MATH 2023 • Spring 2015-16 • Multivariable Calculus

Problem Set #3 • Chain Rule, Directional Derivatives, Gradients

1. (F) Suppose w = f ( x, y, z) where x = g(s), y = h(s, t) and z = k (t). Assume all functions
involved are C1 . Draw the tree diagram to showcase the relations between w, x, y, z, s and
∂w ∂w
t. Hence, write down the chain rule for calculating the partial derivatives: and .
∂s ∂t
Use the symbols ∂ and d appropriately.

Solution:
W
∂w ∂w dx ∂w ∂y
∂s
=
∂x ds
+
∂y ∂s *
∂w
=
∂w ∂y ∂w dz
+
# +
∂t ∂y ∂t ∂z dt

2. (F) Recall that the rectangular-polar coordinates conversion rules are given as follows:

x = r cos q
y = r sin q

∂f
A function f ( x, y) is said to be rotationally/radially symmetric if ∂q = 0, i.e. when
:
regarded as a function of (r, q ), it depends only the radial variable r but not the angular
variable q. For instance, f ( x, y) = x2 + y2 is rotationally symmetric since f (r, q ) = r2 .
Using the chain rule, show that f is rotationally symmetric if and only if:

∂f ∂f
y =x .
∂x ∂y

Solution: By the chain rule:

∂f ∂ f ∂x ∂ f ∂y
= +
∂q ∂x ∂q ∂y ∂q

a.
∂f ∂ ∂f ∂
= (r cos q ) + (r sin q )
∂x ∂q ∂y ∂q
'
∂f ∂f .
= ( r sin q ) + (r cos q )
∂x ∂y
∂f ∂f
= y +x .
∂x ∂y
∂f
Therefore, ∂q = 0 if and only if y ∂∂xf = x ∂∂yf .

Page 1
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

3. (F) Suppose f (u, v) is a C2 function, and u = s2 t and v = s + t2 . Express the second

∂2 f
partial derivative in terms of f uu , f uv , f vv , s and t.
∂s∂t

Solution:
∂f ∂ f ∂u ∂ f ∂v
= +
∂t ∂u ∂t ∂v ∂t
= f u · ( 1) + f v · 2t = f u + 2t f v fax
∂2 f ∂ ∂f
✓ ◆ st
=
∂s∂t ∂s ∂t
∂
= ( f u + 2t f v )
∂s✓ ◆ ✓ ◆
∂ f u ∂u ∂ f u ∂v ∂ f v ∂u ∂ f v ∂v
= + + 2t +
∂u ∂s ∂v ∂s ∂u ∂s ∂v ∂s
= ( f uu · 2s + f uv · 1) + 2t ( f vu · 2s + f vv · 1)
2s f uu + (4t 1) f uv + 2t f vv .
=
* t
Here we have used the fact that f is C2 , and so f uv = f vu .

4. (F) Let f ( x, y, z) be a C1 function of three variables, and z be a C1 function of ( x, y) such

that
f ( x, y, z( x, y)) = 0.
Using the chain rule, show that:

∂z fx ∂z fy
= and = .
∂x fz ∂y fz

Solution:
∂ ∂ f
f ( x, y, z( x, y)) = 0=0
∂x
∂f
∂x
∂ f ∂z
* Z

+ =0 ^
∂x ∂z ∂x

:
∂f
∂z ∂x fx
= ∂f
=
∂x fz
∂z

Similarly,

∂ ∂
f ( x, y, z( x, y)) = 0=0 i
∂y ∂y = Z

∂f ∂ f ∂z ^
+ =0 X T
∂y ∂z ∂y
∂f
∂z ∂y fy
= ∂f
=
∂y fz
∂z

Page 2
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

5. (FF) Let f ( x, y) be a C1 function. Consider two parametric curves r1 (t) = x1 (t)i + y1 (t)j
and r2 (t) = x2 (t)i + y2 (t)j which satisfy:

r1 (0) = r2 (0) and r10 (0) = r20 (0).

(a) Show that

d d
f ( x1 (t), y1 (t)) = f ( x2 (t), y2 (t)).
dt t =0 dt t =0

Solution:
Using the chain rule, we get:

d ∂ f dx ∂ f dy
f ( x1 (t), y1 (t)) = +
dt ∂x dt ∂y dt
∂f 0 ∂f 0
= x1 ( t ) + y (t) (since x = x1 and y = y1 in this case)
∂x ∂y 1

At t = 0, we get:

d ∂f ∂f
f ( x1 (t), y1 (t)) = x10 (0) + y10 (0)
dt t =0 ∂x ( x1 (0),y1 (0)) ∂y ( x1 (0),y1 (0))

Similarly, one can also show:

d ∂f ∂f
f ( x2 (t), y2 (t)) = x20 (0) + y20 (0)
dt t =0 ∂x ( x2 (0),y2 (0)) ∂y ( x2 (0),y2 (0))

It is given from the problem that:

r1 (0) = r2 (0) and r10 (0) = r20 (0).

Therefore, we have:

x1 (0) = x2 (0) x10 (0) = x20 (0)

y1 (0) = y2 (0) y10 (0) = y20 (0)

and so:
d d
f ( x1 (t), y1 (t)) = f ( x2 (t), y2 (t)).
dt t =0 dt t =0

(b) Give a geometric interpretation of the above result.

Solution: This result shows that any two parametric curves with the same posi-
tion and velocity at t = 0 will give the same rate of change of a function f along
these two curves at t = 0.
"

toffee
"

Page 3
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

6. (FF) The wave equation is an important partial differential equation which governs the
propagation of waves. Let u( x, y, z, t) be the displacement of the wave at position ( x, y, z)
at time t. It can be shown by several physical laws (such as the Hooke’s Law) that u
satisfies: ✓ 2 ◆
∂2 u 2 ∂ u ∂2 u ∂2 u
=c + 2+ 2 (1)
∂t2 ∂x2 ∂y ∂z
where c is a constant (which is the wave speed).
In one (spatial) dimension, the wave equation can be stated as:

∂2 u 2
2∂ u
= c . (2)
∂t2 ∂x2

It turns out that the chain rule of several variables has a nice application on solving
the one dimensional wave equation. The following exercise guides you to show that if
u( x, t) is a solution to the one dimensional wave equation, then it must take the form
u( x, t) = F ( x ct) + G ( x + ct) where F and G are arbitrary differentiable functions of
single variable.
Let u( x, t) solve the one dimensional wave equation (2).
(a) Define x = x ct and h = x + ct. Regard u as a function of x and h, and x and h are
functions of x and t. Using the chain rule of multivariable functions, show that:

ut = c(uh ux ) and utt = c2 (uxx 2uxh + uhh ).

Solution: By chain rule:

∂u ∂u ∂x ∂u ∂h
= + x
∂t ∂x ∂t ∂h ∂t
= ux · ( x ct)t + uh · ( x + ct)t m
= ux · ( c) + uh · c
= c ( u h u x ).

tx
∂ut
utt =
∂t
∂
= c ( u h u x ).
∂t

Since uh and ux are also functions of (x, h ), the chain rule applies to uh and ux in
the same way as it does to u:

∂uh ∂uh ∂x ∂uh ∂h

*
= +
∂t ∂x ∂t ∂h ∂t
= uhx · ( x ct)t + uhh · ( x + ct)t
÷
'

cuhx + cuhh .
=
∂ux ∂ux ∂x ∂ux ∂h
= +
∂t ∂x ∂t ∂h ∂t
= uxx · ( x ct)t + uxh · ( x + ct)t
A
×
He

= cuxx + cuxh .

Page 4
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

Substitute these two results into utt , we have:

∂
utt = c (uh ux )
∂t
✓ ◆
∂uh ∂ux
=c
∂t ∂t
= c( cuhx + cuhh + cuxx cuxh )
= c2 (uxx uxh uhx + uhh )
2
= c (uxx 2uxh uhh ).

(b) Using the chain rule agin, show that

u x = ux + uh and u xx = uxx + 2uxh + uhh .

Solution: Apply the chain rule again:

×
u x = ux x x + uh hx
= ux ( x ct) x + uh ( x + ct) x
= ux + uh .
=
Att
u xx = (u x ) x
= (ux + uh ) x
= (ux ) x + (uh ) x
= (ux )x x x + (ux )h hx + (uh )x x x + (uh )h hx
= uxx ( x ct) x + uxh ( x + ct) x + uhx ( x ct) x + uhh ( x + ct) x
= uxx + uxh + uhx + uhh
= uxx + 2uxh + uhh .

(c) Combining the results of (a), (b) and the wave equation, show that uxh = 0.

Solution: Results in (a) and (b) show:

utt = c2 (uxx 2uxh + uhh ), u xx = uxx + 2uxh + uhh .

Substitute them into the wave equation utt = c2 u xx , we have:

c2 (uxx 2uxh + uhh ) = c2 (uxx + 2uxh + uhh )

uxx 2uxh + uhh = uxx + 2uxh + uhh
2uxh = 2uxh
0 = 2uxh + 2uxh = 4uxh
uxh = 0.

Page 5
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

(d) Finally, deduce that u, as a function of x and h, must be in the form of:

u(x, h ) = F (x ) + G (h )

where F and G are arbitrary functions. Hence, in terms of the original variables x
and t, u must take the form u( x, t) = F ( x ct) + G ( x + ct).

Solution: Finally, we have uxh = 0. Therefore,

∂ux
= 0.
∂h

Thus ux is independent of h, so ux is a function of x only. Let f (x ) be this function

and so
∂u
ux = f (x ), or equivalently, = f (x )
∂x
Integrating both sides by x, we get:
ˆ
u= f (x )dx + ‘integration constant’.

Note that the integration ‘constant’ is no longer a constant, but a quantity not de-
pending on the integration variable x. In other words, the ‘integration constant’
now becomes a function of h.
Denote this function by G (h ), then:
ˆ
u(x, h ) = f (x )dx + G (h ).

Since f is an arbitrary function of f (x )dx is arbitrary too. We rewrite

´
( x ) x,
f (x )dx as F (x ), where F is an arbitrary function of x. Finally, we have:
´

u(x, h ) = F (x ) + G (h ).

Since x = x ct and h = x + ct, in terms of the ( x, t)-variables:

u( x, t) = F ( x ct) + G ( x + ct).

FYI: The general solution u( x, t) = F ( x ct) + G ( x + ct) of the wave equation describes
the superposition of two waves – one has a graph given by the function F and it shifts to
the right by c unit lengths per unit time, another has a graph given by the function G and
it shifts to the left by c unit lengths per unit time. The results in this exercise show that
all solutions to the one-dimensional wave equation are superposition of these waves.

7. (FFF) In many physics, geometry and engineering applications, it is often more conve-
nient to use polar or sphereical coordinates since many physical quantities are rotationally
symmetric.
The conversion rule of rectangular and polar coordinates is given by:

x = r cos q
y = r sin q

Page 6
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

Let u be a function of x and y. Since ( x, y) can be converted into (r, q ), we can also regard
u as a function of (r, q ). The chain rule can be used to derive some conversion formulae
between u x , uy and ur , uq .
An important operator in physics, geometry and engineering is called the Laplacian. In
two dimensions, it is defined as:
∂2 u ∂2 u
r2 u = + 2 = u xx + uyy .
∂x2 ∂y
In this exercise, we will show that r2 u can be expressed in polar form as:
∂2 u 1 ∂u 1 ∂2 u
r2 u = + + .
∂r2 r ∂r r2 ∂q 2
The polar form of the Laplacian is often used when dealing with rotationally symmetric
functions, i.e. a function u which does not depend on q but only on r. For such functions,
their Laplacian is simply:
1
r2 u = urr + ur .
r
p y
2 2
(a) Use the fact that r = x + y and tan q = x , show that:
∂r x ∂r y ∂q y ∂q x
= , = , = , = 2.
∂x r ∂y r ∂x r2 ∂y r

Solution: Using the fact that r2 = x2 + y2 , by differentiating both sides with

respect to x, we get:
∂ 2 ∂ 2
r = ( x + y2 )
∂x ∂x
∂r ∂r x
2r = 2x =) =
∂x ∂x r
Similarly, by differentiating both sides with respect to y, we get:
∂ 2 ∂ 2
r = ( x + y2 )
∂y ∂x
∂r ∂r y
2r = 2y =) =
∂y ∂y r
y
To find ∂x
∂q
and ∂y
∂q
, we consider the fact that tan q = x. Differentiate both sides
with respect to x, we get:
∂ ∂ ⇣y⌘
tan q =
∂x ∂x x
2 ∂q y
sec q =
∂x x2
∂q y
=
∂x x sec2 q
2

y y2
Since sec2 q = 1 + tan2 q, and recall that tan q = x, we have sec2 q = 1 + x2
.
Therefore, after simplication, we can obtain:
∂q y y y y
= = ⇣ ⌘ = = .
∂x x2 sec2 q x2 · 1+ y2 x 2 + y2 r2
x2

Page 7
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

y
Similarly, differentiating both sides of tan q = x by y, we get:

∂ ∂ ⇣y⌘ 1
tan q = =
∂y ∂y x x
∂q 1
sec2 q =
∂y x
∂q 1
=
∂y x sec2 q

y2
We have previously derived that sec2 q = 1 + x2
, and so:

∂q 1 1 1 x
= ⇣ ⌘ = ⇣ ⌘ = = .
∂y x 1+
y2
x
x 2 + y2 r2 r2
x2 x2 x

(b) Regard u as a function of (r, q ), and (r, q ) are functions of ( x, y). Sketch a tree
diagram to showcase these relations. Using the chain rule, show that:
xur yuq
ux = ,
r r2
yur xu
uy = + 2q .
r r

Solution:
∂u ∂r ∂u ∂q
ux = +
∂r ∂x ∂q ∂x
⇣ y⌘
x
= ur · + u q · from (a)
r r2
xur yuq
=
r r2
∂u ∂r ∂u ∂q
uy = +
∂r ∂y ∂q ∂y
y x
= ur · + u q · 2 from (a)
r r
yur xu
= + 2q
r r

Page 8
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

(c) Using quotient and product rules, show that:

ur xurx x 2 ur yuqx 2xyuq

u xx = + 3 2
+
r r r r r4
ur yury 2
y ur xuqy 2xyuq
uyy = + 3
+ 2
r r r r r4

Solution:
∂u x ∂ ⇣ xur yuq ⌘
u xx = = from (b)
∂x ∂x r r2
r · ∂x ( xur ) xur ∂x
∂ ∂r
r2 · ∂x
∂ ∂ 2
(yuq ) yuq · ∂x r
= 2 4
quotient rule
r r
x 2
r (ur + xurx ) xur · r r · yuqx yuq · 2r · ∂x
∂r
= product rule
r2 r4
ur xurx 2
x ur yuqx 2ryuq · xr
= + + break it down
r r r3 r2 r4
which is exactly what we need to show after simplifying the last term.
Similarly for uyy :

∂ ∂ ⇣ yur xu ⌘
uyy = uy = + 2q
∂y ∂y r r
r ∂y
∂
(yur ) yur ∂y
∂r
r2 ∂y
∂
( xuq ) xuq · ∂ 2
∂y r
= +
r2 r4
r ur + yury
y
yur · r r2 xuqy xuq · 2r · ∂r
∂y
= +
r2 r4
ur yury y2 ur xuqy 2rxuq · yr
= + +
r r r3 r2 r4
as required (after simplifying the last term).

(d) Since ur and uq are functions of (r, q ), and (r, q ) are functions of ( x, y), they share the
same tree diagram as u in part (b), and hence we have

∂ur ∂ur ∂r ∂ur ∂q

urx = = +
∂x ∂r ∂x ∂q ∂x
and similar for other second derivatives ury , uqx and uqy . Show that:

xurx + yury = rurr

xuqy yuqx = uqq

Solution: We apply the chain rule for each of urx , ury , uqx and uqy :

∂ur ∂ur ∂r ∂ur ∂q

urx = = +
∂x ∂r ∂x ⇣ ∂q⌘∂x
x y
= urr · + urq · from (a)
r r2

Page 9
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

⇣ ⌘
∂ur ∂2 u
Note that ∂r = ∂
∂r
∂u
∂r = ∂r2
= urr . Similar for urq .
Similarly,

∂ur ∂ur ∂r ∂ur ∂q

ury = = +
∂y ∂r ∂y ∂q ∂y
y x
= urr · + urq · 2
r r
Combining these two results, we then have:
⇣ x y⌘ ⇣ y x⌘
xurx + yury = x urr · urq · 2 + y urr · + urq · 2
r r r r
x2 xy y2 xy
= urr · urq · 2 + urr · + urq · 2
r r r r
x 2 + y2
= urr · cancellation
r
r2
= urr · = rurr
r

The second identity can be proven in a similar way:

∂ ∂uq ∂r ∂u ∂q
uqy = uq = + q
∂y ∂r ∂y ∂q ∂y
y x
= uqr · + uqq · 2
r r
∂ ∂uq ∂r ∂u ∂q
uqx = uq = + q
∂x ∂r ∂x⇣ ∂q ∂x
x y⌘
= uqr · + uqq ·
r r2
Therefore, we can show:
⇣ y x⌘ ⇣ x ⇣ y ⌘⌘
xuqy yuqx = x uqr · + uqq · 2 y uqr · + uqq ·
r r r r2
xy x 2 xy y 2
= uqr · + uqq · 2 uqr · + uqq ·
r r r r
x 2 + y2
= uqq ·
r2
r2
= uqq · 2 = uqq
r
as required.

Page 10
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

(e) Combining the results proved in previous parts, show that:

1 1
u xx + uyy = urr + ur + 2 uqq .
r r

Solution: OK! Finally, we can combine everything together:

u xx + uyy
ur xurx x 2 ur yuqx 2xyuq
= + 3 2
+
r r r r r4
ur yury y2 ur xuqy 2xyuq
+ + + from (c)
r r r3 r2 r4
2ur xurx + yury 2 2
( x + y ) ur
= +
r r r3
xuqy yuqx
+ factorization, tidy up
r2
2ur rurr r 2 ur u
= + 3
+ qq from (d)
r r r r2
2ur ur u
= + urr + qq
r r r2
1 1
= urr + ur + 2 uqq
r r
which is exactly as required! Cheers!

8. (F) Compute the directional derivative of the following functions at the given point P in
the direction of the given vector v. Moreover, find the unit direction u along which the
function increases most rapidly.
(a) f ( x, y) = x2 y2 , P( 1, 3), v = 35 i 4
5 j.

Solution:
∂f ∂f
rf = i+ j = 2x i 2y j
∂x ∂y
r f ( P) = r f ( 1, 3) = 2( 1) i 2( 3) j = 2i + 6j
s
✓ ◆2 ✓ ◆
3 4 2
|v| = + =1
5 5
v 3 4
v̂ = =v= i j
|v| 5 5
✓ ◆
3 4
Dv̂ f ( P) = r f ( P) · v̂ = ( 2i + 6j) · i j = 6
5 5
Since Du f ( P) = r f ( P) · u = |r f ( P)| |u| cos q = |r f ( P)| cos q for any unit vector
u. Here q is the angle between r f ( P) and u. It is the largest when q = 0 (i.e.
cos q = 1). Therefore, Du f ( P) achieves its maximum when r f ( P) and u are
parallel. Therefore, the unit direction u along which the function increases most
rapidly is given by:
r f ( P) 2i + 6j
u= = p .
|r f ( P)| 40

Page 11
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

(b) g( x, y) = e x y, P(ln 2, ln 3), v = i + j.

Solution:
x y x y
rg = e i e j
ln 2 ln 3 ln 2 ln 3 1 1
r g( P) = e i e j= i j
6 6
v i+j
v̂ = = p
|v| 2
✓ ◆
1 1 i+j 2
Dv̂ g( P) = r g( P) · v̂ = i j · p = p
6 6 2 6 2
r g( P) 1 1
u= = p i p j
|r g( P)| 2 2

(c) h( x, y) = e xy , P(1, 0), v = 5i + 12j.

Solution: (Answer only)

12
Dv̂ h( P) = , u = j.
13

(d) F ( x, y, z) = xy + yz + zx + 4, P(2, 2, 1), v = j k.

Solution: (Answer only)

3 i + 3j
Dv̂ F ( P) = p , u= p .
2 10

(e) G ( x, y, z) = e xyz 1, P(0, 1, 1), v = 2i + 2j k.

Solution: (Answer only)

2
rv̂ G ( P) = , u= i.
3

9. (F) For each surface and the given point P, find the value a such that P lies on the surface,
and then find an equation of the tangent plane to the surface at the point P:
(a) x2 + y + z = 3, P(2, 0, a)

Solution: To solve for a, we put ( x, y, z) = (2, 0, a) into the given equation:

22 + 0 + a = 3 =) a = 1

To find the equation of the tangent plane at P, we need a normal vector to the
surface:
r( x2 + y + z 3) = 2xi + j + k

Page 12
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

Therefore, n = (2xi + j + k)| P = 2(2)i + j + k = h4, 1, 1i. Equation of the tangent

plane at P is: 4x + 1y + 1z = 4(2) + 1(0) + 1( 1). After simplification:

4x + y + z = 7.

(b) xy sin z = 1, P( a, 2, p/6)

Solution: (Answer only)

a=1
1 p
r ( xy sin z
1)| P = i + j + 3k
2
1 p p
x + y + 3z = 2 + .
2 2

(c) yze xz = 8, P(0, a, 4)

Solution: (Answer only)

a=2
xz
r (yze 8)| P = 32i + 4j + 2k
16x + 2y + z = 8

(d) z = e xy , P(1, 0, a)

Solution: (Answer only)

a=1
xy
r(z e )| P = j+k
y+z = 1

(e) z = ln(1 + xy), P(1, 2, a).

Solution:

a = ln 3
2 1
r(z ln(1 + xy))| P = i j+k
3 3
2 1 4
x y+z = + ln 3.
3 3 3

Page 13
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

10. (F) Let

GMm
V ( x, y, z) = p
x2 + y2 + z2
where G, M and m are constants. Define F( x, y, z) = rV ( x, y, z).
(a) Verify that:
xi + yj + zk
F( x, y, z) = GMm .
( x2 + y2 + z2 )3/2

Solution:
!
∂V ∂ GMm
= p
∂x ∂x x 2 + y2 + z2
∂ 1/2
= GMm x 2 + y2 + z2
∂x✓ ◆
1 2 3/2 ∂ 2
= GMm · x + y2 + z2 ( x + y2 + z2 )
2 ∂x
1 2x
= GMm · ·
2 ( x2 + y2 + z2 )3/2
x
= GMm ·
( x + y + z2 )3/2
2 2

Similarly, we have:

∂V y
= GMm ·
∂y ( x + y + z2 )3/2
2 2

∂V z
= GMm ·
∂z ( x + y + z2 )3/2
2 2

Therefore,
∂V ∂V ∂V
rV = i+ j+ k
∂x ∂y ∂z
xi + yj + zk
= GMm
( x + y2 + z2 )3/2
2

as required. Note that F = rV.

(b) Show that |F( x, y, z)| is inversely proportional to the squared distance from ( x, y, z)
to the origin in R3 .

Page 14
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

Solution:

xi + yj + zk
|F( x, y, z)| = GMm
( x2 + y2 + z2 )3/2
1
= GMm | xi + yj + zk|
( x2 + y2 + z2 )3/2
1/2
x 2 + y2 + z2 1
= GMm = GMm
( x 2 + y2 + z2 )
3/2 x2 + y2 + z2
p
Note that the distance from ( x, y, z) to the origin is given by r = x 2 + y2 + z2 .
Hence, we have:
GMm
|F( x, y, z)| =
r2
as required.

11. (FF) Consider the function

f ( x, y) = cos( x + y)
as well as the plane P given by the equation

x y=0.

The intersection of the graph of f with P is a curve C. Find the slope of the tangent line
to C at the point (p, p ) using directional derivatives. [Hint: First sketch a diagram of the
graph, the plane and the curve.]

Solution: The intersection between the plane P and the xy-plane is the straight-line
yD = x. EThere are two unit vectors
D in the E x-y-plane parallel to P, given by u =
p1 , p1 and its negative u = p1 , p1 . The directional derivative of f in the
2 2 2 2
direction of u is now equal to
⌧
∂f ∂f
rf · u = , ·u
∂x ∂y
1 1
= p sin( x + y) p sin( x + y)
2 2
p
= 2 2 sin( x + y)

So the slope of C at (p, p ) is therefore equal to 0, by evaluating this equation at

( x, y) = (p, p ).
Alternatively one can also work with u instead of u and one would obtain the
negative of the above solution, which is also 0 in our situation.

12. (FF) One approach for finding the normal vector of the tangent plane at a given point
( x0 , y0 ) to a graph z = f ( x, y) is by writing the graph equation as a level surface z
Df ( x, y) = 0 ofE a three-variable function g( x, y, z) := z f ( x, y). Then, the gradient r g =
∂f ∂f
∂x , ∂y , 1 at point ( x0 , y0 , f ( x0 , y0 )) is perpendicular to the level surface { g = 0},

Page 15
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

and so we can take it to be a normal vector of the tangent plane as long as r g 6= 0 at

( x0 , y0 , f ( x0 , y0 )).
D E
∂f ∂f
In fact, it is also possible to show the normal vector is ∂x , ∂y , 1 using a purely two-
variable argument instead of going up one higher dimension.

(a) Consider a given function f ( x, y), and a given point ( x0 , y0 ). Find a parametrization:
r1 (t) =?i+?j+?k
of the curve on the graph z = f ( x, y) travelling in the x-direction while keeping y
fixed at y0 (i.e. the red curve in the diagram). Hence, find the tangent vector of the
curve r1 (t) at the point ( x0 , y0 , f ( x0 , y0 )). Label this tangent vector by T1 .

Solution: The projection of the red curve on the xy-plane is a straight line pass-
ing through ( x0 , y0 ) and is parallel to the x-axis, i.e. vector i. Therefore, the
( x, y)-coordinates of the path is given by:
x ( t ) = x0 + t
y ( t ) = y0
The z-coordinate of the red curve is determined by the function f , i.e. z(t) =
f ( x (t), y(t)) = f ( x0 + t, y0 ). Therefore, the parametrization of the red curve is:
r1 (t) = x (t)i + y(t)j + z(t)k = ( x0 + t)i + y0 j + f ( x0 + t, y0 ) k.
When t = 0, the “particle” is at the point ( x0 , y0 , f ( x0 , y0 )). Therefore, the tangent
vector T1 to the curve at this point is given by:
d
T1 = r10 (0) = i + 0j + f ( x0 + t, y0 ) k.
dt t =0
d
To compute dt f ( x0 + t, y0 ), we apply the chain rule:
d ∂f d ( x0 + t ) ∂ f d ( y0 )
f ( x0 + t, y0 ) = ( x0 + t, y0 ) + ( x0 + t, y0 )
dt ∂x dt ∂y dt
∂f
= ( x0 + t, y0 ).
∂x
Evaluate at t = 0, we get:
d ∂f
f ( x0 + t, y0 ) = ( x0 , y0 ).
dt t =0 ∂x
Therefore, we get:
∂f
T1 = i + ( x0 , y0 ) k.
∂x

Page 16
MATH 2023 Chain Rule, Directional Derivatives, Gradients Problem Set #3

(b) Find a parametrization r2 (t) of the curve on the graph z = f ( x, y) travelling in the y-
direction while keeping x fixed at x0 (i.e. the blue curve in the diagram). Hence, find
the tangent vector of r2 (t) at the point ( x0 , y0 , f ( x0 , y0 )). Label this tangent vector by
T2 .

Solution: Similar to (a), the parametrization is given by:

r2 ( t ) = x0 i + ( y0 + t ) j + f ( x0 , y0 + t ) k

and the tangent vector at ( x0 , y0 , f ( x0 , y0 )) is:

∂f
T2 = r20 (0) = j + ( x0 , y0 ) k
∂y

(c) Since both T1 and T2 are tangent vectors to the graph, they are parallel to the tangent
plane. Therefore, the normal vector to the tangent plane must be perpendicular to
both T1 and T2 . Using this fact, show that the normal vector to the tangent plane is
given by ⌧
∂f ∂f
( x0 , y0 ), ( x0 , y0 ), 1 .
∂x ∂y

Solution: Since the normal vector n to the tangent plane is orthogonal to the
tangent vectors T1 and T2 , it can be computed by taking cross product: n =
T1 ⇥ T2 which will yield the vector stated in the problem.

Optional

13. The spherical coordinates (r, q, f) is another important coordinate system in R3 . We

will learn that in later chapters. The conversion rules between spherical and rectangular
coordinates are given by:

x = r sin f cos q
y = r sin f sin q
z = r cos f

Given a C2 function f ( x, y, z), it can be regarded as a function of (r, q, f) as well under the
above conversion rule. Show that the Laplacian r2 f := f xx + f yy + f zz can be expressed
in spherical coordinates as:
✓ ◆ ✓ ◆
∂2 f ∂2 f ∂2 f 1 ∂ 2∂f 1 ∂ ∂f 1 ∂2 f
+ + = 2 r + 2 sin f + .
∂x 2 ∂y 2 ∂z 2 r ∂r ∂r r sin f ∂f ∂f r2 sin2 f ∂q 2

[Note: It is a very time consuming exercise. It took me 4 hours to do it when I was an

undergraduate.]

Page 17