Gradient, Divergence and Curl in

Orthogonal Curvilinear Co-ordinate

Systems
Written by: Md. Nazmun Sadat Khan (Information Technology Manager, Community of Physics)

Let us consider an orthogonal curvilinear co-ordinate system with basis vectors (𝑢, 𝑣, 𝑤) with its
corresponding scaling factors 𝑓(𝑢, 𝑣, 𝑤), 𝑔(𝑢, 𝑣, 𝑤) and ℎ(𝑢, 𝑣, 𝑤). Therefore, an infinitesimal change of
radius vector 𝑑𝑟⃗ is

𝑑𝑟⃗ = 𝑓 𝑑𝑢 𝑢̂ + 𝑔 𝑑𝑣 𝑣̂ + ℎ 𝑑𝑤 𝑤
̂.

We want to find the expressions for gradient, divergence and curl in this co-ordinate system.

Expression for gradient

To begin with, we note that the gradient of a (scalar) function, say 𝑝(𝑥, 𝑦, 𝑧) can be used to
determine the exact differential change of that function when its input parameters change, i.e.
∂p ∂p ∂p
𝑑𝑝 = 𝑑𝑥 + 𝑑𝑦 + 𝑑𝑧
∂x ∂y ∂z
∂p ∂p ∂p
=( 𝑥̂ + 𝑦̂ + 𝑧̂ ) ∙ (𝑑𝑥 𝑥̂ + 𝑑𝑦 𝑦̂ + 𝑑𝑧 𝑧̂ )
∂x ∂y ∂z
= ∇𝑓 ∙ 𝑑𝑟⃗
∂p ∂p ∂p
where ∇𝑝 = ∂x 𝑥̂ + ∂y 𝑦̂ + ∂z 𝑧̂ in Cartesian co-ordinates. Now if 𝑝(𝑥, 𝑦, 𝑧) is modified to use the
new curvilinear co-ordinates, i.e. 𝑝(𝑢, 𝑣, 𝑤), then the exact differential change becomes
∂p ∂p ∂p
𝑑𝑝 = 𝑑𝑢 + 𝑑𝑣 + 𝑑𝑤
∂u ∂v ∂w
1 ∂p 1 ∂p 1 ∂p
= 𝑓𝑑𝑢 + 𝑔𝑑𝑣 + ℎ𝑑𝑤
𝑓 ∂u 𝑔 ∂v ℎ ∂w
1 ∂p 1 ∂p 1 ∂p
=( 𝑢̂ + 𝑣̂ + ̂) ∙ (𝑓 𝑑𝑢 𝑢̂ + 𝑔 𝑑𝑣 𝑣̂ + ℎ 𝑑𝑤 𝑤
𝑤 ̂)
𝑓 ∂u 𝑔 ∂v ℎ ∂w
which as seen before can be written as

𝑑𝑝 = ∇𝑓 ∙ 𝑑𝑟⃗
where,
 1 ∂p 1 ∂p 1 ∂p
∇𝑝 = 𝑢̂ + 𝑣̂ + 𝑤
̂
𝑓 ∂u 𝑔 ∂v ℎ ∂w
which gives the expression of gradient in an orthogonal curvilinear co-ordinate system.

Expression for divergence

⃗⃗, at some particular point in space, can be interpreted as the ‘flux’
Divergence of a vector field, say 𝐾
⃗⃗ flowing per unit volume, in or out of a tiny infinitesimal volume 𝑑𝑉, centered
of the vector field 𝐾
around that point.

𝑤
̂
𝐷

∆𝑤
𝐺
𝐵
𝑣̂
𝑃(𝑢, 𝑣, 𝑤)

∆𝑢 𝐹
∆𝑣

𝑢̂

Consider a point 𝑃(𝑢, 𝑣, 𝑤) around which we consider an infinitesimal volume 𝑂𝐴𝐵𝐶𝐷𝐸𝐹𝐺, such
that three of its edges align with the orthogonal basis vectors (𝑢̂, 𝑣̂, 𝑤
̂). The volume lies within the
∆𝑢 ∆𝑢 ∆𝑣 ∆𝑣 ∆𝑤 ∆𝑤
region 𝑢 − 2
→ 𝑢+ 2
, 𝑣− 2
→ 𝑣+ 2
and 𝑤 − 2
→ 𝑤+ 2
.

⃗⃗(𝑢, 𝑣, 𝑤), which can be expressed around the point as,

Let the point 𝑃 be taken in the vector field 𝐾

⃗⃗(𝑢, 𝑣, 𝑤) = 𝐾𝑢 (𝑢, 𝑣, 𝑤) 𝑢̂ + 𝐾𝑣 (𝑢, 𝑣, 𝑤) 𝑣̂ + 𝐾𝑤 (𝑢, 𝑣, 𝑤) 𝑤

𝐾 ̂
 ∆𝑢 ∆𝑣 ∆𝑤
Now, consider the flux through the surfaces 𝑂𝐴𝐵𝐶 and 𝐸𝐹𝐺𝐷. Between 𝑂(𝑢 − 2
,𝑣 − 2
,𝑤 − 2
)
∆𝑢 ∆𝑣 ∆𝑤
and 𝐴(𝑢 + 2
,𝑣 − 2
,𝑤 − 2
), lies 𝑂𝐴 within a range of ∆𝑢. Moreover, 𝑂𝐴 lies along 𝑢̂ direction
(where 𝑣 and 𝑤 remains constant), and since

𝑑𝑟⃗ = 𝑓 𝑑𝑢 𝑢̂ + 𝑔 𝑑𝑣 𝑣̂ + ℎ 𝑑𝑤 𝑤
̂.
So, it is evident that,

⃗⃗⃗⃗⃗⃗ = 𝑓 ∆𝑢 𝑢̂
𝑂𝐴
⃗⃗⃗⃗⃗⃗| = 𝑓 ∆𝑢
∴ 𝑂𝐴 = |𝑂𝐴

Similarly, 𝑂𝐺 = 𝑔 ∆𝑣 and 𝑂𝐶 = ℎ ∆𝑤.

Clearly, area of the surface 𝑂𝐴𝐵𝐶 = 𝑂𝐴 ∙ 𝑂𝐶 = 𝑓ℎ ∆𝑢 ∆𝑤, and direction of area vector is along −𝑣̂.
Also, the area of the surface 𝐸𝐹𝐺𝐷 = 𝑓ℎ ∆𝑢 ∆𝑤 except that the direction is along the opposite direction
along +𝑣̂.

Now, we assume that along the surfaces 𝑂𝐴𝐵𝐶 and 𝐸𝐹𝐺𝐵 respective, the field 𝐾⃗⃗ does not vary
⃗⃗ near the center of each surface.
appreciably. We therefore use the value of 𝐾

⃗⃗ ∙ (−𝑣̂)
So, flux along 𝑂𝐴𝐵𝐶 = 𝑓ℎ ∆𝑢 ∆𝑤 𝐾

= −𝑓𝑔 𝐾𝑣 ∆𝑢∆𝑤 ∆𝑣
|(𝑢, 𝑣 − , 𝑤)
2

⃗⃗ ∙ (𝑣̂)
And along 𝐸𝐹𝐺𝐵 = 𝑓ℎ ∆𝑢 ∆𝑤 𝐾

= 𝑓ℎ 𝐾𝑣 ∆𝑢∆𝑤 ∆𝑣
|(𝑢 , 𝑣 + , 𝑤)
2

So net flux along the surfaces perpendicular to 𝑣̂,

𝛷1 = lim 𝑓ℎ 𝐾𝑣 ∆𝑢∆𝑤 ∆𝑣 − 𝑓ℎ 𝐾𝑣 ∆𝑢∆𝑤 ∆𝑣

∆𝑢→0 |(𝑢 , 𝑣 + , 𝑤) |(𝑢, 𝑣 − , 𝑤)
2 2
∆𝑣→0
∆𝑤→0

𝑓ℎ 𝐾𝑣 ∆𝑣 − 𝑓ℎ 𝐾𝑣 ∆𝑣
|(𝑢 , 𝑣 + , 𝑤) |(𝑢, 𝑣 − , 𝑤)
2 2
= lim ∆𝑢 ∆𝑣 ∆𝑤
∆𝑢→0 ∆𝑣
∆𝑣→0
∆𝑤→0

𝜕(𝑓ℎ𝐾𝑣 )
= 𝑑𝑢 𝑑𝑣 𝑑𝑥
𝜕𝑣
⃗⃗ ∙ (𝑢̂)
Similarly, flux along 𝐴𝐵𝐸𝐹 = 𝑔ℎ ∆𝑣 ∆𝑤 𝐾

= 𝑔ℎ 𝐾𝑢 ∆𝑣∆𝑤 ∆𝑢
|(𝑢 + , 𝑣 , 𝑤)
2

⃗⃗ ∙ (−𝑢̂)
And along 𝑂𝐶𝐷𝐺 = 𝑔ℎ ∆𝑣 ∆𝑤 𝐾

= −𝑔ℎ 𝐾𝑢 ∆𝑣∆𝑤 ∆𝑢
|(𝑢 − , 𝑣 , 𝑤)
2
Thus, net flux along the surfaces perpendicular to 𝑢̂,

𝛷2 = lim 𝑔ℎ 𝐾𝑢 ∆𝑣∆𝑤 ∆𝑢 − 𝑔ℎ 𝐾𝑢 ∆𝑣∆𝑤 ∆𝑢

∆𝑢→0 |(𝑢 + , 𝑣 , 𝑤) |(𝑢 − , 𝑣 , 𝑤)
2 2
∆𝑣→0
∆𝑤→0

𝑔ℎ 𝐾𝑢 ∆𝑢 − 𝑔ℎ 𝐾𝑢 ∆𝑢
|(𝑢 + , 𝑣, 𝑤) |(𝑢 − , 𝑣, 𝑤)
2 2
𝛷2 = lim ∆𝑢 ∆𝑣 ∆𝑤
∆𝑢→0 ∆𝑢
∆𝑣→0
∆𝑤→0

𝜕(𝑔ℎ𝐾𝑢 )
= 𝑑𝑢 𝑑𝑣 𝑑𝑥
𝜕𝑢
Identically, flux along the surfaces perpendicular to 𝑤
̂,
𝜕(𝑓𝑔𝐾𝑤 )
𝛷3 = 𝑑𝑢 𝑑𝑣 𝑑𝑥
𝜕𝑤
So, net flux through volume 𝑂𝐴𝐵𝐶𝐷𝐸𝐹𝐺,

𝛷 = 𝛷1 + 𝛷2 + 𝛷3
𝜕(𝑔ℎ𝐾𝑢 ) 𝜕(𝑓ℎ𝐾𝑣 ) 𝜕(𝑓𝑔𝐾𝑤 )
∴ 𝛷=[ + + ] 𝑑𝑢 𝑑𝑣 𝑑𝑥
𝜕𝑢 𝜕𝑣 𝜕𝑤
Now volume of the infinitesimal region around 𝑃(𝑢, 𝑣, 𝑤) is,

𝑑𝑉 = lim 𝑂𝐴 ∙ 𝑂𝐺 ∙ 𝑂𝐶
∆𝑢→0
∆𝑣→0
∆𝑤→0

= lim 𝑓𝑔ℎ ∆𝑢 ∆𝑣 ∆𝑤
∆𝑢→0
∆𝑣→0
∆𝑤→0

= 𝑓𝑔ℎ 𝑑𝑢 𝑑𝑣 𝑑𝑤
⃗⃗, by
Finally, we can now determine the gradient of the vector field 𝐾
𝐹𝑙𝑢𝑥 𝛷 𝑡ℎ𝑟𝑜𝑢𝑔ℎ 𝑖𝑛𝑓𝑖𝑛𝑖𝑡𝑒𝑠𝑖𝑚𝑎𝑙 𝑣𝑜𝑙𝑢𝑚𝑒 𝑑𝑉
⃗⃗ =
∇∙𝐾
𝑑𝑉
𝜕(𝑔ℎ𝐾𝑢 ) 𝜕(𝑓ℎ𝐾𝑣 ) 𝜕(𝑓𝑔𝐾𝑤 )
[ + + 𝑑𝑢 𝑑𝑣 𝑑𝑥
𝜕𝑢 𝜕𝑣 𝜕𝑤 ]
=
𝑓𝑔ℎ 𝑑𝑢 𝑑𝑣 𝑑𝑤
1 𝜕(𝑔ℎ𝐾𝑢 ) 𝜕(𝑓ℎ𝐾𝑣 ) 𝜕(𝑓𝑔𝐾𝑤 )
⃗⃗ =
∴∇∙𝐾 [ + + ]
𝑓𝑔ℎ 𝜕𝑢 𝜕𝑣 𝜕𝑤
which is the form of the divergence of a field in an orthogonal curvilinear co-ordinate system.
Expression for curl
Curl of a vector field 𝐾⃗⃗ describes the infinitesimal rotation of a vector field around a point. The
⃗⃗ is
direction of the curl is the axis of rotation, using the right-hand rule. The curl of the vector field 𝐾
defined as the line integral of the vector field 𝐾 ⃗⃗ about an infinitesimal loop stationed around a point
divided by the area of that loop.

Consider an infinitesimal small loop 𝑂𝐴𝐵𝐶 in the 𝑢𝑣 plane, stationed around the point 𝑃(𝑢, 𝑣, 𝑤)
such that 𝑂𝐴 lies along 𝑢̂ and 𝑂𝐶 lies along 𝑣̂. Assuming that the co-ordinate system is right-handed, we
chose the direction of the area vector of the loop to coincide with the positive direction of 𝑤
̂, i.e.
pointing out of this page.

𝐵
𝑣̂

𝑃(𝑢, 𝑣, 𝑤)

𝑢̂
𝐴

∆𝑢 ∆𝑣 ∆𝑢 ∆𝑣
⃗⃗⃗⃗⃗⃗ lies between 𝑂(𝑢 −
Now as before, 𝑂𝐴 ,𝑣 − , 𝑤) and 𝐴(𝑢 + ,𝑣 − , 𝑤), thus using
2 2 2 2

𝑑𝑟⃗ = 𝑓 𝑑𝑢 𝑢̂ + 𝑔 𝑑𝑣 𝑣̂ + ℎ 𝑑𝑤 𝑤
̂.
we find that
 ⃗⃗⃗⃗⃗⃗
𝑂𝐴 = 𝑓 ∆𝑢 𝑢̂.
∆𝑢 ∆𝑣 ∆𝑢 ∆𝑣
⃗⃗⃗⃗⃗⃗ lies between 𝐴(𝑢 +
Similarly, 𝐴𝐵 ,𝑣 − , 𝑤) and 𝐵 (𝑢 + ,𝑣 + , 𝑤), hence
2 2 2 2

⃗⃗⃗⃗⃗⃗
𝐴𝐵 = 𝑔 ∆𝑣 𝑣̂.

Therefore, the remaining arms of the loop can be expressed as

⃗⃗⃗⃗⃗⃗
𝐵𝐶 = −𝑓 ∆𝑢 𝑢̂, ⃗⃗⃗⃗⃗⃗
𝐶𝑂 = −𝑔 ∆𝑣 𝑣̂.

⃗⃗ as
Again, as done previously, we express the field 𝐾

⃗⃗(𝑢, 𝑣, 𝑤) = 𝐾𝑢 (𝑢, 𝑣, 𝑤) 𝑢̂ + 𝐾𝑣 (𝑢, 𝑣, 𝑤) 𝑣̂ + 𝐾𝑤 (𝑢, 𝑣, 𝑤) 𝑤

𝐾 ̂

Now, assuming 𝐾 ⃗⃗ remains relatively constant along ⃗⃗⃗⃗⃗⃗ ⃗⃗ at the center of the
𝑂𝐴, we use the value of 𝐾
⃗⃗ along ⃗⃗⃗⃗⃗⃗
line. Therefore, the line integral of 𝐾 𝑂𝐴 is

∆𝑣
⃗⃗ ∙ ⃗⃗⃗⃗
∫𝐾 ⃗⃗ (𝑢, 𝑣 −
𝑑𝑙 = 𝐾 , 𝑤) ∙ ⃗⃗⃗⃗⃗⃗
𝑂𝐴 = 𝐾𝑢 𝑓 ∆𝑢 ∆𝑣
2 |(𝑢,𝑣 − ,𝑤)
2
𝑂𝐴

And along ⃗⃗⃗⃗⃗⃗

𝐵𝐶 is

∆𝑣
⃗⃗ ∙ ⃗⃗⃗⃗
∫𝐾 ⃗⃗ (𝑢, 𝑣 +
𝑑𝑙 = 𝐾 , 𝑤) ∙ ⃗⃗⃗⃗⃗⃗
𝐵𝐶 = −𝐾𝑢 𝑓 ∆𝑢 ∆𝑣
2 |(𝑢,𝑣+ ,𝑤)
2
𝐵𝐶

So, line integrals along ⃗⃗⃗⃗⃗⃗

𝑂𝐴 and ⃗⃗⃗⃗⃗⃗
𝐵𝐶 combined, applying limit, gives

⃗⃗ ∙ ⃗⃗⃗⃗
∫𝐾 ⃗⃗ ∙ ⃗⃗⃗⃗
𝑑𝑙 + ∫ 𝐾 𝑑𝑙 = lim −𝐾𝑢 𝑓 ∆𝑢 ∆𝑣 + 𝐾𝑢 𝑓 ∆𝑢 ∆𝑣
∆𝑢→0 |(𝑢,𝑣+ ,𝑤) |(𝑢,𝑣 − ,𝑤)
2 2
𝑂𝐴 𝐵𝐶 ∆𝑣→0

𝐾𝑢 𝑓 ∆𝑣 − 𝐾𝑢 𝑓 ∆𝑣
|(𝑢,𝑣+ ,𝑤) |(𝑢,𝑣 − ,𝑤)
2 2
= lim − ( ) ∆𝑢 ∆𝑣
∆𝑢→0 ∆𝑣
∆𝑣→0

𝜕(𝑓𝐾𝑢 )
=− 𝑑𝑢 𝑑𝑣
𝜕𝑣
⃗⃗⃗⃗⃗⃗ we have
Focusing on the vertical arms of the loop, along 𝐴𝐵

∆𝑢
∫𝐾 ⃗⃗⃗⃗ = 𝐾
⃗⃗ ∙ 𝑑𝑙 ⃗⃗ (𝑢 + ⃗⃗⃗⃗⃗⃗ = 𝐾𝑣 𝑔 ∆𝑣
, 𝑣, 𝑤) ∙ 𝐴𝐵 ∆𝑢
2 |(𝑢+ ,𝑣,𝑤)
2
𝐴𝐵

and along ⃗⃗⃗⃗⃗⃗

𝐶𝑂,

∆𝑢
∫𝐾 ⃗⃗⃗⃗ = 𝐾
⃗⃗ ∙ 𝑑𝑙 ⃗⃗ (𝑢 − ⃗⃗⃗⃗⃗⃗ = −𝐾𝑣 𝑔 ∆𝑣
, 𝑣, 𝑤) ∙ 𝐶𝑂 ∆𝑢
2 |(𝑢− ,𝑣,𝑤)
2
𝐶𝑂

So, combined along ⃗⃗⃗⃗⃗⃗

𝐴𝐵 and ⃗⃗⃗⃗⃗⃗
𝐶𝑂, also applying limit,
 ∫𝐾 ⃗⃗⃗⃗ + ∫ 𝐾
⃗⃗ ∙ 𝑑𝑙 ⃗⃗⃗⃗ = lim 𝐾𝑣 𝑔 ∆𝑣
⃗⃗ ∙ 𝑑𝑙 ∆𝑢 − 𝐾𝑣 𝑔 ∆𝑣 ∆𝑢
∆𝑢→0 |(𝑢+ ,𝑣,𝑤) |(𝑢 − ,𝑣,𝑤)
2 2
𝐴𝐵 𝐶𝑂 ∆𝑣→0

𝐾𝑣 𝑔 ∆𝑢 − 𝐾𝑣 𝑔 ∆𝑢
|(𝑢+ ,𝑣,𝑤) |(𝑢 − ,𝑣,𝑤)
⃗⃗ ∙ ⃗⃗⃗⃗ ⃗⃗ ∙ ⃗⃗⃗⃗ 2 2
∫𝐾 𝑑𝑙 + ∫ 𝐾 𝑑𝑙 = lim ( ) ∆𝑢 ∆𝑣
∆𝑢→0 ∆𝑢
𝐴𝐵 𝐶𝑂 ∆𝑣→0

𝜕(𝑔𝐾𝑣 )
= 𝑑𝑢 𝑑𝑣
𝜕𝑢
⃗⃗,
So, total loop integral of 𝐾

⃗⃗ ∙ ⃗⃗⃗⃗
∮𝐾 ⃗⃗ ∙ ⃗⃗⃗⃗
𝑑𝑙 = ∫ 𝐾 ⃗⃗ ∙ ⃗⃗⃗⃗
𝑑𝑙 + ∫ 𝐾 ⃗⃗ ∙ ⃗⃗⃗⃗
𝑑𝑙 + ∫ 𝐾 ⃗⃗ ∙ ⃗⃗⃗⃗
𝑑𝑙 + ∫ 𝐾 𝑑𝑙
𝑂𝐴 𝐴𝐵 𝐵𝐶 𝐶𝑂

𝜕(𝑔𝐾𝑣 ) 𝜕(𝑓𝐾𝑢 )
=[ − ] 𝑑𝑢 𝑑𝑣
𝜕𝑢 𝜕𝑣
Area of the loop integral,

𝑑𝐴 = lim 𝑂𝐴 ∙ 𝐴𝐵
∆𝑢→0
∆𝑣→0

= 𝑓𝑔 𝑑𝑢 𝑑𝑣
⃗⃗, therefore can be obtained by dividing the loop integral by the infinitesimal
Curl of the vector field 𝐾
area. Since the loop integral is taken in an anti-clockwise direction in the 𝑢𝑣 plane, thus the axis of
rotation and also the direction of the curl will lie along the positive direction of 𝑤
̂. Thus

⃗⃗⃗⃗
⃗⃗ ∙ 𝑑𝑙
∮𝐾
⃗⃗ =
∇×𝐾 𝑤
̂
𝑑𝐴
𝜕(𝑔𝐾𝑣 ) 𝜕(𝑓𝐾𝑢 )
[ − ] 𝑑𝑢 𝑑𝑣
𝜕𝑢 𝜕𝑣
= 𝑤
̂
𝑓𝑔 𝑑𝑢 𝑑𝑣
1 𝜕(𝑔𝐾𝑣 ) 𝜕(𝑓𝐾𝑢 )
⃗⃗ =
∴∇×𝐾 [ − ]𝑤
̂
𝑓𝑔 𝜕𝑢 𝜕𝑣
Similarly, considering loops in 𝑣𝑤 and 𝑢𝑤 planes, we can find the 𝑢 and 𝑣 components of the curl in
the same manner. The resulting form of the curl, all components considered is
1 𝜕(ℎ𝐾𝑤 ) 𝜕(𝑔𝐾𝑣 ) 1 𝜕(𝑓𝐾𝑢 ) 𝜕(ℎ𝐾𝑤 ) 1 𝜕(𝑔𝐾𝑣 ) 𝜕(𝑓𝐾𝑢 )
⃗⃗ =
∇×𝐾 [ − ] 𝑢̂ + [ − ] 𝑣̂ + [ − ]𝑤
̂
𝑔ℎ 𝜕𝑣 𝜕𝑤 𝑓ℎ 𝜕𝑤 𝜕𝑢 𝑓𝑔 𝜕𝑢 𝜕𝑣