Quantum error correction II:
Detecting bit-flips and phase-flips
Austin Fowler
What will the measurement results be?
Make sure you can derive these results
0 1
0 1
A useful definition
0 1
0 1
Repeated detection
Consider a data qubit (top) being checked by a measure qubit (bottom)
Repeated detection: results without errors
0 0 0 0
In the absence of errors, each measurement will report zero
Repeated detection: results with a data error?
0
Now let’s consider the case of a bit flip error occurring on the data qubit
Repeated detection: results with a data error?
0 1 1 1
A single data qubit bit flip changes all future measurements
Repeated detection: what about a measurement error?
0
A measurement error can be modeled as a bit flip before the measurement gate
Repeated detection: what about a measurement error?
0 1 0 0
A measurement error flips one result
Repeated detection: summary
0 1 0 0
● A single data error permanently flips the measurement stream
● A single measurement error just flips a single bit
Repeated detection
● Circuit detects if the top and bottom qubits are the same or different, it
calculates the parity as was done in the classical repetition code
● Can use it to find errors in states without collapse
● Not enough to identify which qubit suffered an error
The quantum repetition code
● The above circuit can monitor states of the form
● Let’s discuss how the above can reliably find any single bit flip
The quantum repetition code: no errors
0 0 0 0
0 0 0 0
● In the absence of errors, each new measurement will be the same as the
previous measurement
● A detector is a set of measurements with an expected parity in the absence of
errors
The quantum repetition code: no errors
0 0 0 0
1 1 1 1
● Every sequential pair of measurements is a detector
● Note that it’s not the specific value of the individual measurements that
matters, just the parity of the set
The quantum repetition code: one error
X
0 0 1 1
0 0 0 0
● Detector: a set of measurements with an expected parity
● Detection event: a detector with an unexpected parity
○ A single error can lead to one detection event
The quantum repetition code: one error
0 0 1 1
X
0 0 1 1
● Detector: a set of measurements with an expected parity
● Detection event: a detector with an unexpected parity
○ A single error can lead to one detection event
○ A single error can lead to two detection events
The quantum repetition code: one error
0 1 1 1
X
0 0 1 1
● Detector: a set of measurements with an expected parity
● Detection event: a detector with an unexpected parity
○ A single error can lead to two detection events in two different detection rounds
○ This is rare as there is very small region in circuit where this can happen
The quantum repetition code: no errors
0 0 0 0
0 0 0 0
● Let’s shrink the detector bubbles so they only touch their measurements
○ Can represent each possible set of detection events with an edge
○ Build a graph, thin edges means few errors lead to that set of detection events
The quantum repetition code: unknown errors
0 1 1 1
0 1 0 0
● Suppose we run an experiment and observed this pattern of measurements
The quantum repetition code: unknown errors
0 1 1 1
0 1 0 0
● Suppose we run an experiment and observed this pattern of measurements
● Highlight the detectors that are detection events
The quantum repetition code: unknown errors
0 1 1 1
0 1 0 0
● Suppose we run an experiment and observed this pattern of measurements
● Highlight the detectors that are detection events
● Use minimum weight perfect matching (maximum probability, wt = -ln pedge)
○ Won’t chose the low probability diagonal edge in this case
The quantum repetition code: inferred errors
X
0 1 1 1
X
0 1 0 0
● Suppose we run an experiment and observed this pattern of measurements
● Highlight the detectors that are detection events
● Use minimum weight perfect matching (maximum probability, wt = -ln pedge)
● Infer the presence of errors
The quantum repetition code: performance
X
0 1 1 1
X
0 1 0 0
● Provided errors are rare, random, and independent, d data qubits enables the
detection and correction (in software) of up to (d-1)/2 X errors
● Logical errors exponentially suppressed with code distance due to low
probability of many errors on a single graph path from top to bottom
The quantum repetition code: no errors
● Can use the Z detection circuit to protect states
● Next time, the math we will need to study more complex codes, namely
stabilizers
