Page 27 -
P. 27

II. The Partition Function
                        20
                                              π 2  +  1  log(1 − e −w )
                                              6w     2

                                                    1      1         1     e −kw

                                             +                   −      +         .
                                                    k   e kw  − 1   kw      2
                           The form of this series is very suggestive. Indeed we recognize any
                                  1              A(kw)
                        series     A(kw)              w as a Riemann sum, approximating
                                  k               kw
                                                ∞ A(t)
                        the Riemann integral        dt for small positive w. It should come
                                              0   t
                        as no surprise then, that such series are estimated rather accurately.
                        So let us review the “Riemann sum story”.
                        Riemann Sums

                        Suppose that φ(x) is a positive decreasing function on (0, ∞) and
                                                         ∞
                        that h> 0. The Riemann sum          φ(kh)h is clearly equal to the
                                                         k 1
                        area of the union of rectangles and so is bounded by the area under
                                            ∞                 ∞
                        y   φ(x). Hence         φ(kh)h ≤      φ(x)dx. On the other hand,
                                            k 1             0

                        the series  ∞   φ(kh)h can be construed as the area of this union of
                                    k 0
                        these rectangles and, as such, exceeds the area under y   φ(x).So
                                                               ∞
                        this time we obtain   ∞  φ(kh)h ≥       φ(x)dx.
                                              k 0            0
                           Combining these two inequalities tells us that the Riemann sum
                        lies within h · φ(0) of the Riemann integral. This is all very nice and
                        ratheraccuratebutitrefersonlytodecreasingfunctions.However,we
                        may easily remedy this restriction by subtracting two such functions.
                        Thereby we obtain
                          ∞                          ∞

                            [φ(kh) − ψ(kh)]h −        [φ(x) − ψ(x)]   h[φ(0) + ψ(0)].
                         k 1                       0
                        Calling φ(x) − ψ(x)   Fàx) and then observing that φ(0) + ψ(0)
                        is the total variation V of Fàx) we have the rather general result

                                       ∞               ∞

                                          F(kh)h −       Fàx)   h · V(F).             (4)
                                      k 1             0
                           To be sure, we have proven this result only for real functions but
                        in fact it follows for complex ones, by merely applying it to the real
                        and imaginary parts.
   22   23   24   25   26   27   28   29   30   31   32