Quantum error correction I:
What are quantum errors?
Austin Fowler
Classical error correction fundamentals
Suppose you wish to store a bit, 0 or 1
Suppose your best bit has a probability p per unit time of flipping
How can you increase the chance of successful storage?
Classical error correction fundamentals
Suppose you wish to store a bit, 0 or 1
Suppose your best bit has a probability p per unit time of flipping
How can you increase the chance of successful storage?
Answer: store multiple copies, periodically take majority vote (instant and perfect)
success! success! fail :-(
0 0 0 0 0 1 1
0 1 0 0 0 0 1
1 error majority 2 errors majority 3 errors majority
0 0 0 1 0 0 1
0 0 0 1 0 1 1
0 0 0 0 0 1 1
Classical error correction fundamentals
The state 00000 is called logical 0
The state 11111 is called logical 1
This is the classical repetition code, one of many possible error correction codes
success! success! fail :-(
0 0 0 0 0 1 1
0 1 0 0 0 0 1
1 error majority 2 errors majority 3 errors majority
0 0 0 1 0 0 1
0 0 0 1 0 1 1
0 0 0 0 0 1 1
Classical error correction fundamentals
The code distance is the number of bits that need to be flipped to convert logical 0
into logical 1
Code distance d = 5 in this example
A distance d code can only fail if at least (d+1)/2 errors occur, p → O(p3)
success! success! fail :-(
0 0 0 0 0 1 1
0 1 0 0 0 0 1
1 error majority 2 errors majority 3 errors majority
0 0 0 1 0 0 1
0 0 0 1 0 1 1
0 0 0 0 0 1 1
Classical error correction fundamentals
Don't actually need to measure bits directly
Pairwise parity measurements are sufficient
Use minimum weight perfect matching (MWPM) to decode instead of majority vote
Don't need to apply corrections, track in software (green)
pairwise pairwise
parity parity
0 0 0 0
1 0
0 1 1 1
1 error 1 graph solution 2 errors 1 graph solution
0 0 1 1
0 0
0 0 1 1
0 1
0 0 0 0
Quantum errors are more complex, eg: leakage
There is a lot more to deal with that just classical bit flips...
arXiv:2211.04728
Leakage in the surface code
Start the center qubit in |2⟩ and observe
One leakage event causes many problems
Spreading: transport in CZ gates, unwanted interactions
Correlated errors across many cycles and qubits
arXiv:2211.04728
Measurement-induced state transitions
Measure/readout
line
|0)
|1)
Resonator |2)
|3)
Qubit
Crank up the resonator probe power
High-energy impacts
Repetitive correlated sampling
Number of errors
Time (ms)
McEwen et al., Nat. Phys. 18 107-111 (2022), arXiv:2104.05219
Real quantum computers are not ideal!
Controlled-X
Hadamard
Real quantum computers are not ideal!
Controlled-X
Hadamard
Real quantum computers are not ideal!
Controlled-X
Hadamard
Real quantum computers are not ideal!
This is the error model we actually want (and all QEC can handle):
Lots of physics and engineering still required to build a device with this error model.
Decomposing errors
● It would be nice if we only need to worry about these errors:
Decomposing errors
● It would be nice if we only need to worry about these errors:
● Can write arbitrary errors as a linear combinations of X and Z errors:
Decomposing errors
Can do something similar with two-qubit noise:
If we can build a quantum device with only 1- and 2-qubit computational
basis errors not handled at the hardware level, then detecting bit- and
phase-flips and tracking them in software is enough.
Decomposing errors
If we can build a quantum device with only 1- and 2-qubit computational
basis errors not handled at the hardware level, then detecting bit- and
phase-flips and tracking them in software is enough.
pairwise pairwise
parity parity
0 0 0 0
1 0
0 1 1 1
1 error 1 graph solution 2 errors 1 graph solution
0 0 1 1
0 0
0 0 1 1
0 1
0 0 0 0
