Mellin Transform of Cosine Analytic Continuation

9.5 Analytic continuation of Mellin transforms

Now M [ e - x ; s ] = 0 x s - 1 e - x d x = Γ ( s )

is analytic in the right-hand plane 0 < ( s ) .

Consider next the MT of the function e - i t . This is given by M [ e - i x ; s ] = 0 x s - 1 e - i x d x .

Note that the integral does not converge absolutely, only conditionally. We can evaluate the integral by considering C z s - 1 e - i z d z

where the contour C is as shown in figure 10.1 .


PIC

Figure 9.16: Contour C = L 1 + C 1 + L 2 + C 2


From Cauchy's theorem C z s - 1 e - i z d z = 0

since the integrand is analytic inside C . Now | I C 2 | = - π 2 0 i δ e i θ ( δ e i θ ) s - 1 e - i δ ( cos θ + i sin θ ) d θ 0 ,

as δ 0 provided ( s ) > 0 . Next with s = σ + i λ | C 1 | = - - π 2 0 ( R e i θ ) s - 1 i R e i θ e - i R cos θ + R sin θ d θ .

So | C 1 | e | λ | π 2 0 π 2 e λ θ R σ e - R sin θ d θ

e | λ | π 2 R σ e | λ | π 2 0 π 2 e - 2 R θ π d θ

< π e | λ | π 2 e | λ | π 2 R σ - 1 .

Thus | C 1 | 0 as R

provided ( s ) < 1 . Hence taking the limit as δ 0 and R we obtain L 1 + L 2 = 0

giving 0 x s - 1 e - i x d x = 0 ( e - i π 2 y ) s - 1 e - i π 2 e - y d y

= e - i s π 2 0 y s - 1 e - y d y = e - i s π 2 Γ ( s )

Hence the Mellin transform of e i x is given by

M [ e - i x ; s ] = Γ ( s ) e - i s π 2 , (9.5.1)

provided 0 < ( s ) < 1 .

We can analytically continue M [ e - i x ; s ] into the right half-plane ( s ) 1 using the formula (9.5.1 ).



Similarly we showed earlier that the function f ( x ) = 1 ( 1 + x ) has the Mellin transform

M [ 1 1 + x ; s ] = π sin ( π s ) , (9.5.2)

again provided 0 < ( s ) < 1 . We can use (9.5.2 ) to analytically continue M [ 1 ( 1 + x ) ; s ] into the right half-plane ( s ) 1 . The analytic continuation gives a function which has simple poles at the positive integers.



Let F ( s ) = M [ f ( x ) ; s ] be the Mellin transform of f ( x ) which is analytic in the strip α < ( s ) < β . Suppose that as x we have f ( x ) ~ e - d x ν m = 0 x - r m n = 0 N ( m ) c m n ( log x ) n ,

where ( d ) 0 , ν > 0 and ( r m ) is an increasing sequence in m , and N ( m ) are non-negative integers. Then F ( s ) can be continued analytically into the right-half plane β < ( s ) . The analytic continuation is such that

  1. If ( d ) > 0 , then F ( s ) is analytic in the right-half plane.
  2. If d = - i ω , ω 0 then F ( z ) is analytic in the right half-plane.
  3. If d = 0 then F ( s ) is analytic in the right-half plane except for poles.

The proofs of the results are given in detail in (Bleistein & Handelsman , 1975, chap. 4). Outline proofs are as follows:

Proof of (1) This follows from the properties of the Mellin transform given earlier. Proof of (2) Write F ( s ) = 0 x s - 1 f ( x ) d x = 0 1 x s - 1 f ( x ) d x + 1 x s - 1 f ( x ) , d x ,

and express the second integral as 1 ( f ( x ) - h k ( x ) ) x s - 1 d x + 1 h k ( x ) x s - 1 d x ,

where h k ( x ) = e i ω x ν ( r m ) < k x - r m n = 0 N ( m ) c m n ( log x ) n .

Note that f ( x ) - h k ( x ) = O ( e i ω x ν x - ( r j ) ( log x ) N ( j ) ) where j is such that ( r j ) k .

So the first term is analytic in the extended strip α < ( s ) < k .

We need to prove that the term 1 h k ( x ) x s - 1 d x ,

can be analytically continued in the extended strip.

Let h k ( x ) = e i ω x ν x - r 0 H k ( x ) , where H k ( x ) = O ( ( log x ) N ( 0 ) ) as x .

Now 1 h k ( x ) x s - 1 d x , = 1 e i ω x ν x s - r 0 - 1 H k ( x ) d x

= 1 e i ω x ν x ν - 1 x s - r 0 - ν H k ( x ) d x

= x s - r 0 - ν e i ω x ν H k ( x ) i ω ν 1

- 1 e i ω x ν x s - r 0 - ν - 1 ( s - r 0 - ν ) x H k ( x ) + x d H k ( x ) d x i ω ν d x

The last term is of the form 1 x s - r 0 - ν - 1 e i ω x ν H ̃ k ( x ) d x ,

and is analytic in the extended region ( z ) < ( r 0 ) + ν .

By integrating by parts q times the integral 1 h k ( x ) x s - 1 d x ,

can be analytically continued into the extended region α < ( s ) < ( r 0 ) + q ν .

Putting it all together this shows that the Mellin transform of f ( x ) can be analytically continued into the right-half plane α < ( s ) , as an analytic function.

Proof of (3) Introduce h k ( x ) = ( r m ) < k x - r m n = 0 N ( m ) c m n ( log x ) n .

and note that the Mellin transform of f ( x ) is 0 x s - 1 f ( x ) d x

= 0 1 f ( x ) x s - 1 d x + 1 ( f ( x ) - h k ( x ) ) d x + 1 x s - 1 h k ( x ) d x .

The first term is analytic in α < ( s ) .

As before the term I 2 = 1 x s - 1 ( f ( x ) - h k ( x ) ) d x

is analytic in the strip α < ( s ) < k .

The last term can be computed directly noting that 1 x s - r m - 1 ( log x ) n d x = n ! ( - 1 ) n + 1 ( s - r m ) n + 1 .

Hence M [ f ; s ] = m ( r m ) < k n = 0 N ( m ) ( - 1 ) n + 1 c m n n ! ( s - r m ) n + 1 + I 2 .

The right hand is analytic except for poles at s = r m in the extended strip α < ( s ) < k , and since k is arbitrary this proves the result.

The analytic continuation of M [ f ; s ] into the left-half plane ( s ) < α can be obtained and depends on the properties of f ( x ) as x 0 +

Suppose that as x 0 + we have f ( x ) ~ e - q x μ m = 0 x a m n = 0 K ( m ) b m n ( log x ) n ,

where ( q ) 0 , μ < 0 and ( a m ) is an increasing sequence in m , and K ( m ) are non-negative integers. Then F ( s ) can be continued analytically into the left-half plane ( s ) < α . The analytic continuation is such that

  1. If ( q ) > 0 , then F ( s ) is analytic in the left-half plane ( s ) < β .
  2. If q = - i ω , ω 0 then F ( z ) is analytic in the left half-plane ( s ) < α .
  3. If q = 0 then F ( s ) is analytic in the left-half plane except for poles at s = - a m with Laurent expansion n = 0 K ( m ) b m n ( - 1 ) m n ! ( s + a m ) n + 1 .

In summary, the above results show that the Mellin transform of a function f ( x ) , which has a particular type of asymptotic behaviour for small or large x , can be analytically continued into the whole complex plane with at worst pole singularities at points which can be predicted by the asymptotic behaviour of f ( x ) for x 0 + and x .

Example Consider f ( x ) = e i x = n = 0 ( i x ) n n ! .

Now

M [ e i x ; s ] = Γ ( s ) e i s π 2 , (9.5.3)

which is analytic in the strip 0 < ( s ) < 1 . The analytic continuation of the transform into the left-half plane using (9.5.3 ) shows that the function is analytic in ( s ) < 0 except for simple poles at s = 0 , 1 , - 2 , - n , . . . , as predicted by the theorem.

Example

Consider E i ( x ) = x e - u u d u

= - log ( x ) - γ - n = 1 ( - 1 ) n x n n n ! .

The Mellin transform of E i ( x ) is given by

M [ E i ( x ) ; s ] = Γ ( s ) s , ( s ) > 0 . (9.5.4)

We can analytically continue (9.5.4 ) into the left-half plane and the analytic continuation gives a function which has simple poles at the negative integers and a double pole at s = 0 .



stampcancest2002.blogspot.com

Source: http://oldwww.ma.man.ac.uk/~gajjar/magicalbooks/methods/notest1_2se39.html

0 Response to "Mellin Transform of Cosine Analytic Continuation"

Post a Comment

Iklan Atas Artikel

Iklan Tengah Artikel 1

Iklan Tengah Artikel 2

Iklan Bawah Artikel