A tough integral from the Berkeley Math Tournament

แชร์
ฝัง
  • เผยแพร่เมื่อ 19 ธ.ค. 2024

ความคิดเห็น • 64

  • @petrie911
    @petrie911 2 หลายเดือนก่อน +17

    Pulling the sum outside the integral, the terms of the sum are half the inner products of sin(x)/x and sin(nx)/x. By plancherel's theorem, this is equal to the inner product of their Fourier transforms. Those transforms are box functions of height sqrt(pi/2) and width 1 and n, respectively. So their product is a box function of height pi/2 and width 1, easily integrated to give pi. The terms of the sum are now pi/(2 n!), which clearly sums to (pi/2)(e-1).

    • @maths_505
      @maths_505  2 หลายเดือนก่อน +3

      Beautiful

  • @joshuaiosevich3727
    @joshuaiosevich3727 2 หลายเดือนก่อน +31

    I think drunkenness helped me solve this. I proved that the integral sin(nx)sin(x)/x^2 is constant, it’s an instant solution.

  • @CM63_France
    @CM63_France 2 หลายเดือนก่อน +1

    Hi,
    6:02 : dx
    "ok, cool" : 2:10 , 4:09 , 6:57 , 7:42 ,
    "terribly sorry about that" : 2:55 , 5:54 , 7:34 , 9:56 , 12:48 .

  • @Woowoo1373
    @Woowoo1373 2 หลายเดือนก่อน +9

    I love Blackpenredpen! He is such a lovely guy and you too! Straight up you two!

  • @darthTwin6
    @darthTwin6 หลายเดือนก่อน

    Amazing generalized solution! You make it look so easy too

  • @venkatamarutiramtarigoppul2078
    @venkatamarutiramtarigoppul2078 2 หลายเดือนก่อน +2

    1/2 cos (k-1)x-1/2 cos k+1x /x2 now use feynmens trick ( both the integrals are convergent ) so it would be integral from (k-1,0) to (k+1,infinty) sinxy/y dy dx

  • @adityagoel5746
    @adityagoel5746 2 หลายเดือนก่อน +7

    Let a denote the positive real root of the polynomial x ^ 2 - 3x - 2 . Compute the remainder when [a ^ 1000] is divided by the prime number 997. [.] denotes the greatest integer function.
    Pls solve this too.

    • @sbares
      @sbares 2 หลายเดือนก่อน +1

      Fun little problem.
      Letting b be the other root, one easily sees that -1 < b < 0, so 0 < b^1000 < 1. Therefore [a^1000] = a^1000 + b^1000 - 1 (EDIT: since a^1000 + b^1000, being a symmetric polynomial in the roots, is an integer, which must then be the next integer up from a^1000). It turns out that the polynomial x^2 - 3x - 2 does not have a root in F997 as can be seen by applying quadratic reciprocity to its discriminant 17. However, we can extend F997 to a larger field F997[a] = F997[X]/(X^2 - 3X - 2) in which it does have two roots (abusively also denoted by a and b here). Since there exists a homomorphism from Z[a] to F997[a] mapping the integers to themselves (modulo 997), a to a and b to b, we may do our calculations in this field.
      Now for the fun part. The Frobenius map x |-> x^997 generates the Galois group Gal(F997[a]/F997) (this is a general fact about finite fields) and hence exchanges the two roots. In F997[a] we thus have
      a^1000 + b^1000 - 1 = a^997 a^3 + b^997 b^3 - 1 = b a^3 + a b^3 - 1 = ab(a^2 + b^2) - 1 = ab((a+b)^2 - 2ab) - 1
      From the coefficients of the polynomial, we can read off ab= -2, a+b = 3, so
      ab((a+b)^2 - 2ab) - 1 = (-2)(3^2 - 2*(-2)) - 1 = -27 = 970.
      Back in Z, we therefore have
      [a^1000] == 970 mod 997.

    • @adityagoel5746
      @adityagoel5746 2 หลายเดือนก่อน

      @@sbaresbro what kind of god are you?
      And do you use whatsapp and can you send me a photo of the solution using pen and paper?

  • @pavlopanasiuk7297
    @pavlopanasiuk7297 2 หลายเดือนก่อน

    I've noticed that the sum under the integral is computable and converges to e^(cosx)*sin(sinx). While it seemed nightmerish,
    int from 0 to \infty sinx * e^(cosx) * sin(sinx) / x^2 dx is an integral of an even, complex-analytic function (one can check x=0 is also regular point). So I integrated f(z) = sinz * e^(cosz) * sin(sinz) / z^2 over a standard countour (-R -> -epsilon, small epsilon arc around z=0, epsilon -> R, large R arc). Arc integrals evaluate to zeros by direct limiting, they behave as O(epsilon), O(1/R). The remainding two integrals are the same and equal to the original one, so the countour integral is twice the original one. Yet the integrand is analytic, so they both equal to zero. That is wrong, of course.
    I fail to see any issues, can anyone assist?

    • @maths_505
      @maths_505  2 หลายเดือนก่อน

      The integral over the small circle should not be zero

    • @pavlopanasiuk7297
      @pavlopanasiuk7297 2 หลายเดือนก่อน

      @@maths_505 I would think the same, but the function is analytic in the neighborhood . Analyzing that circular integral manually I get O(epsilon), do you get some different expr?

    • @maths_505
      @maths_505  2 หลายเดือนก่อน

      @@pavlopanasiuk7297 I have a video on a couple of similar integrals in the contour integration playlist. Perhaps they may help.

  • @Greg852I
    @Greg852I 2 หลายเดือนก่อน +4

    Cool Solution!
    I found a path to the same solution representing the sin(x)sin(kx)/x product as a int(-1,1) sin(kx+ax). By use of a nice substitution t=(k+a)x in the (0,inf) integral we seperate the sum and the two integrals into a Dirichlet integral, a sum giving basically e-1 and an integral over the number one from -1 to 1. Giving the same as your Answer!
    I was so happy that this worked

  • @canoflame6127
    @canoflame6127 2 หลายเดือนก่อน +5

    Hey! Thats! My rule! I rule this kingdom! You do what i say! My rule is your rule... Thats the rule of me... The rules of montather muhamed... Thats canoflames rule.... I own the rule now...

  • @Sugarman96
    @Sugarman96 2 หลายเดือนก่อน +1

    6:09 Papa Fourier comes in handy with those. You can extend the integral over from -inf to inf, then simply replace the cosines with e^j(k±1)x (the imaginary part of the integral is 0). At which point electrical engineers will recognize the denominator as looking like the Fourier transform of e^-a|t|, which makes the integrals the inverse fourier transforms (constants not withstanding), at which point you just plug in (k±1) into the original function.

    • @Dharun-ge2fo
      @Dharun-ge2fo 2 หลายเดือนก่อน

      Just curious, are you a student in college or school preparing for jee

    • @Sugarman96
      @Sugarman96 2 หลายเดือนก่อน +1

      @@Dharun-ge2fo Just a few credits short of getting my BA in EE

  • @ericknutson8310
    @ericknutson8310 2 หลายเดือนก่อน +4

    I was thinking you were gonna use a Lebachebsky ID for pi periodic functions to reduce from 0 to infinity to 0 to pi/2 and also remove a sinx/x.
    I haven’t checked to see if that sum is pi periodic.

  • @krisbrandenberger544
    @krisbrandenberger544 2 หลายเดือนก่อน +1

    @ 2:50 There shouldn't be a k in the exponent.

  • @leeshaocheng239
    @leeshaocheng239 2 หลายเดือนก่อน +1

    glad to involve in your video❤

  • @mcalkis5771
    @mcalkis5771 2 หลายเดือนก่อน +3

    TH-cam forgot the notification but doesn't matter cause I check on you anyways.

  • @alexkaralekas4060
    @alexkaralekas4060 2 หลายเดือนก่อน +1

    Now that was beautiful

    • @maths_505
      @maths_505  2 หลายเดือนก่อน

      Indeed

  • @DiffgaFysgie
    @DiffgaFysgie 2 หลายเดือนก่อน +3

    Umm... I'm sorry to bother you, but I wish to let you know that purple text on black background is somewhat hard to read. Could you possibly use some other colours to increase contrast and readability of your notes? Thank you!

  • @ericthegreat7805
    @ericthegreat7805 2 หลายเดือนก่อน +1

    Is this connected in any way to Fourier series?

    • @petrie911
      @petrie911 2 หลายเดือนก่อน +3

      Yeah, you can use Plancherel's theorem to switch to the Fourier transforms of sin(x)/x and sin(nx)/x, which results in a trivial integral.

  • @192chickenking
    @192chickenking 2 หลายเดือนก่อน +2

    done with sinxsinkx=1/2[cosx(k-1)-cosx(k+1)] and contour integral of complex analysis

  • @Noam_.Menashe
    @Noam_.Menashe 2 หลายเดือนก่อน

    This reminds me that sometime ago I found the integral from 0 to pi of f(e^ix)/(x^2+1-2xcos(a))dx or something like that. It's not that hard using the fourier series but it was nearly two years ago and I forgot the details. I probably have the notes kept somewhere...

    • @maths_505
      @maths_505  2 หลายเดือนก่อน +1

      Would love to hear more about it when you remember the details.

    • @Noam_.Menashe
      @Noam_.Menashe 2 หลายเดือนก่อน

      @@maths_505 I found it. It's in "A treatise of the integral calculus" part II, page 278.
      It goes like this:
      The integral from 0 to pi of:
      [f(a+e^ix)+f(a+e^-ix)]/(1-2*b*cos(x)+b^2)dx
      equals:
      2*pi/(1-c^2)*f(a+c) for |c|1

    • @Noam_.Menashe
      @Noam_.Menashe 2 หลายเดือนก่อน

      Misremembered badly but it still is quite general. Especially the factor of a+
      Derivation really is using Fourier series for both functions and orthogonality of cosines.

  • @lokithe.godofmischief
    @lokithe.godofmischief 2 หลายเดือนก่อน +3

    Ques i found, which seemed impossible
    Let f(x)=(x⁴)+(x³)+(x²)
    Let y depict the inverse of the function f(x)
    Then, evaluate the integral from (lower bound)2A to (upper bound)8A of (1/(y²+y⁴)) with the limit A tends to ♾️infinity
    The variable of integration is obviously dx, as y will be a function of x only

    • @vancedforU
      @vancedforU 2 หลายเดือนก่อน

      I believe that the answer is ln4
      Wait for me to post my process

    • @vancedforU
      @vancedforU 2 หลายเดือนก่อน +1

      You can look at my post
      I think I spent more time writing the details in a formal way than actually solving the problem haha

    • @lokithe.godofmischief
      @lokithe.godofmischief 2 หลายเดือนก่อน

      @@vancedforU alright found it thank you so much🥰🥰

  • @sundaresanabishek5127
    @sundaresanabishek5127 2 หลายเดือนก่อน +1

    Heyy brother fan from Sri Lanka ❤🎉

    • @maths_505
      @maths_505  2 หลายเดือนก่อน +1

      Lovely hearing from you my friend

  • @MrWael1970
    @MrWael1970 2 หลายเดือนก่อน

    Very Nice. Thanks

  • @RandomLogan156
    @RandomLogan156 2 หลายเดือนก่อน +2

    12:01
    My right ear just got tickled through my headphones.

    • @RandomLogan156
      @RandomLogan156 2 หลายเดือนก่อน

      Dude, a hearted comment. I’m honored! I love your videos man!

    • @maths_505
      @maths_505  2 หลายเดือนก่อน +1

      ​@@RandomLogan156thanks bro

  • @vancedforU
    @vancedforU 2 หลายเดือนก่อน +1

    Nice
    I solved it by pulling out the summation operator, then integrating by parts (integrate 1/x^2, differentiate sinx sin(nx))
    Then you can simplify it down to dirichlet integral

  • @Mario_Altare
    @Mario_Altare 2 หลายเดือนก่อน

    Hi 🙂Could we use Lobachevsky's integral to solve this one? I.e.
    sin(nx) = Im e^(inx)/n!
    So:
    I = Im ∫_0^∞ sin⁡x/x^2 ∑_(n=1)^∞ (e^ix)^n/n! dx]
    → I=Im∫_0^∞〖sin⁡x/x^2 (e^cos⁡x e^(i sin x)-1)dx〗
    Now Loba comes in:
    I = Im ∫_0^(π/2)〖(e^cos⁡x e^(i sin x)-1)/sin⁡x dx〗
    At this point, I'd like to use Feynman here, but how? Or is it just a wrong way to try to solve it?

    • @maths_505
      @maths_505  2 หลายเดือนก่อน +2

      Lobachevsky works when f(x) is pi periodic. Here, e^(cos(x)) isn't so Lobachevsky doesn't apply.

    • @Mario_Altare
      @Mario_Altare 2 หลายเดือนก่อน

      @@maths_505 Oops, you're right 👌

  • @joshuaiosevich3727
    @joshuaiosevich3727 2 หลายเดือนก่อน +1

    Lmao this problem admits a ton of different solutions

  • @soreto314
    @soreto314 2 หลายเดือนก่อน

    which app do you use for notes?

  • @WhatToSayWTS
    @WhatToSayWTS 2 หลายเดือนก่อน

    How are you so good at math im only familiar with symbols can you suggest ways for me to understand a tiny proportion if not all what you explain

    • @scrumptious9673
      @scrumptious9673 2 หลายเดือนก่อน

      Maybe take a course? There’s a whole lotta maths right there

  • @AndyBaiduc-iloveu
    @AndyBaiduc-iloveu 2 หลายเดือนก่อน +4

    Do a problem involving quaternions with rotations or vectors. Or mabye integrate wrt quaternions!

    • @AndyBaiduc-iloveu
      @AndyBaiduc-iloveu 2 หลายเดือนก่อน

      By the way , I love your videos

  • @jannesfilgerdamm1419
    @jannesfilgerdamm1419 2 หลายเดือนก่อน

    Do you like the show "the lord if the rings"?

  • @scrumptious9673
    @scrumptious9673 2 หลายเดือนก่อน +1

    Terribly sorry

  • @Tosi31415
    @Tosi31415 2 หลายเดือนก่อน +2

    what the hell i did it in my head in like 5 seconds, using lobachevski's rule, and then with some rough guesses in my head

    • @maths_505
      @maths_505  2 หลายเดือนก่อน +3

      I guessed it to be 3 and I'm glad I was right on the money

    • @vancedforU
      @vancedforU 2 หลายเดือนก่อน

      Right I forgot about that

  • @الاستاذحيدر_كاظم_العمري
    @الاستاذحيدر_كاظم_العمري 2 หลายเดือนก่อน +1

    This is the rule of Muntadhir al-Hasani

  • @anuragkr3026
    @anuragkr3026 2 หลายเดือนก่อน

    I am scrolling through math channel and commnets to find people who would like to help in my math idea its related to 3d geometry
    Any math guy here who thinks he is able to discuss?

  • @خرائطالعالم-ك3ف
    @خرائطالعالم-ك3ف 2 หลายเดือนก่อน

    You are a thief

    • @ralex810
      @ralex810 2 หลายเดือนก่อน +3

      nuh uh