I am studying fluid mechanics and am trying to gain a finer intuition for the meaning behind the material derivative. So, firstly this is defined as $ (\partial_t + \mathbf{u}\cdot\nabla )f$ where $f$ is some multivariable function and $\mathbf{u}$ is the velocity field. I know the material derivative gives the rate of change of $f$ in a particular fluid element as it flows with the fluid. The first term gives the contribution from the temporal rate of change of f at a specific point, and the second term gives the contribution from the movement of the fluid element itself, and is the directional derivative of f in the direction of the velocity. However, it is stated in my notes that the second term, $\mathbf{u}\cdot\nabla(f)$ is the rate of change of f along some streamline multiplied by the magnitude of $\mathbf{u}$, ie. $\mathbf{u}\cdot\nabla(f)= df/ds \vert{\mathbf{u}}\vert$ where s is the distance along the streamline. I dont understand this. We are following the fluid element along a particle path, why is there some random streamline important? Can someone explain?

  • $\begingroup$ It is not "some" streamline but the one that goes through the point where you are evaluating the derivative. Since particles follow streamlines (for time-independent $\mathbf{u}$), $\mathbf{u} \cdot \nabla f$ is the change of $f$ the way a particle carried by $\mathbf{u}$ experiences it, if $f$ is time-independent also. I think it is good to think of the material derivative if you a looking at a property of the fluid (e.g. entropy, temperature, ..) and of the other derivative when you have an external field, like gravity $\endgroup$ – user66081 May 5 '17 at 17:35
  • $\begingroup$ Thanks for the reply, you helped my understanding. It would be really helpful if you have an algebraic way of deriving the fact we are talking about: $ \mathbf{u}\cdot\nabla(f)= df/ds \vert{\mathbf{u}}\vert$. I believe if we split u into its magnitude and directional unit vector, we can then obtain the magnitude of u multiplied by the dot product between this unit vector and the gradient of f, which somehow can be shown as equal to this thing. $\endgroup$ – Max Sykes May 6 '17 at 13:45
  • $\begingroup$ I believe this just requires a simple understanding of directional derivative which I unfortunately lack- any explanation appreciated $\endgroup$ – Max Sykes May 6 '17 at 13:48

This is to answer the question $u \cdot \nabla f = \frac{df}{ds} |u|$, where $u$ is the velocity field. Everything is time-independent (stationary) here.

Let's hypothesize that $\frac{df}{d\tau}$ means the derivative along the stream line. A streamline is an integral line of (i.e. the solution of the ODE, traced out) $X'(\tau) = u(X(\tau))$, and let's say $X(0) = x_0$ is where we evaluate the formula in question. Here, $\tau$ is a fictitious time that parameterizes the streamline (it coincides with the real time if $u$ is stationary). Then $\frac{d f}{d \tau}(x_0)$ is the derivative of the function $h(\tau) := f(X(\tau))$ at $\tau = 0$. By the chain rule and the definitions, $$ h'(\tau) = \nabla f(X(\tau)) \cdot X'(\tau) = \nabla f(X(s)) \cdot u(X(s)) $$ so that $$ \frac{df}{d\tau}(x_0) = h'(0) = \nabla f(x_0) \cdot u(x_0). $$

Ok, that's not what we wanted; the variable $s$ in the formula is apparently not what I called the fictitious time. Let's take $s = \tau |u|$ which has the unit of length. Now we parameterize the streamline by length rather than time. This looks uncomfortable, because $u$ depends on spatial variable; indeed, the streamline is an integral line of the unit-speed $Y'(s) = u(Y(s)) / |u(Y(s))|$ with $Y(0) = x_0$. Let us set $g(s) := f(Y(s))$, and compute the derivative $g'(0)$. We have $$ g'(s) = \nabla f(Y(s)) \cdot Y'(s) = \nabla f(Y(s)) \cdot u(Y(s)) / |u(Y(s))| $$ so that $$ \frac{df}{ds}(x_0) = g'(0) = \nabla f(x_0) \cdot u(x_0) / |u(x_0)| . $$ Multiply by $|u(x_0)|$ to obtain the original formula.


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.