Why a Self-Intersection Changes the Calculation
A simple closed curve has a clear interior and a consistent orientation: as you travel around it, the interior stays on one side. A self-intersecting path can pass through the same point more than once, so its interior is not described by one simple region with one boundary orientation. Some lobes may be traversed counterclockwise and others clockwise.
The useful strategy is to split the path at its crossings. This produces simple closed curves, or loops, to which Green’s theorem applies separately. The original line integral is the sum of the integrals around those loops, with each loop’s actual direction preserved.
Definitions and Green’s Theorem for Each Piece
A simple closed curve is a closed curve that does not cross or touch itself, apart from its starting and ending point. A piecewise-smooth curve consists of finitely many smooth pieces joined end to end. In this tutorial, assume the closed path is piecewise smooth and has finitely many isolated self-intersections, with no overlapping arcs. These assumptions let us split the path at its crossings into finitely many simple loops. A segment that is exactly retraced in the opposite direction contributes zero to a line integral and can be cancelled.
A loop has positive orientation when it is traversed counterclockwise, and negative orientation when it is traversed clockwise. Green’s theorem, in circulation form, states that if \(C\) is a positively oriented, piecewise-smooth simple closed boundary of a region \(R\), and \(P\) and \(Q\) have continuous first partial derivatives on an open set containing \(R\), then
If the same boundary is traversed clockwise, its line integral is the negative of the right-hand side. Write \(\varepsilon=1\) for a counterclockwise loop and \(\varepsilon=-1\) for a clockwise loop. Its contribution is then \(\varepsilon\) times the double integral over the region it encloses.
Why the Loop Contributions Add
The addition rule follows from the additivity of a line integral over consecutive portions of a parametrized path. Splitting a path at a parameter value changes neither the path nor its integral: it only writes the integral as a sum over subintervals. At a crossing, splitting and reconnecting the pieces in closed cycles partitions the same directed path pieces among the loops.
Apply Green’s theorem to each loop, using its actual orientation. This proves the decomposition formula:
The regions \(R_i\) need not be disjoint. If they overlap, the displayed sum counts the curl density once for each positively oriented covering and subtracts it for each negatively oriented covering. Equivalently, define the signed multiplicity \(m(x,y)=\sum_{i=1}^{k}\varepsilon_i\,\mathbf{1}_{R_i}(x,y)\), where \(\mathbf{1}_{R_i}\) is \(1\) inside \(R_i\) and \(0\) outside. The result is the integral of the curl density weighted by \(m\). Thus, a region covered once in each direction cancels rather than contributing twice.
Split the parameterized path at every self-intersection, so each directed arc has crossing points as endpoints.
Follow the directed arcs into closed cycles. If a cycle still self-intersects, split it again until each piece is simple.
For every loop, determine whether it is counterclockwise or clockwise; record the sign before evaluating any area integral.
Use Green’s theorem on each simple loop and add the signed contributions. Check that the pieces trace the original path with no omissions or extra arcs.
Worked Example: A Figure Eight with Oppositely Oriented Lobes
Worked Example: A Figure Eight with Oppositely Oriented Lobes
Let \(C\) trace two triangular lobes meeting at the origin. The first is \(O\to(1,0)\to(0,1)\to O\), and the second is \(O\to(0,-2)\to(-1,0)\to O\). Compute the circulation for \(P=-y/2\) and \(Q=x/2\).
The curl density is
The first triangle is counterclockwise and has area \(\frac12\). Its contribution is \(1\cdot\frac12=\frac12\). The second triangle is clockwise and has area \(1\), so its contribution is \(-1\cdot1=-1\). Therefore, Green’s theorem gives
To check the result directly, on a straight segment from \((x_0,y_0)\) to \((x_1,y_1)\), this field has integral \(\frac12(x_0y_1-y_0x_1)\). For the first lobe, the three segment contributions are \(0\), \(\frac12\), and \(0\), giving \(\frac12\). For the second lobe they are \(0\), \(-1\), and \(0\), giving \(-1\). Their sum is \(-\frac12\), as required.
Worked Example: A Bow-Tie Path with a Nonconstant Curl
Worked Example: A Bow-Tie Path with a Nonconstant Curl
Consider the closed polygonal path with vertices \((-1,-1)\to(1,1)\to(-1,1)\to(1,-1)\to(-1,-1)\). It crosses itself at \(O=(0,0)\). Compute its circulation for \(P=-y\) and \(Q=x^2/2\).
The crossing splits the path into an upper triangular loop \(O\to(1,1)\to(-1,1)\to O\), which is counterclockwise, and a lower triangular loop \(O\to(1,-1)\to(-1,-1)\to O\), which is clockwise. The curl density is
Each triangle has area \(1\) and centroid with \(x\)-coordinate \(0\). Since \(x+1\) is linear, its average over either triangle equals its value at the centroid, namely \(1\). Thus the upper loop contributes \(1\cdot1=1\), while the clockwise lower loop contributes \(-1\cdot1=-1\). The total is \(0\).
Check directly on the split boundary using \(-y\,dx+(x^2/2)\,dy\). On the upper loop, the edge integrals are as follows:
The upper sum is \(-\frac13+2-\frac23=1\). On the lower loop, the three edge integrals are \(\frac13\), \(-2\), and \(\frac23\), giving \(\frac13-2+\frac23=-1\). The two loop integrals therefore sum to \(1+(-1)=0\), confirming the Green’s theorem calculation.
Worked Example: Different Lobes and a Linear Curl Density
Worked Example: Different Lobes and a Linear Curl Density
Let \(C\) consist of two triangular loops meeting at the origin. The first is \(O\to(2,0)\to(0,1)\to O\), counterclockwise. The second is \(O\to(0,-1)\to(-1,0)\to O\), clockwise. Find the circulation for \(P=0\) and \(Q=x^2+xy\).
The curl density is \(2x+y\). The first triangle has area \(1\) and centroid \((2/3,1/3)\), so the integral of this linear function over the triangle is its value at the centroid times the area:
The second triangle has area \(\frac12\) and centroid \((-1/3,-1/3)\). Its positively oriented region integral is \(\left(2(-1/3)-1/3\right)\frac12=-\frac12\). Since its boundary is clockwise, its contribution to the path integral is the negative of that region integral: \(\frac12\). The total circulation is therefore \(\frac53+\frac12=\frac{13}{6}\).
For a direct check, since \(P=0\), each segment contributes \(\int Q\,dy\). On the first loop, the edges \(O\to(2,0)\), \((0,1)\to O\) contribute \(0\), because \(dy=0\) or \(Q=0\). On the middle edge \((2,0)\to(0,1)\), use \(x=2-2t,\ y=t,\ dy=dt\), for \(0\le t\le1\). Then \(Q=4-6t+2t^2\), so the contribution is \(4-3+\frac23=\frac53\).
On the second loop, the first and last edges likewise contribute \(0\). For \((0,-1)\to(-1,0)\), set \(x=-t,\ y=-1+t,\ dy=dt\). Then \(Q=t^2+t-t^2=t\), and the integral is \(\int_0^1t\,dt=\frac12\). The direct total is \(\frac53+\frac12=\frac{13}{6}\), matching the area calculation.
A Common Mistake: Counting Every Lobe as Positive
For the bow-tie example, the two triangular interiors have total geometric area \(1+1=2\). A tempting but incorrect calculation is to multiply that unsigned area by the curl density’s average value \(1\) and conclude that the circulation is \(2\). This ignores that the lower triangle is traversed clockwise.
The correct signed sum is \(1-1=0\). The fact that the two lobes have equal area does not mean their contributions reinforce each other: orientation determines whether each contribution is added or subtracted. More generally, if the curl density varies, the signed integrals over the individual regions must be computed; subtracting unsigned areas alone is not enough.
Why the Decomposition Matters
A self-intersecting path is not, as a whole, the simple boundary of one ordinary region, so the hypotheses of Green’s theorem do not apply to it as though it had one unambiguous interior. Decomposition supplies simple boundaries that do meet those hypotheses. It also preserves information that a sketch of the filled-in interior can hide: which regions are traversed in each direction and whether their contributions cancel.
When different simple loops overlap, use the signed sum over their regions rather than the area of their union. The formula remains valid because it follows from the original path’s line integral, piece by piece. A useful final check is to compare that sum with a direct boundary calculation when the segments or arcs are manageable.
Check Your Understanding
Use the definitions and signed-loop formula to answer these questions.
- What assumptions on the path allow its crossings to be split into finitely many simple loops?
- How does the Green’s theorem contribution change when a loop is traversed clockwise instead of counterclockwise?
- If a point lies inside one counterclockwise loop and one clockwise loop, what is its signed multiplicity?
- Why is the unsigned total area of the bow-tie’s two lobes not enough to compute its circulation?
- What property of line integrals justifies adding the integrals around the loops in a decomposition?