Comprehensive Oxford Mathematics and Physics Online School (COMPOS)

Year 13
Mathematics Assignment 02
Differentiation 2
David Vaccaro, Vladimir Chernov, Alexander Lvovsky

Due 06 November, 2022

This is the second Y13 Mathematics assignment from COMPOS. This assignment is designed to stretch
you and no student is expected to complete all questions on the first attempt. The problems are hard,
but do not let this discourage you. Give each problem a go, and skip to the next one if you are stuck.
The questions in each section are arranged in the order of increasing complexity, so you should try all
sections. Very similar problems will be discussed in tutorials and webinars, so think of the questions
you would like to ask. We hope that eventually you will be able to solve most of the problems. Good
luck!

1 Implicit Differentiation

In previous sheets we have differentiated functions of the explicit form y = f (x), where y is written as a
function of x. However, some functions are defined implicitly by an equation satisfied by y and x but where
y is not the subject. The equation of a circle x2 + y 2 = r2 gives a familiar example of an implicit function.

You can use the standard rules of differentiation to find the derivative of implicit functions. More examples
can be found in these Khan Academy videos: video 1, video 2, video 3.

Example 1. An implicit function is given x2 + y 2 = 4. Find the derivative y 0 of y with respect to x, in

terms of x and y.

Solution. Differentiating both sides with respect to x, we obtain

(x2 )0 + (y 2 )0 = 40

Note that the chain rule must be used for y 2 : (y 2 )0 = 2y × y 0 :

2x + 2yy 0 = 0.

Rearranging:
x
y0 = − (1)
y

If we recognise the previous curve as a circle, then the result makes sense as the tangent1 (with gradient y 0 )
1 Refer to Y12 Mathematics Assignment 5 if you need to refresh your memory about tangents and normals.

1
 y
is perpendicular to the radius (with gradient ).
x

y
gradient= • (x,y)
x
x

x
gradient=−
y

One may argue

√ that the above procedure can be avoided by simply solving the implicit function equation to
obtain y = 4 − x2 , and then finding the derivative using the chain rule:

(4 − x2 )0 −x
y0 = √ =√ ,
2 4−x 2 4 − x2
which is the same as Eq. (1). However, the method for implicit differentiation we just studied obviates
the need
√ for equation solving, which may be challenging or even impossible. Furthermore, the solution
y = 4 − x2 implies a positive y, whereas the result (1) applies for all y’s — positive or negative.

Problem 1 (4 marks). An implicit function is given by

a) xy − sin(x + y) = 0;
b) 4x2 y 2 − 7x + sin y = 4y − 3.

Find the derivative yx0 in terms of x and y.

The next two questions deal with conic sections. If you haven’t studied them at school, you can find
information on the ellipse and hyperbola on this Khan Academy page.

Problem 2 (3 marks). Find the equations of the tangents to the ellipse

2x2 + 3y 2 = 14

at the two points where x = 1. Find the coordinates of their point of intersection. Illustrate your result
graphically.

Problem 3 (3 marks). Find the equations of the tangents to the hyperbola

x2 y2
− =1
a2 b2
at the point (c, d). Show that none of these tangents pass through the origin. Choose a specific set of
parameters {a, b, c, d} and illustrate your result graphically.

2
2 Connected Rates of Change

Many practical situations involve several related variables. For example as the radius of a sphere changes,
so too do the surface area and the volume, or as the temperature of a gas changes, so too do the volume and
the pressure. In these situations the chain rule can be used to find the rates of change of each of the related
variables.

dV dr
Example 2. The volume of a sphere is increasing a a rate of = 20 cm3 s−1 . Find the rate at which
dt dt
the radius is increasing when the radius is r = 6 cm.

Solution. Differentiating V with respect to t using the chain rule yields

dV dV dr
= ×
dt dr dt

dV
V = 43 πr3 ⇒ = 4πr2 . Notice that this is the same expression as the surface area as the sphere — this
dr
is not a coincidence, try to explain why this is so. We have

dr 1 dV
= .
dt 4πr2 dt

Substituting the given values, we find

dr
≈ 0.0442 cm s−1 .
dt

More examples can be found in these Khan Academy videos: video 1, video 2, video 3.

Problem 4 (PAT 2006) (2 marks). The volume of a spherical balloon increases by 1 cm3 every second.
What is the rate of growth of the radius when the surface area of the balloon is 100 cm2 ?

Problem 5 (PAT 2016) (4 marks). A cylinder of dough is squashed such that its height h decreases
linearly with time t as h(t) = h0 − αt for t < h0 /α. Assume that the volume V of the dough remains
constant, and it retains a cylindrical shape. Find an expression for the rate of change of the radius of the
cylinder as a function of time and the parameters h0 , α, and V . Does the rate of change increase or decrease
with time?

3 Derivative of the inverse function

Example 3. Find the derivative of y = ln(x).

Solution. Let us rewrite y = ln(x) as x = ey . We can think of the latter expression as an equation of an
implicit function y(x). Differentiating both sides of this equation by x gives

dey dy
1= .
dy dx

3
 dey
Since = ey , we have
dy
dy 1
= y.
dx e
Since ey = x, we find
dy 1
ln0 (x) = = .
dx x

Problem 6 (4 marks).

Show that

1
a) arcsin0 (x) = √ ;
1 − x2
1
b) arccos0 (x) = − √ .
1 − x2

Let us generalize this argument. Suppose the function y = f (x) has a derivative on some interval a < x < b
and is invertible with g = f −1 (see Y12 Mathematics Assignment 8 for the definition of an inverse function).
Suppose we know the derivative g 0 (y) of the function g(y). Can we use it to find the derivative of f (x)?

Starting with the relationship x = g(y) and differentiating with respect to x gives that

dy
g 0 (y) =1
dx
or
dy 1
f 0 (x) = = 0 . (2)
dx g (y)
Please convince yourself that the results of Example 3 and Problem 6 are consistent with the general expres-
sion above.

To visualize the result (2), note that the graph of the inverse function y = g(x) is the reflection of the graph
of y = f (x) in the line y = x. Under this reflection to the tangent to y = f (x) at (a, b) is reflected onto the
tangent to y = g(x) at (b, a) and the gradients of these two tangents are reciprocal of one another.

4
 y

y=x

(a, b)

y = f (x)
(b, a)

y = g(x) = f −1 (x)

Problem 7 (6 marks). Find the derivatives of the functions y = f (x) that are inverse to the following
functions x = g(y):
ey
a) x = sinh y; b) x = tanh y; c) x = for y > 0.
1 + e2y

In the examples so far we were able to express our answers in terms of x. To that ende, we had to solve the
equation x = g(y) to find the explicit form of the function y = f (x), and then substitute this into Eq. (2).
But sometimes this equation is difficult or even impossible to solve. In this case, like for implicit functions
discussed in the previous section, we may need to keep the dependence on y in our answer for f 0 (x).

Example 4.

a) Show that the function g : R → R defined as x = g(y) = 2y + sin y is invertible, and find the domain
of the inverse.
b) If f = g −1 , find the equation to the tangent to the curve y = f (x) at the point where x = 2π.

c) Is it possible to find a general expression for f 0 (x)?

Solution a) We have f 0 (y) = 2 + cos y. As −1 ≤ cos y ≤ 1 always, we see that g 0 (y) > 0. This implies that
g(y) is strictly increasing, and hence one-to-one and invertible on the entire set R.

We can note that as y → ∞ that f (x) → ∞ and as y → −∞ that f (y) → −∞ so the range of g is R. This
implies that the domain of f = g −1 is also R.

5
b) We note that g(π) = 2π + sin π = 2π. Using the relationship for the derivative of the inverse function f ,
we have
1 1
f 0 (2π) = 0 = = 1.
g (π) 2 + cos π

Since g(π) = 2π, we have that f (2π) = π and hence the tangent passes through (2π, π). The gradient of the
tangent is 1, and we thus have that the tangent is y = x − π.

c) In general we have
1 1
f 0 (x) = =
g 0 (y) 2 + cos y

where y is the unique number satisfying the equation 2y + sin y = x. However, there is no easy way to get y
in terms of x.

Problem 8 (2 marks). Show that x = g(y) = y 3 + 3y has an inverse. If the inverse function is denoted by
f , find the equation of the normal to the curve y = f (x) at the point where x = 4.

*Is it possible to find an expression for f 0 (x) in terms of x?

4 Derivatives of a parametric function

Sometimes both y and x are given in terms of some third variable, known as a parameter. For example, x(t)
and y(t) are the coordinates of a particle dependent on time t. Then y(t) can be differentiated with respect
to time using the chain rule:
dy dy dt
= ×
dx dt dx
dy
dy dt
= dx
dx dt
0 ẏ
y = ,
ẋ
where in the last line ẏ and ẋ represent differentiation with respect to time.  Thinking physically, this last
ẋ(t)
expression makes sense, if we note that the velocity vector v (t) = points in the direction of the
ẏ(t)
ẏ(t)
tangent to the path. The ratio gives the gradient of the velocity vector.
ẋ(t)

dy v (t)
dt
dx
dt

6
Problem 9 (3 marks).

Find the derivative y 0 (x) (assume a > 0, b > 0), if:

p √ p √
a) x = 3 1 − t, y = 1 − 3 t.
b) x = sin2 t, y = cos2 t.
c) x = a cos t, y = b sin t.

Problem 10 (3 marks).

a) Sketch the curve given parametrically by:

x = at2
y = 2at

b) Find the equation of the normal to the curve at t = t0 .

c) Find the values of t0 for which this normal passes through (5a, 2a).

5 Functions of multiple variables

At this point in your studies you are very familiar with functions of a single variable, and know that y = f (x)
can be represented as a graph. It is also possible to define a function of two or more variables. A two-variable
function z = u(x, y) can be represented as a surface of height z, which varies as a function of x and y.

The usual method of representing a two-variable function graphically is the contour plot, which consists of
lines connecting points (x, y) corresponding to the same z — akin to lines showing elevation on topographic
maps. The values of z, for which contours are drawn, are typically chosen at equal distances (i.e. to represent
an arithmetic progression).

Example 5. Sketch the surface

z = x2 + y 2

√
Solution The contour x2 + y 2 = z, which for k > 0 is a circle with radius z. The curve, sketched below, is
a paraboloid.

7
 y

z = x2 + y 2 2
z

4
1
x
3
−2 −1 1 2
√ −1
R= z 2 y

1
−2
x

Problem 11 (8 marks). Sketch the following surfaces, and describe the contour lines:

a) z = 2x + y;
2
+y 2
b) z = e−x
c) z = xy;
1
d) z = x2 +y 2 +1 .

6 Partial derivatives

6.1 First-order partial derivatives

Suppose that u(x, y) is a real valued function and let x increase by a small amount h while y remains
unchanged in value. The average rate of increase of u with respect to x is given by
u(x + h, y) − u(x, y)
.
h
The limit as h → 0, if it exists, is called the partial derivative with respect to x at the point (x, y) and is
∂u(x, y)
written . We can define a similar partial derivative by letting y vary and keeping x constant:
∂x
u(x, y + y) − u(x, y) ∂u(x, y)
lim = . (3)
h→0 h ∂y

∂u
Notice that in order to avoid a surfeit of symbol, one can often write ∂y if it is clear that this is being
evaluated at the point (x, y).

The rules of ordinary differentiation (chain rule, product rule, quotient rule etc) can easily by extended to
partial differentiation, as the only thing one needs to remember is that any variables aside from the one you
are differentiating by need to be treated as if they were a constant.

∂u ∂u
Example 6. Find ∂x and ∂y for the functions.

8
 a) u = x3 y 2 ;
b) u = ln(x2 + y 2 );
c) u = xy .

Solution.

a) To find ∂u 3 2
∂x we differentiate the x term with respect to x and leave the factor of y unchanged, so
∂u 2 2 ∂u 3
∂x = 3x y . Analogously we have ∂y = 2x y.

∂u 2x ∂u 2y
b) In both cases we use the chain rule to differentiate ∂x = x2 +y 2 and ∂y = x2 +y 2 .

c) In this case if y is treated as a constant then the function is simply a polynomial in x, and we have
∂u y−1
∂x = yx . If, however, we treat x as a constant, then the function is an exponential in y and to differentiate
we write it as u = eln xy which differentiates to ∂u
∂y = ln xe
lnxy
= ln xxy .

∂u ∂u
Problem 12 (3 marks). Find ∂x and ∂y in the following cases:

a) u = x4 y;
b) u = xy ;

c) u = arctan xy .

6.2 Second-order partial derivatives

Once you have found a partial derivative, with respect either to x or to y, then it is possible to form second,
or higher, derivatives by computing the partial derivative of the result. The following notation is commonly
used:

∂2u
 
∂ ∂u
=
∂x ∂x ∂x2
∂2u
 
∂ ∂u
=
∂y ∂x ∂y∂x
∂2u
 
∂ ∂u
=
∂x ∂y ∂x∂y
∂2u
 
∂ ∂u
=
∂y ∂y ∂y 2

Example 7. Find the four second-order partial derivatives for u = x2 ln y.

Solution

9
 ∂u
We have = 2x ln y. The partial derivatives of this function are
∂x
∂2u
 
∂ ∂u
= = 2 ln y
∂x2 ∂x ∂x
∂2u
 
∂ ∂u 2x
= =
∂y∂x ∂y ∂x y

∂u x2
On the other hand, = and then .
∂y y

∂2u
 
∂ ∂u 2x
= =
∂x∂y ∂x ∂y y
2
x2
 
∂ u ∂ ∂u
= = −
∂y 2 ∂y ∂y y2

We can see here that the two mixed partial derivatives are both equal regardless of whether we differentiate
∂2u ∂2u
first with respect to x and then with respect to y, or the other way around, i.e. = . This result
∂y∂x ∂x∂y
is known as Clairaut’s theorem.

Let us show this. In spirit of Eq. (3), we write

∂ 2 u(x, y) ∂ ∂u(x, y)
=
∂x∂y ∂x ∂y
∂u(x+hx ,y) ∂u(x,y)
∂y − ∂y
= lim
hx →0 hx
u(x+hx ,y+hy )−u(x+hx ,y) u(x,y+hy )−u(x,y)
hy − hy
= lim .
hx →0,hy →0 hx

If we change the order of x and y, the above expression changes to

u(x+hx ,y+hy )−u(x,y+hy ) u(x+hx ,y)−u(x,y)
∂ 2 u(x, y) hx − hx
= lim .
∂y∂x hx →0,hy →0 hy

We can see by inspection that the two expressions are the same.

6.3 Approximations and errors

In Y12, we learned about Taylor series2 . The first-order Taylor approximation of a one-variable function
f (x) is f (x + ∆x) = f (x) + f 0 (x)∆x. Let us generalize it to multiple variables. If φ(x, y) is a differentiable
function and ∆x and ∆y, are small, then
∂φ ∂φ
φ(x + ∆x, y + ∆y) − φ(x, y) = (x, y)∆x + (x, y)∆y (4)
∂x ∂y
2 If you started COMPOS in Year 13 and are unfamiliar with Taylor series, you can watch this Khan academy video for an

introduction.

10
To see this, let the positive x-axis point east, the positive y-axis point north.

Imagine walking due East in a straight line from (x, y) to (x + ∆x, y). On this line the value of y is constant,
so by the very definition of the partial derivative, we have, for small ∆x that
φ(x + ∆x, y) − φ(x, y) ∂φ
≈ (x, y),
∆x ∂x
which rearranges to
∂φ
φ(x + ∆x, y) ≈ φ(x, y) + (x, y)∆x. (5)
∂x
Note that this is simply the normal approximation of the rate of increase of a function by the gradient of its
tangent.

z = φ(x, y)
• φ(x + ∆x, y)
• φ(x, y) + ∂φ
∂x (x, y)∆x
φ(x, y)
•
x(East)
x x + ∆x

We can now consider walking due North from (x + ∆x, y) to (x + ∆x, y + ∆y). By identical reasoning we
have that
∂φ
φ(x + ∆x, y + ∆y) − φ(x + ∆x, y) ≈ (x + ∆x, y)∆y. (6)
∂y

Notice the partial derivative is computed at (x + ∆x, y) rather than the original point (x, y) because the
second part of the journey starts from where the first part of the journey finished. But if ∆x is small then
∂φ ∂φ
(x + ∆x, y) ≈ (x, y)∆y.
∂y ∂y

Combining Eqs. (5) and (6), and taking the above approximation, we obtain Eq. (4).

11
Equation (4) is often written more concisely as
∂φ ∂φ
∆φ ≈ ∆x + ∆y. (7)
∂x ∂y

Problem 13 (4 marks).

a) For φ = xm y n show that  

∆φ ∆x ∆y
≈ m +n . (8)
φ x y
√
3
1003
b) Find an approximation for √ . Do not use a calculator.
102

Suppose that, as a physicist, you are performing an experiment, in which you measure the variables x and
y, and use your measurements to compute some function φ(x, y). Let the uncertainty of measuring x be ∆x
(i.e. x is known to within ±∆x) and y with the uncertainty ∆y. Equation (7) can then be used to estimate
the uncertainty of φ.

Problem 14 (4 marks).

a) A train is moving at a speed (60±1) km/h. A passenger is walking down the train isle at a speed (5±1)
km/h in the direction opposite to the train’s movement. What is the uncertainty of the passenger’s
speed with respect to the ground?
b) A car is moving from rest with a constant acceleration (5 ± 0.1) m/s2 for (5 ± 0.2) seconds. What is
the uncertainty of the distance that the car has travelled?
c) A voltage (300 ± 1) V is applied to a resistor (600 ± 2) Ohm. What is the uncertainty of the current?

We hope you did not answer “zero” in any of the above questions! A measurement result can deviate from the
true value to both positive and negative sides. When calculating the uncertainly of φ, you need to consider
the worst case scenario, such that the signs of errors in x and y combine to give the highest deviation of φ.

Let us consider two specific cases to obtain the error propagation rules you may be already familiar with.

• Let φ = x ± y. Then Eq. (7) becomes ∆φ = ∆x ± ∆y. This is formulated as the familiar rule that when
you add or subtract two quantities, their absolute uncertainties must be added to obtain the absolute
uncertainty of the result.
∆φ ∆x ∆y
• Let φ = xy or φ = x/y. Then, using Eq. (8), we find = ± . To put it in words, when
φ x y
you multiply or divide two quantities, their relative uncertainties must be added to obtain the relative
uncertainty of the result.

6.4 Conservative forces

In Y12 Physics Assignment 6, you learned the concept of conservative force. The work WAB performed by
a conservative force moving a point object from point A to point B depends only on the location of these

12
points, but not on the path taken. For conservative forces, one can define potential energy: a function V (~r)
such that
WAB = −(VB − VA ),
where VA and VB are the potential energy values at the points A and B, respectively. We have seen that
the weight and spring force are conservative.

The new message we would like to convey in this assignment is as follows. The components of a conservative
force vector are equal to the partial derivatives of the potential energy by the respective coordinates3 :

∂V (~r)
 
 
Fx  ∂x 
F~ (~r) = = −  ∂V (~r)  .
 (9)
Fy
∂y
In other words, a conservative force is the negative gradient of its potential energy.

To see this, let us consider an element ∆~r = (∆x, ∆y) of the path from A to B, which is sufficiently short
to be approximated as straight. The corresponding work equals

W = −∆V = F~ · ∆~r = Fx ∆x + Fy ∆y ,

where ∆V is the change of the potential energies between the ending and starting points of the path.
Comparing this with Eq. (7), we obtain the result (9).

This result is straightforwardly generalized to three dimensions. For example, in the case of gravitational field
near the Earth surface, we have V = mgz. The components of the force are then equal to Fx = −∂V /∂x = 0,
Fy = −∂V /∂y = 0, and Fz = −∂V /∂z = −mg, as expected.

Example 8. The potential energy of an object in three dimensions is given by V = −K/r, where K is a
constant and r is the absolute value of its radius vector. Find the corresponding force vector.
p
Solution. We write r = x2 + y 2 + z 2 . Differentiating the potential energy, we find

−Kx/(x2 + y 2 + z 2 )3/2
     
∂V /∂x x
K
F~ = −  ∂V /∂y  =  −Ky/(x2 + y 2 + z 2 )3/2  = − 2 2 + z 2 )3/2
 y .
∂V /∂z 2 2 2 3/2 (x + y z
−Kz/(x + y + z )

This can be rewritten simply as F~ = −K~r/r3 or F~ = −K r̂/r2 , where r̂ = ~r/r is the vector of magnitude
1 in the direction of ~r. In other words, the force’s direction is towards the origin, and its magnitude is
proportional to the square of the distance to the origin. Examples of such force are electrostatic attraction
of a point charge or gravitational attraction of a point mass.

Given a force vector’s components, can we tell whether it is conservative? For two dimensions, there is
∂Fx ∂2V ∂Fy ∂2V
a simple recipe. From Eq. (9), we can see that = and = . In other words, for a
∂y ∂y∂x ∂x ∂x∂y
3A vector of the form (9) — whose components are partial derivatives of some position-dependent scalar by the respective
~ (read
coordinates — is called the gradient of that scalar. Gradients are important, so they are assigned a special symbol, ∇
“nabla”):  
∂V (~
r)/∂x ~ (~
≡ ∇V r).
∂V (~
r)/∂y
Note that previously we encountered the term “gradient” in the meaning of “derivative” or “slope”. The gradient defined here
~
is the extension of this term to multiple dimensions. For example, if z(x, y) is the height of a hill, the vector ∇z(x, y) shows
the direction of the steepest accent up this hill and the slope steepness.

13
conservative force, the derivative of Fx with respect to y must be equal to the derivative of Fy with respect
to x.

Problem 15 (4 marks). Which of the forces below are conservative? If the force is conservative, find the
corresponding potential energy.

a) Fx = y, Fy = −x;
b) Fx = y, Fy = x;
c) Fx = x, Fy = y, Fz = z.

6.5 Miscellaneous problems

Problem 16 (3 marks). Consider a function u(x, y). Its both arguments x and y are, in turn, dependent
on a single parameter t — so we can also think of u as a function of t. Show that
du ∂u dx ∂u dy
= + .
dt ∂x dt ∂y dt

This is the higher dimensional version of the chain rule. You should think at this point about how it can be
extended to functions of more variables, e.g for the scalar field u(x, y, z).

Problem 17 (4 marks). Let u(x, t) = φ(x + ct) + ψ(x − ct), where φ and ψ are arbitrary functions of a
single argument. Verify that u(x, t) satisfies the differential equation known as the wave equation

∂2u 2
2∂ u
= c .
∂t2 ∂x2

7 Optional extra problems

Problem 18. *

Find the derivative y 0 (x) (assume a > 0, b > 0), if:

a) x = a cosh t, y = b sinh t.
b) x = a cos3 t, y = a sin3 t.
c) x = a(t − sin t), y = a(1 − cos t)
d) x = e2t cos2 t, y = e2t sin2 t.
t 1
e) x = arcsin √ , y = arccos √ .
1+t 2 1 + t2

14