Why Triangular Boundaries Need Careful Setup
A triangular boundary is simple enough to describe with three straight segments, but each segment has its own direction and contribution to the circulation. The total depends on traversing the edges in the order required by the chosen orientation. A sign error on even one edge can spoil the sum.
Stokes’ Theorem provides an independent way to calculate the same circulation: integrate the curl of the field across a surface bounded by the triangle. For a planar triangular patch, both the boundary calculation and the surface calculation can be carried out explicitly. Comparing them is a useful check on parameterization, orientation, and arithmetic.
Setting Up an Oriented Triangle
Let the vertices be \(A\), \(B\), and \(C\). The directed boundary \(A\to B\to C\to A\) has orientation determined by the ordered pair of side vectors \(\overrightarrow{AB}\) and \(\overrightarrow{AC}\). Their cross product points in the direction of the normal consistent with this traversal. Reversing the vertex order reverses both the boundary orientation and the compatible normal.
For a direct circulation calculation, parameterize each edge from its starting vertex to its ending vertex. If an edge is \(\mathbf{r}(t)\), with \(0\le t\le 1\), its contribution is \(\int_0^1 \mathbf{F}(\mathbf{r}(t))\cdot\mathbf{r}'(t)\,dt\). Add the three contributions; do not replace the directed edges by an unsigned perimeter or area.
For the surface calculation, an affine parameterization of the triangular patch is especially convenient. Set \(\mathbf{r}(u,v)=A+u(B-A)+v(C-A)\), where \(u\ge 0\), \(v\ge 0\), and \(u+v\le 1\). The oriented area vector is \(\mathbf{r}_u\times\mathbf{r}_v\), so the curl flux is \(\iint (\nabla\times\mathbf{F})(\mathbf{r}(u,v))\cdot(\mathbf{r}_u\times\mathbf{r}_v)\,du\,dv\). The order of the cross product matters: swapping its factors reverses the normal.
Worked Example: A Horizontal Triangle with Polynomial Curl
Let \(\mathbf{F}(x,y,z)=\langle -y^2,x^2,0\rangle\), and let the boundary be the triangle with vertices \(A=(0,0,0)\), \(B=(2,0,0)\), and \(C=(0,1,0)\), traversed \(A\to B\to C\to A\). Compute its circulation directly and compare it with the upward curl flux.
On edge \(AB\), take \(\mathbf{r}_{AB}(t)=(2t,0,0)\), \(0\le t\le1\). Then \(\mathbf{r}'_{AB}=(2,0,0)\) and \(\mathbf{F}(\mathbf{r}_{AB}(t))=\langle0,4t^2,0\rangle\). Thus \(\mathbf{F}\cdot\mathbf{r}'_{AB}=0\), and the contribution is \(0\).
On edge \(BC\), take \(\mathbf{r}_{BC}(t)=(2-2t,t,0)\). Its derivative is \(\langle-2,1,0\rangle\), and the field along it is \(\langle-t^2,(2-2t)^2,0\rangle\). Therefore, $$ \int_{BC}\mathbf{F}\cdot d\mathbf{r} =\int_0^1\left(2t^2+(2-2t)^2\right)\,dt =\int_0^1(4-8t+6t^2)\,dt =4-4+2=2. $$
On \(CA\), take \(\mathbf{r}_{CA}(t)=(0,1-t,0)\), with derivative \(\langle0,-1,0\rangle\). The field is \(\langle-(1-t)^2,0,0\rangle\), so its dot product with the derivative is \(0\). The circulation is therefore \(0+2+0=2\).
Now \(\nabla\times\mathbf{F}=\langle0,0,2x+2y\rangle\). The triangle lies in the \(xy\)-plane with upward normal, and its region is \(0\le x\le2\), \(0\le y\le1-x/2\). Hence $$ \iint_S(\nabla\times\mathbf{F})\cdot\mathbf{n}\,dS =\int_0^2\int_0^{1-x/2}(2x+2y)\,dy\,dx =\int_0^2\left(1+x-\frac{3x^2}{4}\right)\,dx =2+2-2=2. $$ The flux agrees with the sum of the three directed edge integrals.
Why the Two Calculations Agree
The agreement is not a coincidence specific to a convenient choice of field. On the reference triangle \(D=\{(u,v):u\ge0,\ v\ge0,\ u+v\le1\}\), use the affine parameterization \(\mathbf{r}(u,v)=A+u(B-A)+v(C-A)\). Write \(\mathbf{b}=B-A\) and \(\mathbf{c}=C-A\); then \(\mathbf{r}_u=\mathbf{b}\) and \(\mathbf{r}_v=\mathbf{c}\).
Define \(G_1(u,v)=\mathbf{F}(\mathbf{r}(u,v))\cdot\mathbf{b}\) and \(G_2(u,v)=\mathbf{F}(\mathbf{r}(u,v))\cdot\mathbf{c}\). Along a curve in the parameter triangle, \(\mathbf{F}\cdot d\mathbf{r}=G_1\,du+G_2\,dv\). The parameter triangle’s positively oriented boundary runs from \((0,0)\) to \((1,0)\), then to \((0,1)\), and back to \((0,0)\); under \(\mathbf{r}\), these are exactly the directed edges \(A\to B\to C\to A\).
Green’s Theorem on \(D\) converts the boundary integral into the integral of \(\partial G_2/\partial u-\partial G_1/\partial v\). The chain rule gives the identity $$ \frac{\partial G_2}{\partial u}-\frac{\partial G_1}{\partial v} = (\nabla\times\mathbf{F})(\mathbf{r}(u,v))\cdot(\mathbf{b}\times\mathbf{c}). $$ Indeed, differentiating \(G_2\) in the \(\mathbf{b}\) direction and \(G_1\) in the \(\mathbf{c}\) direction, then subtracting, pairs the component derivatives of \(\mathbf{F}\) in precisely the combination defining curl dotted with \(\mathbf{b}\times\mathbf{c}\). Since \(\mathbf{b}\times\mathbf{c}=\mathbf{r}_u\times\mathbf{r}_v\), integration over \(D\) gives the curl flux through the oriented triangular patch. This proves the correspondence for an affine parameterized triangle, assuming the field is continuously differentiable on a neighborhood of the patch.
A Triangle in a Tilted Plane
A triangle need not lie horizontally. In a tilted plane, the cross product of the side vectors supplies the oriented area vector directly. This avoids having to find a unit normal and a separate surface-area factor.
Worked Example: A Tilted Triangle and a Constant Curl
Let \(\mathbf{F}(x,y,z)=\langle-y/2,x/2,0\rangle\). Take \(A=(0,0,0)\), \(B=(1,0,1)\), and \(C=(0,1,1)\), with boundary \(A\to B\to C\to A\).
On \(AB\), use \(\mathbf{r}(t)=(t,0,t)\). The field is \(\langle0,t/2,0\rangle\), while \(\mathbf{r}'=\langle1,0,1\rangle\), so the dot product and edge integral are both \(0\).
On \(BC\), use \(\mathbf{r}(t)=(1-t,t,1)\). Then \(\mathbf{r}'=\langle-1,1,0\rangle\) and \(\mathbf{F}=\langle-t/2,(1-t)/2,0\rangle\). Thus $$ \int_{BC}\mathbf{F}\cdot d\mathbf{r} =\int_0^1\left(\frac{t}{2}+\frac{1-t}{2}\right)\,dt =\int_0^1\frac12\,dt=\frac12. $$
On \(CA\), use \(\mathbf{r}(t)=(0,1-t,1-t)\). The derivative is \(\langle0,-1,-1\rangle\), and the field is \(\langle-(1-t)/2,0,0\rangle\), giving a zero dot product. The total circulation is \(1/2\).
The curl is \(\nabla\times\mathbf{F}=\langle0,0,1\rangle\). The side vectors are \(B-A=\langle1,0,1\rangle\) and \(C-A=\langle0,1,1\rangle\), so \((B-A)\times(C-A)=\langle-1,-1,1\rangle\). This is the oriented area vector, whose \(z\)-component is \(1\). The parameter triangle has area \(1/2\), so the flux of the constant curl is \(1\cdot(1/2)=1/2\), matching the edge sum.
A Second Direct Check
For a field with more than one nonzero curl component, it is important to use the component of curl normal to the patch. The following horizontal triangle keeps the normal simple while making the curl calculation less immediate.
Worked Example: A Shifted Triangle with Nonzero Edge Contributions
Let \(\mathbf{F}(x,y,z)=\langle-y,x,x+y\rangle\), and use \(A=(1,0,0)\), \(B=(3,0,0)\), and \(C=(1,2,0)\), traversed \(A\to B\to C\to A\).
On \(AB\), \(y=0\) and \(dy=0\), so the first two components contribute \((-y)\,dx+x\,dy=0\). The third component contributes nothing because \(dz=0\). The integral is \(0\).
On \(BC\), take \(\mathbf{r}(t)=(3-2t,2t,0)\). Then \(\mathbf{r}'=\langle-2,2,0\rangle\), and the first two field components are \(\langle-2t,3-2t\rangle\). Their dot product with \(\langle-2,2\rangle\) is \(4t+6-4t=6\). The edge contribution is \(\int_0^1 6\,dt=6\).
On \(CA\), take \(\mathbf{r}(t)=(1,2-2t,0)\). Then \(\mathbf{r}'=\langle0,-2,0\rangle\), and the second component of the field is \(x=1\). The dot product is \(-2\), so the integral is \(-2\). The total circulation is \(0+6-2=4\).
Direct differentiation gives \(\nabla\times\mathbf{F}=\langle1,-1,2\rangle\). The triangle is in the \(xy\)-plane with upward normal, so only the \(z\)-component \(2\) contributes to flux. Its area is \(\frac12(2)(2)=2\). Thus the flux is \(2\cdot2=4\), agreeing with the direct calculation.
Common Mistake: Reversing the Edges but Not the Normal
In the shifted-triangle example, the order \(A\to B\to C\to A\) is counterclockwise when viewed from above, and the compatible normal is upward. If the edges are instead traversed \(A\to C\to B\to A\), every directed edge is reversed. Each line integral changes sign, so the new circulation is \(-4\), not \(4\).
The compatible surface normal must also point downward. With that normal, the curl flux is \(\langle1,-1,2\rangle\cdot\langle0,0,-1\rangle\) times the area \(2\), which is \(-2\cdot2=-4\). Keeping the upward normal while reversing the boundary would pair inconsistent orientations and produce a false comparison.
A Reliable Calculation Order
Write the three vertices in the order the boundary is followed, including the closing edge.
Use a parameter running from \(0\) to \(1\) in the directed edge’s actual direction, then evaluate \(\mathbf{F}\cdot\mathbf{r}'(t)\).
Differentiate the field to obtain its curl, and use the same vertex order to determine the patch’s normal direction.
Evaluate the curl flux over the triangular patch and check that it equals the sum of the three edge contributions.
These steps make a triangle calculation transparent: the edge sum shows how each part of the boundary contributes, while the curl flux checks the result through the interior patch. A disagreement is a reason to inspect edge direction, dot products, curl components, or the orientation pairing before assuming the theorem has failed.
Check Your Understanding
For each question, keep track of both the directed boundary and its compatible surface orientation.
- For an affine parameterization \(\mathbf{r}(u,v)=A+u(B-A)+v(C-A)\), what region in the \(uv\)-plane parameterizes the triangular patch?
- What oriented area vector is determined by the ordered vertices \(A,B,C\)?
- If the boundary traversal is reversed, what happens to the circulation and to the compatible curl flux?
- Why must the three edge integrals be added as directed integrals rather than as unsigned quantities?
- For a horizontal triangle with upward normal, which component of the curl contributes to the flux?