If you're seeing this message, it means we're having trouble loading external resources on our website.

If you're behind a web filter, please make sure that the domains *.kastatic.org and *.kasandbox.org are unblocked.

Main content

Addressing treating differentials algebraically

Addressing treating differentials algebraically.

Want to join the conversation?

  • spunky sam blue style avatar for user Toni Laine
    What is an example of a situation where treating dy,dx,dt,.. algebraically would cause a problem?
    (38 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user aecanales
    Is there anywhere I can read the "mathematically rigorous" explanation on why this works?
    (21 votes)
    Default Khan Academy avatar avatar for user
    • blobby green style avatar for user jandleslap23
      I recommend Keislers 'elementary infinitesimal calculus', it provides a lot of visual intuition and covers single variable, multivariate, and vector calculus as well as some differential equations. It is available for free I believe as well as his follow up 'foundations of infinitesimal calculus' which provides more rigour (which I am not personally interested in).
      (14 votes)
  • blobby green style avatar for user Yanko Mirov
    MIT OCW's course on differential equations is very long - ~40 lectures on the topic whereas KhanAcademy has nowhere near that amount of material. Is KhanAcademy's material enough or should we treat just as a introductory rather than complete course on the topic?
    (5 votes)
    Default Khan Academy avatar avatar for user
  • aqualine ultimate style avatar for user Don Samson
    Is there any other way to find particular solution with more rigorous method?
    (8 votes)
    Default Khan Academy avatar avatar for user
  • piceratops ultimate style avatar for user Benjamin Galbally
    Can you not say that:

    (1/y) dy = (1) dx

    and then integrate both sides?
    (4 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user milad yekani
    I am still not feeling good about this type of treating of dx notation. just symbolic manipulation isn't enough :((
    I want to have a powerful intuition. how can I get it??
    (6 votes)
    Default Khan Academy avatar avatar for user
  • purple pi purple style avatar for user Yevtushenko Oleksandr
    Can't we just do the u-substitution here ?
    ∫(1/y)*(dy/dx)dx, u = y => du/dx = dy/dx, then ∫(1/u)du ?
    I mean, it's the same logic, isn't it ? We can call this not u-substitution but y-substitution, same drill.

    Basically, if we need to find an antiderivative of equation in form y(x) * dy/dx, we just need to throw away the derivative, cause it's a consequence of the chain rule on y(x), and find an antiderivative of y(x) only.

    What's so difficult here that Sal says at "the rigour you need to show that this is ok in this situation is not an easy thing to show" ?
    (4 votes)
    Default Khan Academy avatar avatar for user
    • piceratops ultimate style avatar for user Nolan Ryzen Terrence
      I know it is very late but here is the thing.. dy/dx is the rate of change of some function y(say).Now cancelling/transposing the rate equation doesn't seem mathematically correct right?(I mean dx is infinitesimal change of something how can you just multiply with a variable it kinda looks weird). And regarding the y-substitution you are actually not changing anything. You are putting 'y' as u

      Nolan :)
      (1 vote)
  • duskpin ultimate style avatar for user Diego Oliveira
    It would be great if someone could explain what, exactly, is not mathematically rigorous with this treating of differentials (or, if it is mathematically rigorous, then what did Sal mean in the video), especially given that:
    (1) it apparently works flawlessly, and
    (2) in the 1960's there's been the non-standard analysis approach, which, if I understand correctly, sought to give infinitesimals such a solid foundation that this treating of derivatives as the ratio of infinitesimals could now be considered legitimate.
    (3 votes)
    Default Khan Academy avatar avatar for user
    • piceratops ultimate style avatar for user James Martin
      I am not an expert in nonstandard analysis, so take this answer with a grain of salt; parts could be inaccurate.

      Introductory calculus courses typically study the real numbers, whereas nonstandard analysis studies the hyperreal numbers. Although many of the concepts are the same, you're ultimately studying a different object with different properties.

      Many (all?) of the theorems we care about in Calculus BC carry over by the Transfer Principle, but in higher-level math, they might not. It's not rigorous in that we don't understand, or ask about, when the transfer principle applies; we just hope it'll work.

      Compare, for example, the study of integers vs. rationals. The arithmetical operations (addition, subtraction, multiplication) obey the same rules, but more complicated questions like "what polynomials have zeroes" do not.

      Likewise, if we're just studying introductory calculus, the operations obey the same rules and we don't care whether we're working with the reals or hyperreals, but at a higher level, they're not the same. Additionally, the conventional definitions are different (e.g. are we taking delta-epsilon limits, or dealing with infinitesimals?) which affects how the subject is taught.

      (Also compare how we write divergent limits as lim [x->a] f(x) = infinity whereas in the extended reals, infinity is an actual number, and the limit actually does equal infinity.)
      (1 vote)
  • blobby green style avatar for user Marcus Tornea
    This is how I find it mathematically valid though:
    Say that dy/dx = y, divide both sides by y: (1/y)(dy/dx)=1
    Then integrate both sides with respect to x:
    int[(1/y)(dy/dx)dx] = int(dx)
    And you have y in the integral and its derivative dy/dx. That's simple u substitution.
    (2 votes)
    Default Khan Academy avatar avatar for user
  • aqualine ultimate style avatar for user Nikki
    If you compare dy/dx and dx/dy of an equation, are they called "inverses?" Is there a mathematical term for this?
    (2 votes)
    Default Khan Academy avatar avatar for user

Video transcript

- [Instructor] So when you first learn calculus, you learn that the derivative of some function f, could be written as f prime of x is equal to the limit as, then there's multiple ways of doing this, the change in x approaches zero of f of x plus our change in x, minus f of x, over our change in x. And you learn multiple notations for this. For example, if you that y is equal f of x, you might write this as y prime. You might write this as d-y, d-x, which you'll often hear me say is the derivative of y, with respect to x, and that you can do the derivative of f with respect to x because y is equal to our function. But then later on when you, especially when you start getting into differential equations, you see people start to treat this notation as an actual algebraic expression. For example, you will learn or you might have already seen. If you're trying to solve the differential equation, the derivative of y with respect to x, is equal to y. So the rate of change of y with respect to x is equal to the value of y itself. This is one of, the most basic differential equations you might see. You'll see this technique, where people just say, "Well, let's just multiply both sides by d-x." Just treating d-x like as if it's some algebraic expression. So you multiply both sides by d-x and then you have, so that would cancel out algebraically, and so you see people treat it like that. So you have d-y is equal to y times d-x, and then they'll say, "Okay let's divide both sides by y." Which is a reasonable thing to do. Y is an algebraic expression. So if you divide both sides by y, you get one over y, d-y is equal to d-x. And then folks will integrate both sides to find a general solution to this differential equation. But my point on this video isn't to think about how do you solve a differential equation here, but to think about this notion of using, what we call differentials. So a d-x, or a d-y, and treating them algebraically like this. Treating them as algebraic expressions, where I can just multiply both sides by just d-x or d-y, or divide both sides by d-x or d-y. And I don't normally say this, but the rigor you need to show that this is okay in this situation, is not an easy thing to say. And so to just feel reasonably okay about doing this, this is a little bit hand wavy, it's not super mathematically rigorous. But it has proven to be a useful tool for us, to find these solutions. And conceptually the way that I think about a d-y, or a d-x, is this is the super small change in y, in response to a super small change in x. And that's essentially what this definition of the limit is telling us. Especially as Delta x approaches zero, we're going to have a super small change in x, as Delta x approaches zero. And then we're going to have a resulting super small change in y. So that's one way that you can feel a little bit better, of... And this is actually one of the justifications for this type of notation. As you can view this, what's the resulting super small, or what's the super small change in y, for a given super small change in x? Which is giving us the sense of what's the limiting value of the slope? As we go from the slope of the secant line to a tangent line? And if you view it that way you might feel a little bit better about using the differentials, or creating them algebraically. Let me just multiply both sides by that super small change in x. So the big picture is, this is a technique that you'll often see, in introductory differential equations classes. Introductory multi-variable classes and introductory calculus classes. But it's not very mathematically rigorous, to just treat differentials like algebraic expressions. But even thought it's not very mathematically rigorous to do it willy nilly like that. It has proven to be very useful. Now as you get more sophisticated in your mathematics there are rigorous definitions of a differential. Where you can get a better sense of where it is mathematically rigorous to use it and where it isn't. But the whole point here is, if you felt a little weird feeling about multiplying both sides by d-x or dividing both sides by d-x or d-y. Your feeling was mathematically justified, because it's not a very rigorous thing to do. At least until you have more rigor behind it, but I will tell you that if you're an introductory student it is a reasonable thing to do as you explore and manipulate some of these basic differential equations.