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

3d curl formula, part 2

This finishes the demonstration of how to compute three-dimensional curl using a certain determinant. Created by Grant Sanderson.

Want to join the conversation?

  • leafers seedling style avatar for user rja42
    Another question: Before I watched these videos I tried to do some 3D curl exercises, and I got the formula wrong: instead of crossing nabla with the vector valued function, I crossed nabla with the divergence of the vector valued function. Now I realize that this is not what 3d curl is, that is, after watching this video, but still: does crossing nabla with the divergence of a 3d vector valued function has any mathematical/physical meaning at all? Thank you for your time.
    (5 votes)
    Default Khan Academy avatar avatar for user
    • winston baby style avatar for user Andrew
      crossing nabla with the divergence of a 3d vector valued function is basically taking the curl of the divergence, which I don't think has any physical meaning on its own.
      An interesting point: the reverse--taking the divergence of the curl--is always equal to zero. ∇⋅ (∇ x F) = 0
      (5 votes)
  • leaf green style avatar for user Arthur Benson
    What are the units of measurement for curl? Is there a way to tell based on the units of measurement for the vector field? For example, what if the vector field measures velocity?
    (4 votes)
    Default Khan Academy avatar avatar for user
  • leaf green style avatar for user First Aid Sac
    why are we subtracting j?
    (3 votes)
    Default Khan Academy avatar avatar for user
    • male robot hal style avatar for user Alan Lorenzato
      We are subtracting j because of how the determinant of a matrix is computed. You could as well write +(dP/dz-dR/dx)j instead of -(dR/dx-dP/dz)j and it would have the same meaning, it's just that in the second case you grouped a -1 that in the first case is instead distributed inside the parentheses, thus changing the order of the two partial derivatives.

      To add to this, the method used in the video to compute the determinant is called the method of Laplace. If instead you used Sarrus' method, that only works for 3x3 matrices, you would find your j already tied up and with the -1 already distributed inside the parentheses. By the way you would get the same result with both methods, apart for the differently distributed -1.
      (3 votes)
  • leafers seedling style avatar for user Benjamin Langlois
    For the j column, why do dR/dx - dP/dz rather than dP/dz - dR/dx?
    (2 votes)
    Default Khan Academy avatar avatar for user
  • hopper cool style avatar for user Elliott Chen
    Is there a difference between a dot product and a cross product?

    If yes, then what is the difference?
    (1 vote)
    Default Khan Academy avatar avatar for user
    • primosaur seed style avatar for user Ian Pulizzotto
      Yes, there's a big difference. The dot product of two vectors is a scalar. The cross product of two vectors is a vector.

      Also, for two nonzero vectors, the dot product is zero if and only if the two vectors are perpendicular, but the cross product is the zero vector if and only if the two vectors are parallel or antiparallel.

      Have a blessed, wonderful day!
      (3 votes)
  • blobby green style avatar for user jp2338
    (2 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user s1155092086
    I think it's worth pointing out that the reason why i j k terms are similar to the 2d formula is that when we try to represent 3d rotation with a vector, we are actually trying to represent it using a linear combination of rotations in the positive direction of x, y and z. And as we accept the that 3d rotations are very similar to 3d vectors, that we can add rotations the same way we add vectors is also an interesting and potentially puzzling fact.
    (1 vote)
    Default Khan Academy avatar avatar for user
  • piceratops ultimate style avatar for user Evan Indge
    How do you extend cross product into higher dimensional Cartesian space? Like, if I wanted to calculate the curl of a function with an R4 output space, what exactly would I do to calculate that? And how exactly does the determinant relate to the theory behind the cross product? Thank you!
    (1 vote)
    Default Khan Academy avatar avatar for user
    • aqualine tree style avatar for user Tanishq Tejaswi
      Curl, as defined here, is only defined in 3-dimensional Euclidean Space. It can be generalized to lower dimensions like how Grant did it for 2-dimensional space in a previous video on 2D curl, but it can't be done for higher dimensions.

      The determinant is simply a way to easily remember how a cross product is computed.
      (1 vote)
  • blobby green style avatar for user ashanbuke007
    What does the magnitude of curl vector represents? And does it follows all the properties of cross product? Please Explain with some Physical example.
    (1 vote)
    Default Khan Academy avatar avatar for user
    • eggleston orange style avatar for user el_dAna
      The magnitude of the curl is the magnitude of angular velocity (in a physical instance ). Remember the situation of an airflow in a room with a ping pond "magically suspended" at a point in the flow? The ball would rotate about an axis either slowly of fast. Computing the curl for the motion of the airflow at that point where the ball is suspended gives the "speed" at which the ball would be rotating about its axis (Speed of the airflow at that point). The axis (direction) about which it rotates is given by the x, y ,z components of the curl.


      Reiterating his explanation on cross products.

      Now the cross product(c) between two vectors in 2d lets say a(on the x-axis) and b(angle theta to the positive x-axis) is given be |c| = |a||b|sin( theta).
      c is perpendicular to both a and b so on the 'z' axis.
      Intuitively |c| = the 2d area between a and b.
      Thus area = area of the parallelogram = base * height
      base = |a| and height = bsin(theta)....the y component of b.

      Therefor area = |a||b|sin(theta) = |c|.

      The components of the a and b were used directly and the outcome is the area (physically speaking).

      if a and b are differentiable functions and their derivatives used instead then the cross product is the rate of rotation. (physically it may be the angular velocity of the ping pong)

      Things get quite 'messy' in 3d but i hope this 2d intuition was useful to build up to 3d... would be glad if someone could offer some help.

      Maybe in 3d, the curl on the x-y plane is on the z plane(as explained above where the vectors only had a and y components) Since they have three components here then the curl for the x-z and y-z planes have to also be computed and then to find the resultant rotation, they are computed arithmetically.(since they are all vectors - cross products are vectors)

      Happy learning
      (1 vote)

Video transcript

- [Voiceover] So I'm explaining the formula for three-dimensional curl and where we left off, we have this determinant of a three-by-three matrix, which looks absurd because none of the individual components are actual numbers, but nevertheless, I'm about show how when you kind of go through the motions of taking a determinant, you get a vector-valued function that corresponds to the curl. So let me show you what I mean by that. If you're computing the determinant, of the guy that we have pictured there in the upper-right, you start by taking this upper-left component, and then multiplying it by the determinant of the sub-matrix, the sub-matrix whose rows are not the row of I and whose columns are not the row of I. So what that looks like over here, is we're gonna take that unit vector I, and then multiple it by a certain little determinant, and what this sub-determinant involves is multiplying this partial-partial Y by R, which means taking the partial derivative with respect to Y of the multi-variable function R and then subtracting off the partial derivative with respect to Z of Q. So we're subtracting off partial derivative with respect to Z of the multi-variable function Q, and then that, so that's the first thing that we do, and then as a second part, we take this J and we're gonna subtract off, so you're kind of thinking plus, minus, plus, for the elements in this top row, so we're gonna subtract off J, multiplied by another sub-determinant, and then this one is gonna involve, you know, this column that it's not part of and this column that it's not part of, and you imagine those guys as a two-by-two matrix, and its determinant involves taking the partial derivative with respect to X of R, so that's kind of the diagonal, partial-partial X of R, and then subtracting off the partial derivative with respect to Z of P, so partial-partial Z of P, and then that's just two out of three of the things we need to do for our overall determinant, because the last part we're gonna add, we're gonna add that top-right component, K, multiplied by the sub-matrix whose columns involve the column it's not part of and whose rows involve the rows that it's not part of, so K multiplied by the determinant of this guy is going to be, let's see, partial-partial X of Q, so that's partial-partial X of Q minus partial-partial Y of P, so partial derivative with respect to Y of the multi-variable function P, and that entire expression is the three-dimensional curl of the function whose components are P, Q, and R. So here we have our vector function, vector-valued function V whose components are P, Q, and R, and when you go through this whole process of imagining the cross-product between the Del operator, this Nabla symbol, and the vector output P, Q, and R, what you get is this whole expression, and, you know, here we're writing with I, J, K notation, if you're writing it as a column vector, I guess I didn't erase some of these guys, but if you're writing this as a column vector, it would look like saying the curl of your vector valued function V as a function of X, Y, and Z is equal to, and then what I'd put in for this first component would be what's up there, so that would be your partial with respect to Y of R minus partial of Q with respect to Z, so partial of Q with respect to Z, and I won't copy it down for all of the other ones but in principle, you know, you'd kind of, whatever this J component is, and I guess we're subtracting it so you'd subtract there, you'd copy that as the next component and then over here. But often times when you're computing curl, you kinda switch to using this IJK notation. My personal preference, I typically default to column vectors and other people will write in terms of I, J, and K, it doesn't really matter as long as you know how to go back and forth between the two. One really quick thing that I wanna highlight before doing an example of this is that the K-component here, the Z-component of the output, is exactly the two-dimensional curl formula. If you kind of look back to the videos on 2-D curl and what its formula is, that is what we have here. And in fact all the other components kind of look like mirrors of that but you're using slightly different operators and slightly different functions but if you think about rotation that happens purely in the XY plane, just two-dimensional rotation, and how in three dimensions that's described with a vector in the K direction and again, if that doesn't quite seem clear, maybe look back at the video on describing rotation with a three-dimensional vector and the right-hand rule, but vector is pointing in the pure Z direction, describe rotation in the XY plane, and what's happening with these other guys is kind of similar, right? Rotation that happens purely in the XZ plane is gonna correspond with a rotation vector in the Y direction, the direction perpendicular to the X, let's see, so the XZ plane over here. And then similarly this first component kind of tells you all the rotation happening in the YZ plane and the vectors in the I direction, the X direction of the output, kind of corresponds to rotation in that plane. Now when you compute it, you're not always thinking about oh, you know, this corresponds to rotation in that plane and this corresponds to rotation in that plane, you're just kind of computing it to get a formula out, but I think it's kind of nice to recognize that all the intuition that we put into the two-dimensional curl does show up here and another thing I wanna emphasize is this is not a formula to be memorized. I would not, if I were you, try to sit down and memorize this long expression. The only thing that you need to remember, the only thing, is that curl is represented as this Del cross V, this Nabla symbol cross product with a vector-valued function V, because from there, whatever your components are, you can kind of go through the process that I just did and the more you do it, the quicker it becomes, it's kind of long but it doesn't take that long, and it's certainly much more fault-tolerant than trying to remember something that has as many moving parts as the formula that you see here, and in the next video I'll go through an actual example of that. I'll have functions for P, Q, and R and walk through that process in a more concrete context. I'll see you then!