∫ 0 1 x Li 4 2 ( x ) d x = B A − E C π D − G F ζ ( H ) + K I π J − M π L ζ ( N ) − P O ζ ( Q ) + T R π S + W ζ U ( V ) − Y π X ζ ( Z ) + γ α π β
The equation is true for positive integers A , B , C , … , Z , α , β , γ . Find the minimum value of the sum of those integers.
Notations :
Li n ( a ) denotes the polylogarithm function, Li n ( a ) = k = 1 ∑ ∞ k n a k .
ζ ( ⋅ ) denotes the Riemann zeta function .
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.
Alternative method: (I did not solve the question because my method was too tedious)
Expressing the integral as a sum and integrating gives:
n = 1 ∑ ∞ n 4 1 m = 1 ∑ ∞ m 4 ( m + n + 2 ) 1
Using Partial Fractions, we get:
n = 1 ∑ ∞ n 4 1 m = 1 ∑ ∞ ( ( n + 2 ) m 4 1 − ( n + 2 ) 2 m 3 1 + ( n + 2 ) 3 m 2 1 + ( n + 2 ) 4 ( m + n + 2 ) 1 − ( n + 2 ) 4 m 1 )
n = 1 ∑ ∞ n 4 1 ( ζ ( 4 ) n + 2 1 − ζ ( 3 ) ( n + 2 ) 2 1 + ζ ( 2 ) ( n + 2 ) 3 1 − ( n + 2 ) 4 H n + 2 )
From here onwards, it should be easy, but very tedious to calculate the exact value.
I think ishan did a fantastic job. I also find this in the paper as Mark :)
That certainly works but, as you say, would be difficult to generalise.
Log in to reply
Not difficult, but tedious. It's just a matter of taking partial fractions, which can be handled rather neatly using Heaviside Cover Up Method.
Log in to reply
@Mark Hennings I will delve deeper into it in the near future and provide any updates that I find.
Looking a little closer, there is a typo in what you have written here. Dividing out by ( n + 1 ) s should give the formula J ( s ) = ( − 1 ) s + 1 ( n + 1 ) s H s + 1 + r = 2 ∑ s ( − 1 ) s − r ( n + 1 ) s − r + 1 1 ζ ( q ) (you forgot to divide out by ( n + 1 ) s within the sum) which makes the general formula ∫ 0 1 x m L i p ( x ) L i q ( x ) d x = r = 1 ∑ ∞ r p 1 { k = 2 ∑ q ( m + r + 1 ) q − k + 1 ( − 1 ) q − k ζ ( k ) + ( m + r + 1 ) q ( − 1 ) q − 1 H m + r + 1 } for p , q ≥ 2 and m ≥ 0 .
The real challenge will be to obtain a closed formula for the RHS of this expression. If nothing else, we need a formula which makes it obvious that this integral is a symmetric function of p and q .
Log in to reply
@Mark Hennings I've fixed the typo. I think the denominators in the last summation for the general result can be simplified using partial fractions by Heaviside Cover Up Method. The wikipedia page for Heaviside Cover Up doesn't provide a method for the general case. I'm talking about a general case for repeated factors. As an explicit example, the case for distinct factors has been used by Mr. Robjohn here .
Log in to reply
The typo still persists in your "In general" formula. I'm thinking about partial fractions...
Log in to reply
@Mark Hennings – I had edited it several times, I hope it has been fixed now. You can refresh your webpage to see it.
Log in to reply
@Ishan Singh – I have to work now, but I think I have cracked the partial fractions business. I will post this later. That should handle the first term of the integral (the sum involving Zeta functions) in terms of Harmonic numbers and Zeta functions. The last term will be expressible in terms of Euler sums and Zeta functions.
Log in to reply
@Mark Hennings – Nice! Do you think it is different from the paper already available? Is it worth publishing?
Log in to reply
@Ishan Singh – A simple induction on m shows that r m ( r + a ) n a m + n − 1 = j = 1 ∑ m ( − 1 ) m − j ( n − 1 m + n − j − 1 ) r j a j − 1 + ( − 1 ) m j = 1 ∑ n ( m − 1 m + n − j − 1 ) ( r + a ) j a j − 1 for all positive integers m , n , a . This means, for example, that r = 1 ∑ ∞ r m ( r + a ) n a m + n − 1 = j = 2 ∑ m ( − 1 ) m − j ( n − 1 m + n − j − 1 ) a j − 1 ζ ( j ) + ( − 1 ) m j = 2 ∑ n ( m − 1 m + n − j − 1 ) a j − 1 ζ ( j ) + ( − 1 ) m + 1 j = 1 ∑ n ( m − 1 m + n − j − 1 ) H a ( j ) for all positive integers m , n , a .
Thus the first of the two terms (my version of the formula) can be expressed in terms of rational multiples of Zeta functions and products of pairs of Zeta functions.
The second term of the formulae will be expressible in terms of the sums r = 1 ∑ ∞ r b H r + a r = 1 ∑ ∞ ( r + a ) n H r + a which (after another dose of partial fractions) can be written in terms of Euler sums and Zeta functions, as I thought.
Pushing this all through, and getting a nice closed expression for the integral is still a major undertaking in combinatorics!
Log in to reply
@Mark Hennings – What you did with Partial Fractions is remarkable! Has it been done before, or what we have here is a new result?
Log in to reply
@Ishan Singh – I don't know, but I would be very surprised if it were original. All I did was to work it out by hand for m = 1 , and then (inductively) work out the result for m = 2 , 3 , at which point I guessed the pattern (and then wasted a lot of time finding the easiest expression for the pattern), and proved it by induction. There are plenty of people out there who are much better at combinatorics than I am, and I am sure the question has been asked before!
Log in to reply
@Mark Hennings – Nevertheless, it is interesting and beautiful. Is there a way to know whether it is original? (I'm asking since the paper you provided didn't derive a closed form and even their method was different than mine).
@Mark Hennings – @Mark Hennings Also, won't the method (of taking partial fractions) for the second summation (involving H_n) be exactly similar as for the first? If yes, then we'd have a beautiful closed form. So why did you say that "it is still a major undertaking in combinatorics" ?
Log in to reply
@Ishan Singh – It depends on what we are prepared to be satisfied with. I guess we could push this through, and end up with an expression involving about six complicated sums (some of them double sums).
I would not regard this as an elegant closed form without some tidying. It ought/might well be possible to simplify/combine these sums, and end up with a neater expression. As previously mentioned, it would be nice if the final formula was visibly symmetric in p and q (my notation), for example. Tidying up the elements of the sums would be quite a challenge!
I will have a shot at getting an expression, and seeing whether it will tidy up (you are right in that the same partial fraction result can be reused).
Log in to reply
@Mark Hennings – Okay. Meanwhile I'll also try that.
Log in to reply
@Ishan Singh – I am almost there, I think (although I have gone back to the recurrence relations in places, and don't need the full-blown partial fractions result). I will keep you posted.
Log in to reply
@Mark Hennings – @Mark Hennings Thanks. Looking forward to it.
Log in to reply
@Ishan Singh – OK, I have a formula which works (I am still checking it against the results in the Freitas paper I linked to, but all is well so far). The various coefficients are at worst finite sums of binomial coefficients and harmonic numbers.
Since I cannot find a reference to a concrete formula, it may be that this could be published. Therefore I do not think it is sensible to post the result here just yet (anyway, my solution runs to about 7 A4 pages of typescript), since that would just make the result public! I have e-mailed Prof. Freitas (one of whose areas of specialism is Special Functions) to ask if this result is new. If it isn't, I will simply post the result. If it is, we will have to think about publication.
Log in to reply
@Mark Hennings – That is awesome! Hope Prof. Freitas sends good news :)
@Ishan Singh – If I can make it work, it is certainly different. Whether it is new is another matter. I would have to try to find out. First things first, though.
Awesome and that is an understatement.
According to this paper , we have ∫ 0 1 x L i 4 ( x ) 2 d x = = 2 5 6 3 6 1 − 6 4 5 9 ζ ( 2 ) − 3 2 3 ζ ( 3 ) + 3 2 1 9 ζ ( 4 ) − 4 1 ζ ( 2 ) ζ ( 3 ) − 8 3 ζ ( 5 ) + 1 6 7 ζ ( 6 ) + 4 1 ζ ( 3 ) 2 − 2 1 ζ ( 3 ) ζ ( 4 ) + 1 2 7 ζ ( 8 ) 2 5 6 3 6 1 − 3 8 4 5 9 π 2 − 3 2 3 ζ ( 3 ) + 2 8 8 0 1 9 π 4 − 2 4 1 π 2 ζ ( 3 ) − 8 3 ζ ( 5 ) + 2 1 6 0 1 π 6 + 4 1 ζ ( 3 ) 2 − 1 8 0 1 π 4 ζ ( 3 ) + 1 6 2 0 0 1 π 8 Mathematica confirms this answer numerically to 5 0 decimal places.
Thus we have A = 3 6 1 , B = 2 5 6 , C = 5 9 , D = 2 , E = 3 8 4 , F = 3 , G = 3 2 , H = 3 , I = 1 9 , J = 4 , K = 2 8 8 0 , L = 2 , M = 2 4 , N = 3 , O = 3 , P = 8 , Q = 5 , R = 1 , S = 6 , T = 2 1 6 0 , U = 2 , V = 3 , W = 4 , X = 4 , Y = 1 8 0 , Z = 3 , α = 1 , β = 8 , γ = 1 6 2 0 0 , making the answer 2 2 6 2 0 .
@Mark Hennings ,we really liked your comment, and have converted it into a solution. If you subscribe to this solution, you will receive notifications about future comments.
Log in to reply
This is not the complete solution. I want someone to post the complete solution. Please can you delete it?
Log in to reply
I am afraid that it is a complete solution. Using what has already been discovered is part of any scientific endeavour. I could type out a number of recurrence relations from the paper, which between them are sufficient to calculate the integral, but these formulae are already well-expressed in this paper, and I have provided a link so that the interested can read them.
You and @Ishan Singh have alternative solutions; let us see one of them!
I have an original proof of this. (Although I haven't seen what's in the paper, may or may not be the same method)
Log in to reply
@Ishan Singh – The paper uses integration by parts to consider the integrals J ( m , p , q ) = ∫ 0 1 x m L i p ( x ) L i q ( x ) d x and uses integration by parts (integrate x m , differentiate L i p ( x ) L i q ( x ) ) to find a recurrence relation J ( m , p , q ) = m + 1 1 [ ζ ( p ) ζ ( q ) − J ( m , p − 1 , q ) − J ( m , p , q − 1 ) ] for m ≥ 0 and p , q ≥ 2 which enables us to express J ( m , p , q ) in terms of zeta functions and the integrals of the form J ( m , 1 , r ) .
It then uses different integration by parts strategies to define more recurrence relations which enable the integrals J ( m , 1 , q ) = − ∫ 0 1 x m ln ( 1 − x ) L i q ( x ) d x to be evaluated. To begin with, see what happens if you integrate ln ( 1 − x ) and differentiate x m L i q ( x ) .
Is this your technique?
Log in to reply
@Mark Hennings – I have posted my method. Would you mind checking it out?
@Ishan Singh – You are obligated to post a proof! =D
What was your method?
Log in to reply
@Ishan Singh – My method was same as that of Mark Hennings. I used the recurrance relation to evaluate it.
Problem Loading...
Note Loading...
Set Loading...
Proof : Let, J ( s ) = ∫ 0 1 x n L i s ( x )
Using I.B.P.,
J ( s ) = [ n + 1 x n + 1 L i s ( x ) ] 0 1 − n + 1 1 ∫ 0 1 x n L i s − 1 ( x ) d x
= n + 1 ζ ( s ) − n + 1 1 J ( s − 1 )
⟹ J ( s ) = n + 1 ζ ( s ) − n + 1 1 J ( s − 1 )
⟹ J ( r ) = n + 1 ζ ( r ) − n + 1 1 J ( r − 1 )
Multiplying both sides with ( − 1 ) r ( n + 1 ) r , we have,
( − 1 ) r ( n + 1 ) r J ( r ) = ( − 1 ) r ( n + 1 ) r − 1 ζ ( r ) + ( − 1 ) r − 1 ( n + 1 ) r − 1 J ( r − 1 )
⟹ ( − 1 ) r ( n + 1 ) r J ( r ) − ( − 1 ) r − 1 ( n + 1 ) r − 1 J ( r − 1 ) = ( − 1 ) r ( n + 1 ) r − 1 ζ ( r )
Taking summation on both sides, we have,
r = 2 ∑ s [ f ( r ) − f ( r − 1 ) ] = r = 2 ∑ s [ ( − 1 ) r ( n + 1 ) r − 1 ζ ( r ) ] ( ∗ )
where f ( r ) = ( − 1 ) r ( n + 1 ) r J ( r ) . Clearly, ( ∗ ) telescopes. Evaluating it, we have,
f ( s ) = f ( 1 ) + r = 2 ∑ s [ ( − 1 ) r ( n + 1 ) r − 1 ζ ( r ) ]
Note that,
J ( 1 ) = − ∫ 0 1 x n ln ( 1 − x ) d x
= r = 1 ∑ ∞ ∫ 0 1 r x n + r d x
= r = 1 ∑ ∞ r ( r + n + 1 ) 1
= n + 1 1 r = 1 ∑ ∞ ( r 1 − r + n + 1 1 )
= n + 1 1 r = 1 ∑ ∞ ( ( r 1 − r + 1 1 ) + ( r + 1 1 − r + 2 1 ) + … + ( r + n 1 − r + n + 1 1 ) )
= n + 1 H n + 1
⟹ f ( 1 ) = − H n + 1
Putting these values in the recurrence relation, we have,
J ( s ) = ( − 1 ) s + 1 ( n + 1 ) s H n + 1 + ( − 1 ) s r = 2 ∑ s [ ( − 1 ) r ( n + 1 ) s − r + 1 ζ ( r ) ] □
Now,
I = ∫ 0 1 x [ L i 4 ( x ) ] 2 d x
= r = 1 ∑ ∞ r 4 1 ∫ 0 1 x r + 1 L i 4 ( x ) d x
Using the proposition,
= r = 1 ∑ ∞ r 4 ( r + 2 ) 3 ζ ( 2 ) − r = 1 ∑ ∞ r 4 ( r + 2 ) 2 ζ ( 3 ) + r = 1 ∑ ∞ z 4 ( r + 2 ) ζ ( 4 ) − r = 1 ∑ ∞ r 4 ( r + 2 ) 4 H r + 2
Now, using partial fractions and the elementary result r = 1 ∑ ∞ r p H r = ( 1 + 2 p ) ζ ( p + 1 ) − 2 1 k = 1 ∑ p − 2 ζ ( k + 1 ) ζ ( q − k ) , it simplifies to,
I = 2 5 6 3 6 1 − 3 8 4 5 9 π 2 − 3 2 3 ζ ( 3 ) + 2 8 8 0 1 9 π 4 − 2 4 1 π 2 ζ ( 3 ) − 8 3 ζ ( 5 ) + 2 1 6 0 1 π 6 + 4 1 ζ ( 3 ) 2 − 1 8 0 1 π 4 ζ ( 3 ) + 1 6 2 0 0 1 π 8
⟹ Ans. = 2 2 6 2 0
In general,
∫ 0 1 x n L i s ( x ) L i q ( x ) d x = r = 1 ∑ ∞ r s 1 ( ( n + r + 1 ) q ( − 1 ) q + 1 H n + r + 1 + ( − 1 ) q k = 2 ∑ q ( − 1 ) k ( n + r + 1 ) q − k + 1 ζ ( k ) )
which can be evaluated similarly by using partial fractions and the elementary Euler Sum.