4

Tried this simple example where initializing two qubits in $|0\rangle$ state, performing biased bit flip erasure on qubit 1 using HERALDED_PAULI_CHANNEL_1 while measuring qubit 0:

import stim
    import numpy as np
    circuit = stim.Circuit()
    circuit.append_operation("R", (0, 1))
    p = 0.2
    circuit.append_operation("HERALDED_PAULI_CHANNEL_1", 1, (p, p, 0, 0))
    circuit.append_operation("M",0)
    N = 1000
    result = circuit.compile_sampler().sample(shots = N)
    one_num = np.sum([int(True == j[-1]) for j in result])
    print('%s/%s' %(one_num, N))
    display(circuit.diagram("timeline-svg"))

I get around 200/1000 measurement error on the qubit 0 as if the error is on qubit 0. What is wrong here? Thank you.

FDGod
  • 2,278
  • 2
  • 4
  • 29
Xiao Xiao
  • 43
  • 2

1 Answers1

4

Update: This is fixed in stim~=1.12.1 and in stim~=1.13.dev1701377008. The only stable version affected was 1.12.0, since HERALDED_PAULI_CHANNEL_1 was released in 1.12.

This is a bug. Good catch. I've written a fix and am merging it: https://github.com/quantumlib/Stim/pull/676. Also added some tests that verify it's fixed. I also checked HERALDED_ERASE, which historically was added at the same time but didn't have the same bug.

enter image description here

Within two hours you should be able to pip install stim~=1.13.dev and get a version with the bug fixed. I'll also backport the fix to 1.12 and release a version 1.12.1. That should be done before the end of the day.

Craig Gidney
  • 36,389
  • 1
  • 29
  • 95