CHAPTER 5

DIV, GRAD, AND CURL

1. The operator and the gradient:
Recall that the gradient of a dierentiable scalar eld on an open
set D in R
n
is given by the formula:
(1) =
_

x
1
,

x
2
, . . . ,

x
n
_
.
It is often convenient to dene formally the dierential operator in
vector form as:
(2) =
_

x
1
,

x
2
, . . . ,

x
n
_
.
Then we may view the gradient of , as the notation suggests,
as the result of multiplying the vector by the scalar eld . Note
that the order of multiplication matters, i.e.,

x
j
is not

x
j
.
Let us now review a couple of facts about the gradient. For any
j n,

x
j
is identically zero on D i (x
1
, x
2
, . . . , x
n
) is independent
of x
j
. Consequently,
(3) = 0 on D = constant.
Moreover, for any scalar c, we have:
(4) is normal to the level set L
c
().
Thus gives the direction of steepest change of .
2. Divergence
Let F : D R
n
, D R
n
, be a dierentiable vector eld. (Note that
both spaces are n-dimensional.) Let F
1
, F
2
, . . . , F
n
be the component
(scalar) elds of f. The divergence of F is dened to be
1
2 DIV, GRAD, AND CURL
(5) div(F) = F =
n

j=1
F
j
x
j
.
This can be reexpressed symbolically in terms of the dot product as
(6) F = (

x
1
, . . . ,

x
n
) (F
1
, . . . , F
n
).
Note that div(F) is a scalar eld.
Given any n n matrix A = (a
ij
), its trace is dened to be:
tr(A) =
n

i=1
a
ii
.
Then it is easy to see that, if DF denotes the Jacobian matrix of F,
i.e., the n n-matrix (F
i
/x
j
), 1 i, j n, then
(7) F = tr(DF).
Let be a twice dierentiable scalar eld. Then its Laplacian is
dened to be
(8)
2
= ().
It follows from (1),(5),(6) that
(9)
2
=

2

x
2
1
+

2

x
2
2
+ +

2

x
2
n
.
One says that is harmonic i
2
= 0. Note that we can formally
consider the dot product
(10) = (

x
1
, . . . ,

x
n
) (

x
1
, . . . ,

x
n
) =
n

j=1

2
x
2
j
.
Then we have
(11)
2
= ( ).
CHAPTER 5 3
Examples of harmonic functions:
(i) D = R
2
; (x, y) = e
x
cos y.
Then

x
= e
x
cos y,

y
= e
x
sin y,
and

2

x
2
= e
x
cos y,

2

y
2
= e
x
cos y. So,
2
= 0.
(ii) D = R
2
{0}. (x, y) = log(x
2
+y
2
) = 2 log(r).
Then

x
=
2x
x
2
+y
2
,

y
=
2y
x
2
+y
2
,

2

x
2
=
2(x
2
+y
2
)2x(2x)
(x
2
+y
2
)
2
=
2(x
2
y
2
)
(x
2
+y
2
)
2
, and

y
2
=
2(x
2
+y
2
)2y(2y)
(x
2
+y
2
)
2
=
2(x
2
y
2
)
(x
2
+y
2
)
2
. So,
2
= 0.
(iii) D = R
n
{0}. (x
1
, x
2
, ..., x
n
) = (x
2
1
+x
2
2
+ +x
2
n
)
/2
= r

for
some xed R.
Then

x
i
= r
1 x
i
r
= r
2
x
i
, and

x
2
i
= ( 2)r
4
x
i
x
i
+r
2
1.
Hence
2
=

n
i=1
(( 2)r
4
x
2
i
+r
2
) = ( 2 + n)r
2
.
So is harmonic for = 0 or = 2 n ( = 1 for n = 3).
3. Cross product in R
3
The three-dimensional space is very special in that it admits a vec-
tor product, often called the cross product. Let i,j,k denote the
standard basis of R
3
. Then, for all pairs of vectors v = xi + yj + zk
and v

= x

i +y

j +z

k, the cross product is dened by

(12) vv

= det
_
i j k
x y z
x

_
= (yz

z)i(xz

z)j+(xy

y)k.
Lemma 1. (a) v v

= v

v (anti-commutativity)
(b) i j = k, j k = i, k i = j
(c) v (v v

) = v

(v v

) = 0.
Corollary: v v = 0.
Proof of Lemma (a) v

v is obtained by interchanging the second

and third rows of the matrix whose determinant gives v v

. Thus
v

v=v v

.
(b) i j = det
_
i j k
1 0 0
0 1 0
_
, which is k as asserted. The other two iden-
tities are similar.
(c) v(vv

) = x(yz

z)y(xz

z)+z(xy

y) = 0. Similarly
for v

(v v

).
Geometrically, v v

can, thanks to the Lemma, be interpreted as

follows. Consider the plane P in R
3
dened by v,v

. Then v v

will
lie along the normal to this plane at the origin, and its orientation is
4 DIV, GRAD, AND CURL
given as follows. Imagine a corkscrew perpendicular to P with its tip
at the origin, such that it turns clockwise when we rotate the line Ov
towards Ov

in the plane P. Then v v

will point in the direction in

which the corkscrew moves perpendicular to P.
Finally the length ||v v

|| is equal to the area of the parallelogram

spanned by v and v

. Indeed this area is equal to the volume of the

parallelepiped spanned by v, v

and a unit vector u = (u

x
, u
y
, u
z
) or-
thogonal to v and v

. We can take u = v v

/||v v

|| and the (signed)

volume equals
det
_
_
u
x
u
y
u
z
x y z
x

_
_
=u
x
(yz

z) u
y
(xz

z) + u
z
(xy

y)
=||v v

|| (u
2
x
+u
2
y
+u
2
z
) = ||v v

||.
4. Curl of vector fields in R
3
Let F : D R
3
, D R
3
be a dierentiable vector eld. Denote by
P,Q,R its coordinate scalar elds, so that F = Pi + Qj + Rk. Then
the curl of F is dened to be:
(13) curl(F) = F = det
_
i j k

z
P Q R
_
.
Note that it makes sense to denote it F, as it is formally the
cross product of with f. Explicitly we have
F = (R/y Q/z) i(R/x P/z) j+(Q/x P/y) k
If the vector eld F represents the flow of a uid, then the curl
measures how the ow rotates the vectors, whence its name.
Proposition 1. Let h (resp. F) be a C
2
scalar (resp. vector) eld.
Then
(a): (h) = 0.
(b): (F) = 0.
Proof: (a) By denition of gradient and curl,
(h) = det
_
i j k

z
h
x
h
y
h
z
_
=
_

2
h
yz


2
h
zy
_
i +
_

2
h
zx


2
h
xz
_
j +
_

2
h
xy


2
h
yx
_
k.
CHAPTER 5 5
Since h is C
2
, its second mixed partial derivatives are independent of
the order in which the partial derivatives are computed. Thus,
(fh) = 0.
(b) By the denition of divergence and curl,
(F) =
_

x
,

y
,

z
_

_
R
y

Q
z
,
R
x
+
P
z
,
Q
x

P
y
_
=
_

2
R
xy


2
Q
xz
_
+
_

2
R
yx
+

2
P
yz
_
+
_

2
Q
zx


2
P
zy
_
.
Again, since F is C
2
,

2
R
xy
=

2
R
yx
, etc., and we get the assertion.

Warning: There exist twice dierentiable scalar (resp. vector)

elds h (resp. F), which are not C
2
, for which (a) (resp. (b)) does not
hold.
When the vector eld F represents uid ow, it is often called ir-
rotational when its curl is 0. If this ow describes the movement of
water in a stream, for example, to be irrotational means that a small
boat being pulled by the ow will not rotate about its axis. We will
see later that the condition F = 0 occurs naturally in a purely
mathematical setting as well.
Examples: (i) Let D = R
3
{0} and F(x, y, z) =
y
(x
2
+y
2
)
i
x
(x
2
+y
2
)
j.
Show that F is irrotational. Indeed, by the denition of curl,
F = det
_
i j k

z
y
(x
2
+y
2
)
x
(x
2
+y
2
)
0
_
=

z
_
x
x
2
+y
2
_
i+

z
_
y
x
2
+y
2
_
j+
_

x
_
x
x
2
+y
2
_

y
_
y
x
2
+y
2
__
k
=
_
(x
2
+y
2
) + 2x
2
(x
2
+y
2
)
2

(x
2
+y
2
) 2y
2
(x
2
+y
2
)
2
_
k = 0.
(ii) Let m be any integer = 3, D = R
3
{0}, and
F(x, y, z) =
1
r
m
(xi +yj +zk), where r =
_
x
2
+y
2
+z
2
. Show that F
is not the curl of another vector eld. Indeed, suppose F = G.
Then, since F is C
1
, G will be C
2
, and by the Proposition proved
above, F = (G) would be zero. But,
F =
_

x
,

y
,

z
_

_
x
r
m
,
y
r
m
,
z
r
m
_
6 DIV, GRAD, AND CURL
=
r
m
2x
2
(
m
2
)r
m2
r
2m
+
r
m
2y
2
(
m
2
)r
m2
r
2m
+
r
m
2z
2
(
m
2
)r
m2
r
2m
=
1
r
2m
_
3r
m
m(x
2
+y
2
+z
2
)r
m2
_
=
1
r
m
(3 m).
This is non-zero as m = 3. So F is not a curl.
Warning: It may happen that the divergence of F is zero, but F
is still not a curl. In fact this happens in example (ii) above if we
allow m = 3. We cannot treat this case, however, without establishing
Stokes theorem (later).