MATH2023 Multivariable Calculus Problem Set 3

L2/L3 (Fall 2019)

Solution to Extra Problem Set 3

𝑦
1. (a) The natural domain of 𝑓 is the subset of ℝ2 consisting of all (𝑥, 𝑦) such that > 0. This inequality holds
𝑥
if and only if either 𝑥 and 𝑦 are both positive or 𝑥 and 𝑦 are both negative. Therefore the natural
domain of 𝑓 is the union of the open first quadrant and the open third quadrant.
𝑦
For each 𝑘 ∈ ℝ, the curve ln = 𝑘 is a level curve of 𝑓. Rearranging this we get
𝑥
𝑦 = 𝑒 𝑘 𝑥,
so the level curves of 𝑓 are all the lines in ℝ2 passing through the origin with positive slope (with the origin
removed).

(b) The natural domain of 𝑔 is the subset of ℝ2 consisting of all (𝑥, 𝑦) such that 𝑦 − 𝑥 ≥ 0. This inequality
2
holds if and only if 𝑦 ≥ 𝑥. Therefore the natural domain of 𝑔 is the closed region in ℝ above the line
𝑦 = 𝑥.
For each 𝑘 ≥ 0, the curve √𝑦 − 𝑥 = 𝑘 is a level curve of 𝑔. Rearranging this we get
𝑦 = 𝑥 + 𝑘2,
so the level curves of 𝑔 are all the lines in ℝ2 with slope 1 and with non-negative 𝑦-intercept.

(c) The natural domain of ℎ is the subset of ℝ2 consisting of all (𝑥, 𝑦) such that 𝑥 2 + 𝑦 2 ≠ 0 . This
requirement holds if and only if (𝑥, 𝑦) ≠ (0, 0). Therefore the natural domain of ℎ is the whole ℝ2 with
the origin removed.
𝑥
− 2 2
For each 𝑘 > 0 and 𝑘 ≠ 1, the curve 𝑒 𝑥 +𝑦 = 𝑘 is a level curve of ℎ. Rearranging this we get
1 2 1 2
(𝑥 + ) + 𝑦2 = ( ) ,
2 ln 𝑘 2 ln 𝑘
so the level curves of ℎ are all the circles in ℝ2 centered at some point on the 𝑥-axis and passing through
the origin (with the origin removed).
𝑦 𝑦 𝑦

𝑥 𝑥 𝑥

Domain and level curves of 𝑓 Domain and level curves of 𝑔 Domain and level curves of ℎ

2. (a) The natural domain of 𝑓 is the subset of ℝ3 consisting of all (𝑥, 𝑦, 𝑧) such that 𝑥 2 − (𝑦 + 𝑧)𝑥 + 𝑦𝑧 ≠ 0.
Noting that 𝑥 2 − (𝑦 + 𝑧)𝑥 + 𝑦𝑧 = (𝑥 − 𝑦)(𝑥 − 𝑧), this requirement holds if and only if 𝑥 − 𝑦 and 𝑥 − 𝑧
are both non-zero. Therefore the natural domain of 𝑓 is the whole space ℝ3 with the two planes
𝑥−𝑦=0 and 𝑥−𝑧 =0
removed.

Page 1 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)

(b) The natural domain of 𝑔 is the subset of ℝ2 consisting of all (𝑥, 𝑦) such that (𝑥 − 𝑦)2 ∈ [−1, 1]. This
requirement holds if and only if −1 ≤ 𝑥 − 𝑦 ≤ 1, i.e. 𝑦 ≥ 𝑥 − 1 and 𝑦 ≤ 𝑥 + 1. Therefore the natural
domain of 𝑔 is the closed region in ℝ2 above the line 𝑦 = 𝑥 − 1 and below the line 𝑦 = 𝑥 + 1.

(c) The natural domain of ℎ is the subset of ℝ3 consisting of all (𝑥, 𝑦, 𝑧) such that 𝑧 − 𝑥 2 − 𝑦 2 + 2𝑥 + 3 > 0.
This requirement holds if and only if 𝑧 > 𝑥 2 + 𝑦 2 − 2𝑥 − 3, i.e. 𝑧 + 4 > (𝑥 − 1)2 + 𝑦 2 . Therefore the
natural domain of ℎ is the open region in ℝ3 above the circular paraboloid 𝑧 + 4 > (𝑥 − 1)2 + 𝑦 2 with
vertex (1, 0, −4) opening towards the positive 𝑧-axis.
𝑦 𝑧
𝑧
−√3 −1
1 𝑦=𝑥−1 𝑦
𝑥−𝑧=0 3 √3
−1
𝑥
0 1
𝑦 𝑦=𝑥+1 −1
𝑥

𝑥 𝑧 + 4 = (𝑥 − 1)2 + 𝑦 2
𝑥−𝑦 = 0

3. (a)

√ 𝑥 − √𝑦 + 1 √ 𝑥 − √𝑦 + 1 √ 𝑥 + √𝑦 + 1
lim = lim ⋅
(𝑥,𝑦)→(4,3) 𝑥 − 𝑦 − 1 (𝑥,𝑦)→(4,3) 𝑥 − 𝑦 − 1
√ 𝑥 + √𝑦 + 1
𝑥−𝑦−1
= lim
(𝑥,𝑦)→(4,3) (𝑥 − 𝑦 − 1)(√𝑥 + √𝑦 + 1)
1 11
= lim = = .
(𝑥,𝑦)→(4,3)
√ 𝑥 + √𝑦 + 1 √4 + √3 + 1 4

(b) Along the 𝑥-axis, we have

𝑥 4 − 𝑦2 𝑥4 − 0
lim = lim = lim 𝑥 2 = 0,
(𝑥,𝑦)→(0,0) 𝑥 2 + 𝑦 2 𝑥→0 𝑥 2 + 0 𝑥→0
𝑦=0

but along the 𝑦-axis, we have

𝑥 4 − 𝑦2 0 − 𝑦2
lim 2 2
= lim = lim −1 = −1.
(𝑥,𝑦)→(0,0) 𝑥 + 𝑦 𝑦→0 0 + 𝑦 2 𝑦→0
𝑥=0

𝑥 4 −𝑦 2
The limit is different along two different paths of approach, so lim does not exist.
(𝑥,𝑦)→(0,0) 𝑥 2 +𝑦 2

(c)
𝑥 − 𝑦 + 2√ 𝑥 − 2√ 𝑦 (√𝑥 − √𝑦)(√𝑥 + √𝑦 + 2)
lim = lim
(𝑥,𝑦)→(0,0) √ 𝑥 − √𝑦 (𝑥,𝑦)→(0,0) √ 𝑥 − √𝑦

= lim (√𝑥 + √𝑦 + 2) = √0 + √0 + 2 = 2.
(𝑥,𝑦)→(0,0)

Page 2 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)
1
(d) Note that −1 ≤ sin ≤ 1 for every 𝑥 ≠ 0, so
𝑥

1
−|𝑦| ≤ 𝑦 sin ≤ |𝑦|
𝑥
for every 𝑥 ≠ 0. Now lim −|𝑦| = 0 and lim |𝑦| = 0, so by Squeeze theorem we also have
(𝑥,𝑦)→(0,0) (𝑥,𝑦)→(0,0)

1
lim 𝑦 sin = 0.
(𝑥,𝑦)→(0,0) 𝑥

(e) Along the line 𝑦 = 1, we have

𝑦 ln 𝑦 1 ln 1
lim = lim = lim 0 = 0,
(𝑥,𝑦)→(0,1) 𝑥 𝑥→0 𝑥 𝑥→0
𝑦=1

but along the line 𝑦 = 𝑥 + 1, we have

𝑦 ln 𝑦 (𝑥 + 1) ln(𝑥 + 1) ln(𝑥 + 1) + 1
lim = lim = lim (ľHôpital's rule)
(𝑥,𝑦)→(0,1) 𝑥 𝑥→0 𝑥 𝑥→0 1
𝑦=𝑥+1

= 1.

𝑦 ln 𝑦
The limit is different along two different paths of approach, so lim does not exist.
(𝑥,𝑦)→(0,1) 𝑥

(f) Along the 𝑥-axis, we have

𝑥 2 + sin2 𝑦 𝑥 2 + sin2 0 1 1
lim 2 2
= lim 2 2
= lim = ,
(𝑥,𝑦)→(0,0) 2𝑥 + 𝑦 𝑥→0 2𝑥 + 0 𝑥→0 2 2
𝑦=0

but along the 𝑦-axis, we have

𝑥 2 + sin2 𝑦 02 + sin2 𝑦 sin 𝑦 2
lim = lim = (lim ) = 12 = 1.
(𝑥,𝑦)→(0,0) 2𝑥 2 + 𝑦 2 𝑦→0 2(02 ) + 𝑦 2 𝑦→0 𝑦
𝑥=0

𝑥 2 +sin2 𝑦
The limit is different along two different paths of approach, so lim does not exist.
(𝑥,𝑦)→(0,0) 2𝑥 2 +𝑦 2

sin 𝑢 ≤ 𝑢 for 𝑢 ≥ 0.
𝑦 2 −sin2 𝑦 𝑑
(g) We first consider lim . Since (𝑡 2 − sin2 𝑡) = 2𝑡 − 2 sin 𝑡 cos 𝑡 = 2𝑡 − sin 2𝑡 ≥ 0 for 𝑡 ≥ 0,
(𝑥,𝑦)→(0,0) 𝑥 2 +𝑦 2 𝑑𝑡

we see that 𝑡 2 − sin2 𝑡 is an increasing function for 𝑡 ≥ 0. As a result, for every (𝑥, 𝑦) ≠ (0, 0), we have
the inequalities

|𝑦|2 − sin2 |𝑦| (𝑥 2 + 𝑦 2 ) − sin2 √𝑥 2 + 𝑦 2 |𝑦| ≤ √𝑥 2 + 𝑦 2

sin 𝑢 ≤ 𝑢 for 𝑢 ≥ 0. 0≤ ≤ .
𝑥 2 + 𝑦2 𝑥 2 + 𝑦2
Now since lim 0 = 0 and
(𝑥,𝑦)→(0,0)

(𝑥 2 + 𝑦 2 ) − sin2 √𝑥 2 + 𝑦 2 𝑟 2 − sin2 𝑟 sin 𝑟 2

lim = lim = 1 − ( lim ) = 1 − 12 = 0,
(𝑥,𝑦)→(0,0) 𝑥 2 + 𝑦2 𝑟→0+ 𝑟2 𝑟→0+ 𝑟

𝑦 2 −sin2 𝑦 |𝑦|2 −sin2 |𝑦|

we have lim = lim = 0 by Squeeze theorem. Therefore the required limit is
(𝑥,𝑦)→(0,0) 𝑥 2 +𝑦 2 (𝑥,𝑦)→(0,0) 𝑥 2 +𝑦 2

𝑥 2 + sin2 𝑦 𝑥 2 + 𝑦2 𝑦 2 − sin2 𝑦
lim 2 2
= lim 2 2
− lim = 1 − 0 = 1.
(𝑥,𝑦)→(0,0) 𝑥 + 𝑦 (𝑥,𝑦)→(0,0) 𝑥 + 𝑦 (𝑥,𝑦)→(0,0) 𝑥 2 + 𝑦 2

Page 3 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)

(h) Let’s convert the expression into polar coordinates. We have

|𝑥| + |𝑦| 𝑟|cos 𝜃| + 𝑟|sin 𝜃| 1
= = (|cos 𝜃| + |sin 𝜃|).
𝑥 2 + 𝑦2 𝑟2 𝑟
1 1
Since |cos 𝜃| + |sin 𝜃| ≥ cos 2 𝜃 + sin2 𝜃 = 1 for every 𝜃, we always have (|cos 𝜃| + |sin 𝜃|) ≥ . Now
𝑟 𝑟

1 |𝑥|+|𝑦| 1
since lim+ = +∞, we have lim = lim+ (|cos 𝜃| + |sin 𝜃|) = +∞ by Squeeze Theorem. So
𝑟→0 𝑟 (𝑥,𝑦)→(0,0) 𝑥 2 +𝑦 2 𝑟→0 𝑟

|𝑥| + |𝑦| 𝜋
lim arctan 2 2
= lim arctan 𝑡 = .
(𝑥,𝑦)→(0,0) 𝑥 +𝑦 𝑡→+∞ 2

(i) Let’s convert the expression into polar coordinates centered at (1, 0), i.e.

𝑥 − 1 = 𝑟 cos 𝜃
{ .
𝑦 = 𝑟 sin 𝜃

Then (𝑥, 𝑦) → (1, 0) is equivalent to 𝑟 → 0+ , and

𝑥𝑦 2 − 𝑦 2 (𝑥 − 1)𝑦 2 (𝑟 cos 𝜃)(𝑟 sin 𝜃)2
= = = 𝑟 cos 𝜃 sin2 𝜃 .
𝑥2 2 2
+ 𝑦 − 2𝑥 + 1 (𝑥 − 1) + 𝑦 2 𝑟2
Since −1 ≤ cos 𝜃 sin2 𝜃 ≤ 1 for every 𝜃, we always have −𝑟 ≤ 𝑟 cos 𝜃 sin2 𝜃 ≤ 𝑟. Now lim+ −𝑟 = 0 and
𝑟→0
lim+ 𝑟 = 0, so by Squeeze theorem we have
𝑟→0

𝑥𝑦 2 − 𝑦 2
lim = lim 𝑟 cos 𝜃 sin2 𝜃 = 0.
(𝑥,𝑦)→(1,0) 𝑥 2 + 𝑦 2 − 2𝑥 + 1 𝑟→0+

(j) Along the 𝑧-axis, we have

𝑥𝑦 + 𝑦𝑧 2 + 𝑥𝑧 2 (0)(0) + 0𝑧 2 + 0𝑧 2
lim = lim = lim 0 = 0,
(𝑥,𝑦,𝑧)→(0,0,0) 𝑥 2 + 𝑦 2 + 𝑧 4 𝑧→0 02 + 02 + 𝑧 4 𝑧→0
𝑥=0 and 𝑦=0

𝑥 = 𝑡2
but along the curve with parametric equation {𝑦 = 𝑡 2 (or the curve of intersection of the surfaces 𝑥 = 𝑧 2
𝑧=𝑡
and 𝑦 = 𝑧 2 ), we have
𝑥𝑦 + 𝑦𝑧 2 + 𝑥𝑧 2 (𝑧 2 )(𝑧 2 ) + (𝑧 2 )𝑧 2 + (𝑧 2 )𝑧 2
lim = lim = lim 1 = 1.
(𝑥,𝑦,𝑧)→(0,0,0) 𝑥 2 + 𝑦 2 + 𝑧 4 𝑧→0 (𝑧 2 )2 + (𝑧 2 )2 + 𝑧 4 𝑧→0
𝑥=𝑧 2 and 𝑦=𝑧 2

𝑥𝑦+𝑦𝑧 2 +𝑥𝑧 2
The limit is different along two different paths of approach, so lim does not exist.
(𝑥,𝑦,𝑧)→(0,0,0) 𝑥 2 +𝑦 2 +𝑧 4

(k) Converting into spherical coordinates, we have

𝑥 2𝑦2𝑧2 (𝜌 sin 𝜙 cos 𝜃)2 (𝜌 sin 𝜙 sin 𝜃)2 (𝜌 cos 𝜙)2

= = 𝜌4 sin4 𝜙 cos 2 𝜙 cos 2 𝜃 sin2 𝜃.
𝑥 2 + 𝑦2 + 𝑧2 𝜌2
Since 0 ≤ sin4 𝜙 cos 2 𝜙 cos 2 𝜃 sin2 𝜃 ≤ 1 for every 𝜙 and 𝜃, we always have
0 ≤ 𝜌4 sin4 𝜙 cos 2 𝜙 cos 2 𝜃 sin2 𝜃 ≤ 𝜌4 .
Now lim+ 0 = 0 and lim+ 𝜌4 = 0, so by Squeeze theorem we have
𝜌→0 𝜌→0

𝑥 2𝑦2𝑧2
lim = lim 𝜌4 sin4 𝜙 cos 2 𝜙 cos 2 𝜃 sin2 𝜃 = 0.
(𝑥,𝑦,𝑧)→(0,0,0) 𝑥 2 + 𝑦 2 + 𝑧 2 𝜌→0+

Page 4 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)

4. Since the paraboloid 𝑧 = 𝑥 2 + 𝑦 2 is the graph of the function 𝑓: ℝ2 → ℝ defined by

𝑓(𝑥, 𝑦) = 𝑥 2 + 𝑦 2 ,

𝜕𝑓
the required slope of tangent is represented by the partial derivative (1, 2), which is given by
𝜕𝑦

𝜕𝑓 𝑑 𝑑 2
(1, 2) = 𝑓(1, 𝑦)| = (1 + 𝑦 2 )| = 2(2) = 4.
𝜕𝑦 𝑑𝑦 𝑦=2
𝑑𝑦 𝑦=2

1
𝑥 2 sin if 𝑥 ≠ 0
5. Since 𝑓(𝑥, 0) = { 𝑥 , we have
0 if 𝑥 = 0
1
𝑑 𝑓(ℎ, 0) − 𝑓(0, 0) ℎ2 sin − 0 1
𝑓𝑥 (0, 0) = 𝑓(𝑥, 0)| = lim = lim ℎ = lim ℎ sin = 0
𝑑𝑥 𝑥=0 ℎ→0 ℎ ℎ→0 ℎ ℎ→0 ℎ
by Squeeze theorem arguments. On the other hand, since 𝑓(0, 𝑦) = 0 for every 𝑦, it follows trivially that
𝑑 𝑑
𝑓𝑦 (0, 0) = 𝑓(0, 𝑦)| = 0| = 0.
𝑑𝑦 𝑦=0
𝑑𝑦 𝑦=0
1
Finally since |sin | ≤ 1 for every 𝑥 ≠ 0, we have
𝑥
=0 =0 =0

|𝑓(𝑥, 𝑦) − ⏞
𝑓(0, 0) − ⏞
𝑓𝑥 (0, 0) (𝑥 − 0) − ⏞
𝑓𝑦 (0, 0) (𝑦 − 0)|
0≤
√(𝑥 − 0)2 + (𝑦 − 0)2
|𝑓(𝑥, 𝑦)| 2 2
1
= = {√𝑥 + 𝑦 |sin 𝑥 | if 𝑥 ≠ 0 ≤ √𝑥 2 + 𝑦 2
√𝑥 2 + 𝑦 2 0 if 𝑥 = 0

for every (𝑥, 𝑦) ≠ (0, 0). Since lim √𝑥 2 + 𝑦 2 = 0, we have

(𝑥,𝑦)→(0,0)

|𝑓(𝑥, 𝑦) − 𝑓(0, 0) − 𝑓𝑥 (0, 0)(𝑥 − 0) − 𝑓𝑦 (0, 0)(𝑦 − 0)|

lim =0
(𝑥,𝑦)→(0,0)
√(𝑥 − 0)2 + (𝑦 − 0)2
by Squeeze Theorem, and so 𝑓 is differentiable at (0, 0).

𝜕𝑓 𝑦
6. (a) For every (𝑥, 𝑦) we have (𝑥, 𝑦) = (𝑒 𝑥 )(𝑦𝑥 𝑦−1 ).
𝜕𝑥
𝜕𝑓 𝑦 −1
(b) For every (𝑥, 𝑦) we have (𝑥, 𝑦) = 𝑒 𝑦 𝑥 𝑒 .
𝜕𝑥
𝜕𝑓 𝑥𝑒
(c) For every (𝑥, 𝑦) we have (𝑥, 𝑦) = (𝑦 ln 𝑦)(𝑒𝑥 𝑒−1 ).
𝜕𝑥
(d) For every (𝑥, 𝑦) we have
𝜕𝑓 𝜕 cos 𝑥 3
(𝑥, 𝑦) = − ∫ √𝑡 + 1𝑑𝑡 = −√(cos 𝑥)3 + 1(− sin 𝑥) = (sin 𝑥)√cos 3 𝑥 + 1.
𝜕𝑥 𝜕𝑥 𝑦2
𝜕𝑓
(e) We compute at each point (𝑎, 𝑏) ∈ ℝ2 by considering the following three cases.
𝜕𝑥
(i) If 𝑎 < 𝑏, then 𝑓(𝑥, 𝑏) = 𝑒 𝑏−𝑥 for every 𝑥 near 𝑎, so
𝜕𝑓 𝑑 𝑑 𝑏−𝑥
(𝑎, 𝑏) = 𝑓(𝑥, 𝑏)| = 𝑒 | = −𝑒 𝑏−𝑎 .
𝜕𝑥 𝑑𝑥 𝑥=𝑎 𝑑𝑥 𝑥=𝑎

(ii) If 𝑎 > 𝑏, then 𝑓(𝑥, 𝑏) = 𝑒 𝑥−𝑏 for every 𝑥 near 𝑎, so

𝜕𝑓 𝑑 𝑑 𝑥−𝑏
(𝑎, 𝑏) = 𝑓(𝑥, 𝑏)| = 𝑒 | = 𝑒 𝑎−𝑏 .
𝜕𝑥 𝑑𝑥 𝑥=𝑎 𝑑𝑥 𝑥=𝑎

Page 5 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)

(iii) If 𝑎 = 𝑏, then we compute the partial derivative from definition. We have

𝜕𝑓 𝑓(𝑎 + ℎ, 𝑏) − 𝑓(𝑎, 𝑏) 𝑒 |ℎ| − 1

(𝑎, 𝑏) = lim = lim
𝜕𝑥 ℎ→0 ℎ ℎ→0 ℎ
which does not exist since

𝑒 |ℎ| − 1 𝑒ℎ − 1 𝑒 |ℎ| − 1 𝑒 −ℎ − 1
lim+ = lim+ =1 and lim− = lim− = −1.
ℎ→0 ℎ ℎ→0 ℎ ℎ→0 ℎ ℎ→0 ℎ
Combining all cases, we see that

𝜕𝑓 −𝑒 𝑦−𝑥 if 𝑥 < 𝑦
(𝑥, 𝑦) = { 𝑒 𝑥−𝑦 if 𝑥 > 𝑦.
𝜕𝑥
undefined if 𝑥 = 𝑦

𝜕𝑓
(f) We compute at each point (𝑎, 𝑏) ∈ ℝ2 by considering the following two cases.
𝜕𝑥
(i) If 𝑎 ≠ 𝑏, then

sin(𝑥 − 𝑏)
𝑓(𝑥, 𝑏) = for every 𝑥 near 𝑎,
𝑥−𝑏
so
𝜕𝑓 𝑑 sin(𝑥 − 𝑏) (𝑎 − 𝑏) cos(𝑎 − 𝑏) − sin(𝑎 − 𝑏) cos(𝑎 − 𝑏) sin(𝑎 − 𝑏)
(𝑎, 𝑏) = | = = − .
𝜕𝑥 𝑑𝑥 𝑥 − 𝑏 𝑥=𝑎 (𝑎 − 𝑏)2 𝑎−𝑏 (𝑎 − 𝑏)2

(ii) If 𝑎 = 𝑏, then we compute the partial derivative from definition. We have

sin ℎ
𝜕𝑓 𝑓(𝑎 + ℎ, 𝑏) − 𝑓(𝑎, 𝑏) −1
(𝑎, 𝑏) = lim = lim ℎ
𝜕𝑥 ℎ→0 ℎ ℎ→0 ℎ
sin ℎ − ℎ cos ℎ − 1 − sin ℎ
= lim = lim = lim =0
ℎ→0 ℎ2 ℎ→0 2ℎ ℎ→0 2
by ľHôpital’s rule.
Combining all cases, we see that

cos(𝑥 − 𝑦) sin(𝑥 − 𝑦)
− if 𝑥 ≠ 𝑦
𝑓𝑥 (𝑥, 𝑦) = { 𝑥 − 𝑦 (𝑥 − 𝑦)2 .
0 if 𝑥 = 𝑦

7. The partial derivative of 𝑓 with respect to 𝑦 at (𝑎, 𝑏, 𝑐) is defined by

𝜕𝑓 𝑓(𝑎, 𝑏 + ℎ, 𝑐) − 𝑓(𝑎, 𝑏, 𝑐)
(𝑎, 𝑏, 𝑐) ≔ lim .
𝜕𝑦 ℎ→0 ℎ
Applying this definition to the function 𝑓(𝑥, 𝑦, 𝑧) = 𝑥 2 𝑦 2 𝑧, we have

𝜕𝑓 𝑓(1, 2 + ℎ, 3) − 𝑓(1, 2, 3) (1)2 (2 + ℎ)2 (3) − (1)2 (2)2 (3)

(1, 2, 3) = lim = lim
𝜕𝑦 ℎ→0 ℎ ℎ→0 ℎ

3(2 + ℎ)2 − 3(2)2 3(ℎ2 + 4ℎ)

= lim = lim = lim 3(ℎ + 4)
ℎ→0 ℎ ℎ→0 ℎ ℎ→0

= 3(0 + 4) = 12.

Page 6 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)

8. Near (𝑎, 𝑏), we have the following observations:

 The value of 𝑓 decreases as the 𝑥-coordinate increases, so 𝑓𝑥 (𝑎, 𝑏) < 0. 𝑦 𝑓𝑦
 The value of 𝑓 increases as the 𝑦-coordinate increases, so 𝑓𝑦 (𝑎, 𝑏) > 0.
 The level curves are getting sparser as we move from left to right, so 7
6 𝑓𝑥
5
𝑓𝑥 gets less negative as the 𝑥-coordinate increases, and so 𝑓𝑥𝑥 (𝑎, 𝑏) > 0;
4
𝑓𝑦 gets less positive as the 𝑥-coordinate increases, and so 𝑓𝑦𝑥 (𝑎, 𝑏) < 0.
3
 The level curves are getting denser as we move from bottom to top, so
2 𝑥
𝑓𝑥 gets more negative as the 𝑦-coordinate increases, and so 𝑓𝑥𝑦 (𝑎, 𝑏) < 0; 1 0
𝑓𝑦 gets more positive as the 𝑦-coordinate increases, and so 𝑓𝑦𝑦 (𝑎, 𝑏) > 0.

9. (a) We have 𝑓(0, 0) = 0. For every (𝑥, 𝑦) ≠ (0, 0), converting into polar coordinates we have
𝑥𝑦(𝑥 2 − 𝑦 2 ) (𝑟 cos 𝜃)(𝑟 sin 𝜃)(𝑟 2 cos 2 𝜃 − 𝑟 2 sin2 𝜃) 𝑟 2
𝑓(𝑥, 𝑦) = = = sin 4𝜃 .
𝑥 2 + 𝑦2 𝑟2 4

𝑟2 𝑟2 𝑟2 𝑟2 𝑟2
Since −1 ≤ sin 4𝜃 ≤ 1 for every 𝜃, we always have − ≤ sin 4𝜃 ≤ . Now lim+ − = lim+ = 0,
4 4 4 𝑟→0 4 𝑟→0 4

so by Squeeze Theorem we also have

𝑟2
lim 𝑓(𝑥, 𝑦) = lim+ sin 4𝜃 = 0.
(𝑥,𝑦)→(0,0) 𝑟→0 4
The above shows that
lim 𝑓(𝑥, 𝑦) = 𝑓(0, 0),
(𝑥,𝑦)→(0,0)

so 𝑓 is continuous at (0, 0).

(b) Since 𝑓(𝑥, 0) = 0 for every 𝑥 and 𝑓(0, 𝑦) = 0 for every 𝑦, it follows that
𝑑 𝑑 𝑑 𝑑
𝑓𝑥 (0, 0) = 𝑓(𝑥, 0)| = 0| =0 and 𝑓𝑦 (0, 0) = 𝑓(0, 𝑦)| = 0| = 0.
𝑑𝑥 𝑥=0 𝑑𝑥 𝑥=0 𝑑𝑦 𝑦=0
𝑑𝑦 𝑦=0

(c) For every (𝑥, 𝑦) ≠ (0, 0), converting into polar coordinates we have
=0 =0 =0

|𝑓(𝑥, 𝑦) − ⏞
𝑓(0, 0) − ⏞
𝑓𝑥 (0, 0) (𝑥 − 0) − ⏞
𝑓𝑦 (0, 0) (𝑦 − 0)|

√(𝑥 − 0)2 + (𝑦 − 0)2

|𝑥𝑦(𝑥 2 − 𝑦 2 )| |(𝑟 cos 𝜃)(𝑟 sin 𝜃)(𝑟 2 cos 2 𝜃 − 𝑟 2 sin2 𝜃)| 𝑟
= 3 = = |sin 4𝜃|.
(𝑥 2 + 𝑟3 4
𝑦 2 )2
𝑟 𝑟 𝑟
Since 0 ≤ |sin 4𝜃| ≤ for every 𝜃 and since lim+ = 0, we have
4 4 𝑟→0 4
|𝑓(𝑥, 𝑦) − 𝑓(0, 0) − 𝑓𝑥 (0, 0)(𝑥 − 0) − 𝑓𝑦 (0, 0)(𝑦 − 0)| 𝑟
lim = lim+ |sin 4𝜃| = 0
(𝑥,𝑦)→(0,0) √(𝑥 − 0)2 + (𝑦 − 0)2 𝑟→0 4

by Squeeze Theorem, and so 𝑓 is differentiable at (0, 0).

Page 7 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)

(d) For every (𝑥, 𝑦) ≠ (0, 0) we have

(3𝑥 2 𝑦 − 𝑦 3 )(𝑥 2 + 𝑦 2 ) − (𝑥 3 𝑦 − 𝑥𝑦 3 )(2𝑥) 𝑥 4 𝑦 + 4𝑥 2 𝑦 3 − 𝑦 5
𝑓𝑥 (𝑥, 𝑦) = = ,
(𝑥 2 + 𝑦 2 )2 (𝑥 2 + 𝑦 2 )2

(04 )𝑦+4(02 )𝑦 3 −𝑦 5
so 𝑓𝑥 (0, 𝑦) = (02 +𝑦 2 )2
= −𝑦 for every 𝑦 ≠ 0. Combining with 𝑓𝑥 (0, 0) = 0 as obtained in (b), we

have 𝑓𝑥 (0, 𝑦) = −𝑦 for every 𝒚. Therefore

𝑑 𝑑
𝑓𝑥𝑦 (0, 0) = 𝑓 (0, 𝑦)| = (−𝑦)| = −1.
𝑑𝑦 𝑥 𝑦=0
𝑑𝑦 𝑦=0

On the other hand, for every (𝑥, 𝑦) ≠ (0, 0) we have

(𝑥 3 − 3𝑥𝑦 2 )(𝑥 2 + 𝑦 2 ) − (𝑥 3 𝑦 − 𝑥𝑦 3 )(2𝑦) 𝑥 5 − 4𝑥 3 𝑦 2 − 𝑥𝑦 4
𝑓𝑦 (𝑥, 𝑦) = = ,
(𝑥 2 + 𝑦 2 )2 (𝑥 2 + 𝑦 2 )2

𝑥 5 −4𝑥 3 (02 )−𝑥(04 )

so 𝑓𝑦 (𝑥, 0) = (𝑥 2 +02 )2
= 𝑥 for every 𝑥 ≠ 0. Combining with 𝑓𝑦 (0, 0) = 0 as obtained in (b), we

have 𝑓𝑦 (𝑥, 0) = 𝑥 for every 𝒙. Therefore

𝑑 𝑑
𝑓𝑦𝑥 (0, 0) = 𝑓𝑦 (𝑥, 0)| = 𝑥| = 1.
𝑑𝑥 𝑥=0 𝑑𝑥 𝑥=0
Now 𝑓𝑥𝑦 (0, 0) ≠ 𝑓𝑦𝑥 (0, 0), but this does not contradict the mixed derivatives theorem because 𝑓𝑥𝑦 and 𝑓𝑦𝑥
are not continuous at (0, 0). To see this, we check that for every (𝑥, 𝑦) ≠ (0, 0) we have

𝑥 6 + 9𝑥 4 𝑦 2 − 9𝑥 2 𝑦 4 − 𝑦 6
𝑓𝑥𝑦 (𝑥, 𝑦) = = 𝑓𝑦𝑥 (𝑥, 𝑦),
(𝑥 2 + 𝑦 2 )3

which has limit 1 as (𝑥, 𝑦) → (0, 0) along the 𝑥-axis but has limit −1 as (𝑥, 𝑦) → (0, 0) along the 𝑦-axis.

10. (a) We simplify the expression first. We have

1
𝑓(𝑥, 𝑦) = sin(𝑥 + 𝑦) cos(𝑥 − 𝑦) = (sin 2𝑥 + sin 2𝑦).
2
So

𝜕2𝑓 𝜕 𝜕𝑓 𝜕 1 𝜕
(𝑥, 𝑦) = [ (𝑥, 𝑦)] = [ (0 + 2 cos 2𝑦)] = (cos 2𝑦) = 0.
𝜕𝑥𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 2 𝜕𝑥

(b) Using the mixed derivatives theorem, we have

𝜕2𝑓 𝜕2𝑓 𝜕 𝜕𝑓 𝜕
(𝑥, 𝑦) = (𝑥, 𝑦) = [ (𝑥, 𝑦)] = (−𝑦 sin 𝑥𝑦 + 0) = − sin 𝑥𝑦 − 𝑥𝑦 cos 𝑥𝑦.
𝜕𝑥𝜕𝑦 𝜕𝑦𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦

(c) We simplify the expression before taking partial derivatives. We have

𝑒 𝑥+𝑦 + 𝑒 𝑥−𝑦 𝑒 𝑦 + 𝑒 −𝑦
𝑓(𝑥, 𝑦) = = .
𝑒 𝑥+𝑦 − 𝑒 𝑥−𝑦 𝑒 𝑦 − 𝑒 −𝑦
Therefore using the mixed derivatives theorem, we have

𝜕2𝑓 𝜕2𝑓 𝜕 𝜕𝑓 𝜕
(𝑥, 𝑦) = (𝑥, 𝑦) = [ (𝑥, 𝑦)] = (0) = 0.
𝜕𝑥𝜕𝑦 𝜕𝑦𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦

Page 8 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)

11. Suppose that there exists a function 𝑓: ℝ2 → ℝ which satisfies

𝑓𝑥 (𝑥, 𝑦) = 𝑦 and 𝑓𝑦 (𝑥, 𝑦) = 𝑥 2 for every (𝑥, 𝑦) ∈ ℝ2 .
Then
𝑓𝑥𝑦 (𝑥, 𝑦) = 1 and 𝑓𝑦𝑥 (𝑥, 𝑦) = 2𝑥 for every (𝑥, 𝑦) ∈ ℝ2 ,
so 𝑓𝑥𝑦 and 𝑓𝑦𝑥 are continuous on ℝ2 . So by mixed derivatives theorem, we have
𝑓𝑥𝑦 (𝑥, 𝑦) = 𝑓𝑦𝑥 (𝑥, 𝑦) for every (𝑥, 𝑦) ∈ ℝ2 .
But 𝑓𝑥𝑦 (0, 0) = 1 ≠ 0 = 𝑓𝑦𝑥 (0, 0), which is a contradiction. Therefore such a function 𝑓 does not exist. ∎

12. For every (𝑥, 𝑡) ∈ ℝ2 we have

𝜕 𝜕
𝑓𝑡 (𝑥, 𝑡) = 𝑔′ (𝑥 + 𝑐𝑡) (𝑥 + 𝑐𝑡) + ℎ′ (𝑥 − 𝑐𝑡) (𝑥 − 𝑐𝑡)
𝜕𝑡 𝜕𝑡
= 𝑐𝑔′ (𝑥 + 𝑐𝑡) − 𝑐ℎ′ (𝑥 − 𝑐𝑡),
and similarly

𝜕 𝜕 𝜕
𝑓𝑡𝑡 (𝑥, 𝑡) = 𝑓 (𝑥, 𝑡) = 𝑐𝑔′′ (𝑥 + 𝑐𝑡) (𝑥 + 𝑐𝑡) − 𝑐ℎ′′ (𝑥 − 𝑐𝑡) (𝑥 − 𝑐𝑡)
𝜕𝑡 𝑡 𝜕𝑡 𝜕𝑡
= 𝑐 2 𝑔′′ (𝑥 + 𝑐𝑡) + 𝑐 2 ℎ′′ (𝑥 − 𝑐𝑡).
On the other hand, we have

𝜕 𝜕
𝑓𝑥 (𝑥, 𝑡) = 𝑔′ (𝑥 + 𝑐𝑡) (𝑥 + 𝑐𝑡) + ℎ′ (𝑥 − 𝑐𝑡) (𝑥 − 𝑐𝑡) = 𝑔′ (𝑥 + 𝑐𝑡) + ℎ′ (𝑥 − 𝑐𝑡)
𝜕𝑥 𝜕𝑥
and so

𝜕 𝜕 𝜕
𝑓𝑥𝑥 (𝑥, 𝑡) = 𝑓𝑥 (𝑥, 𝑡) = 𝑔′′ (𝑥 + 𝑐𝑡) (𝑥 + 𝑐𝑡) + ℎ′′ (𝑥 − 𝑐𝑡) (𝑥 − 𝑐𝑡) = 𝑔′′ (𝑥 + 𝑐𝑡) + ℎ′′ (𝑥 − 𝑐𝑡).
𝜕𝑥 𝜕𝑥 𝜕𝑥
Now for every (𝑥, 𝑡) ∈ ℝ2 we have
𝑐 2 𝑓𝑥𝑥 (𝑥, 𝑡) = 𝑐 2 𝑔′′ (𝑥 + 𝑐𝑡) + 𝑐 2 ℎ′′ (𝑥 − 𝑐𝑡) = 𝑓𝑡𝑡 (𝑥, 𝑡),
so 𝑓 is a solution to the wave equation 𝑓𝑡𝑡 = 𝑐 2 𝑓𝑥𝑥 .

13. For every (𝑥, 𝑡) ∈ ℝ × (0, +∞) we have

1 1 3 𝑥2 1 𝑥2 𝑥 2 1 𝑥2 𝑥2 1
𝑓𝑡 (𝑥, 𝑡) = [(− 𝑡 −2 ) (𝑒 −4𝑘𝑡 ) + (𝑡 −2 ) (𝑒 −4𝑘𝑡 2
)] = 𝑒 −4𝑘𝑡 ( 2
− ).
√4𝜋𝑘 2 4𝑘𝑡 √4𝜋𝑘𝑡 4𝑘𝑡 2𝑡

𝑥2 𝑥2
1 2𝑥 1 1 𝑥
On the other hand, we have 𝑓𝑥 (𝑥, 𝑡) = 𝑒 −4𝑘𝑡 (− )= (− ) 𝑒 −4𝑘𝑡 and so
√4𝜋𝑘𝑡 4𝑘𝑡 𝑘 √4𝜋𝑘𝑡 2𝑡

1 1 1 𝑥2 𝑥 𝑥2 2𝑥 1 1 𝑥2 𝑥2 1
𝑓𝑥𝑥 (𝑥, 𝑡) = [(− ) (𝑒 −4𝑘𝑡 ) + (− ) (𝑒 −4𝑘𝑡 ) (− )] = 𝑒 −4𝑘𝑡 ( − ).
𝑘 √4𝜋𝑘𝑡 2𝑡 2𝑡 4𝑘𝑡 𝑘 √4𝜋𝑘𝑡 4𝑘𝑡 2 2𝑡

Now for every (𝑥, 𝑡) ∈ ℝ × (0, +∞) we have

1 𝑥2 𝑥2 1
𝑘𝑓𝑥𝑥 (𝑥, 𝑡) = 𝑒 −4𝑘𝑡 ( 2
− ) = 𝑓𝑡 (𝑥, 𝑡),
√4𝜋𝑘𝑡 4𝑘𝑡 2𝑡

so 𝑓 is a solution to the heat equation 𝑓𝑡 = 𝑘𝑓𝑥𝑥 .

Page 9 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)
14. (a) Taking partial derivatives with respect to 𝑥, we have 𝑢𝑥 (𝑥, 𝑦) = 2𝑥. So the first equation 𝑢𝑥 = 𝑣𝑦 gives
𝑣𝑦 (𝑥, 𝑦) = 2𝑥.
Taking antiderivatives on both sides we have

𝑣(𝑥, 𝑦) = ∫ 2𝑥 𝑑𝑦 = 2𝑥𝑦 + 𝐶(𝑥)

where 𝐶 is an arbitrary function that depends on 𝑥 only. Now taking partial derivatives with respect to 𝑥,
we have 𝑣𝑥 (𝑥, 𝑦) = 2𝑦 + 𝐶 ′ (𝑥). So the second equation 𝑣𝑥 = −𝑢𝑦 gives
𝑢𝑦 (𝑥, 𝑦) = −2𝑦 − 𝐶 ′ (𝑥).
But from 𝑢(𝑥, 𝑦) = 𝑥 2 − 𝑦 2 we have 𝑢𝑦 (𝑥, 𝑦) = −2𝑦 , so 𝐶 ′ (𝑥) = 0 and 𝐶 becomes a constant.
Therefore 𝑣(𝑥, 𝑦) = 2𝑥𝑦 + 𝐶, where 𝐶 is a constant.

(b) Taking partial derivatives of 𝑢 with respect to 𝑥, we have 𝑢𝑥 (𝑥, 𝑦) = 2 − 3𝑥 2 + 3𝑦 2 . So the first equation
𝑢𝑥 = 𝑣𝑦 gives
𝑣𝑦 (𝑥, 𝑦) = 2 − 3𝑥 2 + 3𝑦 2 .
Taking antiderivatives on both sides we have

𝑣(𝑥, 𝑦) = ∫(2 − 3𝑥 2 + 3𝑦 2 )𝑑𝑦 = 2𝑦 − 3𝑥 2 𝑦 + 𝑦 3 + 𝐶(𝑥)

where 𝐶 is an arbitrary function that depends on 𝑥 only. Now taking partial derivatives with respect to 𝑥,
we have 𝑣𝑥 (𝑥, 𝑦) = −6𝑥𝑦 + 𝐶 ′ (𝑥). So the second equation 𝑣𝑥 = −𝑢𝑦 gives
𝑢𝑦 (𝑥, 𝑦) = 6𝑥𝑦 − 𝐶 ′ (𝑥).
But from 𝑢(𝑥, 𝑦) = 2𝑥 − 𝑥 3 + 3𝑥𝑦 2 we have 𝑢𝑦 (𝑥, 𝑦) = 6𝑥𝑦, so 𝐶 ′ (𝑥) = 0 and 𝐶 becomes a constant.
Therefore 𝑣(𝑥, 𝑦) = 2𝑦 − 3𝑥 2 𝑦 + 𝑦 3 + 𝐶, where 𝐶 is a constant.

(c) Taking partial derivatives with respect to 𝑥, we have 𝑢𝑥 (𝑥, 𝑦) = 𝑒 𝑥 cos 𝑦. So the first equation 𝑢𝑥 = 𝑣𝑦
gives
𝑣𝑦 (𝑥, 𝑦) = 𝑒 𝑥 cos 𝑦.
Taking antiderivatives on both sides we have

𝑣(𝑥, 𝑦) = ∫ 𝑒 𝑥 cos 𝑦 𝑑𝑦 = 𝑒 𝑥 sin 𝑦 + 𝐶(𝑥)

where 𝐶 is an arbitrary function that depends on 𝑥 only. Now taking partial derivatives with respect to 𝑥,
we have 𝑣𝑥 (𝑥, 𝑦) = 𝑒 𝑥 sin 𝑦 + 𝐶 ′ (𝑥). So the second equation 𝑣𝑥 = −𝑢𝑦 gives
𝑢𝑦 (𝑥, 𝑦) = −𝑒 𝑥 sin 𝑦 − 𝐶 ′ (𝑥),
But from 𝑢(𝑥, 𝑦) = 𝑒 𝑥 cos 𝑦 we have 𝑢𝑦 (𝑥, 𝑦) = −𝑒 𝑥 sin 𝑦, so 𝐶 ′ (𝑥) = 0 and 𝐶 becomes a constant.
Therefore 𝑣(𝑥, 𝑦) = 𝑒 𝑥 sin 𝑦 + 𝐶, where 𝐶 is a constant.

Page 10 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)

15. The given circle can be parametrized by 𝑥 = cos 𝑡 and 𝑦 = sin 𝑡 where 𝑡 ∈ [0, 2𝜋], so we may let 𝑓(𝑡) be the
temperature at the point (cos 𝑡 , sin 𝑡) on the circle, so that
𝑓(𝑡) = 𝑇(cos 𝑡 , sin 𝑡)
for 𝑡 ∈ [0, 2𝜋]. Now
𝜕𝑇 𝑑𝑥 𝜕𝑇 𝑑𝑦
𝑓 ′ (𝑡) = +
𝜕𝑥 𝑑𝑡 𝜕𝑦 𝑑𝑡

= (8 cos 𝑡 − 4 sin 𝑡)(− sin 𝑡) + (8 sin 𝑡 − 4 cos 𝑡)(cos 𝑡)

= 4 sin2 𝑡 − 4 cos 2 𝑡 ,
𝜋 3𝜋 5𝜋 7𝜋
so 𝑓 ′ (𝑡) = 0 only when 4 sin2 𝑡 = 4 cos 2 𝑡, i.e. when 𝑡 = or 𝑡 = or 𝑡 = or 𝑡 = . Since
4 4 4 4
′′ (𝑡)
𝑓 = 16 sin 𝑡 cos 𝑡 = 8 sin 2𝑡 ,
𝜋 5𝜋 3𝜋 7𝜋
it follows that 𝑓 ′′ ( ) = 𝑓 ′′ ( ) > 0 and 𝑓 ′′ ( ) = 𝑓 ′′ ( ) < 0. Therefore by the second derivative test (for
4 4 4 4
3𝜋 7𝜋 𝜋 5𝜋
functions of one real variable), 𝑓 attains local maximum at and , and attains local minimum at and .
4 4 4 4

Since 𝑓 ′ is a periodic function with period 𝜋 and 𝑓(0) = 𝑓(2𝜋), it follows that 𝑓 also has period 𝜋 (how to
3𝜋 7𝜋 𝜋 5𝜋
prove this?) and so 𝑓 ( ) = 𝑓 ( ) and 𝑓 ( ) = 𝑓 ( ). Since [0, 2𝜋] is a closed and bounded interval, we
4 4 4 4
3𝜋 7𝜋 𝜋 5𝜋
conclude that 𝑓 attains absolute maximum at and , and attains absolute minimum at and .
4 4 4 4

1 1 1 1
Therefore the temperature on the circle is the highest at the points (− , ) and ( ,− ), and is the lowest at
√2 √2 √2 √2

1 1 1 1
the points ( , ) and (− ,− ).
√2 √2 √2 √2

16. Differentiating both sides of the equation 𝑓(𝑥, 𝑦, 𝑧) = 0 with respect to 𝑥 using chain rule, we get
𝑓𝑥 (𝑥, 𝑦, 𝑧) ⋅ 1 + 𝑓𝑦 (𝑥, 𝑦, 𝑧) ⋅ 0 + 𝑓𝑧 (𝑥, 𝑦, 𝑧)𝑧𝑥 (𝑥, 𝑦) = 0,

𝑓𝑥 (𝑥,𝑦,𝑧)
so provided that 𝑓𝑧 (𝑥, 𝑦, 𝑧) ≠ 0, we have 𝑧𝑥 (𝑥, 𝑦) = − . Similarly, differentiating both sides of the
𝑓𝑧 (𝑥,𝑦,𝑧)

equation 𝑓(𝑥, 𝑦, 𝑧) = 0 with respect to 𝑦 using chain rule, we get

𝑓𝑥 (𝑥, 𝑦, 𝑧) ⋅ 0 + 𝑓𝑦 (𝑥, 𝑦, 𝑧) ⋅ 1 + 𝑓𝑧 (𝑥, 𝑦, 𝑧)𝑧𝑦 (𝑥, 𝑦) = 0,

𝑓𝑦 (𝑥,𝑦,𝑧)
so provided that 𝑓𝑧 (𝑥, 𝑦, 𝑧) ≠ 0, we have 𝑧𝑦 (𝑥, 𝑦) = − . ∎
𝑓𝑧 (𝑥,𝑦,𝑧)

17. Differentiating both sides of the equation 𝐹(𝑠, 𝑡) = 𝑓(𝑥, 𝑦) with respect to 𝑠 using chain rule, we have
𝜕𝐹 𝜕𝑓 𝜕𝑥 𝜕𝑓 𝜕𝑦 𝜕𝑓 𝜕𝑓
= + = 2𝑠 + 2𝑡 .
𝜕𝑠 𝜕𝑥 𝜕𝑠 𝜕𝑦 𝜕𝑠 𝜕𝑥 𝜕𝑦

Differentiating both sides of the above equation with respect to 𝑡 using product rule, we have

𝜕2𝐹 𝜕 𝜕𝑓 𝜕𝑓 𝜕 𝜕𝑓 𝜕𝑓 𝜕 𝜕𝑓
= (2𝑠 + 2𝑡 ) = 2𝑠 ( ) + [2 + 2𝑡 ( )] .
𝜕𝑡𝜕𝑠 𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑡 𝜕𝑦

Page 11 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)

Now since chain rule gives

𝜕 𝜕𝑓 𝜕 𝜕𝑓 𝜕𝑥 𝜕 𝜕𝑓 𝜕𝑦 𝜕2𝑓 𝜕2𝑓
( ) = [ ( )] + [ ( )] = −2𝑡 2 + 2𝑠
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑡 𝜕𝑦 𝜕𝑥 𝜕𝑡 𝜕𝑥 𝜕𝑦𝜕𝑥
and
𝜕 𝜕𝑓 𝜕 𝜕𝑓 𝜕𝑥 𝜕 𝜕𝑓 𝜕𝑦 𝜕2𝑓 𝜕2𝑓
( ) = [ ( )] + [ ( )] = −2𝑡 + 2𝑠 2 ,
𝜕𝑡 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑡 𝜕𝑦 𝜕𝑦 𝜕𝑡 𝜕𝑥𝜕𝑦 𝜕𝑦
we obtain
𝜕2𝐹 𝜕2𝑓 𝜕2𝑓 𝜕𝑓 𝜕2𝑓 𝜕2𝑓
= 2𝑠 (−2𝑡 2 + 2𝑠 )+2 + 2𝑡 (−2𝑡 + 2𝑠 2 )
𝜕𝑡𝜕𝑠 𝜕𝑥 𝜕𝑦𝜕𝑥 𝜕𝑦 𝜕𝑥𝜕𝑦 𝜕𝑦

𝜕2𝑓 𝜕2𝑓 𝜕2𝑓 𝜕𝑓

= −4𝑠𝑡 2
+ 4𝑠𝑡 2 + (4𝑠 2 − 4𝑡 2 ) +2
𝜕𝑥 𝜕𝑦 𝜕𝑦𝜕𝑥 𝜕𝑦

𝜕2𝑓 𝜕2𝑓 𝜕2𝑓 𝜕𝑓

= −2𝑦 2
+ 2𝑦 2 + 4𝑥 +2 .
𝜕𝑥 𝜕𝑦 𝜕𝑦𝜕𝑥 𝜕𝑦

18. (a) In Example 3.60 we have obtained

𝜕𝑓 𝜕𝐹 sin 𝜃 𝜕𝐹 𝜕𝑓 𝜕𝐹 cos 𝜃 𝜕𝐹
= (cos 𝜃) − and = (sin 𝜃) + .
𝜕𝑥 𝜕𝑟 𝑟 𝜕𝜃 𝜕𝑦 𝜕𝑟 𝑟 𝜕𝜃

So

𝜕2𝑓 𝜕 𝜕𝑓 sin 𝜃 𝜕 𝜕𝑓
2
= (cos 𝜃) ( ) − ( )
𝜕𝑥 𝜕𝑟 𝜕𝑥 𝑟 𝜕𝜃 𝜕𝑥

𝜕 𝜕𝐹 sin 𝜃 𝜕𝐹 sin 𝜃 𝜕 𝜕𝐹 sin 𝜃 𝜕𝐹

= (cos 𝜃) [(cos 𝜃) − ]− [(cos 𝜃) − ]
𝜕𝑟 𝜕𝑟 𝑟 𝜕𝜃 𝑟 𝜕𝜃 𝜕𝑟 𝑟 𝜕𝜃

𝜕 2 𝐹 sin 𝜃 𝜕𝐹 sin 𝜃 𝜕 2 𝐹
= (cos 𝜃) [(cos 𝜃) + 2 − ]
𝜕𝑟 2 𝑟 𝜕𝜃 𝑟 𝜕𝑟𝜕𝜃

sin 𝜃 𝜕𝐹 𝜕 2 𝐹 cos 𝜃 𝜕𝐹 sin 𝜃 𝜕 2 𝐹

− [(− sin 𝜃) + (cos 𝜃) − − ]
𝑟 𝜕𝑟 𝜕𝜃𝜕𝑟 𝑟 𝜕𝜃 𝑟 𝜕𝜃 2

𝜕 2 𝐹 2 sin 𝜃 cos 𝜃 𝜕𝐹 2 sin 𝜃 cos 𝜃 𝜕 2 𝐹 sin2 𝜃 𝜕𝐹 sin2 𝜃 𝜕 2 𝐹

= (cos 2 𝜃) + − + + 2 ,
𝜕𝑟 2 𝑟2 𝜕𝜃 𝑟 𝜕𝑟𝜕𝜃 𝑟 𝜕𝑟 𝑟 𝜕𝜃 2
and similarly

𝜕2𝑓 𝜕 2 𝐹 2 sin 𝜃 cos 𝜃 𝜕𝐹 2 sin 𝜃 cos 𝜃 𝜕 2 𝐹 cos 2 𝜃 𝜕𝐹 cos 2 𝜃 𝜕 2 𝐹

2
= (sin2 𝜃) 2 − + + + .
𝜕𝑦 𝜕𝑟 𝑟2 𝜕𝜃 𝑟 𝜕𝑟𝜕𝜃 𝑟 𝜕𝑟 𝑟 2 𝜕𝜃 2
Therefore

𝜕 2 𝑓 𝜕 2 𝑓 𝜕 2 𝐹 1 𝜕𝐹 1 𝜕 2 𝐹
+ = + + .
𝜕𝑥 2 𝜕𝑦 2 𝜕𝑟 2 𝑟 𝜕𝑟 𝑟 2 𝜕𝜃 2

(b) We just outline the steps and omit the tedious calculation details. First we apply the chain rule to get

𝐹𝜌 = (sin 𝜙 cos 𝜃)𝑓𝑥 + (sin 𝜙 sin 𝜃)𝑓𝑦 + (cos 𝜙)𝑓𝑧

{𝐹𝜙 = (𝜌 cos 𝜙 cos 𝜃)𝑓𝑥 + (𝜌 cos 𝜙 sin 𝜃)𝑓𝑦 + (−𝜌 sin 𝜙)𝑓𝑧 ,
𝐹𝜃 = (−𝜌 sin 𝜙 sin 𝜃)𝑓𝑥 + (𝜌 sin 𝜙 cos 𝜃)𝑓𝑦

Page 12 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)
and next we solve for 𝑓𝑥 , 𝑓𝑦 and 𝑓𝑧 in the above system of linear equations to obtain

cos 𝜙 cos 𝜃 sin 𝜃

𝑓𝑥 = (sin 𝜙 cos 𝜃)𝐹𝜌 + ( ) 𝐹𝜙 + (− )𝐹
𝜌 𝜌 sin 𝜙 𝜃
cos 𝜙 sin 𝜃 cos 𝜃
𝑓𝑦 = (sin 𝜙 sin 𝜃)𝐹𝜌 + ( ) 𝐹𝜙 + ( )𝐹 .
𝜌 𝜌 sin 𝜙 𝜃
sin 𝜙
𝑓 = (cos 𝜙)𝐹𝜌 + (− ) 𝐹𝜙
{𝑧 𝜌

Therefore
𝜕 𝜕𝑓 cos 𝜙 cos 𝜃 𝜕 𝜕𝑓 sin 𝜃 𝜕 𝜕𝑓
𝑓𝑥𝑥 = (sin 𝜙 cos 𝜃) ( )+( ) ( ) + (− ) ( )
𝜕𝜌 𝜕𝑥 𝜌 𝜕𝜙 𝜕𝑥 𝜌 sin 𝜙 𝜕𝜃 𝜕𝑥

𝜕 cos 𝜙 cos 𝜃 sin 𝜃

= (sin 𝜙 cos 𝜃) [(sin 𝜙 cos 𝜃)𝐹𝜌 + ( ) 𝐹𝜙 + (− )𝐹 ]
𝜕𝜌 𝜌 𝜌 sin 𝜙 𝜃

cos 𝜙 cos 𝜃 𝜕 cos 𝜙 cos 𝜃 sin 𝜃

+( ) [(sin 𝜙 cos 𝜃)𝐹𝜌 + ( ) 𝐹𝜙 + (− )𝐹 ]
𝜌 𝜕𝜙 𝜌 𝜌 sin 𝜙 𝜃

sin 𝜃 𝜕 cos 𝜙 cos 𝜃 sin 𝜃

+ (− ) [(sin 𝜙 cos 𝜃)𝐹𝜌 + ( ) 𝐹𝜙 + (− )𝐹 ]
𝜌 sin 𝜙 𝜕𝜃 𝜌 𝜌 sin 𝜙 𝜃

=⋯
sin2 𝜃 + cos 2 𝜙 cos 2 𝜃 sin2 𝜃 − 2 sin2 𝜙 cos 2 𝜃
=( ) 𝐹𝜌 + (sin2 𝜙 cos 2 𝜃)𝐹𝜌𝜌 + ( ) 𝐹𝜙
𝜌 𝜌2 tan 𝜙

cos 2 𝜙 cos 2 𝜃 2 sin 𝜃 cos 𝜃 sin2 𝜃 2 sin 𝜙 cos 𝜙 cos 2 𝜃

+( ) 𝐹𝜙𝜙 + ( ) 𝐹𝜃 + ( ) 𝐹𝜃𝜃 + ( ) 𝐹𝜌𝜙
𝜌2 𝜌2 sin2 𝜙 𝜌2 sin2 𝜙 𝜌

2 sin 𝜃 cos 𝜃 2 sin 𝜃 cos 𝜃

+ (− ) 𝐹𝜌𝜃 + (− ) 𝐹𝜙𝜃 .
𝜌 𝜌2 tan 𝜙

In a similar way, we obtain

cos 2 𝜃 + cos 2 𝜙 sin2 𝜃 cos 2 𝜃 − 2 sin2 𝜙 sin2 𝜃
𝑓𝑦𝑦 = ( ) 𝐹𝜌 + (sin2 𝜙 sin2 𝜃)𝐹𝜌𝜌 + ( ) 𝐹𝜙
𝜌 𝜌2 tan 𝜙

cos 2 𝜙 sin2 𝜃 2 sin 𝜃 cos 𝜃 cos 2 𝜃

+( 2
) 𝐹𝜙𝜙 + (− 2 2 ) 𝐹𝜃 + ( 2 2 ) 𝐹𝜃𝜃
𝜌 𝜌 sin 𝜙 𝜌 sin 𝜙

2 sin 𝜙 cos 𝜙 sin2 𝜃 2 sin 𝜃 cos 𝜃 2 sin 𝜃 cos 𝜃

+( ) 𝐹𝜌𝜙 + ( ) 𝐹𝜌𝜃 + ( 2 ) 𝐹𝜙𝜃
𝜌 𝜌 𝜌 tan 𝜙

and
sin2 𝜙 2 sin 𝜙 cos 𝜙 sin2 𝜙 2 sin 𝜙 cos 𝜙
𝑓𝑧𝑧 = ( ) 𝐹𝜌 + (cos 2 𝜙)𝐹𝜌𝜌 + ( 2
) 𝐹𝜙 + ( 2
) 𝐹𝜙𝜙 + (− ) 𝐹𝜌𝜙 .
𝜌 𝜌 𝜌 𝜌

Summing these three expressions yields

2 1 1 1
𝑓𝑥𝑥 + 𝑓𝑦𝑦 + 𝑓𝑧𝑧 = 𝐹𝜌 + 𝐹𝜌𝜌 + 2 𝐹𝜙 + 2 𝐹𝜙𝜙 + 2 2 𝐹𝜃𝜃
𝜌 𝜌 tan 𝜙 𝜌 𝜌 sin 𝜙

1 𝜕 1 𝜕 1
= (𝜌2 𝐹𝜌 ) + 2 (sin 𝜙 𝐹𝜙 ) + 2 2 𝐹𝜃𝜃
𝜌2 𝜕𝜌 𝜌 sin 𝜙 𝜕𝜙 𝜌 sin 𝜙

as desired. ∎

Page 13 of 14
MATH2023 Multivariable Calculus Problem Set 3
L2/L3 (Fall 2019)
𝑢
19. (a) Let 𝐺(𝑢, 𝑥) = ∫0 √𝑡 4 + 𝑥 3 𝑑𝑡. Then

𝜕𝐺
= √𝑢 4 + 𝑥 3
𝜕𝑢
by the Fundamental Theorem of Calculus, and
𝜕𝐺 𝑢
𝜕 𝑢
3𝑥 2
= ∫ ( √𝑡 4 + 𝑥 3 ) 𝑑𝑡 = ∫ 𝑑𝑡 .
𝜕𝑥 0 𝜕𝑥 4
0 2√𝑡 + 𝑥
3

Now let 𝑢 = 𝑥 2 . Then 𝐹(𝑥) = 𝐺(𝑢(𝑥), 𝑥), so by the chain rule we have
𝜕𝐺 𝑑𝑢 𝜕𝐺 𝑢
3𝑥 2
𝐹 ′ (𝑥) = + = (√𝑢4 + 𝑥 3 ) (2𝑥) + ∫ 𝑑𝑡
𝜕𝑢 𝑑𝑥 𝜕𝑥 4
0 2√𝑡 + 𝑥
3

2
3𝑥 2 𝑥 1
= 2𝑥 √𝑥 8 + 𝑥3 + ∫ 𝑑𝑡.
2 0 √𝑡 4 + 𝑥 3

1
(b) Let 𝐺(𝑢, 𝑥) = ∫𝑢 √𝑡 3 + 𝑥 2 𝑑𝑡. Then

𝜕𝐺
= −√𝑢3 + 𝑥 2
𝜕𝑢
by the Fundamental Theorem of Calculus, and
1 1
𝜕𝐺 𝜕 𝑥
= ∫ ( √𝑡 3 + 𝑥 2 ) 𝑑𝑡 = ∫ 𝑑𝑡 .
𝜕𝑥 𝑢 𝜕𝑥
3
𝑢 √𝑡 + 𝑥
2

Now let 𝑢 = 𝑥 2 . Then 𝐹(𝑥) = 𝐺(𝑢(𝑥), 𝑥), so by the chain rule we have
1
𝜕𝐺 𝑑𝑢 𝜕𝐺 𝑥
𝐹 ′ (𝑥) = + = (−√𝑢3 + 𝑥 2 ) (2𝑥) + ∫ 𝑑𝑡
𝜕𝑢 𝑑𝑥 𝜕𝑥 3
𝑢 √𝑡 + 𝑥
2

1
1
= −2𝑥 √𝑥 6 + 𝑥 2 + 𝑥 ∫ 𝑑𝑡 .
𝑥2 √𝑡 3 + 𝑥2

Page 14 of 14