(O::AbstractQuantumObject{Operator}) =
+    get_typename_wrapper(O)(_lindblad_dissipator(O.data), SuperOperator(), O.dimensions)
 
 # It is already a SuperOperator
-lindblad_dissipator(O::AbstractQuantumObject{SuperOperator}, Id_cache = nothing) = O
+lindblad_dissipator(O::AbstractQuantumObject{SuperOperator}) = O
 
 @doc raw"""
     liouvillian(
         H::AbstractQuantumObject,
-        c_ops::Union{Nothing,AbstractVector,Tuple}=nothing,
-        Id_cache=I(prod(H.dimensions));
+        c_ops::Union{Nothing,AbstractVector,Tuple}=nothing;
         assume_hermitian::Union{Bool,Val} = Val(true),
     )
 
@@ -179,54 +167,44 @@ where
 \mathcal{D}(\hat{C}_n) [\cdot] = \hat{C}_n [\cdot] \hat{C}_n^\dagger - \frac{1}{2} \hat{C}_n^\dagger \hat{C}_n [\cdot] - \frac{1}{2} [\cdot] \hat{C}_n^\dagger \hat{C}_n
 ```
 
-The optional argument `Id_cache` can be used to pass a precomputed identity matrix. This can be useful when the same function is applied multiple times with a known Hilbert space dimension.
-
 See also [`spre`](@ref), [`spost`](@ref), and [`lindblad_dissipator`](@ref).
 
 !!! warning "Beware of type-stability!"
     If you want to keep type stability, it is recommended to use `assume_hermitian = Val(true)` instead of `assume_hermitian = true`. See [this link](https://docs.julialang.org/en/v1/manual/performance-tips/#man-performance-value-type) and the [related Section](@ref doc:Type-Stability) about type stability for more details.
 """
 function liouvillian(
     H::AbstractQuantumObject{OpType},
-    c_ops::Union{Nothing,AbstractVector,Tuple} = nothing,
-    Id_cache::Diagonal = I(prod(H.dimensions));
+    c_ops::Union{Nothing,AbstractVector,Tuple} = nothing;
     assume_hermitian::Union{Bool,Val} = Val(true),
 ) where {OpType<:Union{Operator,SuperOperator}}
-    L = liouvillian(H, Id_cache; assume_hermitian = assume_hermitian)
+    L = liouvillian(H; assume_hermitian = assume_hermitian)
     if !isnothing(c_ops)
-        return L + _sum_lindblad_dissipators(c_ops, Id_cache)
+        return L + _sum_lindblad_dissipators(c_ops)
     end
     return L
 end
 
-liouvillian(H::Nothing, c_ops::Union{AbstractVector,Tuple}, Id_cache::Diagonal = I(prod(c_ops[1].dims)); kwargs...) =
-    _sum_lindblad_dissipators(c_ops, Id_cache)
+liouvillian(H::Nothing, c_ops::Union{AbstractVector,Tuple}; kwargs...) = _sum_lindblad_dissipators(c_ops)
 
 liouvillian(H::Nothing, c_ops::Nothing; kwargs...) = 0
 
-liouvillian(
-    H::AbstractQuantumObject{Operator},
-    Id_cache::Diagonal = I(prod(H.dimensions));
-    assume_hermitian::Union{Bool,Val} = Val(true),
-) = get_typename_wrapper(H)(_liouvillian(H.data, Id_cache, makeVal(assume_hermitian)), SuperOperator(), H.dimensions)
-
-liouvillian(H::AbstractQuantumObject{SuperOperator}, Id_cache::Diagonal; kwargs...) = H
+liouvillian(H::AbstractQuantumObject{Operator}; assume_hermitian::Union{Bool,Val} = Val(true)) =
+    get_typename_wrapper(H)(_liouvillian(H.data, makeVal(assume_hermitian)), SuperOperator(), H.dimensions)
 
-_sum_lindblad_dissipators(c_ops::Nothing, Id_cache::Diagonal) = 0
+liouvillian(H::AbstractQuantumObject{SuperOperator}; kwargs...) = H
+_sum_lindblad_dissipators(c_ops::Nothing) = 0
 
-function _sum_lindblad_dissipators(c_ops::AbstractVector, Id_cache::Diagonal)
-    return sum(op -> lindblad_dissipator(op, Id_cache), c_ops; init = 0)
-end
+_sum_lindblad_dissipators(c_ops::AbstractVector) = sum(op -> lindblad_dissipator(op), c_ops; init = 0)
 
 # Help the compiler to unroll the sum at compile time
-@generated function _sum_lindblad_dissipators(c_ops::Tuple, Id_cache::Diagonal)
+@generated function _sum_lindblad_dissipators(c_ops::Tuple)
     N = length(c_ops.parameters)
     if N == 0
         return :(0)
     end
-    ex = :(lindblad_dissipator(c_ops[1], Id_cache))
+    ex = :(lindblad_dissipator(c_ops[1]))
     for i in 2:N
-        ex = :($ex + lindblad_dissipator(c_ops[$i], Id_cache))
+        ex = :($ex + lindblad_dissipator(c_ops[$i]))
     end
     return ex
 end

src/spectrum.jl

@@ -140,7 +140,7 @@ function _spectrum(
     _tr = SparseVector(D^2, [1 + n * (D + 1) for n in 0:(D-1)], ones(_complex_float_type(L), D)) # same as vec(system_identity_matrix)
     _tr_A = transpose(_tr) * spre(A).data
 
-    Id = I(D^2)
+    Id = Eye(D^2)
 
     # DO the idx = 1 case
     ω = ωList[1]

src/spin_lattice.jl

@@ -53,11 +53,11 @@ function multisite_operator(dims::Union{AbstractVector,Tuple}, pairs::Pair{<:Int
 
     _dims[sites] == [get_dimensions_to(op)[1].size for op in ops] || throw(ArgumentError("The dimensions of the operators do not match the dimensions of the lattice."))
 
-    data = kron(I(prod(_dims[1:(sites[1]-1)])), ops[1].data)
+    data = kron(Eye(prod(_dims[1:(sites[1]-1)])), ops[1].data)
     for i in 2:length(sites)
-        data = kron(data, I(prod(_dims[(sites[i-1]+1):(sites[i]-1)])), ops[i].data)
+        data = kron(data, Eye(prod(_dims[(sites[i-1]+1):(sites[i]-1)])), ops[i].data)
     end
-    data = kron(data, I(prod(_dims[(sites[end]+1):end])))
+    data = kron(data, Eye(prod(_dims[(sites[end]+1):end])))
 
     return QuantumObject(data; type = Operator(), dims = dims)
 end

src/time_evolution/brmesolve.jl

@@ -126,7 +126,6 @@ function _brterm(
     _check_br_spectra(spectra)
 
     U = rst.vectors
-    Id = I(prod(rst.dimensions))
 
     skew = @. rst.values - rst.values' |> real
     spectrum = spectra.(skew)
@@ -166,7 +165,7 @@ function _brterm(
         tidyup!(bd_term)
     end
 
-    out = (_sprepost(A_mat_spec_t, A_mat) + _sprepost(A_mat, A_mat_spec) - _spost(ac_term, Id) - _spre(bd_term, Id)) / 2
+    out = (_sprepost(A_mat_spec_t, A_mat) + _sprepost(A_mat, A_mat_spec) - _spost(ac_term) - _spre(bd_term)) / 2
 
     (sec_cutoff != -1) && (out .*= M_cut)