I came across this equation $$F=-\nabla V$$ where $V$ is potential energy. I do understand that $$F(r)=-\frac{dV}{dr}.$$ Hence does this mean the nabla operator in this case means derivative? Because I checked up the definition of $\nabla$ and none of them suits this definition.

  • 16
    $\begingroup$ If you checked the definition of $\nabla$, but none of them was the derivative, you checked the wrong definition. $\endgroup$ – ACuriousMind Jul 5 '15 at 23:19
  • 1
    $\begingroup$ When the nabla $\nabla$ is written right next to its argument (without dot, cross, etc), it is usually known as the "grad" operator. Gradient... $\endgroup$ – Floris Jul 6 '15 at 0:02

On spherical coordinates, the gradient of a general function $V$ is: $$ \nabla V = \frac{\partial V}{\partial r}\mathbf e_r + \frac{1}{r}\frac{\partial V}{\partial\theta}\mathbf e_\theta + \frac{1}{r\sin\theta}\frac{\partial V}{\partial\phi}\mathbf e_\phi $$

If $V(r, \theta, \phi)$ only depends on $r$, that is $V = V(r)$, which is exactly the case of the gravitational potential, then the partial derivatives with respect to $\theta$ and $\phi$ are zero, and therefore the $\nabla$ resumes to: $$ \nabla V = \frac{\partial V}{\partial r}\mathbf e_r = \frac{dV}{dr}\mathbf e_r $$


Given your question, it seems likely that your misunderstanding comes from a limited sense of vectors, fields, and partial derivatives. So there's a lot of education that we have to cover in a very short time.

Multivariate functions

When we transition from a function $f(x)$ to a field, which is a function of many variables $f(x, y, z)$, we suddenly have many ways to take the derivative. Instead of our iconic equation valid for small differences in x $\delta x$,$$f(x + \delta x) \approx f(x) + f'(x)\cdot\delta x$$we now have to write$$f(x + \delta x, y + \delta y, z + \delta z) \approx f(x, y, z) + f_x(x, y, z)\cdot\delta x + f_y(x, y, z)\cdot\delta y + f_z(x, y, z)\cdot\delta z.$$The principle is the same, but each variation that we're making must be made independently. Instead of us saying "when you zoom in far enough, the graph of $f(x)$ looks like a straight line," we're saying, "when you zoom in far enough, $f(x, y, z)$ looks like the expression $a x + b y + c z + d$." In this case the full name of $a = f_x(x,y,z)$ is "the partial derivative of $f$ with respect to its first variable, holding its second and third variables constant, evaluated at the point $(x, y, z)$."

All of that technicality is needed because you could hold other things constant, too. For example, you could ask how $f$ varies with $x$ when $y - x$ is held constant. Clearly this would correspond to the above case where $\delta y = \delta x$ while $\delta z = 0$. To succinctly express this kind of stuff, it's useful sometimes to write $(\partial f/\partial u)_{P,Q}$ as "the partial derivative of $f$ with respect to $u$ holding the expressions $P$ and $Q$ constant." So for example our observation about $y - x$ being held constant would be phrased as$$\left(\frac{\partial f}{\partial x}\right)_{y-x,z} = \left(\frac{\partial f}{\partial x}\right)_{y,z} + \left(\frac{\partial f}{\partial y}\right)_{x,z}.$$But there are some very common operators which we might leave out the extra detail. You will variously see all of these conventions: $$\partial_x f = \frac{\partial f}{\partial x} = f_x = \left(\frac{\partial f}{\partial x}\right)_{y,z}.$$Similarly, when we unqualifiedly talk about $\partial_r$ we're talking about keeping either, in cylindrical coordinates, $\theta$ and $z$ fixed, or else, in spherical coordinates, $\theta$ and $\phi$ fixed.

Work out some examples for yourself, like determining that the partial derivatives for $f(x, y, z) = x^3 + y^2 z^2 + x y z - x^2 z$ are $f_x = 3x^2 + y z - 2 x z$, $f_y = 2 y z^2 + x z$, $f_z = 2 y^2 z + x y - x^2$.


If you're doing physics with derivatives you've probably already seen vectors. The technical definition of vector in physics is, "a vector anything which transforms like a vector does under coordinate transformations." Lest that sound circular, it comes with an example: any arrow between two points in space. With that example, this definition extends that definition to anything which essentially "does the same thing" as a displacement vector does in space. One particular vector, the position vector, is usually denoted $\vec r = \left[x; y; z\right]$, and it points outward from some "origin" (0, 0, 0) to some point $(x, y, z)$.

One nice thing about vectors is that we have worked out which expressions are "coordinate independent" with them. Supposing that $\vec a$ is given by the coordinates $[a_x; a_y; a_z]$ and $\vec b$ is, in those same coordinates, $[b_x; b_y; b_z]$. Assume that the coordinate directions are all orthogonal. Then the number $\vec a \cdot \vec b = a_x b_x + a_y b_y + a_z b_z$ is known to be coordinate independent. It is 0 if and only if two vectors are perpendicular, having an alternate expression as $|\vec a| |\vec b| \cos \psi$ where $\psi$ is the angle between the two vectors in the plane that they both occupy. Furthermore, the "length" or "magnitude" of a vector, $|\vec a| = +\sqrt{\vec a \cdot \vec a}$ is also coordinate-independent as a special case. An almost-coordinate-independent "pseudovector" in 3D is given by $\vec a \times \vec b = \left[a_y b_z - a_z b_y;~a_z b_x - a_x b_z;~ a_x b_y - a_y b_x\right]$. It is "pseudo" because it changes sign when the coordinates change their "orientation" as in $(x,y,z)\rightarrow(x, z, y)$ or $(x, y, z)\rightarrow(-x, y, z)$.

When we see these properties we can immediately rewrite the above expression as$$f(\vec r + \delta\vec r) \approx f(\vec r) + \left[\begin{array}{c} f_x(\vec r) \\ f_y(\vec r) \\ f_z(\vec r)\end{array}\right] \cdot \delta r$$When we see this we see that we can write the del operator $\nabla = [\partial_x; \partial_y; \partial_z]$. Taking a scalar field $f(\vec r)$, the operation which converts it to the vector field $\nabla f = [f_x, f_y, f_z]$ is called the gradient operation. At each point this gives you a vector, so it is a vector field.

By its nature, the gradient points, at any point it's evaluated, in the local direction of greatest increase of its scalar function, and perpendicular to a hyperplane $\nabla f \cdot \vec r = C$ wherein that function is approximately constant for small deviations from the point where the gradient was taken.

Conservative fields

Combining our "almost vector" $\nabla$ and our "almost vector product" $\times$, it turns out in 3D that if a vector field $\vec v(\vec r)$ obeys $\nabla \times v = 0$ then it can be written as $\vec v = \nabla \phi$ for some function $\phi(\vec r)$. Usually for forces we instead choose $\vec F = -\nabla U$, with a negative sign. We do this because it fits our intuition that your potential energy U gets larger as you travel higher, fighting "against" the force of gravity. (Basically it makes a potential energy function for a hill look like the profile of that hill.) There is a similar law which is a little less useful, saying that if $\vec v$ obeys $\nabla\cdot v = 0$ then $\vec v = \nabla\times u$ for some $u$.

It is in general a little harder to find expressions for $\nabla$ in the different coordinates; this article is pretty useful to see what it looks like in non-Cartesian coordinates. In the special case for a spherically symmetric potential $U(r, \theta, \phi) = U(r)$ we do indeed discover that $\vec F(\vec r) = -\nabla U = - U'(r) ~\hat r$, where $\hat r = \vec r / |\vec r|$ is the unit vector in the positive $r$-direction. This is really helpful for a lot of different force laws.

  • $\begingroup$ Gotta laugh and cry at this: "a vector anything which transforms like a vector does under coordinate transformations." I prefer the definition as a quantity with a direction and magnitude that can be calculated from a set of numbers called its components. And the components transforming under rotations such that its direction and magnitude remains invariant. $\endgroup$ – Larry Harson Jul 6 '15 at 3:00
  • 1
    $\begingroup$ That line is shamelessly stolen from Prof. Bruce Kusse at Cornell. He said, "A tensor, of course, is anything that transforms like a tensor," sparking some laughs from the audience. He gave them a look and said, "don't laugh, this is science," getting more laughs, then explained that it's not really a joke, here's how we apply the coordinate-change matrix to these tensors and those tensors... and anything which transforms under coordinate changes like those entities is, of course, a tensor. $\endgroup$ – CR Drost Jul 6 '15 at 3:41
  • $\begingroup$ I thought that quote came from Zee $\endgroup$ – Larry Harson Jul 6 '15 at 21:57
  • $\begingroup$ Kusse might have stolen it too; I don't know. Or, intriguingly, maybe Kusse is the scientist mentioned by Zee. $\endgroup$ – CR Drost Jul 6 '15 at 22:35
  • $\begingroup$ Nobody stole it, I've heard it used many times... and it's not wrong either, but it does require some elaboration, of course. $\endgroup$ – user20250 Aug 23 '15 at 4:02

The work-energy theorem leads us to the following result; \begin{equation} \oint \vec F\cdot d\vec s=0 \end{equation}

\begin{equation} \oint \vec F\cdot d\vec s=\underbrace{\int \int }_{\text{surface}}(\nabla \times \vec F)\cdot d\vec n \end{equation} Using the rules of vector calculus there must exist some scalar function such that;

\begin{equation} \vec F=-\nabla U(r) \end{equation} Since, \begin{equation} \nabla \times \nabla U(r)=0 \end{equation} The negative sign is just for convention it ensures that a given force accelerates the particle to a lower potential energy, so that the work is done against the force. Note that any $\nabla \times \nabla \phi$ is equal to zero!


Let $F$ be a force field. Assuming that the force field is a conservative vector field, then it follows that the line integral of the force field is zero $$\oint_{O} F \cdot dr = 0$$ The del operator $\nabla$ is defined in 3 dimensions as $$\nabla \equiv \frac{\partial}{\partial x}\hat{e_x}+\frac{\partial}{\partial y}\hat{e_y}+ \frac{\partial}{\partial z}\hat{e_z}$$ Furthermore from Stokes theorem we have that $$\oint_{O} F \cdot dr = \iint_{B}\nabla \times F \cdot dB$$ Therefore for a conservative vector field the curl of the force field is also zero $$\nabla \times F = 0$$ Let $V$ be a function called the potential such that $$V = -\oint_{O} F\cdot dr$$ Then it follows that $F = -\nabla{V}$, which in the 1 dimensional case reduces to $$F = -\frac{\partial{V}}{\partial{x}}\hat{i} = -\frac{dV}{dx}$$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.