4.0 A better way to understand Differential Equations | Nonlinear Dynamics | Index Theory

Поділитися
Вставка
  • Опубліковано 5 січ 2025

КОМЕНТАРІ • 122

  • @sergpodolnii3962
    @sergpodolnii3962 2 роки тому +108

    These ideas give vibes of complex residue theorem

    • @nicholasandrzejkiewicz
      @nicholasandrzejkiewicz 2 роки тому +42

      That's because complex structures, not just in the complex plane but on general complex geometries, are combinations of the "symplectic structure" which is used here to divide coordinates into position and momentum phase spaces and metric structures, ie dot products. It just turns out you do not need the dot product to take the index (symplectic spaces are all even dimensional like complex spaces, and complex maps can be seen as vector fields just as they are here - hamiltonian vector fields). This is realized more abstractly in "kahler geometry," but you can explore the index ideas here further in the work of the classical mechanist Vladimir Arnold, he has good books on differential equations and mechanics.

    • @mastershooter64
      @mastershooter64 2 роки тому +6

      @@nicholasandrzejkiewicz woah

    • @h-a-y-k4149
      @h-a-y-k4149 2 роки тому +4

      I thought "I forgot much of complex analysis, I guess I won't notice that". And then I noticed it lol

    • @strangerrshade3347
      @strangerrshade3347 2 роки тому +1

      Ah yes yes nerd Group Theorums .. i feel at🏡

    • @manicmadpanickedman2249
      @manicmadpanickedman2249 2 роки тому

      I already explained this on my chan with less jargon and more prop

  • @guessam_aliu5964
    @guessam_aliu5964 Рік тому +4

    I find it very very amizing series that talk on some concepts
    associated with clair explanations.
    thank you so much.

  • @tomosmallpond9306
    @tomosmallpond9306 2 роки тому +22

    As a PhD working with nonlinear dynamics, this video was very entertaining to watch!

    • @namanjain5700
      @namanjain5700 2 роки тому

      Sir I want to see difference between
      1)linear - non linear slope fields
      2)1st order- 2nd order slope fields
      3)1st degree - 2nd degree slope fields
      4)homogeneous - Non-Homogeneous slope fields.
      I can't find them anywhere...
      Can you please help??

    • @whatitmeans
      @whatitmeans 2 роки тому

      Maybe you could find this interesting: recently I have figure out that no linear differential equation solution could achieve by itself the value zero and stays there forever (will violate the conditions for uniqueness), so I have been trying to find a physical model that could have a finite extinction time were the dynamics dies.
      So far I have made this modified version of the nonlinear damped pendulum with a sublinear friction term for some positive constants a and b:
      x'' + a*sgn(x')*sqrt(|x'|)*(sqrt(2)/4+|x'|^(3/2))+b*sin(x)=0, x(0)=pi/2, x'(0)=0
      But I have not being able to prove that they indeed have a finite ending time.
      Do you know if its plausible to prove it through modern nonlinear analysis tools?

    • @siguc
      @siguc 2 роки тому +1

      @@whatitmeans Re: "I have been trying to find a physical model that could have a finite extinction time were the dynamics dies": I'm not sure I follow. How about the motion of a brick on a horizontal surface with constant friction between the brick and the surface? Assuming the brick moves along the surface at t=0, it'll stop eventually. Newton's law: x"=-k*g*sign(x'), where g is 9.8m/s^2 and k is the friction coef.

    • @whatitmeans
      @whatitmeans 2 роки тому

      @@siguc Thanks, I have look for those examples a lot without finding any: the nearest thing was the Norton's Dome example (which I don't know if its a reallistic example), and the Coulomb Force... Does that kind of system have an specific name?(that maybe I am missing), Do you have any source for them?

    • @whatitmeans
      @whatitmeans 2 роки тому

      @@siguc I have searched for your example on google but I didn't find much info so I hope you could share some sources to it...
      Previously I have found that:
      x'=-sgn(x)*sqrt(|x|), x(0)=1
      can stand the solution:
      x(t)=1/4*(1-t/2+|1-t/2|)^2
      which has t=2 as ending time.
      For your example, asuming that all constants equals one, and with initial conditions y(0)=2 and y'(0)=-2 (so, the brick starts instantly reducing the speed after a kick with |y'|=2), in a similar way (through trial and error in Wolfram-Alpha), I think
      y(t)=1/2*(1-t/2+|1-t/2|)^2
      is a solution for your example
      y''=-sgn(y'), y(0)=2, y'(0)=-2
      Achieving and ending time at t=2.
      Hope you could share any source to equations with their solutions, so far the only thing I found is: en.m.wikipedia.org/wiki/Contact_dynamics#Non-smooth_approach
      Thank you very much.

  • @oliverhees4076
    @oliverhees4076 2 роки тому +22

    As an undergrad taking a nonlinear dynamics class this semester, these videos are super relevant and clearly explained. Great job!

    • @jacobguerreso675
      @jacobguerreso675 Рік тому

      thank you for choosing that pfp. always a joy to see Yoshi commenting on an ODE video

  • @ddx240p
    @ddx240p Рік тому

    Extremely underrated channel!!!
    Kep doing ao and channel will grow to millions :)❤

  • @doBobro
    @doBobro 2 роки тому +7

    Thank you for marvelous series. Visual explanations are on spot!

  • @MrLimpingTwerp
    @MrLimpingTwerp 2 роки тому +3

    A grade stuff mate, some serious talent here! Subscribed instantly!

  • @fredericoamigo
    @fredericoamigo 2 роки тому +6

    What a brilliant video! An incredible explanation with really really good visualizations. Keep up the good work!

  • @mnada72
    @mnada72 2 роки тому +2

    Waiting eagerly for the next video. Cheers

  • @DarkePacific
    @DarkePacific 2 роки тому

    I honestly was super lost halfway through but the conclusion summed it up really well that it instantly all became clear again.

  • @edzion6433
    @edzion6433 2 роки тому +29

    This appears to me to be quite related to complex analysis and poles... would be interesting to know how.

    • @pedrofonseca3432
      @pedrofonseca3432 2 роки тому +3

      It is, it's generalization of the specific result you have in complex analysis, search for winding number in Wikipedia.

  • @AMANPATEL-tp6hh
    @AMANPATEL-tp6hh Рік тому

    just wow..amazing..keep uploading such stuff..it's very helpful

  • @robertschlesinger1342
    @robertschlesinger1342 2 роки тому +3

    Very interesting, informative and worthwhile video.

  • @魏寅生
    @魏寅生 2 роки тому +1

    Wish I had seen this video when I was learning Field Theory!

  • @RuanD
    @RuanD 2 роки тому

    Amazing series! Can't wait for the next video

  • @vikrantvijit1436
    @vikrantvijit1436 2 роки тому

    Just right perfect elegant articulation of complex concepts in field of non linear dynamics applied mathematics Modelled on Quantum information Processing universe visualised both waveforms and particles simultaneously unfolding hidden connections, patterns and flows States Chained Phases Transitions Surfacing Planes Traced Source Energy Packets Collapsed TRUTHS functions Vectors Scalars Spacetime Ordered Resonant Frequency Following Responses Induced real world controlled Eco-Systems.

  • @MaxxTosh
    @MaxxTosh 2 роки тому

    See this shit? THIS is why I love math. You tackle one topic, then months later you learn something else and you see these huge connections. This was screaming residue theorem from the moment you had the 1/2pi integral!

  • @MrGeometres
    @MrGeometres 2 роки тому

    I would have a few distinct questions:
    - How does this scale to higher dimensional spaces?
    - Is this computationally feasible?
    - What if there is some uncertainty in the parameters of the differential equation?
    - In particular, when varying the parameters of the DE, the closed curves should appear / disappear. How robust is this method with respect to the uncertainty of the parameters?
    Ideally, one would like some computationally feasible, quantifiable estimate of "how far" the DE is away from one with/without a closed orbit. Is that feasible at all?

  • @anywallsocket
    @anywallsocket 2 роки тому +2

    Isn't this just Stokes' theorem, saying the curl is the integral of the divergence of the path from the vector field? I.e., orbits require integer curls, and stable orbits require positive curls. If so it seems like this can be mapped to just a statement about having zero net curvature for the path in the field, which would be pretty obvious for physical trajectories -- if an orbit experiences a non-zero force integration within one period, it will not follow the same path again, i.e., is an unstable orbit.

  • @kosmokritikos9299
    @kosmokritikos9299 2 роки тому

    If only I could have had you as a math teacher, _I could've been somebody._

  • @atlasxatlas
    @atlasxatlas 2 роки тому +5

    Awesome video!
    Great explanations and visualizations

  • @pseudolullus
    @pseudolullus 2 роки тому +2

    Great one, the whole thing reminded me of topological invariants and curvatures of fields

    • @leostein128
      @leostein128 2 роки тому +4

      It very much is a topological invariant! As we're only interested in vectors' directions, we are replacing nonzero vectors with unit vectors (mapping to the "unit tangent bundle"). Since we're in the plane, the space of unit vectors is S^1 (a topological circle). The space of loops in phase space is also S^1, so we're thinking about maps from S^1 -> S^1. The *index* of a fixpoint is the *degree* of this map from S^1 -> S^1, and the degree is a topological invariant.

  • @Khitiara_
    @Khitiara_ 2 роки тому

    vibes of gauss law in electromagnetism, where the surface integral of the field is proportional to the sum of charges (which as a bonus are integer because of quantum mechanics)

  • @ivanperezavellaneda4031
    @ivanperezavellaneda4031 2 роки тому

    I'm looking forward to the next video

  • @vtrandal
    @vtrandal 2 роки тому

    Is there a problem with your development of "Fixed Point" for I at about 7:50? I ask this because the expression inside the integral is 0/0 (zero divided by zero) . No! You have made your point with a bit of drama that the index is undefined for fixed points. Yes? Brilliantly done. Thank you!

    • @virtually_passed
      @virtually_passed  2 роки тому

      Hi, thanks for the question. Fixed points (x*,y*) are defined as the points where dx/dt = dy/dt = 0. If we plug a fixed point into the expression for the Index (the integral eqn) then we will get an expression divided by 0 which is undefined. This means the index is undefined at fixed points. If you want to learn more about fixed points I define them from scratch in this video:
      ua-cam.com/video/6mLCFyEv3Z0/v-deo.html

  • @zakzqk8485
    @zakzqk8485 2 роки тому +2

    Thank you so much
    That was awesome

  • @collegemathematics6698
    @collegemathematics6698 2 роки тому +3

    Interesting! 🤔

  • @nickc5374
    @nickc5374 2 роки тому +1

    Very well done video

  • @paulschrum4727
    @paulschrum4727 2 роки тому

    I suspect the sign of the index is connected to the sign of the Gaussian curvature of the surface generated by integrating the vectors as if they were slopes on a surface. (Think of the vectors as water flow on the surface of a mountain ridge to grok this.)

  • @joeeeee8738
    @joeeeee8738 2 роки тому +3

    Great! Can't wait for the next episode? When is it coming?

    • @virtually_passed
      @virtually_passed  2 роки тому +1

      Hi, I just managed to squeeze this video out before work got too overwhelming haha. I don't think I'll be able to produce another video until early next year sadly. The next video will likely be about limit cycles. Although, I might take a different path and talk about Floquet theory. Unsure at this point :)

  • @dmitry.shpakov
    @dmitry.shpakov 2 роки тому +1

    Wow! Awesome! Thanks

  • @ahvavee
    @ahvavee Рік тому

    As someone with no PhD, no masters, or bachelors, i say this is good. If you don’t believe me then just ask me.

  • @rubenvorster7186
    @rubenvorster7186 Рік тому

    simple yet cool

  • @SuperMaDBrothers
    @SuperMaDBrothers 2 роки тому

    the eerie connection between this and complex residue means there must be a deeper connection

  • @ryanpmcguire
    @ryanpmcguire 2 роки тому

    I sort of wish 3b1b’s videos looked this good

  • @MegaJani
    @MegaJani 2 роки тому +1

    It's just a hunch, but is this related to the residue theorem?

  • @whatitmeans
    @whatitmeans 2 роки тому +1

    Recently I found that every differential equation that have as solution a non-piecewise Power Series can have only never-ending solutions in time, and even worst, if a solution has a finite extinction time, then the differential equation cannot fulfill uniqueness of solutions! (it does only before the ending time).
    So all things we were teached on physics cannot accurately represent a phenomena with dynamics that by itself becomes zero and stays there forever (it violates the Identity Theorem for Power Series).
    So, those sinks point aren't really sinks, the curves will never reach those points in a finite time.
    But this doesn't mean that they cannot be modeled, but instead that maybe we have be using the mistaken tools: as a simple example, think in the following equation
    x'=-sgn(x)*sqrt(|x|), x(0)=1
    which can stand the solution
    x(t) = 1/4*(1-t/2+|1-t/2|)^2
    which becomes zero at t=2 and stays there forever.
    But it also stands solutions that rises from zero at t=2, which could represent the same kind of issues as the Norton's Dome example, but fortunately between initial conditions and final time t=2 the uniqueness is granted.
    It could be interesting to see some analysis of the nonlinear dynamics of these kind of finite time systems, as example, think in this modified version of the nonlinear damped pendulum with a sublinear friction term for some positive constants a and b:
    x'' + a*sgn(x')*sqrt(|x'|)*(sqrt(2)/4+|x'|^(3/2))+b*sin(x)=0, x(0)=pi/2, x'(0)=0
    How are its dynamics?
    It really exists a finite extinction time T_f where it becomes zero forever after?
    It exist a time T_n where the solution stops oscillating? (since it is going to become zero it should exist)
    Hope you could use this example to your video series and analyse its behavior with current tools for nonlinear dynamics.

    • @namanjain5700
      @namanjain5700 2 роки тому

      Sir I want to see difference between
      1)linear - non linear slope fields
      2)1st order- 2nd order slope fields
      3)1st degree - 2nd degree slope fields
      4)homogeneous - Non-Homogeneous slope fields.
      I can't find them anywhere...
      Can you please help??

  • @federicomagnani1954
    @federicomagnani1954 2 роки тому

    Is this a sort of dynamical interpretation of the residue theorem?

  • @neyhmor
    @neyhmor 2 роки тому

    Beautiful!

  • @AbElsaied
    @AbElsaied 2 роки тому +1

    Great video as always thank you.
    Which software or library do you use to generate and simulate these animations?

    • @virtually_passed
      @virtually_passed  2 роки тому +2

      Thanks :) I coded each of these simulations using Python. In particular, I used a library called "manim". Good luck!

  • @kodirovsshik
    @kodirovsshik 2 роки тому

    Gives very strong complex analysis vibes

  • @despicableme7081
    @despicableme7081 2 роки тому

    Make a video on method of characteristics, partial differential equation

  • @snnwstt
    @snnwstt 2 роки тому

    That seems ok in 2D, but ... imagine an 8 curve, but in 3D, where the middle "intersection" point is at different level (z). We have a close orbit, but isn't it that the index is ... eventually, two? (inspired from some "orbit" of an electron). Does the "index" works only with 2 variables?

  • @Filup
    @Filup 2 роки тому +2

    Last semester I took an ODE course where we were introduced to Lyapunov stability. Are there connections between this and Index Theory?

    • @virtually_passed
      @virtually_passed  2 роки тому +1

      As far as I know, there are no connections between Lyapunov stability and Index theory :)

  • @1495978707
    @1495978707 2 роки тому

    Good shit man

  • @cliffanderson720
    @cliffanderson720 2 роки тому

    Do the equations you listed for cell division ODEs have a name?

  • @judeugwu4987
    @judeugwu4987 2 роки тому

    At 13:31, what if you drew a closed orbit around the saddle in the middle? Would the index also be equal to +1?

    • @virtually_passed
      @virtually_passed  2 роки тому

      Hey Jude, good question. Any simple closed curve that surrounds only the saddle in the middle will have an index of -1. This can also be seen at 5:20.
      Because of this fact, it's impossible for a closed orbit (ie a periodic solution) to exist that surrounds only the saddle fixed point. This is because a closed orbit can only exist if it has an index of +1.

  • @accountname1047
    @accountname1047 2 роки тому

    Nice video

  • @Robinson8491
    @Robinson8491 2 роки тому +1

    Is this the basis for topology?

    • @virtually_passed
      @virtually_passed  2 роки тому +1

      I'm unsure if this is the basis of topology (I doubt it), but some of the ideas expressed in this video at 8:47 use formulas from topology. In particular we make use of a homotopy parameter to describe the "deformation" of the curve.

  • @蒋正-k6u
    @蒋正-k6u 2 роки тому

    this is very very very similar to the residue theorem!

  • @RoboticusMusic
    @RoboticusMusic 2 роки тому +1

    Does this relate to electrical engineering?

    • @virtually_passed
      @virtually_passed  2 роки тому +1

      What a great question! Sadly, I'm not an electrical engineer so I can't answer that question definitively. However, I wouldn't be surprised if nonlinear dynamics plays a large role in this field (at least for specific applications). For what it's worth, the Van der Pol oscillator was originally a model for a certain electrical circuit and can be viewed using an oscilloscope (I found this video here showing that ua-cam.com/video/ZeIxnYLHJuo/v-deo.html)

  • @erikanderson1402
    @erikanderson1402 2 роки тому

    What happens in 3 dimensions, N dimensions?

    • @virtually_passed
      @virtually_passed  2 роки тому

      that's a really great question! I've never heard of this theory being extended to 3D so I'm unsure, sorry!

  • @LaxmikantKachhap
    @LaxmikantKachhap 2 роки тому

    Does that mean if I slap you multiple times then this can predict at what time my hand will touch your face?
    That's freakin amazing.

  • @tomoki-v6o
    @tomoki-v6o 2 роки тому

    Some of these closed curves simular to hysteris loop

  • @albertrotaru8001
    @albertrotaru8001 2 роки тому +1

    Hello there! Can you give me some books on the topic?

    • @virtually_passed
      @virtually_passed  2 роки тому

      I use Nonlinear Dynamics and Chaos by Steven H strogatz

  • @user-sl6gn1ss8p
    @user-sl6gn1ss8p 2 роки тому

    Any (general) way to calculate the indexes other than by sampling?

    • @virtually_passed
      @virtually_passed  2 роки тому +1

      Hi, thanks for the question. What do you mean by sampling? The index can easily be calculated by adding the indices of the fixed points inside the curve.

    • @user-sl6gn1ss8p
      @user-sl6gn1ss8p 2 роки тому

      @@virtually_passed Sure, but say I don't know what the fixed points are, and I want to calculate the index, how's that done?
      By sampling I meant numerically sampling points along a curve to estimate the index.
      Thanks for the reply by the way : )

    • @virtually_passed
      @virtually_passed  2 роки тому

      @@user-sl6gn1ss8p As far as I know there are only 4 ways:
      1) Find the fixed points inside the curve (by solving for dx/dt = dy/dt = 0) and characterize them from their eigenvalues and then determine their index. (this is probably the easiest way)
      2) Calculate the index analytically using the integral expressed at 7:12
      3) Calculate the index numerically using the same integral at 7:12. This can be done using a riemann integral.
      4) 'Sampling' phi at various points and counting the rotations.
      I'd recommend option 1. Failing that I'd recommend option 3. Hope that helps. I'd strongly advise not doing 4 because near a fixed point you can get all types of numerical rounding that could mess up your measurement.

    • @user-sl6gn1ss8p
      @user-sl6gn1ss8p 2 роки тому

      @@virtually_passed makes sense. Thanks : )

  • @neyhmor
    @neyhmor 2 роки тому

    Do you mind if I recommend this Video in my class?

  • @ahmedgharieb5252
    @ahmedgharieb5252 2 роки тому

    When the next episode appears?

    • @virtually_passed
      @virtually_passed  2 роки тому

      I plan to make an entire series consisting of 6 more videos this year. The next one won't be released for a while though. Likely in 2-3 months.

  • @TheMichaelmorad
    @TheMichaelmorad 2 роки тому

    but what about a line of fixed points?

    • @virtually_passed
      @virtually_passed  2 роки тому

      That's a really interesting question! Let's consider
      dx/dt = y
      dy/dt = 0
      There is a line of fixed points along the x axis (because y = 0 satisfies dx/dt = dy/dt = 0)
      In this case, you can't draw a simple closed curve that goes through the x axis because the curve cannot pass through a fixed point (because phi is undefined at a fixed point). So the index will = 0 for any other curve you draw. As a result, you can say this system will never have any periodic orbits.

  • @AstroTibs
    @AstroTibs 2 роки тому

    14:08 More like: It's impossible for a closed orbit to exist here which encircles these poles.

  • @joeeeee8738
    @joeeeee8738 2 роки тому

    Another thing, everytime I see the thumbnail of the 2nd video of this series, I imagine if it's possible to calculate things like the time it takes to make 3 revolutions (like back and forth), or calculate how many revolutions until it stops or some other calculations. Is it possible? I imagine it being useful as an applied maths exercise

    • @virtually_passed
      @virtually_passed  2 роки тому +1

      Hi :) I assume you mean the thumbnail for this video, right?
      ua-cam.com/video/I9UEBRya4X0/v-deo.html&ab_channel=VirtuallyPassed
      This is for a spring mass damper system. This spring mass damper system is very nice because it's fully linear (m,c,k are all constant) so it's possible to derive an expression for its motion analytically with an equation!
      You can derive from first principles that the frequency of this system is:
      f = w_n*√(1-ζ^2)/(2*π)
      Where w_n = √(k/m)
      And 2*ζ*w_n = c/m
      This means that the time it takes to make 3 (incomplete) revolutions is:
      T = 3 / f
      ----------- in case you're interested -----------
      It's comparably quite easy to find the period of periodic orbits for linear systems. For nonlinear systems its waaay harder. In fact, just proving that periodic orbits exist is quite hard for nonlinear systems.

    • @joeeeee8738
      @joeeeee8738 2 роки тому

      @@virtually_passed yeah that's the one I meant! I knew about that dynamic system (not the actual formulas you mentioned) but wanted to know if there were some approaches in general to other dynamical systems. Now I know they are difficult lol. But hey! That could be a great idea for a video if you find examples for it!

    • @virtually_passed
      @virtually_passed  2 роки тому +1

      @@joeeeee8738 A crazy fact about nonlinear systems is that some of them generally have no analytical solution! That means it's impossible to get an equation to predict their motion using 'normal' operators like sqrt(), exp(), +, -, /, * . For this reason, a lot more emphasis is placed on finding out whether a periodic orbit exists or not, and how stable it is, rather than finding an equation that determines its exact period. That being said, there is a method called "Harmonic Balancing" which can roughly approximate the period of a periodic solution.

  • @virtually_passed
    @virtually_passed  Рік тому

    Next video is now available! ua-cam.com/video/-I5VPE1Hbwg/v-deo.html

  • @erikeriknorman
    @erikeriknorman 2 роки тому

    3-Body Systems have entered chat...

  • @halmirofigliolo1637
    @halmirofigliolo1637 2 роки тому

    I thought you had a million subscribers

  • @xs6819
    @xs6819 2 роки тому

    WTH DOES THIS ALWAYS WORK!?

  • @plekkchand
    @plekkchand 2 роки тому

    Don't call us " guys". Just say hello, for God's sake.

  • @robbiecee2
    @robbiecee2 2 роки тому

    Wot

  • @SuperMaDBrothers
    @SuperMaDBrothers 2 роки тому +1

    bro taking strogatz notes and putting them into video format is not education

    • @hridoyranjankalita4251
      @hridoyranjankalita4251 7 місяців тому

      Bro he is not a scientist that he will teach us his own theory. Every teacher do the same,they help us to understand things that's difficult to study by ourselves.

  • @namanjain5700
    @namanjain5700 2 роки тому

    Sir please make video to differentiate between linear and nonlinear direction fields.

    • @virtually_passed
      @virtually_passed  2 роки тому

      Hi, I've already made a video about linear 2nd order differential equations & Vector Fields here: ua-cam.com/video/I9UEBRya4X0/v-deo.html
      and an intro to nonlinear 2nd order differential equations & Vector Fields here: ua-cam.com/video/6mLCFyEv3Z0/v-deo.html
      Is that what you're looking for? :)