Canonical approach is a way to map QCD at imaginary chemical potential regions to the real ones. The essential idea is simple: Z(¥mu,T) = ¥sum_n Z_n (exp(¥mu/T))^n, where Z_n, the canonical partition functions, do not depend on ¥mu. We construct Zn in the imaginary ¥mu regions, and we can calculate Z(¥mu,T) at any real ¥mu.
Since Z_n drops very fastly as |n| increases, and we must fight against the overlap problem, we need several algorithmic and computational tricks, such as a multi-precision calculation.
In Vladivostok, we have investigated the sources of unstable Z_n, and developed methods to overcome them. I will report these studies and the outcomes, especially those for comparison of heavy ion collision experiment.