Main content

### Course: Multivariable calculus > Unit 2

Lesson 8: Differentiating vector-valued functions (articles)# Multivariable chain rule, simple version

The chain rule for derivatives can be extended to higher dimensions. Here we see what that looks like in the relatively simple case where the composition is a single-variable function.

## What we're building to

- Given a multivariable function
, and two single variable functions$f(x,y)$ and$x(t)$ , here's what the multivariable chain rule says:$y(t)$

- Written with vector notation, where
, this rule has a very elegant form in terms of the gradient of$\overrightarrow{\mathbf{\text{v}}}(t)=\left[\begin{array}{c}x(t)\\ y(t)\end{array}\right]$ and the vector-derivative of$f$ .$\overrightarrow{\mathbf{\text{v}}}(t)$

## A more general chain rule

As you can probably imagine, the multivariable chain rule generalizes the chain rule from single variable calculus. The single variable chain rule tells you how to take the derivative of the composition of two functions:

What if instead of taking in a one-dimensional input, $t$ , the function $f$ took in a two-dimensional input, $(x,y)$ ?

Well, in that case, it wouldn't make sense to compose it with a scalar-valued function $g(t)$ . Instead, let's say there are two separate scalar-valued functions $x(t)$ and $y(t)$ , and we plug these in as the coordinates of $f$ . The overall composition will be a single variable function, with a single-number input $t$ , and a single-number output $f(x(t),y(t))$ , as shown in this diagram:

There is still a chain rule that lets you compute the derivative of this new single-variable function $f(x(t),y(t))$ , and it involves the partial derivatives of $f$ :

Keep in mind, an expression like $\frac{\partial f}{\partial x}}{\displaystyle \frac{dx}{dt}$ is shorthand for

That is, both are functions of $t$ , but $\frac{\partial f}{\partial x}$ is evaluated via the intermediate functions $x(t)$ and $y(t)$ .

## Written with vector notation

Rather than thinking of $x(t)$ and $y(t)$ as being separate functions, it's common to package them together into a single, vector-valued function:

Then instead of writing the composition as $f(x(t),y(t))$ , you can write it as $f(\overrightarrow{\mathbf{\text{v}}}(t))$ .

With this notation, the multivariable chain rule can be written more compactly as a dot product between the gradient of $f$ and the vector-derivative of $\overrightarrow{\mathbf{\text{v}}}(t)$ :

Written like this, the analogy with the single-variable derivative is clearer.

The gradient $\mathrm{\nabla}f$ plays the role of the derivative of $f$ , and the vector derivative ${\overrightarrow{\mathbf{\text{v}}}}^{\prime}(t)$ plays the role as the ordinary derivative of $g$ .

## Intuition for why the chain rule works

As a warm up, consider the single variable chain rule for a composition like $f(g(t))$ . Here's how I like to understand that composition:

- First,
maps a point$g$ on the number line to another point$t$ the number line.$g(t)$ - Then
comes in and maps the point$f$ to yet another point on the number line,$g(t)$ $f(g(t))$

Understanding the derivative of $f(g(t))$ requires understanding how a tiny change in $t$ changes the final output.

So let's dive into what the chain rule is really saying.

- The term
represents how a tiny change in${{\displaystyle \frac{dg}{dt}}}$ influences the intermediate output,$t$ .$g(t)$

- The term
represents how a tiny change in${{\displaystyle \frac{df}{dg}}}$ influences the final output$g$ .$f(g(t))$

- The total change in
due to a small change in$f$ is then the product of both these influences.$t$

## Extend this intuition to more dimensions

The intuition is similar for the multivariable chain rule. You can think of $\overrightarrow{\mathbf{\text{v}}}$ as mapping a point on the number line to a point on the $xy$ -plane, and $f(\overrightarrow{\mathbf{\text{v}}}(t))$ as mapping that point back down to some place on the number line. The question is, how does a small change in the initial input $t$ change the total output $f(\overrightarrow{\mathbf{\text{v}}}(t))$ ?

Let's break down what the multivariable chain rule is saying, spelling it out in terms of the component functions $x(t)$ and $y(t)$ :

- The term
represents how a tiny change in${{\displaystyle \frac{dx}{dt}}}$ influences the intermediate output$t$ .$x(t)$

- Likewise the term
represents how a tiny change in${{\displaystyle \frac{dy}{dt}}}$ influences the second intermediate output$t$ .$y(t)$

- The term
represents how a tiny change${{\displaystyle \frac{\partial f}{\partial x}}}$ **to the**of an input to -component$x$ influences its output, and similarly the term$f$ accounts for how a small change${{\displaystyle \frac{\partial f}{\partial y}}}$ **to the**of the input changes -component$y$ .$f$

- One way a small change to
influences$t$ is that it first changes$f(x(t),y(t))$ , which in turn changes$x(t)$ . This effect is captured in the product$f$ .${{\displaystyle \frac{\partial f}{\partial x}}}{{\displaystyle \frac{dx}{dt}}}$

- The other way a change to
changes the output of$t$ is by first changing the second intermediate output$f(x(t),y(t))$ , which in turn affects the output of$y(t)$ . This effect is captured in the product$f$ .${{\displaystyle \frac{\partial f}{\partial y}}}{{\displaystyle \frac{dy}{dt}}}$ - Adding these two products gives the total change in
.$f$

## Connection with directional derivative

You might notice that the dot product expression for the multivariable chain rule looks a lot like a directional derivative:

In fact, that's exactly what it is! The derivative ${\overrightarrow{\mathbf{\text{v}}}}^{\prime}({t}_{0})$ at a particular value ${t}_{0}$ gives a vector in the input space of $f$ :

If $\overrightarrow{\mathbf{\text{v}}}(t)$ is interpreted as a parametric path inside this space, perhaps thought of as the trajectory of a particle, the derivative at a particular point in time ${t}_{0}$ gives the velocity vector of this particle at that time.

With this interpretation, the chain rule tells us that the derivative of the composition $f(\overrightarrow{\mathbf{\text{v}}}(t))$ is the

**directional derivative of**$f$ along the derivative of $\overrightarrow{\mathbf{\text{v}}}(t)$ .This should make sense, because a tiny change by "$dt$ " to $t$ should, by the meaning of the derivative, cause a tiny change $d\overrightarrow{\mathbf{\text{v}}}$ to the output of $\overrightarrow{\mathbf{\text{v}}}(t)$ . And the point of the directional derivative is that a tiny change $d\overrightarrow{\mathbf{\text{v}}}$ to the input of $f$ should cause a change $df$ as determined by $\frac{\partial f}{\partial \overrightarrow{\mathbf{\text{v}}}}}={\mathrm{\nabla}}_{\overrightarrow{\mathbf{\text{v}}}}f$ .

## Example 1: With and without the new chain rule

Define $f(x,y)$ like this:

And define $\overrightarrow{\mathbf{\text{v}}}(t)$ like this:

Find the derivative $\frac{d}{dt}}f(\overrightarrow{\mathbf{\text{v}}}(t))$ .

**Solution without chain rule**:

Before throwing our fancy new tool at the problem, it's worth pointing out that this is something we can solve by first writing out the composition as a single variable function of $t$ :

Now you can take the ordinary derivative:

But of course, the purpose of this example is to get a feel for what the chain rule feels like.

**Solution using chain rule**:

First, let's explicitly state the component functions of $\overrightarrow{\mathbf{\text{v}}}(t)$ :

According to the chain rule,

Taking the partial derivatives of $f(x,y)={x}^{2}y$ and the ordinary derivatives of $x(t)=\mathrm{cos}(t)$ , $y(t)=\mathrm{sin}(t)$ , we get

We want everything in terms of $t$ , so we plug in $x=\mathrm{cos}(t)$ and $y=\mathrm{sin}(t)$ .

Reassuringly, this is the same as the answer we got without using the chain rule. You might be thinking that this new chain rule makes things unnecessarily complicated, and the dirty little secret is that for concrete computations like this one, it is often not needed.

However, it is useful for writing equations in terms of an unknown function, as the next example shows.

## Example 2: Unknown function

Suppose the temperature across a two-dimensional region varies according to a function $T(x,y)$ , which we do not know. You wander throughout this region, sampling temperatures as you go, and your $x$ and $y$ coordinates as functions of time are

In taking your measurements, you notice that the temperature never changes along your path. What can you say about the partial derivatives of $T$ ?

## Summary

- Given a multivariable function
, and two single variable functions$f(x,y)$ and$x(t)$ , here's what the multivariable chain rule says:$y(t)$

- Written with vector notation, where
, this rule has a very elegant form in terms of the gradient of$\overrightarrow{\mathbf{\text{v}}}(t)=\left[\begin{array}{c}x(t)\\ y(t)\end{array}\right]$ and the vector-derivative of$f$ .$\overrightarrow{\mathbf{\text{v}}}(t)$

## Want to join the conversation?

- 3rd paragraph (A more general chain rule), 1st equation states: derivative of f with respect to x ... I think it should be with respect to t: d/dt f(g(t)).(5 votes)
- yeh this part got me like "wat" but i guess its just a typo(4 votes)

- If I have something like P(s(t),t), then would the derivative of P with respect to t be just the partial derivative of P with respect to s multiplied by the derivative of s with respect to t, or is it a bit more complicated than that?(3 votes)
- Hi Rayden, I had a similar question.

Suppose that instead of P(s(t),t) you had P(s(t),r(t)), being r(t)=t. Now, try to apply the chain rule and you will get a sum of two terms. The term concerning s(t) is straightforward. The term concerning r(t) results in (dP(t)/dr).(dr/dt) = (dP(t)/dr).(dt/dt) = (dP(t)/dr).1 = dP(t)/dr = dP(t)/dr(t) = dP(t)/dt (remember that r(t)=t ). So, I guess the final result would be: Dp(s(t),t) = dP(t)/dt + (dP/ds).(ds/dt).(4 votes)

- Why this is not possible?

total change in f(x,y)= lim ( f(x(t+h),y(t+h)) - f(x(t),y(t)) ) / h

h->0

Also,what this expression tells ?

Thank you.(3 votes)- I'm not sure, but I guess what you've wrote is just another valid way of looking at it. If you consider x(t), y(t) as parametric path on a plane, you're calculating the variation of f on each point x,y depending on t. It seems to lead to kind of directional derivative of f along the path. Someone correct me if I am wrong, please.(2 votes)

- Could someone please tell me how to do the challenge question in the section "Intuition for why the chain rule works" (in the additional section at the end of the last bullet point)? Thank you.(2 votes)
- lim ℎ→0 (𝑓(𝑔(𝑡₀ + ℎ)) − 𝑓(𝑔(𝑡₀)))∕ℎ

= lim ℎ→0 (𝑓(𝑔(𝑡₀ + ℎ) − 𝑔(𝑡₀) + 𝑔(𝑡₀)) − 𝑓(𝑔(𝑡₀)))∕ℎ

= lim ℎ→0 (𝑓(ℎ⋅(𝑔(𝑡₀ + ℎ) − 𝑔(𝑡₀))∕ℎ + 𝑔(𝑡₀)) − 𝑓(𝑔(𝑡₀)))∕ℎ

= lim ℎ→0 (𝑓(ℎ⋅𝑔′(𝑡₀) + 𝑔(𝑡₀)) − 𝑓(𝑔(𝑡₀)))∕ℎ

= lim ℎ→0 (𝑓(𝑔(𝑡₀) + ℎ⋅𝑔′(𝑡₀)) − 𝑓(𝑔(𝑡₀)))∕(ℎ⋅𝑔′(𝑡₀))⋅𝑔′(𝑡₀)

Let 𝑘 = ℎ⋅𝑔′(𝑡₀)

Notice that ℎ → 0 ⇒ 𝑘 → 0

(assuming 𝑔′(𝑡₀) exists)

Thereby, we can write the limit as

lim 𝑘→0 (𝑓(𝑔(𝑡₀) + 𝑘) − 𝑓(𝑔(𝑡₀)))∕𝑘⋅𝑔′(𝑡₀)

= 𝑓′(𝑔(𝑡₀))⋅𝑔′(𝑡₀)(3 votes)

- This is a bit confusing. Say f(x, y)=xy, x(t)=t and y(t)=t. If we want to find partial diff. df/dx, can't we substitute y=x in the expression for f and say that it is equal to 2x(=2t) instead of y(=t)?(1 vote)
- You could do that, but regardless, you would still have to find dx/dt (after writing out the chain rule). There are plenty of examples of chain rule where you could substitute functions like x(t) or y(t) into another function like f(x,y), yes it would make life easier and avoids chain rule altogether, however that doesn't teach you chain rule or the importance of it. Learning chain rule is important because you aren't always given explicit function definitions. Sometimes you are only given one multivariable function and are asked to find the partial of one variable with respect to another (for example, the partial of pressure with respect to volume in the van der Waals equation) and would need to know how to write out and do a multivariable chain rule.(4 votes)

- In the last equation under “Connection with the directional derivative” shouldn’t it be ∇_v’_f?(2 votes)
- True! It's the derivative of f evaluated at v(t) and taking the direction of v'(t).(2 votes)

- What is the answer to the challenge question?(2 votes)
- Consider f(g(t). g takes single value to n dimensional vector and f takes n dimensional vector to single value. So f(g(t)) is a single variable function from t.

To compute the derivative of f(g(t)), we use the multivariable chain rule.

d (f(g(t))) / dt = g'(t) * ∇f(g(t))

This is very similar to the single variable chain rule

d (f(g(t))) / dt = g'(t) * f'(g(t)

df(g(t))/dt captures how f(g(t)) changes with t in both cases.

g'(t), in both cases, captures how g changes with respect to t.

In single variable case, f'(g(t)) captures how f changes with respect to g, thereby making the whole expression capture how f changes with respect to t.

So how does ∇f capture how f changes with respect to g in multivariable case?(1 vote)