Page 623 - Handbook Of Integral Equations
P. 623

12.3-2. The Hermite Interpolation Polynomial

               The Hermite interpolation polynomial is used for the construction of the canonical function of the
               nonhomogeneous Riemann problem in Subsections 10.4-7 and 12.3-9.
                                                                       (j)
                   Let distinct points z k (k =1, ... , m) be given, and a number ∆  (j =0, 1, ... , n k – 1) be
                                                                       k
               assigned to each point z k , where the n k are given positive integers. It is required to construct a
               polynomial U p (z) of the least possible degree such that
                                          (j)
                                 (j)
                                U (z k )= ∆ ,  k =1, ... , m,  j =0, 1, ... , n k – 1,
                                 p
                                          k
                         (j)
               where the U (z k ) are the values of the jth-order derivatives of the polynomial at the points z k . The
                         p
               numbers z k are called the interpolation nodes and n k the interpolation multiplicities at the nodes
               z k .
                   There exists a unique polynomial with these properties. It has the form (e.g., see V. I. Smirnov
               and N. A. Lebedev (1964))
                                      m           n k –1               m
                                           ζ(z)                r
                               U p (z)=              A k,r (z – z k ) ,  p =  n k – 1,
                                         (z – z k ) n k
                                      k=1         r=0                 k=1
                                 m                  r      (j)     r–j
                                                         ∆ k     d    (z – z k ) n k
                           ζ(z)=   (z – z k ) ,  A k,r =                           ,
                                         n k
                                                       j!(r – j)! dz r–j  ζ(z)
                                k=1                 j=0                        z=z k
                                      k =1, ... , m,  r =0, 1, ... , n k – 1;
               and this polynomial is unique.
                   The interpolation polynomial U p (z) constructed for some function f(z) must satisfy the following
               conditions at the points z k :
                             (j)
                                           (j)
                            U (z k )= ∆ (j)  = f (z k ),  k =1, ... , m,  j =0, 1, ... , n k – 1,
                             p        k
                      (j)
               where f (z k ) is the value of the jth-order derivative of f(z) at the point z k .
                 12.3-3. Notion of the Index
               Let L be a smooth closed contour, and let D(t) be a continuous nowhere vanishing function on this
               contour.
                   The index ν of the function D(t) with respect to the contour L is the increment of the argument
               of D(t) along L (traversed in the positive direction) divided by 2π:
                                                        1
                                          ν = Ind D(t)=   [arg D(t)] L .                    (1)
                                                       2π
                   Since ln D(t)=ln |D(t)| + i arg D(t) and since after the traverse the function |D(t)| returns to its
               original value, it follows that [ln D(t)] L = i[arg D(t)] L , and hence
                                                    1
                                               ν =    [ln D(t)] L .                         (2)
                                                   2πi
               The index can be expressed in the form of an integral as follows:
                                               1               1
                                  ν = Ind D(t)=      d ln D(t)=     d arg D(t).             (3)
                                              2πi  L          2π  L
               If the function D(t) is not differentiable but has bounded variation, then the integral is regarded as
                                                                ˘
               the Stieltjes integral. Since D(t) is continuous, the image Γ of the closed contour L is a closed
               contour as well, and the increment of the argument D(t) along L is a multiple of 2π. Hence, the
               following assertions hold.




                 © 1998 by CRC Press LLC








               © 1998 by CRC Press LLC
                                                                                                             Page 606
   618   619   620   621   622   623   624   625   626   627   628