Error as incompatible function arguments. The following argument types are supported: 1. (self: dolfin.cpp.la.BlockMatrix, arg0: tuple, arg1: dolfin.cpp.la.GenericMatrix)

1

My code is as follows: I am getting the error:
ypeError: setitem(): incompatible function arguments. The following argument types are supported:
1. (self: dolfin.cpp.la.BlockMatrix, arg0: tuple, arg1: dolfin.cpp.la.GenericMatrix) -> None

Invoked with: <dolfin.cpp.la.BlockMatrix object at 0x7f8658e08a78>, (0, 2), array([[0., 0., 0., …, 0., 0., 0.],
[0., 0., 0., …, 0., 0., 0.],
[0., 0., 0., …, 0., 0., 0.],
…,
[0., 0., 0., …, 0., 0., 0.],
[0., 0., 0., …, 0., 0., 0.],
[0., 0., 0., …, 0., 0., 0.]])

mesh = UnitSquareMesh(10,10)
RT1 = FiniteElement("RT", mesh.ufl_cell(), 1)
V1 = FunctionSpace(mesh, "RT", 1)
CG1 = FiniteElement("CG", mesh.ufl_cell(), 1)
mixed_A = MixedElement([RT1, CG1, RT1])
mixed_B = MixedElement([RT1, CG1])
W_A = FunctionSpace(mesh, mixed_A)
W_B = FunctionSpace(mesh, mixed_B)
(vpsi, vphi, vbigphi) = TestFunctions(W_A)
(vq, vu ) = TestFunctions(W_B)
trials_1A = TrialFunction(W_A)
trials_1B = TrialFunction(W_B) 
Sol_A, Sol0_A = Function(W_A) , Function(W_A)
Sol_B, Sol0_B = Function(W_B) , Function(W_B)
(psi, phi, bigphi) = split(Sol_A)
(q, u) = split(Sol_B)
(psi0, phi0, bigphi0) = split(Sol0_A)
(q0, u0) = split(Sol0_B)
(psi_1, phi_1, bigphi_1) = split(trials_1A)
(q_1, u_1) = split(trials_1B)
def sigma(u):
   return u**2

def beta_new(u):
    return (1/(u**2))


uinitial = Expression("exp(x[1] + x[0] - t)", t = 0, degree = 2)
u0 = interpolate(uinitial, W_B.sub(1).collapse())
t = 0
dt = 0.1
T = 1
t = dt
A11 = assemble(inner(psi_1, vbigphi)*dx)
A23 = assemble(inner(bigphi_1, vpsi)*dx)
A33 = assemble(div(bigphi_1)*vphi*dx)
A12 = assemble(div(vbigphi)*phi_1*dx) 
A_21_mat = - inner(sigma(u0)*psi_1, vpsi)*dx
nn = FacetNormal(mesh)
nrelems = mesh.num_cells()  
nrsides = mesh.num_edges() 
pp = Expression(" sin(x[1] + x[0] - t )", degree = 2, t = 0)
f1 = Expression(" (- 3*exp(x[0] - t + x[1])  + 1)", degree = 3, t=0)
f2 = Expression("2*sin(t + x[0] + x[1])", degree = 2, t=0 )
LHS1_mat = BlockMatrix(3,3)
while t <= T :
 A21 = assemble(A_21_mat)
 f2.t = t
 b1_3 = assemble(inner(f2, vphi)*dx)
 pp.t = t
 b1_1 = assemble(-pp*dot(nn, vbigphi)*ds)
 A13 = np.zeros((nrsides,nrsides))
 A22 = np.zeros((nrsides, nrelems))
 A31 = np.zeros((nrelems,nrsides))
 A32 = np.zeros((nrelems, nrelems))
 LHS1_mat[0,0] = A11
 LHS1_mat[0,1] = A12
 LHS1_mat[0,2] = A13   ##### Error here
 LHS1_mat[1,0] = A21
 LHS1_mat[1,1] = A22
 LHS1_mat[1,2] = A23
 LHS1_mat[2,0] = A31
 LHS1_mat[2,1] = A32
 LHS1_mat[2,2] = A33
 LHS1_mat = np.block([[A11, A12, A13], [A21, A22, A23],  [A31, A32, A33]])
2

You should not use numpy matrices in a assembly. First of all,they are dense structures, which should never be used on large problems. Secondly, as the error message tells you, you have to send ina matrix type supported by dolfin (should be a Petsc matrix).

You could have a look at CBC-block: