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


𝘶-Substitution essentially reverses the chain rule for derivatives. In other words, it helps us integrate composite functions.
When finding antiderivatives, we are basically performing "reverse differentiation." Some cases are pretty straightforward. For example, we know the derivative of x2 is 2x, so 2xdx=x2+C. We can use this straightforward reasoning with other basic functions, like sin(x), ex, 1x, etc.
Other cases, however, are not that simple. For example, what is cos(3x+5)dx? Hint: it's not sin(3x+5)+C. Try differentiating that and you will see why.
One method that can be very useful is u-substitution, which basically reverses the chain rule.

Using u-substitution with indefinite integrals

Imagine we are asked to find 2xcos(x2)dx. Notice that 2x is the derivative of x2, which is the "inner" function in the composite function cos(x2). In other words, letting u(x)=x2 and w(x)=cos(x), we have:
This suggests that u-substitution is called for. Let's see how it's done.
First, we differentiate the equation u=x2 according to x, while treating u as an implicit function of x.
In that last row we multiplied the equation by dx so du is isolated. That's somewhat unorthodox, but useful for our next step. So we have u=x2 and du=2xdx. Now we can perform a substitution in the integral:
After the substitution we are left with an expression for the antiderivative of cos(u) in terms of u. How convenient! cos(u) is a basic function so we can find its antiderivative in a straightforward way. The only thing left to do is return the function to be in terms of x:
In conclusion, 2xcos(x2)dx is sin(x2)+C. You can differentiate sin(x2)+C to verify that this is true.
Key takeaway #1: u-substitution is really all about reversing the chain rule:
  • According to the chain rule, the derivative of w(u(x)) is w(u(x))u(x).
  • In u-substitution, we take an expression of the form w(u(x))u(x) and find its antiderivative w(u(x)).
Key takeaway #2: u-substitution helps us take a messy expression and simplify it by making the "inner" function the variable.
Problem 1.A
Problem set 1 will walk you through all the steps of finding the following integral using u-substitution.
How should we define u?
Choose 1 answer:

Common mistake: getting incorrect expressions for u or du

Choosing the wrong expression for u will result in a wrong answer. For example, in Problem set 1, u must be defined as 2x3+5. Letting u be 6x2 or (2x3+5)6 will never work.
Remember: For u-substitution to apply, we must be able to write the integrand as w(u(x))u(x). Then, u must be defined as the inner function of the composite factor.
Another crucial step in this process is finding du. Make sure you are differentiating u correctly, because a wrong expression for du will also result in a wrong answer.
Problem 2
Tim was asked to find cos(5x7)dx. This is his work:
Is Tim's work correct? If not, what is his mistake?
Choose 1 answer:

Common mistake: not realizing u-substitution is called for

Remember: When integrating a composite function, we can't simply take the antiderivative of the outer function. We need to use u-substitution.
Letting W be an antiderivative of w, this point can be expressed mathematically as follows:

Another common mistake: confusing the inner function and its derivative

Imagine you're trying to find x2cos(2x)dx. You might say "since 2x is the derivative of x2, we can use u-substitution." Actually, since u-substitution requires taking the derivative of the inner function, x2 must be the derivative of 2x for u-substitution to work. Since that's not the case, u-substitution doesn't apply here.

Sometimes we need to multiply/divide the integral by a constant.

Imagine we are asked to find sin(3x+5)dx. Notice that while we have a composite function sin(3x+5), it is not multiplied by anything. That might seem weird at first, but let's proceed and see what happens.
We let u=3x+5, then du=3dx. Now we substitute u into the integral, not before we perform this clever manipulation:
See what we did there? In order to have 3dx in the integrand, we multiplied the entire integral by 13. That way we allowed for u-substitution while keeping the value of the integral the same.
Let's continue with the substitution:
Key takeaway: Sometimes we need to multiply or divide the entire integral by a constant, so we can achieve the appropriate form for u-substitution without changing the value of the integral.
Problem 3
Choose 1 answer:

Want more practice? Try this exercise.

Want to join the conversation?

  • male robot johnny style avatar for user W1LDK1D_K19
    who's confused cos I'm definitely baffled o_0
    (50 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user Thomas Evans
    Repeatedly the idea of multiplying the dx out is being referred to as unorthodox. Is there a more formal or mainstream way of doing or thinking about this? It makes intuitive sense to me, but I'm wondering if there's something I should know about the "proper" method.
    (44 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user minsunk
    Would you please help me to solve ∫6x^2e^2xdx?
    (1 vote)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user Jennifer Hall
    Is there a video somewhere that goes over just u, du, dx and how they are chosen or derived?
    (5 votes)
    Default Khan Academy avatar avatar for user
    • female robot grace style avatar for user loumast17
      u is just the variable that was chosen to represent what you replace.

      du and dx are just parts of a derivative, where of course u is substituted part fo the function. u will always be some function of x, so you take the derivative of u with respect to x, or in other words du/dx.

      There should be videos on this playlist such as "u-substitution intro" and a few others I would watch, but you would also want to watch chain rule videos as well, though here is my explanation. am assuming familiarity with the chain rule.

      The chain rule starts with a composite function f(g(x)). Such as sin(x^2), where one function is the sine operation and the other is the squared operation. For the sake of clarity g(f(x)) would be sin^2(x).

      Anyway, the chain rule says if you take the derivative with respect to x of f(g(x)) you get f'(g(x))*g'(x). That means if you have a function in THAT form, you can take the integral of it to look like f(g(x)). The process of doing this is traditionally u substitution.

      So you start with f'(g(x))*g'(x). the first step is to make u=g(x) that way, when you take the derivative of u with respect to x (in other words du/dx) this gets you g'(x) So now you know what g(x) is in f(g(x)). since you start with f'(g(x))*g'(x) you ust have to take the integral of f'(g(x)) to get f(g(x)), though it's easier if g(x) is just a single variable, so we substitute in u for g(x). of course at the end you need to re-substitute g(x) for it.

      For notation's sake, and to ensure everything works out logically du/dx = g'(x) is changed to du = g'(x) dx because when you take the integral of something you add dx (as long as it is with respect to x.) so the integral of f'(g(x))*g'(x) dx gets g'(x) dx replaced with du because f'(g(x)) becomes f'(u). so now you have the integral of f'(u) du which of course becomes f(u), then you replace u with g(x) to get f(g(x)) effectively undoing the chain rule.

      Let me know if this did not help. And of course, I should mention with indefinite integrals you always put a + C at the end of your answer.
      (11 votes)
  • male robot hal style avatar for user James
    if trigonometry enters this field, am I gone?
    (5 votes)
    Default Khan Academy avatar avatar for user
    • male robot donald style avatar for user Venkata
      Pretty vague question. Trigonometry is extensively integrated into calculus, so you'll eventually encounter it. That said, if you know your trig identities, it's really easy. Trust me, going forward, the easiest thing about integrals is doing the integral itself, be it trig, logs or anything else. Once you learn about double and triple integrals (yeah they exist), you'll see that the integration is the easiest part of it.
      (5 votes)
  • blobby green style avatar for user robert zitelli
    Can I use u-substitution to find anti-derivative of (x^2 + 1)^2? u = (x^2 + 1) and du/dx = 2x
    (3 votes)
    Default Khan Academy avatar avatar for user
  • leafers ultimate style avatar for user Samuel Rodriguez
    If u substitution does not apply in ∫x^2 cos(2x)dx which approach should we use?
    (4 votes)
    Default Khan Academy avatar avatar for user
  • leaf green style avatar for user Yu Aoi
    can we multiply/divide by a variable?
    (3 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user Isabella Mora
    Hi, would you please help me solve (2x^3)/(1+x^2) using you-substitution? I don't quite understand why the answer I am looking at online calls for a subtraction of two integrals. Thanks!
    (3 votes)
    Default Khan Academy avatar avatar for user
  • leaf yellow style avatar for user Lightning Driver, Esq.
    So I have this question that has me baffled. It is: find the integral of 2sec^2⋅(x)tan(x)*

    Now to solve it, I need u substitution. Well, when I saw the question, I found that I have two options for u, and both will work. Here's what I did:

    *u = tan(x)*
    du = sec^2(x) dx
    dx = du/sec^2(x)

    Putting this in the question, I get:
    int(2sec^2(x)⋅u du/sec^2(x)) = int(2u du)
    = 2int(u du) = 2×1/2 u^2 = u^2 = tan^2(x) + C

    So here we got the integral is *tan^2(x) + C

    Now the second option

    u = sec(x)*
    du = sec(x)tan(x) dx
    dx = du/sec(x)tan(x)

    Putting this in the question, *but only substituting for one sec(x)*, I get:
    int(2u⋅sec(x)tan(x) du/sec(x)tan(x)) = int(2u du) = u^2 (as done above) = sec^2(x) + C

    Here, I get *sec^2(x) + C

    How's that possible??
    (2 votes)
    Default Khan Academy avatar avatar for user
    • male robot donald style avatar for user Venkata
      Both are correct answers. The fun part about integrals is that you can have several antiderivatives.

      If you want, differentiate both of them. For tan^(2)(x), you get 2tan(x)sec^(2)(x) = 2sec^(2)(x)tan(x). For sec^(2)(x), you get 2sec(x)sec(x)tan(x) = 2sec^(2)(x)tan(x). Same answer.
      (3 votes)