a1 Department of Mathematics, Kansas State University, Manhattan, Kansas 66502, U.S.A.
a2 Department of Mathematics and Statistics, Lancaster University, Lancaster, LA1 4YF
Let G be a matrix function of type m × n and suppose that G is expressible as the sum of an H∞ function and a continuous function on the unit circle. Suppose also that the (k – 1)th singular value of the Hankel operator with symbol G is greater than the kth singular value. Then there is a unique superoptimal approximant to G in : that is, there is a unique matrix function Q having at most k poles in the open unit disc which minimizes s∞(G – Q) or, in other words, which minimizes the sequence
with respect to the lexicographic ordering, where
and Sj(·) denotes the jth singular value of a matrix. This result is due to the present authors [PY1] in the case k = 0 (when the hypothesis on the Hankel singular values is vacuous) and to S. Treil[T2] in general. In this paper we give a proof of uniqueness by a diagonalization argument, a high level algorithm for the computation of the superoptimal approximant and a recursive parametrization of the set of all optimal solutions of a matrix Nehari—Takagi problem.
(Received July 11 1994)
(Revised September 26 1994)
† V. V. Peller'research was supported by an NSF grant in modern analysis