No label defined (Q4037772)

Van der Corput's method of exponential sums has been developed into a delicate theory of exponent pairs due to \textit{J. van der Corput} [Math. Ann. 87, 39-65 (1922; JFM 48.0181.01)] and \textit{E. Phillips} [Q. J. Math., Oxf., Ser. 4, 209-225 (1933; Zbl 0007.29801)], whereby a given exponent pair \((\lambda,\mu)\) can be transformed into another by applying the \(A\)- and \(B\)-processes. \textit{S. W. Graham} [J. Lond. Math. Soc., II. Ser. 33, 203-218 (1986; Zbl 0561.10018)] regarded such processes as linear transformations on projective space, and gave the first published account of computing appropriate optimal exponent pairs from the processes. In specific applications a matrix \({\displaystyle{\theta={{a b c} \choose {d e f}}}}\) is given, and the best exponent pair available for the method is related to the infimum of the ratio \(\theta(\lambda,\mu)={{a\lambda+b\mu+c} \over {d\lambda+e\mu+f}}\), taken over all exponent pairs \((\lambda,\mu)\) produced by the said processes. Graham gave an algorithm for the computation of this infimum, but the scheme may have to be aborted for some \(\theta\) when the indeterminate case in the algorithm is reached. The case arises when there is no sufficient condition to specify whether \(\inf \theta=\inf \theta A\) or \(\inf \theta=\inf \theta BA\). In the present paper the author gives an improvement by having other conditions and procedures on how to proceed when one needs to abort. He also states a conjecture giving a sufficient condition for \(\inf \theta=\inf \theta BA\).