Spatially coupled LDPC were introduced by FelstrÃ¶m and Zigangirov in 1999. They might be viewed in the following way, take several several instances of a certain LDPC code family, arrange them in a row and then mix the edges of the codes randomly among neighboring layers. Moreover fix the bits of the first and last layers to zero. It has soon been found out that iterative decoding behaves much better for this code than for the original LDPC code. A breakthrough occurred when it was proved by Kudekar, Richardson and Urbanke that these codes attain the capacity of all binary input memoryless output-symmetric channels.

All these nice features of classical spatially coupled LDPC codes suggest to study whether they have a quantum analogue. The fact that spatially coupled LDPC codes may afford to have large degrees and still perform well under iterative decoding would be quite interesting in the quantum setting, since by the very nature of the quantum construction of stabilizer codes the rows of the parity-check matrix of the quantum code have to belong to the code which is decoded by the iterative decoder. This implies that we should have rather large row weights to avoid severe error-floor phenomena and/or oscillatory behavior of iterative decoding which degrades significantly its performance.

With Andriyanova and Maurice, I showed last year that it is possible to come up with coupled versions of quantum LDPC codes that perform excellently under iterative decoding. For instance we have constructed a spatially coupled LDPC code family of rate $\approx \frac{1}{4}$ which performs well under iterative decoding even for noise values close to the hashing bound $p \approx 0.127$.

This represents a tremendous improvement over all previous known families of quantum LDPC codes of the same rate.

I will discuss in this talk what can be expected from this approach when these spatially coupled LDPC codes are used for performing fault tolerant computation.

Speaker(s):

Collection/Series:

PIRSA Number:

14070015