University of the Witwatersrand

School of Mathematics

MATH2007 (Multivariable Calculus)

Tutorial solutions

Note: there may be more than one way to solve a problem, and there may be many equivalent
solutions. If you are not sure whether your own attempt is correct, please consult a tutor or lecturer!

Section 1.1: Derivatives and Differentials

1. This question revolves around “lifting” partial derivatives to ordinary derivatives in order to
show that some basic properties of ordinary derivatives also hold for partial derivatives (i.e.
linearity and the product rule):

∂ product rule ∂f ∂g
fg g +f
∂x1 ∂x1 ∂x1

d product rule dϕ dψ
ϕψ ψ +ϕ
dt dt dt

(a) Since ϕ(t) = f (t, x2 , . . . , xn ) = f (x)x1 =t , differentiation of ϕ with respect to t (where x2 ,
. . . , xn are treated as constant) is the same as partial differentiation of f with respect to
x1 :
ϕ(t + h) − ϕ(t)
ϕ0 = lim
h→0 h
f (t + h, x2 , . . . , xn ) − f (t, x2 , . . . , xn )
= lim
h→0
 h 
f (x1 + h, x2 , . . . , xn ) − f (x1 , x2 , . . . , xn )
= lim
h→0 h x1 =t

∂f 
= .
∂x1  x1 =t

The equality shown above is independent of the choice of f . So, for example, we alse have

0 ∂g 
ψ = .
∂x1 x1 =t
We will use this fact in the following answers.
(b) In the same way as part (a), we have

∂0

(aϕ + bψ) = (af + bg) (1)
∂x1 x1 =t

where (af +bg)(x) := af (x)+bg(x) is the usual point-wise addition and scalar multiplication
of functions. On the other hand
 
0 0 0 ∂f  ∂g 
(aϕ + bψ) = aϕ + bψ = a +b (2)
∂x1  ∂x1  x1 =t x1 =t

from part (a). The left hand sides of (1) and (2) are identical, and equating the right hand
sides yields   
∂  ∂f  ∂g 
(af + bg)
 =a +b .
∂x1 x1 =t ∂x1 x1 =t ∂x1 x1 =t

1
 Since t is arbitrary, we may take t = x1 so that
∂ ∂f ∂g
(af + bg) = a +b .
∂x1 ∂x1 ∂x1

(c) Similar to part (b), we have


0 ∂ 
(ϕψ) = f g  ,
∂x1 x1 =t
 
0 0 ∂f 
0 ∂g 
(ϕψ) = ψϕ + ϕψ = g +f .
∂x1 x1 =t ∂x1 x1 =t
Equating right hand sides of the above two equations and taking t = x1 yields
∂ ∂f ∂g
fg = g +f .
∂x1 ∂x1 ∂x1

2. (a) We have two curves, r1 (s) = (s, s2 )T and r2 (t) = (t2 , t)T . The curves intersect when
(s, s2 ) = (t2 , t), in other words t = s2 = (t2 )2 = t4 . Thus we have the equation t − t4 =
t(1 − t3 ) = 0 which has the solution t = 0 or t = 1 (we consider only the real valued
solutions). When t = 0, we have s = t2 = 0 and s = 0. When t = 1, we have s = 1. Thus
the two curves intersect at (0, 0)T (i.e. when t = s = 0) and at (1, 1)T (i.e. when t = s = 1).
If the curves have perpendicular tangents at the points of intersection, then the curves are
orthogonal at the points of intersection. The tangent directions are:
   
0 1 0 2t
r1 (s) = , r2 (t) = .
2s 1
At the intersection point (0, 0)T we have s = t = 0 and the dot product
   
1 0
r01 (0) · r02 (0) = · =0
0 1
is zero, i.e. the two curves are orthogonal at (0, 0)T . At the intersection point (1, 1)T we
have s = t = 1 and the dot product
   
1 2
r01 (1) · r02 (1) = · = 4 6= 0
2 1
is non-zero, i.e. the two curves are not orthogonal at (1, 1)T .
(b) We have two curves, r1 (s) = (s, s2 )T and r2 (t) = (t, 1/2 − t2 )T . The curves intersect when
(s, s2 ) = (t, 1/2 − t2 ), in other words s = t and s2 = t2 = 12 − t2 . Thus s = t = 1/2 or
s = t = −1/2. Thus the two curves intersect at (1/2, 1/4)T (i.e. when t = s = 1/2) and
at (−1/2, 1/4)T (i.e. when t = s = −1/2). If the curves have perpendicular tangents at
the points of intersection, then the curves are orthogonal at the points of intersection. The
tangent directions are:    
1 1
r01 (s) = , r02 (t) = .
2s −2t
At the intersection point (1/2, 1/4)T we have s = t = 1/2 and the dot product
   
0 0 1 1
r1 (1/2) · r2 (1/2) = · =0
1 −1
is zero, i.e. the two curves are orthogonal at (1/2, 1/4)T . At the intersection point
(−1/2, −1/2)T we have s = t = −1/2 and the dot product
   
0 0 1 1
r1 (−1/2) · r2 (−1/2) = · =0
−1 1
is zero, i.e. the two curves are orthogonal at (−1/2, 1/4)T .

2
3. A circle with radius 5 centered at the origin has the parameterization (5 cos t, 5 sin t)T . The
same circle, but with center (8, −3) is obtained by adding (8, −3) to every point on the circle (a
translation). Thus the parameterized circle with radius 5 and center (8, −3) is
 
5 cos t + 8
r(t) = .
5 sin t − 3

The point (5, 1) is on the circle, since

(5 − 8)2 + (1 − (−3))2 = 52 .

Let the point (5, 1) be obtained at t = t∗ . Our parameterization yields 5 = 5 cos t∗ + 8 and
1 = 5 sin t∗ − 3, or cos t∗ = −3/5 and sin t∗ = 4/5. A tangent vector at (5, 1) is
   
0 −5 sin t∗ −4
r (t∗ ) = = .
5 cos t∗ −3

The tangent line at (5, 1) is (where s is a real parameter)

   
0 5 −4
T(5,1) (s) = r(t∗ ) + sr (t∗ ) = +s .
1 −3

4. We have
 
0 ∂f ∂x

f (x1 , x2 ) = = −x2 sin x1 cos x1
∂x1 ∂x2
0


f (2, 7) = −7 sin 2 cos 2
   
0 h1
h1
df [(2, 7); h] = f (2, 7) = −7 sin 2 cos 2 = −(7 sin 2)h1 + (cos 2)h2
h2 h2

where h = (h1 , h2 )T ∈ R2 .
5. We have
 
0 ∂f ∂x ∂x
= ey (x − 2z)ey −2ey


f (x, y, z) =
∂x ∂y ∂z
   
h1 h1
df [(x, y, z); h] = f 0 (x, y, z) h2  = ey (x − 2z)ey −2ey h2  = ey h1 + (x − 2z)ey h2 − 2ey h3

h3 h3

where h = (h1 , h2 , h3 )T ∈ R3 .
6. (a) First, note that v is a tangent vector to r(t) (at r(t)) if and only if v is non-zero and parallel
to r0 (t). The derivative is
 
∂F1 ∂F1
1
0
 
 ∂x1 ∂x2  x
  1

0
 ∂F2 ∂F2   
F (x1 , x2 ) = 
  =  2x2 e2x1 e 2x1 
 ∂x1 ∂x2 
  
 ∂F3 ∂F3  2
−x2 sin x1 2x2 cos x1
∂x1 ∂x2
so that
 1   h1 
    a 0   a
a1 h 1 h1 1
dF ; 1 =  2a2 e2a1 e2a1  = 2a2 h1 e2a1 + e2a1 h2 .
a2 h2 2 h2 2
−a2 sin a1 2a2 cos a1 −a2 h1 sin a1 + 2a2 h2 cos a1

3
 If v is a tangent vector for curve r(t) at r(t), then v is parallel to r0 (t), i.e. v = αr0 (t) =
(0, α)T for some α ∈ R. A tangent vector for the curve
 
  0
1
F(r(t)) = F =  te2 
t
t2 cos 1

at F(r(t)) is  
0
(F(r(t)))0 =  e2  .
2t cos 1
Now,    
    0 0
1 0
dF [(r(t); v] = dF ; = αe2  = α  e2 
t α
2αt cos 1 2t cos 1
which is a tangent vector (parallel to (F(r(t)))0 ) to the curve F(r(t)) at F(r(t)).
(b) The derivative is
 
∂F1 ∂F1 ∂F1  
1 0 0
∂x ∂y ∂z   y
F0 (x, y, z) = 
 
= x 2xyz 
 ∂F2 ∂F2 ∂F1  −
1 + z2 1 + z2 (1 + z 2 )2
∂x ∂y ∂z

so that
      
a1 h1 1 0 0 h1
dF a2 ; h2  =  a2
  a1 2a1 a2 a3  h2 
−
a3 h3 1 + a23 1 + a23 (1 + a23 )2 h3
 
h1
=  (a2 h1 + a1 h2 )(1 + a23 ) − 2a1 a2 a3 h3  .
(1 + a23 )2

If v is a tangent vector for curve r(t) at r(t), then v is parallel to r0 (t), i.e. v = αr0 (t) =
(0, 2αt, 0)T for some α ∈ R. A tangent vector for the curve
 
2  
1
F(r(t)) = F t2  = 2
t
1

at F(r(t)) is  
0 0
(F(r(t))) = .
2t
Now,    
2 0    
2 0 0
dF [(r(t); v] = dF  t ; 2αt
   = =α
2αt 2t
1 0
which is a tangent vector (parallel to (F(r(t)))0 ) to the curve F(r(t)) at F(r(t)).
7. We write F(x) = (F1 (x), . . . , Fm (x))T and G(x) = (G1 (x), . . . , Gm (x))T .

4
 (a) By linearity of the partial derivatives (see question 1)

∂ ∂
 
(αF1 + βG1 ) · · · (αF1 + βG1 )
 ∂x1 ∂xn 
0
(αF + βG) = 

 .
.. . .. .
..



 ∂ ∂ 
(αFm + βGm ) · · · (αFm + βGm )
∂x ∂xn
 1
∂F1 ∂G1 ∂F1 ∂G1

α +β ··· α +β
 ∂x1 ∂x1 ∂xn ∂xn 
=

 .
.. . .. .
..



 ∂Fm ∂Gm ∂Fm ∂Gm 
α +β ··· α +β
∂x1 ∂x1 ∂xn ∂xn
∂F1 ∂F1 ∂G1 ∂G1
   
··· ···
 ∂x1 ∂xn   ∂x1 ∂xn 
 . . .   . . .. 
= α  ..
 .. ..  + β  ..
  .. . 

 ∂Fm ∂Fm   ∂Gm ∂Gm 
··· ···
∂x1 ∂xn ∂x1 ∂xn
= αF0 + βG0 .

(b) By the product rule for partial derivatives (see question 1)

(gF)0 = (gF1 , gF2 , . . . , gFm )0

∂ ∂
 
(gF1 ) · · · (gF1 )
 ∂x1 ∂xn 
=
 .. .. .. 
 . . .


 ∂ ∂ 
(gFm ) · · · (gFm )
∂x ∂xn
 1
∂g ∂F1 ∂g ∂F1

F +g · · · F1 +g
 1 ∂x1 ∂x1 ∂xn ∂xn 
=
 .. .. .. 
 . . .


 ∂g ∂Fm ∂g ∂Fm 
Fm +g · · · Fm +g
∂x1 ∂x1 ∂xn ∂xn

8. (a) We have
 
0 ∂ϕ ∂ϕ ∂ϕ
= −2 2x2 x3 x22 ,


ϕ =
∂x1 ∂x2 ∂x3
∂F1 ∂F1 ∂F1
 
−2e−2x1 5x53 25x2 x43
 
0  ∂x1 ∂x2 ∂x3 
F =  ∂F ∂F2 ∂F2  = .
2 −2 1 cos x3
∂x1 ∂x2 ∂x3

(b) Theorem 1.1.4 provides

(ϕF)0 = ϕF0 + Fϕ0

−2e−2x1 5x53 25x2 x43 e−2x1 + 5x2 x53
   
= (x3 x22 − 2x1 ) −2 2x2 x3 x22 .


+
−2 1 cos x3 x2 − 2x1 + sin x3

Now try calculate (ϕF)0 directly and compare your answer with this result!

5
 9. Here we consider linearity in the second argument (i.e. h):

dF[a; αh + βg] = F0 (a)(αh + βg)

= αF0 h + βF0 g (matrix multiplication is linear)
= αdF[a; h] + βdF[a; g]

for all α, β ∈ R and h, g ∈ Rn .

10. The results follows immediately from Theorem 1.1.4:

d(gF)[a; h] = (gF)0 (a)h

= (g(a)F0 (a) + F(a)g 0 (a))h
= g(a)F0 (a)h + F(a)g 0 (a)h
= g(a)dF[a; h] + F(a)dg[a; h].

n
X
11. (a) Noting that u · v = uj vj , we obtain
j=1
 0
n
X n
X n
X n
X n
X
0 0 0 0 0
(u · v) =  uj vj  = (uj vj ) = (uj vj + uj vj ) = uj vj + uj vj0 = u0 · v + u · v0 .
j=1 j=1 j=1 j=1 j=1

(b) We wish to make a statement about v · p(t) for all t ∈ R. It is given that v · p0 (t) = 0 for
all t ∈ R. Since v is constant

(v · p(t))0 = v0 · p(t) + v · p0 (t) = v · p0 (t) = 0,

it follows that v · p(t) is constant (i.e. the same value for all t ∈ R). It is also given that
v · p(0) = 0, which is the (constant) value v · p(t) = 0 for all t ∈ R. Thus v and p(t) are
orthogonal for all t ∈ R.
p
(c) Since u(t) is a unit vector, ku(t)k = u(t) · u(t) = 1. It suffices to consider u(t) · u(t) = 1
for all t (can you explain why?). Then we have, for all t,

(u(t) · u(t))0 = 10

if and only if
u0 (t) · u(t) + u(t) · u0 (t) = 0

if and only if
2u0 (t) · u(t) = 0.

Since u0 (t) · u(t) = 0 for all t, u(t) and u0 (t) are orthogonal for all t.
(d) Notice that u(t) is a unit vector for all t, hence part (c) applies. Straightforward calculation
yields
2 2
ϕ0 = e−t − 2t2 e−t
 
0 −2 sin 2t
u =
2 cos 2t
2 2
(ϕu)0 = uϕ0 + ϕu0 = (e−t − 2t2 e−t )u + tet u0

and the last derivative (by part (c)) is a sum of two orthogonal vectors.
 
z cos θ −rz sin θ r cos θ
12. (a) We have F0 =  z sin θ rz cos θ r sin θ .
0 0 1

6
 (b) It follows that det(F0 ) = rz 2 cos2 θ + rz 2 sin2 θ = rz 2 .
   
4t 4
0 00 0 0
13. (a) The velocity is r (t) = 2t − 4 and the acceleration is r (t) = (r (t)) = 2.
  
3 0
(b) The component of acceleration in the direction of the velocity is given by a projection
   
4t 4
r0 (t) 00 1 2t − 4 · 2 = √ 20t − 8
· r (t) = .
kr0 (t)k
p
16t2 + (2t − 4)2 + 9 3 0 20t2 − 16t + 25

14. We note that r(t) = ρ(t)u(t), where u(t) = (cos t, sin t)T is a unit vector (so that question 11 (c)
applies).
   
0 0 0 0 cos t − sin t
(a) We apply the product rule, r (t) = u(t)ρ (t) + ρ(t)u (t) = ρ (t) + ρ(t) .
sin t cos t
(b) Linearity of differentiation yields
       
00 0 0 00 cos t 0 − sin t 0 − sin t − cos t
r (t) = (r (t)) = ρ (t) + ρ (t) + ρ (t) + ρ(t) .
sin t cos t cos t − sin t

Grouping the u and u0 terms yields the desired result

   
cos t − sin t
r00 (t) = (ρ00 (t) − ρ(t)) + 2ρ0 (t) .
sin t cos t

(c) For all a, b ∈ R, we have that

p
kau + bu0 k =
p
(au + bu0 ) · (au + bu0 ) = a2 + b2

since u and u0 are unit vectors and orthogonal to each other. It follows from r0 (t) =
ρ0 (t)u(t) + ρ(t)u0 (t), that
kr0 (t)k = (ρ0 (t))2 + (ρ(t))2 .
p

15. If r(t) lies on the circle of radius R with center 0 for all t ∈ R, then kr(t)k = R is constant.
Differentiating r(t) · r(t) = R2 twice yields

(r(t) · r(t))0 = 2r0 (t) · r(t) = 0

(r(t) · r(t))00 = 2r00 (t) · r(t) + 2r0 (t) · r0 (t) = 0

Solving the last equation for kr0 (t)k2 = r0 (t) · r0 (t) yields

kr0 (t)k = −r00 (t) · r(t).

p

16. Applying Theorem 1.1.5, we find

ω 0 = (r × r0 )0 = r0 × r0 + r × r00 = r × r00

where we used that r0 × r0 = 0 (the cross product of any vector with itself is the zero vector).
17. (a) In terms of limits,
h(t + s) − h(t)
h0 (t) = lim
s→0 s
f (x1 + t + s, x2 ) − f (x1 + t, x2 )
= lim
s→0
 s
∂f  ∂f
= = (x1 + t, x2 ).
∂x1 x1 →x1 +t ∂x1

7
 ∂f
(b) If = 0 for all x ∈ R2 , then h0 (t) = 0 for all t ∈ R. It follows that
∂x1
Z b
h(b) − h(0) = h0 (t) dt = 0.
0

Clearly, h(b) = h(0). In other words, f (x1 + b, x2 ) = f (x1 , x2 ) for all b ∈ R including
b = −x1 which provides f (0, x2 ) = f (x1 , x2 ).
18. Since F · F = 1 for all x ∈ Rn , we have (Theorem 1.1.4)

(F · F)0 = 2FT F0 = 0.

Now assume that det(F0 ) 6= 0 so that (F0 )−1 exists. Then,

2FT F0 (F0 )−1 = 0 (F0 )−1 = 0.

It follows that FT = 0 and kFk = 0. But kFk = 1, so our assumption that det(F0 ) 6= 0 was
false; i.e. det(F0 ) = 0.

Section 1.2: Vector analysis


  
∂f 2x1
  2
  x + x22 
 
1. (a) ∇f =  ∂x1  =  1 .

 ∂f   2x2 
∂x2 x21 + x22
   
∂f
 −x2 ex1 x2 + cos(x1 − x2 )
 ∂x1  


   
(b) ∇f =  ∂f  = −x1 ex1 x2 − cos(x1 − x2 ) .
   
 ∂x2   
   
 ∂f   
2x3
∂x3
 
x1 x3 − x2 cos x3
2. (a) G = curl F =  −x2 x3  and div F = ex2 + sin x3 + x1 x2 .
−x1 ex2
(b) Theorem 1.2.7 provides ∇ · G = ∇ · (∇ × F) = 0. Check this fact by direct calculation! (See
also Section 1.2, question 7).
3.  
1
curl F = x1 − 1 ,
0
div F = x3 + ex2 ,
∇ · (∇ × F) = div curl F = 0,
   
x1 x3 1
F · ((∇ × F) =  ex2  · x1 − 1 = x1 x3 + (x1 − 1)ex2 ,
x1 + x2 0
 
0
∇ × (∇F) = 0 .
1

∂2u ∂2u ∂2u ∂2u

4. (i) Since ∇2 u = + + + = 0 − ex3 sin x2 + ex3 sin x2 + 0 = 0, u is harmonic.
∂x1 ∂x2 ∂x3 ∂x4

8
 ∂2u ∂2u ∂2u
(ii) Since ∇2 u = + + = 6x3 + 0 − 6x3 = 0, u is harmonic.
∂x1 ∂x2 ∂x3
∂2u ∂2u
(iii) Since ∇2 u = + = 2 + 2 6= 0, u is not harmonic.
∂x1 ∂x2
5. By straightforward application of linearity and the product rule for partial derivatives.
6. (a) We have

fh (y + k) − fh (y) = ψ(x + h, y + k) − ψ(x, y + k) − ψ(x + h, y) + ψ(x, y)

and
gk (x + h) − gk (x) = ψ(x + h, y + k) − ψ(x + h, y) − ψ(x, y + k) + ψ(x, y).
Clearly fh (y + k) − fh (y) = gk (x + h) − gk (x).
(b) The first mean value theorem states that there exists a K1 ∈ [0, k] such that
fh (y + k) − fh (y)
= fh0 (y + K1 )
(y + k) − y
or equivalently
fh (y + k) − fh (y) = kfh0 (y + K1 ).
Similarly, there exists H1 ∈ [0, h] such that

gk (x + h) − gk (x) = hgk0 (x + H1 ).

(c) Since (part (a)) fh (y + k) − fh (y) = gk (x + h) − gk (x), we have from part (b)

fh (y + k) − fh (y) = kfh0 (y + K1 ) = hgk0 (x + H1 ) = gk (x + h) − gk (x).

(d) Now (see question 1 of Section 1.1),

  !
∂ψ  ∂ψ 
kfh0 (y + K1 ) = k −
∂y (x,y)→(x+h,y+K1 ) ∂y (x,y)→(x,y+K1 )

and   !
∂ψ  ∂ψ 
hgk0 (x + H1 ) = h − .
∂x (x,y)→(x+H1 ,y+k) ∂x (x,y)→(x+H1 ,y)
Equating the right hand sides of these two equations (which are equal by part (c)) yields
that
   
∂ψ ∂ψ ∂ψ ∂ψ
k (x + h, y + K1 ) − (x, y + K1 ) = h (x + H1 , y + k) − (x + H1 , y) .
∂y ∂y ∂x ∂x
(3)
(e) Similar to part (b), we apply the first mean value theorem to each side of (3):
 
∂ψ ∂ψ ∂ ∂ψ
k (x + h, y + K1 ) − (x, y + K1 ) = kh (x + H2 , y + K1 )
∂y ∂y ∂x ∂y
for some H2 ∈ [0, h], and
 
∂ψ ∂ψ ∂ ∂ψ
h (x + H1 , y + k) − (x + H1 , y) = hk (x + H1 , y + K2 )
∂x ∂x ∂y ∂x
for some K2 ∈ [0, k]. Thus
∂ ∂ψ ∂ ∂ψ
kh (x + H2 , y + K1 ) = hk (x + H1 , y + K2 ).
∂x ∂y ∂y ∂x

9
 (f) Dividing by hk in the previous equation yields As h → 0 and k → 0, we must have
H1 , H2 , K1 , K2 → 0.
∂ ∂ψ ∂ ∂ψ
(x + H2 , y + K1 ) = (x + H1 , y + K2 )
∂x ∂y ∂y ∂x
and taking the limits h → 0 and k → 0 (and H1 , H2 , K1 , K2 → 0) yields
∂ ∂ψ ∂ ∂ψ
(x, y) = (x, y).
∂x ∂y ∂y ∂x
Here we have assumed that all of the partial derivatives exist, and are continuous so that
we can take the limits and equate the limits to the function values.
7. We need to prove that ∇ · (∇ × F) ≡ 0. We have
∂F3 ∂F2
 
 ∂x2 − ∂x3 
 ∂F
 1 ∂F3 

∇ · (∇ × F) = ∇ ·  −
 ∂x3 ∂x 

 ∂F2 ∂F11 
−
∂x ∂x2
 1     
∂ ∂F3 ∂F2 ∂ ∂F1 ∂F3 ∂ ∂F2 ∂F1
= − + − + −
∂x1 ∂x2 ∂x3 ∂x2 ∂x3 ∂x1 ∂x3 ∂x1 ∂x2
2
∂ F3 2
∂ F3 2
∂ F2 2
∂ F3 2
∂ F1 2
∂ F3
= − + − + −
∂x1 ∂x2 ∂x2 ∂x1 ∂x3 ∂x1 ∂x1 ∂x3 ∂x2 ∂x3 ∂x3 ∂x2
and applying Theorem 1.2.6 yields ∇ · (∇ × F) = 0.
8. Applying Theorem 1.2.5(b) and Theorem 1.2.5(d) yields

∇2 (φψ) = ∇ · (∇(φψ)) = ∇ · (ψ∇φ + φ∇ψ) = (∇ψ) · (∇φ) + ψ∇2 φ + (∇φ) · (∇ψ) + φ∇2 ψ.

Since ∇2 φ = ∇2 ψ = 0, we find
∇2 (φψ) = 2(∇ψ) · (∇φ).

9. If f is harmonic, then ∇2 f = 0. We have by Theorem 1.2.5(b)

∇2 f 2 = ∇ · (∇f 2 ) = ∇ · (2f ∇f ) = (∇(2f )) · (∇f ) + 2f (∇ · ∇f ) = 2k∇f k2

where we used Theorem 1.2.5(d) and (a).

10. Theorem 1.2.5(f) and Theorem 1.2.7, provides

∇ × (f ∇g) = (∇f ) × (∇g) + f (∇ × (∇g)) = (∇f ) × (∇g).

11. Since
∇ × (g∇f ) = (∇g) × (∇f ) + g(∇ × (∇f )) = (∇g) × (∇f )
by Theorem 1.2.5(f) and Theorem 1.2.7, and

∇ × (f ∇g) = (∇f ) × (∇g) + f (∇ × (∇g)) = −(∇g) × (∇f )

we have ∇ × (g∇f ) = ∇ × (f ∇g) if and only if

(∇g) × (∇f ) = −(∇g) × (∇f )

i.e. (∇g) × (∇f ) = 0 (which means that ∇g and ∇f are parallel vectors for all x ∈ R3 ).
12. Setting ∇2 f = 4 − c cos x + 2b = 0 yields b = 2c cos x − 2. Since b must be constant, we have
c = 0 and b = −2. This is the only constraint, i.e. f (x, y) = ax + 2x2 − 2y 2 is harmonic for all
a ∈ R.

10
13. Theorem 1.2.5(d) provides

∇ · (f ∇f ) = (∇f ) · (∇f ) + f ∇ · (∇f ) = k∇f k2 + f (∇ · (∇f )).

Thus ∇ · (f ∇f ) = k∇f k2 if and only if f (∇ · (∇f )) = 0. If f = 0 on some region of R2 , then

∇ · (∇f ) = 0 identically in that region. If f 6= 0 on some region of R2 , then ∇ · (∇f ) = 0. It
remains to discuss the boundary between regions where f = 0 and ∇ · (∇f ) = 0.

11