Coordinate Singularities in Curl and Divergence

Why Coordinate Formulas Can Look Singular

Cylindrical and spherical coordinates are useful because they can make a vector field or its derivatives easier to describe. But their coordinate formulas contain factors such as \(1/\rho\), \(1/r\), and \(1/\sin\theta\). These factors may look infinite at the axis, origin, or poles, even when the original vector field is perfectly smooth there.

The reverse can also happen: a formula may be easy to use away from an exceptional point while the original field itself has no consistent value or derivative at that point. The reliable test is to return to Cartesian components and examine the field and its Cartesian derivatives.

Key distinction: A coordinate singularity occurs where a coordinate description degenerates. A genuine field singularity occurs where the original vector field fails to have the required value or derivatives. The two can occur at the same location, but neither implies the other.

Where Cylindrical and Spherical Coordinates Degenerate

In cylindrical coordinates, \(\rho=\sqrt{x^2+y^2}\), and the angle \(\phi\) describes direction in the \(xy\)-plane. At \(\rho=0\), that angle has no unique value. The cylindrical unit vectors \(\mathbf e_\rho\) and \(\mathbf e_\phi\) therefore have no unique direction on the \(z\)-axis.

In spherical coordinates, \(r\) is distance from the origin, \(\theta\) is measured from the positive \(z\)-axis, and \(\phi\) is the azimuthal angle. At \(r=0\), neither angular coordinate identifies a direction. At the poles, where \(\sin\theta=0\), the azimuth \(\phi\) is undefined. Thus the spherical coordinate description is regular only when \(r>0\) and \(0<\theta<\pi\).

These failures concern the coordinates and their basis vectors. They do not, by themselves, say whether a vector field is defined or differentiable at those points. For example, the zero vector can be specified at the origin even though no radial direction is defined there.

Check the Field in Cartesian Coordinates

Write a vector field as \(\mathbf F=(P,Q,R)\) in Cartesian coordinates. Where the required partial derivatives exist, its divergence and curl are

\nabla\cdot\mathbf F = \frac{\partial P}{\partial x} +\frac{\partial Q}{\partial y} +\frac{\partial R}{\partial z}, \qquad \nabla\times\mathbf F = \left( \frac{\partial R}{\partial y}-\frac{\partial Q}{\partial z}, \frac{\partial P}{\partial z}-\frac{\partial R}{\partial x}, \frac{\partial Q}{\partial x}-\frac{\partial P}{\partial y} \right).

The Cartesian formulas provide a direct way to test a point where a curvilinear coordinate formula is not valid. If the field has continuously differentiable Cartesian components in a neighborhood of the point, then its divergence and curl are defined there and vary continuously there. In particular, their limits from nearby points agree with their values at the point.

A limit of the divergence or curl calculated away from a point is not, by itself, proof that the field is differentiable at that point. The field must also have an appropriate value and Cartesian derivatives there. When in doubt, check the original components and their derivatives rather than substituting an exceptional coordinate value into a formula that assumes a regular coordinate chart.

Why the Cylindrical Formula Agrees Away from the Axis

A short derivation shows how an apparent coordinate denominator can arise from a regular Cartesian calculation. Consider a field with cylindrical form \(\mathbf F=f(\rho)\mathbf e_\rho\), with no \(z\)-component. For \(\rho>0\), its Cartesian components are

P=f(\rho)\frac{x}{\rho}, \qquad Q=f(\rho)\frac{y}{\rho}, \qquad R=0.

Since \(\partial\rho/\partial x=x/\rho\) and \(\partial\rho/\partial y=y/\rho\), differentiating the first component gives

\frac{\partial P}{\partial x} = f'(\rho)\frac{x^2}{\rho^2} + f(\rho)\left(\frac{1}{\rho}-\frac{x^2}{\rho^3}\right).

Differentiating the second component in the same way gives

\frac{\partial Q}{\partial y} = f'(\rho)\frac{y^2}{\rho^2} + f(\rho)\left(\frac{1}{\rho}-\frac{y^2}{\rho^3}\right).

Adding these derivatives and using \(x^2+y^2=\rho^2\) yields

\nabla\cdot\mathbf F = f'(\rho)\frac{x^2+y^2}{\rho^2} + \frac{2f(\rho)}{\rho} - f(\rho)\frac{x^2+y^2}{\rho^3} = f'(\rho)+\frac{f(\rho)}{\rho} = \frac{1}{\rho}\frac{d}{d\rho}\bigl(\rho f(\rho)\bigr).

This is the cylindrical divergence formula for this field, derived directly from Cartesian differentiation. The derivation assumes \(\rho>0\); it does not settle what happens on the axis. At the axis, the original Cartesian field must be checked separately.

Worked Examples

Worked Example: A Cylindrical Formula That Extends Smoothly to the Axis

Let \(\mathbf F=\rho\mathbf e_\rho\) for \(\rho>0\). Its cylindrical expression involves a unit vector that is undefined on the axis. Determine whether the field itself extends smoothly there, and find its divergence and curl.

Because \(\mathbf e_\rho=(x/\rho,y/\rho,0)\) away from the axis,

\mathbf F = \rho\left(\frac{x}{\rho},\frac{y}{\rho},0\right) = (x,y,0).

The Cartesian expression gives a smooth extension to the axis: set \(\mathbf F(0,0,z)=(0,0,0)\). Its divergence is

\nabla\cdot\mathbf F = \frac{\partial x}{\partial x} + \frac{\partial y}{\partial y} + \frac{\partial 0}{\partial z} = 1+1+0=2.

Its curl is

\nabla\times\mathbf F = \left( \frac{\partial 0}{\partial y}-\frac{\partial y}{\partial z}, \frac{\partial x}{\partial z}-\frac{\partial 0}{\partial x}, \frac{\partial y}{\partial x}-\frac{\partial x}{\partial y} \right) = (0-0,0-0,0-0) = \mathbf 0.

For \(\rho>0\), the cylindrical divergence formula gives the same result: \(\frac{1}{\rho}\frac{\partial}{\partial\rho}(\rho\cdot\rho) =\frac{1}{\rho}(2\rho)=2\). The Cartesian derivatives verify the values on the axis as well. Thus the undefined direction of \(\mathbf e_\rho\) is a coordinate issue here, not a singularity of the field.

Worked Example: A Genuine Singularity at the Cylindrical Axis

Consider \(\mathbf F=\mathbf e_\rho\) for \(\rho>0\), with zero \(z\)-component. Determine whether it can be extended continuously to the axis, and calculate its divergence and curl away from the axis.

In Cartesian coordinates,

\mathbf F = \left(\frac{x}{\rho},\frac{y}{\rho},0\right), \qquad \rho=\sqrt{x^2+y^2}.

Approach an axis point \((0,0,z_0)\) along the positive \(x\)-direction. The field approaches \((1,0,0)\). Approach along the positive \(y\)-direction and it approaches \((0,1,0)\). Since these limits differ, the field has no continuous extension to that point.

For \(\rho>0\), direct differentiation gives

\frac{\partial}{\partial x}\left(\frac{x}{\rho}\right) = \frac{1}{\rho}-\frac{x^2}{\rho^3}, \qquad \frac{\partial}{\partial y}\left(\frac{y}{\rho}\right) = \frac{1}{\rho}-\frac{y^2}{\rho^3}.

Therefore,

\nabla\cdot\mathbf F = \frac{2}{\rho}-\frac{x^2+y^2}{\rho^3} = \frac{2}{\rho}-\frac{\rho^2}{\rho^3} = \frac{1}{\rho}.

The only potentially nonzero curl component is the \(z\)-component:

\frac{\partial}{\partial x}\left(\frac{y}{\rho}\right) - \frac{\partial}{\partial y}\left(\frac{x}{\rho}\right) = -\frac{xy}{\rho^3} - \left(-\frac{xy}{\rho^3}\right) = 0.

The other two curl components are zero because the field has no \(z\)-component and its first two components are independent of \(z\). Thus \(\nabla\times\mathbf F=\mathbf 0\) away from the axis, while \(\nabla\cdot\mathbf F=1/\rho\). Neither calculation assigns a classical divergence or curl on the axis: the original field is not even continuous there.

Worked Example: A Smooth Field at a Spherical Pole

Consider \(\mathbf F=r\sin\theta\,\mathbf e_\phi\) in spherical coordinates away from the polar axis. The factor \(\mathbf e_\phi\) has no unique direction at a pole. Check the Cartesian field and its divergence and curl.

Using \(x=r\sin\theta\cos\phi\) and \(y=r\sin\theta\sin\phi\), the azimuthal unit vector is \(\mathbf e_\phi=(-\sin\phi,\cos\phi,0)\). Consequently,

\mathbf F = r\sin\theta(-\sin\phi,\cos\phi,0) = (-y,x,0).

This expression is smooth at the \(z\)-axis, including the origin, so it defines the field there without requiring an azimuthal direction. Its divergence is

\nabla\cdot\mathbf F = \frac{\partial(-y)}{\partial x} + \frac{\partial x}{\partial y} + \frac{\partial 0}{\partial z} = 0+0+0 = 0.

Its curl is

\nabla\times\mathbf F = \left( 0-0,\, 0-0,\, \frac{\partial x}{\partial x} - \frac{\partial(-y)}{\partial y} \right) = (0,0,1-(-1)) = (0,0,2).

Thus the spherical basis vector is ambiguous at the poles, but the vector field and its derivatives are not. The Cartesian check establishes the values on the axis; the spherical expression alone is not a valid formula to evaluate at the poles.

Worked Example: A Spherical Expression with a Genuine Origin Singularity

For \(r>0\), let \(\mathbf F=\mathbf e_r\). Determine whether it extends continuously to the origin and find its divergence and curl away from the origin.

In Cartesian coordinates, \(\mathbf e_r=(x/r,y/r,z/r)\), where \(r=\sqrt{x^2+y^2+z^2}\). Along the positive \(x\)-axis the field is \((1,0,0)\), while along the positive \(z\)-axis it is \((0,0,1)\). These different directional limits show that no continuous extension exists at the origin.

For \(r>0\), differentiating the components gives

\nabla\cdot\mathbf F = \left(\frac{1}{r}-\frac{x^2}{r^3}\right) + \left(\frac{1}{r}-\frac{y^2}{r^3}\right) + \left(\frac{1}{r}-\frac{z^2}{r^3}\right) = \frac{3}{r}-\frac{x^2+y^2+z^2}{r^3} = \frac{2}{r}.

Each curl component cancels. For instance, \(\partial(z/r)/\partial y=-yz/r^3\) and \(\partial(y/r)/\partial z=-yz/r^3\), so their difference is zero. The same cancellation holds for the other two components. Hence \(\nabla\times\mathbf F=\mathbf 0\) for \(r>0\). The field still has a genuine singularity at the origin because it has no continuous extension there.

A Practical Test at an Exceptional Point

When a cylindrical or spherical calculation approaches a coordinate exception, separate the question about the field from the question about the formula. The following sequence keeps those questions distinct.

Locate the degeneracy.
Identify whether the point lies on the cylindrical axis, at the spherical origin, or at a spherical pole.

Rewrite the field in Cartesian components.
Use the coordinate relations and basis vectors to express the original vector without relying on an ambiguous direction.

Check the field itself.
Test whether the Cartesian components have a well-defined limit and whether the proposed value agrees with it.

Check Cartesian derivatives.
Differentiate the Cartesian components where possible. Use these derivatives to decide whether divergence and curl exist at the point.

Compare with the curvilinear result.
Use the coordinate formula only on its regular domain, then check whether the Cartesian result agrees there and extends appropriately to the exceptional point.

Common Mistake: Substituting into a Formula Outside Its Domain

A tempting but invalid step is to take a cylindrical formula such as \(\frac{1}{\rho}\frac{\partial}{\partial\rho}(\rho F_\rho)\), set \(\rho=0\), and treat the resulting expression as the divergence on the axis. The formula was derived where \(\rho>0\); substitution at \(\rho=0\) is not justified. For \(\mathbf F=\mathbf e_\rho\), it produces \(1/\rho\), which is not a finite value at the axis, and the Cartesian field has no continuous extension there.

The correction is not to cancel or discard a denominator automatically. Instead, first rewrite the field in Cartesian components. If those components extend smoothly, differentiate the extension at the point. If they do not have a consistent limit, there is no continuous field there whose classical curl or divergence can be found by substitution.

Takeaway: A bad coordinate value is a warning to inspect the field, not a verdict about it. Cartesian smoothness establishes that an apparent coordinate singularity is harmless; failure of the original field or its derivatives to exist indicates a genuine singularity.

Check Your Understanding

For each question, distinguish what the coordinates fail to describe from what the original field does or does not define.

Why is the cylindrical angle \(\phi\) undefined on the \(z\)-axis, and does that alone imply a vector field is singular there?
Rewrite \(\rho\mathbf e_\rho\) in Cartesian components. What does that expression imply about its extension to the axis?
For \(\mathbf F=\mathbf e_\rho\), why do different approach directions rule out a continuous extension to the axis?
At which spherical locations is the azimuth \(\phi\) undefined? What Cartesian check can determine whether a field is nevertheless smooth there?
Why is a limit of a coordinate-form divergence away from an exceptional point not, by itself, proof that the divergence exists at that point?