A natural continuous interpolating structure for continued fractions (Q1301970)

Any construction associated with a sequence naturally offers an interpolation problem: \( C(n)\) \((n \geq 0) \) being the given sequence, it is that of constructing a function \( C \) of a real variable \( t \geq 0 \) for which \( C(t) \) assumes the given values when \( t=0,1, \ldots \). The solution to this problem is by no means unique: \( D(t) \) being a function whose value is zero when \( t=0,1, \ldots \) (of which the theory of Fourier series provides many examples) and \( E(t) \) being finite for \( t \geq 0 \), \( C(t) + D(t)E(t) \) is a further solution to the problem upon offer. For the case in which \( C(n)\) \((n \geq 0) \) are successive convergents of a continued fraction, suitable assumptions being made, the author constructs a function \( C \) which is continuous for nonnegative real values of its argument and has certain piecewise differentiability properties. Some consequences of this construction are investigated. In the degenerate case in which the \( C(n) \) for \( n \geq 0 \) are the successive convergents of a continued fraction whose partial numerators are all equal, the same being true of the partial denominators, \( C(t) \) has a simple form involving an exponential function, and is differentiable for \( t \geq 0 \). The suggestion prompts itself that other cases may exhibit the same sort of behaviour. A convergent of a continued fraction having a certain form may be expressed as a quotient whose denominator is an orthogonal polynomial and whose numerator is an associated orthogonal polynomial of corresponding order. Many orthogonal polynomials may be expressed as integrals whose contours may so be deformed that the polynomials in question are defined not only for integer orders but also for nonnegative real orders: quotients of such polynomials may thus be defined for nonnegative real orders, as is required.