Find n → ∞ lim n ∫ 0 1 cos ( n x ) cosh − 1 ( x 1 ) d x
This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try
refreshing the page, (b) enabling javascript if it is disabled on your browser and,
finally, (c)
loading the
non-javascript version of this page
. We're sorry about the hassle.
nice solution
Thanks^^ Just noticed the wiki-page on "Differentiation under the integral sign" lists the first integral as a counter-example (which is true for the wrong parameter choice), so I'll add a rough sketch of how to calculate the "Dirichlet-Integral": I ( t ) : = ∫ 0 ∞ y sin ( y ) e − y t d y = : ∫ 0 ∞ f ( y , t ) d y , t → ∞ lim I ( t ) = 0 , I ( 0 ) = ? ? For t > 0 , one can use differentiation under the integral sign: I ( 1 ) ( t ) = − ∫ 0 ∞ sin ( y ) e − y t d y = − ℑ { ∫ 0 ∞ e ( i − t ) y d y } = − ℑ { i − t − 1 } = − 1 + t 2 1 The remaining integral is simple, and the boundary condition at t → + ∞ determines the integration constant: I ( t ) = − arct g ( t ) + 2 π ⇒ I ( 0 ) = 2 π
Rem.: The above is just a rough sketch - there are (at least) two details missing:
Suppose that ∣ h ∣ < 2 1 t , where t > 0 , then ∣ ∣ ∣ ∣ h e − y ( t + h ) − e − y t ∣ ∣ ∣ ∣ = ∣ ∣ ∣ − y e − y ( t + θ h ) ∣ ∣ ∣ ≤ y e − 2 1 y t y > 0 (for some 0 < θ < 1 , while h → 0 lim h e − y ( t + h ) − e − y t = − y e − y t y > 0 Thus the DCT implies that h → 0 lim h I ( t + h ) − I ( t ) = h → 0 lim ∫ 0 ∞ h e − y ( t + h ) − e − y t y sin y d y = − ∫ 0 ∞ sin y e − y t d y which proves differentiation under the integral sign here. As you say, this still leaves the matter of the continuity of I at 0 .
To be honest, it is easier to evaluate the improper integral x sin x by integrating z e i z around the almost semicircular contour of radius R in the upper half plane, with a small semicircular contour of radius ε avoiding the singularity at 0 . The fact that
∫ 0 ∞ x sin x d x = ε → 0 R → ∞ lim ∫ ε R x sin x d x = 2 1 π is a matter of standard contour integration.
Continuity from above at t = 0 follows from "Leibniz' Convergence Criterium". Let ε > 0 and choose
Then we use the "triangle inequality" to estimate ∣ I ( 0 ) − I ( t ) ∣ ≤ ∣ ∣ ∣ ∣ ∫ n π ∞ x sin ( x ) d x ∣ ∣ ∣ ∣ + ∣ ∣ ∣ ∣ ∫ n π ∞ x sin ( x ) e − x t d x ∣ ∣ ∣ ∣ + ∫ 0 n π ∣ ∣ ∣ ∣ x sin ( x ) ∣ ∣ ∣ ∣ ( 1 − e − x t ) d x With ∣ x sin ( x ) ∣ ≤ 1 and the choice for t the third integral is smaller than 3 ε . We rewrite the first two integrals as alternating sums. A substitution y : = x − k π yields ∫ n π ∞ x sin ( x ) e − x t d x 0 = k = n ∑ ∞ ∫ k π ( k + 1 ) π x sin ( x ) e − x t d x = k = n ∑ ∞ ( − 1 ) k ∫ 0 π y + k π sin ( y ) e − ( y + k π ) t d y = : k = n ∑ ∞ ( − 1 ) k a k < a k < ∫ 0 π k π sin ( y ) ⋅ 1 d y = k π 2 The sequence a k is monotonically decreasing and converges to zero. By "Leibniz' Convergence Criterium" for alternating series, we get t ≥ 0 : ∣ ∣ ∣ ∣ ∫ n π ∞ x sin ( x ) e − x t d x ∣ ∣ ∣ ∣ = ∣ ∣ ∣ ∣ ∣ k = n ∑ ∞ ( − 1 ) k a k ∣ ∣ ∣ ∣ ∣ < ∣ a n ∣ < n π 2 < 3 ε Putting everything together, we have ∣ I ( 0 ) − I ( t ) ∣ < ε
Like I said, rather than all this, just use contour integration.
Log in to reply
Contour Integration is very much simple for this integral
By integration by parts ∫ 0 1 n cos ( n x ) cosh − 1 ( x 1 ) d x = sin ( n x ) lo g ( x 1 + 1 − x 2 ) ∣ ∣ ∣ ∣ ∣ 0 + 1 + ∫ 0 1 x 1 − x 2 sin ( n x ) d x that equals ∫ 0 n x 1 − n 2 x 2 sin ( x ) d x and by the dominated convergence theorem, as n → ∞ the last integral approaches ∫ 0 ∞ x sin ( x ) d x = 2 π as expected.
The Dominated Convergence Theorem seems to work only if the integration domain does not dependent on n . How did you solve that issue in the second line from the bottom?
Log in to reply
The DCT can apply when the limit varies, but you still need the limit integral to be Lebesgue integrable, and x sin x is not Lebesgue integrable over ( 0 , ∞ ) .
Log in to reply
I see now - with variable integration domains S n you could just define f ~ n ( x ) : = f n ( x ) χ S n ( x ) and rewrite the integral over S n as an integral over R . But then the (point-wise) limit of f ~ n ( x ) has to be Lebesgue integrable over R for a dominating function to exist, which is not the case here.
Thanks for the hint!
Problem Loading...
Note Loading...
Set Loading...
Here is another solution without dominated convergence. We use integration by parts to simplify the integral: I n : = n ∫ 0 1 cos ( n x ) a r c o s h ( x 1 ) d x = sin ( n x ) a r c o s h ( x 1 ) ∣ ∣ ∣ ∣ 0 1 − ∫ 0 1 sin ( n x ) ⋅ x 2 x 2 1 − 1 − 1 d x l’H. = 0 − 0 + ∫ 0 1 x 1 − x 2 sin ( n x ) d x We add the zero ± x 1 to split the pole from the remaining integral. Then we substitute y : = n x in the first part: I n = ∫ 0 1 x sin ( n x ) d x + ∫ 0 1 = : f ( x ) ( x 1 − x 2 1 − x 1 ) sin ( n x ) d x = : ∫ 0 n y sin ( y ) d y + ∫ 0 1 f ( x ) sin ( n x ) d x ( ∗ ) We take a closer look at the remainder f ( x ) to check its behavior on the integration domain: x ∈ [ 0 ; 1 ) : ∣ f ( x ) ∣ = ∣ x ∣ 1 − x 2 1 − 1 − x 2 = ∣ x ∣ 1 − x 2 ( 1 − x 2 + 1 ) x 2 = 1 − x 2 ( 1 − x 2 + 1 ) ∣ x ∣ ≤ 1 − x 2 1 absolutely integrable As f ( x ) is absolutely integrable on [ 0 ; 1 ] , the second integral in ( ∗ ) converges to zero by Riemann-Lebesgue's Lemma (the version concerning Fourier-Coefficients). The first integral on the other hand can be solved by Differentiation under the integral sign or contour integration: n → ∞ lim I n = ∫ 0 ∞ y sin ( y ) d y + n → ∞ lim ∫ 0 1 f ( x ) sin ( n x ) d x = 2 π + 0 = 2 π