pycasino

Use this skill when implementing or working on the Omega (Ω) term of the backflow transformation in PyCasino — the electron-electron-electron (e-e-e) three-body backflow displacement. Covers the ωijk polynomial (manual Eq. 253), the triplet displacement (Eq. 247), the cutoff, the no-cusp + particle-symmetry constraints, spin triplets, the correlation.data format, and the mapping of CASINO's pbackflow.f90 onto casino/backflow.py + casino/readers/backflow.py. The reader is implemented; the evaluation (value/gradient/laplacian) is NOT yet in casino/backflow.py. See also the backflow skill (eta/mu/Phi/Theta) and the qmc skill.

Use this skill when working on the backflow transformation in casino/backflow.py or casino/readers/backflow.py: the eta/mu/Phi/Theta displacement terms, the cutoff and truncation order C, AE-cutoff multiplier g(r), Kato cusp constraints, irrotational backflow, construct_c_matrix, or the backflow Jacobian used in the kinetic energy. This is the physics/math reference (López Ríos et al. 2006) mapped to the code. See also the qmc skill for how backflow enters the wave function and local energy.

Use this skill when profiling or benchmarking pycasino (casino/profiling.py), measuring timings of slater/jastrow/backflow/cusp/markovchain kernels, or when running/writing the test suite (casino/tests/, pytest). Trigger on: profiling, benchmark, timing, performance, slow kernel, run tests, pytest, test_slater, test_backflow, numerical derivative test, CI.

Use this skill when working with any file in pycasino/casino/: wfn.py, slater.py, jastrow.py, backflow.py, dmc.py, vmc.py, cusp.py, harmonics.py, ppotential.py. Trigger on: local energy, kinetic energy, Slater determinant, Jastrow factor, backflow transformation, drift velocity, branching, trial wave function, VMC, DMC, wavefunction optimization, varmin, emin, linear method, stochastic reconfiguration, cusp condition, pseudopotential, T-move.

Use this skill whenever working with Numba JIT compilation, @nb.njit, @nb.jit, parallel=True/False, prange, numba.typed, numba.cuda, numba.vectorize, numba.guvectorize, or debugging Numba compilation errors. Trigger on: slow numpy loops being accelerated, nopython mode errors, Numba type inference failures, reflected list warnings, cache=True, fastmath, nogil, Numba + NumPy interop, or any task involving @njit on scientific/numerical code.