Why Flux Is Invariant Under Reparametrization

Why a Change of Parameters Should Not Change the Surface

A surface patch can be described by many different pairs of parameters. One parametrization may make a calculation convenient, while another may stretch or compress parameter coordinates in a different way. These choices change the coordinate description, but they need not change the oriented surface itself.

Flux is defined using both the vector field and an oriented area element. That area element depends on the parameter derivatives, so it is not enough to observe that two parametrizations trace the same points. We must also account for how the parameter change affects tangent vectors and orientation.

The central fact is that a regular change of parameters multiplies the oriented area element by the Jacobian determinant. The usual change-of-variables theorem then cancels the corresponding change in parameter area. If the determinant is positive, the orientation is preserved and the flux is unchanged. If it is negative, the orientation is reversed and the flux changes sign.

Flux on a Parametrized Surface

Let $D$ be a parameter region in the $uv$-plane, and let $\mathbf r:D\to\mathbb R^3$ be a continuously differentiable parametrization, written as $\mathbf r(u,v)$. The patch is regular when $\mathbf r_u\times\mathbf r_v\ne\mathbf 0$ throughout its interior. The order $u,v$ specifies its orientation.

Definition: If $\mathbf F$ is a continuous vector field on the surface, the flux through the oriented patch parametrized by $\mathbf r$ is $$ \iint_D \mathbf F(\mathbf r(u,v))\cdot (\mathbf r_u\times\mathbf r_v)\,du\,dv. $$

Assume the parameter region has a piecewise smooth boundary, and that the displayed integrand is integrable. These conditions ensure that the flux integral and the change-of-variables formula used below are well-defined. Reversing the order of the two parameters reverses the cross product, and therefore reverses the flux sign.

The Tangent Cross-Product Identity

Suppose $\mathbf q:E\to D$ is a continuously differentiable, one-to-one change of parameters with a continuously differentiable inverse. Write $\mathbf q(s,t)=(u(s,t),v(s,t))$, and define the new surface parametrization by composition: $\mathbf R(s,t)=\mathbf r(u(s,t),v(s,t))$.

The Jacobian determinant of the parameter change is

J_{\mathbf q}(s,t) = \frac{\partial(u,v)}{\partial(s,t)} = u_s v_t-u_t v_s.

By the chain rule, $\mathbf R_s$ and $\mathbf R_t$ are linear combinations of the old tangent vectors. Taking their cross product gives the crucial identity.

Key identity: For $\mathbf R=\mathbf r\circ\mathbf q$, $$ \mathbf R_s\times\mathbf R_t = J_{\mathbf q}(s,t)\, \bigl(\mathbf r_u\times\mathbf r_v\bigr) \big|_{(u,v)=\mathbf q(s,t)}. $$

The determinant appears with its sign, not its absolute value. Its magnitude records the local change in parameter area; its sign records whether the parameter order preserves or reverses orientation.

Deriving the Identity

The chain rule gives the two tangent vectors explicitly:

\mathbf R_s = \mathbf r_u u_s+\mathbf r_v v_s, \qquad \mathbf R_t = \mathbf r_u u_t+\mathbf r_v v_t.

Expand their cross product using bilinearity. Terms that cross a vector with itself vanish, and reversing the order of a cross product changes its sign. Thus

\begin{aligned} \mathbf R_s\times\mathbf R_t &=(\mathbf r_u u_s+\mathbf r_v v_s) \times(\mathbf r_u u_t+\mathbf r_v v_t)\\ &=u_s v_t(\mathbf r_u\times\mathbf r_v) +v_s u_t(\mathbf r_v\times\mathbf r_u)\\ &=u_s v_t(\mathbf r_u\times\mathbf r_v) -v_s u_t(\mathbf r_u\times\mathbf r_v)\\ &=(u_s v_t-u_t v_s)(\mathbf r_u\times\mathbf r_v). \end{aligned}

All old tangent vectors in this calculation are evaluated at $(u,v)=\mathbf q(s,t)$. The scalar factor is exactly $J_{\mathbf q}$, proving the identity. In particular, if the original patch is regular and $J_{\mathbf q}\ne 0$, the new patch is regular as well.

The Change-of-Parameters Rule for Flux

Substitute the cross-product identity into the flux integral for $\mathbf R$. The integrand becomes the old oriented flux density, evaluated at the new parameters, and multiplied by the signed Jacobian. The ordinary change-of-variables theorem, by contrast, uses the absolute value of the Jacobian.

Theorem: Let $\mathbf r:D\to\mathbb R^3$ be a regular $C^1$ parametrization, let $\mathbf F$ be continuous on its patch, and suppose $\mathbf q:E\to D$ is a $C^1$ bijection with a $C^1$ inverse and $J_{\mathbf q}\ne0$. Assume the regions and integrands satisfy the usual conditions for change of variables. If $J_{\mathbf q}>0$ throughout $E$, then $\mathbf R=\mathbf r\circ\mathbf q$ preserves orientation and $$ \iint_E \mathbf F(\mathbf R(s,t))\cdot (\mathbf R_s\times\mathbf R_t)\,ds\,dt = \iint_D \mathbf F(\mathbf r(u,v))\cdot (\mathbf r_u\times\mathbf r_v)\,du\,dv. $$ If $J_{\mathbf q}<0$ throughout $E$, the left-hand integral is the negative of the right-hand integral.

To see both signs directly, set $A(u,v)=\mathbf F(\mathbf r(u,v))\cdot(\mathbf r_u\times\mathbf r_v)$. The identity already proved shows that the new flux is $\iint_E A(\mathbf q(s,t))J_{\mathbf q}(s,t)\,ds\,dt$. For a positive Jacobian, this is the ordinary change-of-variables formula for $\iint_D A(u,v)\,du\,dv$. For a negative Jacobian, $J_{\mathbf q}=-|J_{\mathbf q}|$, so the integral is the negative of that formula. This proves the theorem and shows why orientation is essential to the statement.

Worked Example: A Positive Rescaling of Parameters

Let $\mathbf r(u,v)=(u,v,0)$ on $D=[0,1]\times[0,2]$, oriented by $\mathbf r_u\times\mathbf r_v$. Let $\mathbf F(x,y,z)=(0,0,x+y)$. Reparametrize using $E=[0,\tfrac12]\times[0,2]$ and $\mathbf q(s,t)=(2s,t)$.

First compute the original tangent vectors and their cross product:

\mathbf r_u=(1,0,0),\qquad \mathbf r_v=(0,1,0),\qquad \mathbf r_u\times\mathbf r_v=(0,0,1).

On the surface, $\mathbf F(\mathbf r(u,v))=(0,0,u+v)$, so the original flux is

\begin{aligned} \iint_D (u+v)\,du\,dv &=\int_0^1\left(\int_0^2(u+v)\,dv\right)du\\ &=\int_0^1\left(2u+\frac{2^2}{2}\right)du\\ &=\int_0^1(2u+2)\,du\\ &=\left[u^2+2u\right]_0^1=3. \end{aligned}

The Jacobian of the change is $J_{\mathbf q}=2\cdot1-0\cdot0=2>0$, so it preserves orientation. The composed parametrization and its tangent cross product are

\mathbf R(s,t)=(2s,t,0),\qquad \mathbf R_s=(2,0,0),\qquad \mathbf R_t=(0,1,0),\qquad \mathbf R_s\times\mathbf R_t=(0,0,2).

Since $\mathbf F(\mathbf R(s,t))=(0,0,2s+t)$, the reparametrized flux is

\begin{aligned} \int_0^{1/2}\int_0^2 2(2s+t)\,dt\,ds &=\int_0^{1/2}\int_0^2(4s+2t)\,dt\,ds\\ &=\int_0^{1/2}(8s+4)\,ds\\ &=\left[4s^2+4s\right]_0^{1/2}\\ &=1+2=3. \end{aligned}

The two flux calculations agree, as the positive-Jacobian rule predicts. The transformed cross product has the factor $2$, and the new parameter rectangle has half the width; together, these changes preserve the integral.

Worked Example: A Nonlinear Orientation-Preserving Change

Use the same plane parametrization $\mathbf r(u,v)=(u,v,0)$, now on $D=\{(u,v):0\le u\le1,\ 0\le v\le1+u\}$. Let $\mathbf F(x,y,z)=(0,0,1)$. Define $\mathbf q(s,t)=(s,t(1+s))$ on the unit square $E=[0,1]\times[0,1]$.

The map covers the stated region because $u=s$ and $v=t(1+s)$ range from $0$ to $1+u$. Its Jacobian is

J_{\mathbf q} = \begin{vmatrix} 1&0\\ t&1+s \end{vmatrix} =1+s>0.

For the original parametrization, the oriented normal is $(0,0,1)$. Since the field is also $(0,0,1)$, the flux density is $1$. Therefore,

\begin{aligned} \iint_D 1\,du\,dv &=\int_0^1\int_0^{1+u}1\,dv\,du\\ &=\int_0^1(1+u)\,du\\ &=\left[u+\frac{u^2}{2}\right]_0^1 =\frac32. \end{aligned}

For the new parametrization, $\mathbf R(s,t)=(s,t(1+s),0)$, so

\mathbf R_s=(1,t,0),\qquad \mathbf R_t=(0,1+s,0),\qquad \mathbf R_s\times\mathbf R_t=(0,0,1+s).

The transformed flux is consequently

\begin{aligned} \int_0^1\int_0^1(1+s)\,dt\,ds &=\int_0^1(1+s)\,ds\\ &=\left[s+\frac{s^2}{2}\right]_0^1 =\frac32. \end{aligned}

The result agrees with the original flux. Here the local area factor varies with $s$, rather than being constant, but it is exactly the Jacobian required by the change-of-variables theorem.

Worked Example: A Parameter Swap Reverses Orientation

Let $\mathbf r(u,v)=(u,v,0)$ on the unit square, and let $\mathbf F(x,y,z)=(0,0,1)$. The original cross product is $(0,0,1)$, so the original flux is

\int_0^1\int_0^1 1\,dv\,du=1.

Now take $\mathbf q(s,t)=(t,s)$, which maps the unit square onto itself. Its Jacobian is $J_{\mathbf q}=0\cdot0-1\cdot1=-1$. Thus the change reverses orientation. Indeed, the new parametrization is $\mathbf R(s,t)=(t,s,0)$, and

\mathbf R_s=(0,1,0),\qquad \mathbf R_t=(1,0,0),\qquad \mathbf R_s\times\mathbf R_t=(0,0,-1).

The new flux is therefore

\int_0^1\int_0^1(-1)\,dt\,ds=-1.

This verifies the sign rule: the points of the surface are unchanged, but their orientation is reversed, so the flux changes from $1$ to $-1$.

A Common Mistake: Dropping the Jacobian Sign

A frequent error is to use $|J_{\mathbf q}|$ in the tangent cross-product identity. Absolute values are appropriate in the scalar change-of-variables formula for ordinary area, but the cross product represents an oriented area element. Its transformation uses the signed determinant.

For the swap in the previous example, the incorrect calculation would replace $J_{\mathbf q}=-1$ by $|J_{\mathbf q}|=1$, and conclude that the new oriented flux is $1$. But direct differentiation gives $\mathbf R_s\times\mathbf R_t=(0,0,-1)$, so the flux is $-1$. The correction is to retain the signed determinant in the vector identity, then apply the absolute-value change-of-variables theorem separately.

Takeaway: A reparametrization preserves flux when it is a regular bijective change of parameters with $J_{\mathbf q}>0$. If $J_{\mathbf q}<0$, it reverses orientation and changes the sign of flux. A zero Jacobian is not a regular change of parameters and does not satisfy this rule.

Why This Invariance Matters

The flux integral describes a geometric quantity attached to an oriented surface and a vector field, not to a particular choice of parameter coordinates. The change-of-parameters rule makes that principle precise. It permits a calculation to use whichever regular, orientation-compatible parametrization is most convenient, while guaranteeing that the answer is unchanged.

That freedom is useful only when the hypotheses are respected. The change must cover the same patch without duplicating or omitting points, its Jacobian must be nonzero, and its sign must be tracked. In practical work, checking the sign of $J_{\mathbf q}$ and confirming the direction of the resulting cross product are reliable safeguards against an orientation error.

Check Your Understanding

Use the definitions and sign rule above to answer these questions.

For $\mathbf R=\mathbf r\circ\mathbf q$, what factor multiplies $\mathbf r_u\times\mathbf r_v$ in the identity for $\mathbf R_s\times\mathbf R_t$?
Why does a positive Jacobian preserve the oriented flux under the change-of-variables theorem?
If $J_{\mathbf q}<0$ throughout the parameter region, how does the new flux compare with the original flux?
For the parameter swap $\mathbf q(s,t)=(t,s)$, compute the Jacobian and state what happens to orientation.
Why is replacing the signed Jacobian by its absolute value in the cross-product identity incorrect?

Why Flux Is Invariant Under Reparametrization

What You'll Learn

Why a Change of Parameters Should Not Change the Surface

Flux on a Parametrized Surface

The Tangent Cross-Product Identity

Deriving the Identity

The Change-of-Parameters Rule for Flux

Worked Example: A Positive Rescaling of Parameters

Worked Example: A Positive Rescaling of Parameters

Worked Example: A Nonlinear Orientation-Preserving Change

Worked Example: A Nonlinear Orientation-Preserving Change

Worked Example: A Parameter Swap Reverses Orientation

Worked Example: A Parameter Swap Reverses Orientation

A Common Mistake: Dropping the Jacobian Sign

Why This Invariance Matters

Check Your Understanding

Previous

Next