We shall use the following notation: x will always denote a variable defined on (-co, co), y will always denote a variable defined on (0, co), ]/f(x)]], will stand for the norm off in I,,(--co, co) and ]]f(y)]], for the norm of f in L,(O, co); for p > 0, W,(x) = (1 + x*)O’* exp(-x2/2) and V,(y) = (1 + y)“‘* exp(--y/2); n will denote a strictly positive integer, and q, an arbitrary polynomial of degree at most n; by c we shall denote positive numbers depending at most on /3, and by c(s) positive numbers depending at most on j3 and on the variables enclosed by the parentheses, but not necessarily the same positive number if they appear more than once in the same formula. This paper is a sequel to [ 11, and like it has been deeply influenced by the ideas of G. Freud. The first five theorems below present polynomial inequalities on (-co, co) involving the weight W,(x); the case p = 0 of these results was proved in [ 11. The remaining five theorems present polynomial inequalities on [0, co) involving the weight V,(y). The functions W,(x) were introduced by Freud in [2]. Note that if Qo(x) = ln[ WD(x)] and p > 16[exp(1/16) l] > 1.04, then Q0[(p/16)“‘] 1, W,(x) is neither very strongly regular nor superregular in the sense of Mhaskar [4. 51. Hence the theorems in this paper are not contained in, nor can be trivially inferred from, the results of these authors. We start with:
Read full abstract