3

When I use stim to simulate d=3 Bacon-Shor's code, I can't find the problem with my circuits resulting in the weird DetectorErrorModel. The initialization and stabilizer measurement circuits are as follow: layout

circuits

I focus on $|0\rangle _L$ memory simulation and choose not to use reset after measurement. So with three rounds of stabilizer measurements, the detectors are

...
DETECTOR(0, 0, 0) rec[-4]
DETECTOR(1, 0, 0) rec[-3]
...
SHIFT_COORDS(0, 0, 1)
DETECTOR(0, 0, 0) rec[-4]
DETECTOR(1, 0, 0) rec[-3]
...
SHIFT_COORDS(0, 0, 1)
DETECTOR(0, 0, 0) rec[-4] rec[-8]
DETECTOR(1, 0, 0) rec[-3] rec[-7]
...
DETECTOR(0, 0, 1) rec[-9] rec[-6] rec[-8] rec[-5] rec[-7] rec[-4] rec[-13] rec[-17]
DETECTOR(1, 0, 1) rec[-6] rec[-3] rec[-5] rec[-2] rec[-4] rec[-1] rec[-12] rec[-16]

some error I can't understand appear in auto-generated DetectorErrorModel like D0 D4 L0 ^ D1 D5 ^ L0.

Here is my stim.Circuit file and generated error_model:

circuit.txt

stim.Circuit('''
    R 0 1 2 3 4 5 6 7 8 9 10 11 12
    H 0
    DEPOLARIZE1(0.001) 0
    CX 0 1
    DEPOLARIZE2(0.02) 0 1
    CX 1 2
    DEPOLARIZE2(0.02) 1 2
    H 3
    DEPOLARIZE1(0.001) 3
    CX 3 4
    DEPOLARIZE2(0.02) 3 4
    CX 4 5
    DEPOLARIZE2(0.02) 4 5
    H 6
    DEPOLARIZE1(0.001) 6
    CX 6 7
    DEPOLARIZE2(0.02) 6 7
    CX 7 8
    DEPOLARIZE2(0.02) 7 8
    H 0
    DEPOLARIZE1(0.001) 0
    H 1
    DEPOLARIZE1(0.001) 1
    H 2
    DEPOLARIZE1(0.001) 2
    H 3
    DEPOLARIZE1(0.001) 3
    H 4
    DEPOLARIZE1(0.001) 4
    H 5
    DEPOLARIZE1(0.001) 5
    H 6
    DEPOLARIZE1(0.001) 6
    H 7
    DEPOLARIZE1(0.001) 7
    H 8
    DEPOLARIZE1(0.001) 8 0 1 2 3 4 5 6 7 8
    H 11
    DEPOLARIZE1(0.001) 11
    H 12
    DEPOLARIZE1(0.001) 12
    CX 11 0
    DEPOLARIZE2(0.02) 11 0
    CX 12 1
    DEPOLARIZE2(0.02) 12 1
    DEPOLARIZE1(0.002) 2 3 4 5 6 7 8
    CX 11 1
    DEPOLARIZE2(0.02) 11 1
    CX 12 2
    DEPOLARIZE2(0.02) 12 2
    DEPOLARIZE1(0.002) 0 3 4 5 6 7 8
    CX 11 3
    DEPOLARIZE2(0.02) 11 3
    CX 12 4
    DEPOLARIZE2(0.02) 12 4
    DEPOLARIZE1(0.002) 0 1 2 5 6 7 8
    CX 11 4
    DEPOLARIZE2(0.02) 11 4
    CX 12 5
    DEPOLARIZE2(0.02) 12 5
    DEPOLARIZE1(0.002) 0 1 2 3 6 7 8
    CX 11 6
    DEPOLARIZE2(0.02) 11 6
    CX 12 7
    DEPOLARIZE2(0.02) 12 7
    DEPOLARIZE1(0.002) 0 1 2 3 4 5 8
    CX 11 7
    DEPOLARIZE2(0.02) 11 7
    CX 12 8
    DEPOLARIZE2(0.02) 12 8
    DEPOLARIZE1(0.002) 0 1 2 3 4 5 6
    DEPOLARIZE1(0.001) 0 1 2 3 4 5 6 7 8
    H 11
    DEPOLARIZE1(0.001) 11
    H 12
    DEPOLARIZE1(0.001) 12
    CX 0 9
    DEPOLARIZE2(0.02) 0 9
    CX 3 10
    DEPOLARIZE2(0.02) 3 10
    DEPOLARIZE1(0.002) 1 2 4 5 6 7 8 11 12
    CX 3 9
    DEPOLARIZE2(0.02) 3 9
    CX 6 10
    DEPOLARIZE2(0.02) 6 10
    DEPOLARIZE1(0.002) 0 1 2 4 5 7 8 11 12
    CX 1 9
    DEPOLARIZE2(0.02) 1 9
    CX 4 10
    DEPOLARIZE2(0.02) 4 10
    DEPOLARIZE1(0.002) 0 2 3 5 6 7 8 11 12
    CX 4 9
    DEPOLARIZE2(0.02) 4 9
    CX 7 10
    DEPOLARIZE2(0.02) 7 10
    DEPOLARIZE1(0.002) 0 1 2 3 5 6 8 11 12
    CX 2 9
    DEPOLARIZE2(0.02) 2 9
    CX 5 10
    DEPOLARIZE2(0.02) 5 10
    DEPOLARIZE1(0.002) 0 1 3 4 6 7 8 11 12
    CX 5 9
    DEPOLARIZE2(0.02) 5 9
    CX 8 10
    DEPOLARIZE2(0.02) 8 10
    DEPOLARIZE1(0.002) 0 1 2 3 4 6 7 11 12
    DEPOLARIZE1(0.01) 0 1 2 3 4 5 6 7 8
    X_ERROR(0.05) 9
    M 9
    X_ERROR(0.05) 10
    M 10
    X_ERROR(0.05) 11
    M 11
    X_ERROR(0.05) 12
    M 12
    DETECTOR(0, 0, 0) rec[-4]
    DETECTOR(1, 0, 0) rec[-3]
    DEPOLARIZE1(0.001) 0 1 2 3 4 5 6 7 8 9 10
    H 11
    DEPOLARIZE1(0.001) 11
    H 12
    DEPOLARIZE1(0.001) 12
    CX 11 0
    DEPOLARIZE2(0.02) 11 0
    CX 12 1
    DEPOLARIZE2(0.02) 12 1
    DEPOLARIZE1(0.002) 2 3 4 5 6 7 8 9 10
    CX 11 1
    DEPOLARIZE2(0.02) 11 1
    CX 12 2
    DEPOLARIZE2(0.02) 12 2
    DEPOLARIZE1(0.002) 0 3 4 5 6 7 8 9 10
    CX 11 3
    DEPOLARIZE2(0.02) 11 3
    CX 12 4
    DEPOLARIZE2(0.02) 12 4
    DEPOLARIZE1(0.002) 0 1 2 5 6 7 8 9 10
    CX 11 4
    DEPOLARIZE2(0.02) 11 4
    CX 12 5
    DEPOLARIZE2(0.02) 12 5
    DEPOLARIZE1(0.002) 0 1 2 3 6 7 8 9 10
    CX 11 6
    DEPOLARIZE2(0.02) 11 6
    CX 12 7
    DEPOLARIZE2(0.02) 12 7
    DEPOLARIZE1(0.002) 0 1 2 3 4 5 8 9 10
    CX 11 7
    DEPOLARIZE2(0.02) 11 7
    CX 12 8
    DEPOLARIZE2(0.02) 12 8
    DEPOLARIZE1(0.002) 0 1 2 3 4 5 6 9 10
    DEPOLARIZE1(0.001) 0 1 2 3 4 5 6 7 8 9 10
    H 11
    DEPOLARIZE1(0.001) 11
    H 12
    DEPOLARIZE1(0.001) 12
    CX 0 9
    DEPOLARIZE2(0.02) 0 9
    CX 3 10
    DEPOLARIZE2(0.02) 3 10
    DEPOLARIZE1(0.002) 1 2 4 5 6 7 8 11 12
    CX 3 9
    DEPOLARIZE2(0.02) 3 9
    CX 6 10
    DEPOLARIZE2(0.02) 6 10
    DEPOLARIZE1(0.002) 0 1 2 4 5 7 8 11 12
    CX 1 9
    DEPOLARIZE2(0.02) 1 9
    CX 4 10
    DEPOLARIZE2(0.02) 4 10
    DEPOLARIZE1(0.002) 0 2 3 5 6 7 8 11 12
    CX 4 9
    DEPOLARIZE2(0.02) 4 9
    CX 7 10
    DEPOLARIZE2(0.02) 7 10
    DEPOLARIZE1(0.002) 0 1 2 3 5 6 8 11 12
    CX 2 9
    DEPOLARIZE2(0.02) 2 9
    CX 5 10
    DEPOLARIZE2(0.02) 5 10
    DEPOLARIZE1(0.002) 0 1 3 4 6 7 8 11 12
    CX 5 9
    DEPOLARIZE2(0.02) 5 9
    CX 8 10
    DEPOLARIZE2(0.02) 8 10
    DEPOLARIZE1(0.002) 0 1 2 3 4 6 7 11 12
    DEPOLARIZE1(0.01) 0 1 2 3 4 5 6 7 8
    X_ERROR(0.05) 9
    M 9
    X_ERROR(0.05) 10
    M 10
    X_ERROR(0.05) 11
    M 11
    X_ERROR(0.05) 12
    M 12
    SHIFT_COORDS(0, 0, 1)
    DETECTOR(0, 0, 0) rec[-4]
    DETECTOR(1, 0, 0) rec[-3]
    DEPOLARIZE1(0.001) 0 1 2 3 4 5 6 7 8 9 10
    H 11
    DEPOLARIZE1(0.001) 11
    H 12
    DEPOLARIZE1(0.001) 12
    CX 11 0
    DEPOLARIZE2(0.02) 11 0
    CX 12 1
    DEPOLARIZE2(0.02) 12 1
    DEPOLARIZE1(0.002) 2 3 4 5 6 7 8 9 10
    CX 11 1
    DEPOLARIZE2(0.02) 11 1
    CX 12 2
    DEPOLARIZE2(0.02) 12 2
    DEPOLARIZE1(0.002) 0 3 4 5 6 7 8 9 10
    CX 11 3
    DEPOLARIZE2(0.02) 11 3
    CX 12 4
    DEPOLARIZE2(0.02) 12 4
    DEPOLARIZE1(0.002) 0 1 2 5 6 7 8 9 10
    CX 11 4
    DEPOLARIZE2(0.02) 11 4
    CX 12 5
    DEPOLARIZE2(0.02) 12 5
    DEPOLARIZE1(0.002) 0 1 2 3 6 7 8 9 10
    CX 11 6
    DEPOLARIZE2(0.02) 11 6
    CX 12 7
    DEPOLARIZE2(0.02) 12 7
    DEPOLARIZE1(0.002) 0 1 2 3 4 5 8 9 10
    CX 11 7
    DEPOLARIZE2(0.02) 11 7
    CX 12 8
    DEPOLARIZE2(0.02) 12 8
    DEPOLARIZE1(0.002) 0 1 2 3 4 5 6 9 10
    DEPOLARIZE1(0.001) 0 1 2 3 4 5 6 7 8 9 10
    H 11
    DEPOLARIZE1(0.001) 11
    H 12
    DEPOLARIZE1(0.001) 12
    CX 0 9
    DEPOLARIZE2(0.02) 0 9
    CX 3 10
    DEPOLARIZE2(0.02) 3 10
    DEPOLARIZE1(0.002) 1 2 4 5 6 7 8 11 12
    CX 3 9
    DEPOLARIZE2(0.02) 3 9
    CX 6 10
    DEPOLARIZE2(0.02) 6 10
    DEPOLARIZE1(0.002) 0 1 2 4 5 7 8 11 12
    CX 1 9
    DEPOLARIZE2(0.02) 1 9
    CX 4 10
    DEPOLARIZE2(0.02) 4 10
    DEPOLARIZE1(0.002) 0 2 3 5 6 7 8 11 12
    CX 4 9
    DEPOLARIZE2(0.02) 4 9
    CX 7 10
    DEPOLARIZE2(0.02) 7 10
    DEPOLARIZE1(0.002) 0 1 2 3 5 6 8 11 12
    CX 2 9
    DEPOLARIZE2(0.02) 2 9
    CX 5 10
    DEPOLARIZE2(0.02) 5 10
    DEPOLARIZE1(0.002) 0 1 3 4 6 7 8 11 12
    CX 5 9
    DEPOLARIZE2(0.02) 5 9
    CX 8 10
    DEPOLARIZE2(0.02) 8 10
    DEPOLARIZE1(0.002) 0 1 2 3 4 6 7 11 12
    DEPOLARIZE1(0.01) 0 1 2 3 4 5 6 7 8
    X_ERROR(0.05) 9
    M 9
    X_ERROR(0.05) 10
    M 10
    X_ERROR(0.05) 11
    M 11
    X_ERROR(0.05) 12
    M 12
    SHIFT_COORDS(0, 0, 1)
    DETECTOR(0, 0, 0) rec[-4] rec[-8]
    DETECTOR(1, 0, 0) rec[-3] rec[-7]
    X_ERROR(0.05) 0
    M 0
    X_ERROR(0.05) 1
    M 1
    X_ERROR(0.05) 2
    M 2
    X_ERROR(0.05) 3
    M 3
    X_ERROR(0.05) 4
    M 4
    X_ERROR(0.05) 5
    M 5
    X_ERROR(0.05) 6
    M 6
    X_ERROR(0.05) 7
    M 7
    X_ERROR(0.05) 8
    M 8
    DEPOLARIZE1(0.01) 9 10 11 12
    DETECTOR(0, 0, 1) rec[-9] rec[-6] rec[-8] rec[-5] rec[-7] rec[-4] rec[-13] rec[-17]
    DETECTOR(1, 0, 1) rec[-6] rec[-3] rec[-5] rec[-2] rec[-4] rec[-1] rec[-12] rec[-16]
    OBSERVABLE_INCLUDE(0) rec[-9] rec[-8] rec[-7]
''')

error_model.txt

stim.DetectorErrorModel('''
    error(0.07819) D0 D2
    error(0.102128) D0 D4 L0
    error(0.102128) D0 D4 L0 ^ D1 D5 ^ L0
    error(0.0313222) D0 D4 L0 ^ D3 D5 ^ L0
    error(0.07819) D1 D3
    error(0.105303) D1 D5
    error(0.0860003) D2
    error(0.108379) D2 D4 L0
    error(0.117794) D2 D4 L0 ^ D3 D5 ^ L0
    error(0.0159144) D2 D4 L0 ^ D5 ^ L0
    error(0.0860003) D3
    error(0.108379) D3 D5
    error(0.117794) D4 D5
    error(0.0159144) D4 D5 ^ D2
    error(0.07239) D4 D6
    error(0.0313222) D4 D7
    error(0.120844) D4 L0
    error(0.0159144) D4 L0 ^ D2
    error(0.120844) D5
    error(0.07239) D5 D7
    error(0.0159144) D5 ^ D3
    error(0.166418) D6 D7
    error(0.169079) D6 L0
    error(0.0159144) D6 L0 ^ D4 L0
    error(0.166418) D7
    error(0.0159144) D7 ^ D5
    detector(0, 0, 0) D0
    detector(1, 0, 0) D1
    shift_detectors(0, 0, 1) 0
    detector(0, 0, 0) D2
    detector(1, 0, 0) D3
    shift_detectors(0, 0, 1) 0
    detector(0, 0, 0) D4
    detector(1, 0, 0) D5
    detector(0, 0, 1) D6
    detector(1, 0, 1) D7
''')

(Sorry for I did not combine some same-type gates into one line so it's a little uncomfortable to read.) I'm not sure whether I made some mistakes in writing the circuits or there is any error mechanism I'm not aware of.

Inm
  • 505
  • 3
  • 8

1 Answers1

2

Now that v1.8.0 is out I can use the new explain_detector_error_model_errors feature to find the circuit error causing the symptoms D0 D4 L0 ^ D1 D5 ^ L0:

import stim
circuit = ...your_circuit...

explained = circuit.explain_detector_error_model_errors(
    dem_filter=stim.DetectorErrorModel("""
        error(1) D0 D4 L0 ^ D1 D5 ^ L0
    """),
    reduce_to_one_representative_error=True,
)


print(explained[0].circuit_error_locations[0])

And the simplest way to cause those symptoms is...

CircuitErrorLocation {
    flipped_pauli_product: Z3
    Circuit location stack trace:
        (after 0 TICKs)
        at instruction #9 (DEPOLARIZE1) in the circuit
        at target #1 of the instruction
        resolving to DEPOLARIZE1(0.001) 3
}

A Z error on qubit 3 right after it is hit by a Hadamard in the first layer of Hadamards.

This isn't a logical error, by the way. Keep in mind the two L0 in D0 D4 L0 ^ D1 D5 ^ L0 cancel out. It's a not-so-well decomposed D0 D1 D4 D5.

Craig Gidney
  • 36,389
  • 1
  • 29
  • 95
  • 'Also, you appear to be directly measuring the stabilizers instead of measuring the edges that you're supposed to measure.', Sorry I don't quite understand the meaning of this. Could you explain a bit more? – Inm Jan 29 '22 at 07:59
  • @Inm The Bacon-Shor code is made up of 2-body parity measurements. The measurements don't all commute with each other, but you can multiply combinations of them together to recover the stabilizers that are actually used. – Craig Gidney Jan 29 '22 at 08:02
  • I used the protocol that using one ancilla qubit to make CNOT with other 6 qubits and measure a stabilizer at one time. – Inm Jan 29 '22 at 08:05
  • Besides, the encoding procedure should be fault-tolerant. The point is that I can't think of an error can result in symptoms D0 D1 D4 D5. When I set only 2 detectors every cycle for Z stabilizer measurement, I think the number of symptoms caused by any single error should be no more than 2. – Inm Jan 29 '22 at 08:51
  • 1
    @Inn The latest dev release of stim has a match_errors command line mode where it lists the circuit errors that cause dem errors. They'll be a python method for it, soon. Keep in mind you have DEPOLARIZE2 operations, which have a YY term and each Y term can causes two detection events. And furthermore, using six cnots to measure a stabilizer means that if error occurs on the measurement qubit halfway through that process, it will kick out onto three separate data qubits. The only reason you don't have worse errors is there's only four stabilizers to flip! – Craig Gidney Jan 29 '22 at 14:12
  • @Inn The encoding procedure you used is not fault tolerant. The Bacon-Shor code does have fault tolerant initialization... but you didn't use that. – Craig Gidney Jan 29 '22 at 14:13
  • I noticed you have added explain_error in python api, it's really helpful, Thanks! – Inm Jan 30 '22 at 06:48
  • The encoding procedure is adopt from here, it says it's FT. – Inm Jan 30 '22 at 07:05
  • Huh. Yes, figure 2 (a) does claim that it's fault tolerant. The figure shows an arbitrary state going in, which would not be fault tolerant. But the caption is correct that specifically a Z eigenstate is protected by the time the blue region begins. And I think that's the relevant case for you. – Craig Gidney Jan 30 '22 at 08:29
  • @Inm updated the answer now that I have hard data on the error – Craig Gidney Jan 30 '22 at 08:53
  • I check the circuits again and find I set the detector xor between n/n+1 rather than n/n+2 cycle... I debug this and now that it works well. Thanks again for adding the method to locate error in stim, it should be of help in some situations. – Inm Jan 30 '22 at 09:14
  • That results in the detectors being sensitive to errors spanning all the way back to the start of time, right? One way to detect that kind of thing while coding is to make the circuit unnecessarily deep and check if the detector error model has huge terms (terms growing linearly with depth). An early warning sign is decompose_errors=True starts failing when you're confident it should succeed. – Craig Gidney Jan 30 '22 at 18:26