Antiderivatives and Path Independence

Suppose you are handed a holomorphic function on some domain. You are asked to integrate it along a wiggly path from a point aa to a point bb. Must you know the exact shape of the path to compute the answer? Or does the value depend only on the endpoints? This is exactly the question calculus students ask about ∫abf′(x) dx\int_a^b f'(x)\,dx on the real line. Complex analysis gives it a strikingly clean answer, but only under the right hypotheses. The naive guess is that every holomorphic ff has an antiderivative FF with F′=fF' = f. This would mean integrals become endpoint evaluations. But this guess is false in general. The standard counterexample is f(z)=1/zf(z) = 1/z on the punctured plane. The resolution of this tension is one of the structural pillars of complex analysis. Path independence of contour integrals is equivalent to the existence of a primitive. Both are governed by the topology of the domain in question. Specifically, both depend on whether closed curves in it bound. This connects the purely analytic question ("does an antiderivative exist?") to the purely topological one ("is the domain simply connected?"). The bridge between them is built by Cauchy's Theorem. The big picture is that holomorphic functions are extraordinarily rigid. Once we know ∮γf dz=0\oint_\gamma f\,dz = 0 for every closed contour γ\gamma in a domain DD, we can manufacture a primitive FF. We do this by integrating from a fixed basepoint, exactly as in single-variable calculus. The fundamental theorem of calculus for contours falls out immediately. This article develops that machinery precisely. It states and proves the equivalence between path independence and primitive existence. It proves the fundamental theorem for contour integrals. It shows exactly how simple connectivity enters through Cauchy's theorem. This guarantees primitives on well-behaved domains. This material is a hinge point in the subject. It converts Cauchy's theorem — a statement about vanishing integrals — into a genuine calculus of antiderivatives for complex functions. It also sets up further machinery: Cauchy's integral formula, and then Power Series and Analyticity. That machinery shows every holomorphic function is in fact given locally by a convergent power series.

Key takeaways
  • A primitive (antiderivative) of ff on a domain DD is a holomorphic FF with F′=fF'=f throughout DD.
  • Path independence of ∫γf dz\int_\gamma f\,dz (depending only on endpoints, for paths in DD) is logically equivalent to the existence of a primitive of ff on DD — no extra topological hypothesis is needed for this equivalence itself.
  • Equivalently, path independence holds iff ∮γf dz=0\oint_\gamma f\,dz = 0 for every closed contour γ\gamma in DD.
  • When a primitive exists, the fundamental theorem for contour integrals gives ∫γf dz=F(b)−F(a)\int_\gamma f\,dz = F(b) - F(a) for any contour from aa to bb.
  • On a simply connected domain, Cauchy's Theorem guarantees ∮γf dz=0\oint_\gamma f \, dz = 0 automatically for every holomorphic ff, hence guarantees a primitive exists — this is where topology enters.
  • The function 1/z1/z on C∖{0}\mathbb{C}\setminus\{0\} has no primitive there, showing simple connectivity (or some substitute) cannot be dropped.
  • Key idea: existence of an antiderivative and independence of path are two faces of the same fact, and simple connectivity is the topological guarantee that turns "holomorphic" into "has a primitive."

Where this lives today

The primitive/path-independence circle of ideas is the direct ancestor of residue calculus: since 1/z1/z has no primitive on the punctured disk, its nonzero closed-loop integral (2πi2\pi i) becomes the seed of the residue theorem, used to evaluate real integrals, sum series, and invert Laplace and Fourier transforms in engineering and physics. In complex dynamics and Riemann surface theory, the failure of a single-valued primitive for 1/z1/z is exactly why log⁡z\log z is genuinely multivalued, motivating the construction of Riemann surfaces as the natural domain on which such "primitives" become well-defined and single-valued. In potential theory and electrostatics, holomorphic primitives correspond to complex potentials for two-dimensional irrotational, incompressible flows and electrostatic fields; path independence is the mathematical expression of a field being conservative. In algebraic topology, the equivalence between vanishing loop integrals and simple connectivity is a two-dimensional shadow of de Rham cohomology, where closed forms are exact precisely when the relevant cohomology group vanishes — the complex-analytic story here is the historically first and most concrete instance of that general phenomenon. Finally, this machinery is the direct stepping stone to Power Series and Analyticity: Cauchy's integral formula, proved via the same contour-integral techniques, shows every holomorphic function has a locally convergent power series, and much of that proof reuses the estimates on contour integrals developed here.

Ask the tutor: Why does the existence of a primitive for 1/z21/z^2 on C∖{0}\mathbb C \setminus \{0\} not contradict the fact that this domain isn't simply connected?

DefinitionPrimitive, Contour Integral Path Independence

Let D⊆CD \subseteq \mathbb{C} be a domain (open and connected), and let f:D→Cf : D \to \mathbb{C} be continuous.

Primitive. A function F:D→CF: D \to \mathbb{C} is a primitive (or antiderivative) of ff on DD if FF is holomorphic on DD and

F′(z)=f(z)for all z∈D.F'(z) = f(z) \quad \text{for all } z \in D.

Contour. A contour γ:[a,b]→D\gamma:[a,b]\to D is a piecewise C1C^1 path (continuous, with a continuous piecewise derivative γ′\gamma', finitely many corners allowed).

Contour integral. For ff continuous on DD and γ\gamma a contour in DD,

∫γf(z) dz:=∫abf(γ(t)) γ′(t) dt.\int_\gamma f(z)\, dz := \int_a^b f(\gamma(t))\,\gamma'(t)\, dt.

Path independence. We say ∫γf dz\int_\gamma f\,dz is path-independent in DD if for every two contours γ1,γ2\gamma_1, \gamma_2 in DD with the same initial point pp and same terminal point qq,

∫γ1f dz=∫γ2f dz.\int_{\gamma_1} f\, dz = \int_{\gamma_2} f\, dz.

Intuition

Think of f(z) dzf(z)\,dz as a "1-form" you integrate along a path, exactly like a work integral ∫F⋅dr\int \mathbf F \cdot d\mathbf r in vector calculus. In physics, a force field is conservative — path-independent, equivalently has a potential — precisely when its circulation around every closed loop vanishes. Complex antiderivatives play the same role as potentials: FF is a potential for f dzf\,dz, and F′=fF'=f says f dz=dFf\,dz = dF is an exact differential. The reason 1/z1/z fails to have a global primitive on C∖{0}\mathbb C\setminus\{0\} is topological: the domain has a "hole" at the origin, and a loop around that hole picks up a nonzero circulation 2πi2\pi i. No potential function can be single-valued around such a loop, because going once around and returning to your starting point should return FF to its original value, yet the "would-be" primitive log⁡z\log z increases by 2πi2\pi i. Locally, near any point away from 00, a single-valued branch of log⁡z\log z exists and is a primitive — this is the local/global tension that simple connectivity resolves.

Figure
Two homotopic paths in a simply connected slit domain give equal integrals; a loop around the puncture in C∖{0}\mathbb C\setminus\{0\} does not.
D: slit plane (simply connected) a b γ₁ γ₂ ∫_γ₁ f dz = ∫_γ₂ f dz = F(b) − F(a) ℂ∖{0}: loop around puncture 0 γ: |z|=1 ∮_γ (1/z) dz = 2πi ≠ 0
TheoremPath Independence ⇔ Existence of a Primitive

Let D⊆CD \subseteq \mathbb{C} be a domain and let f:D→Cf: D \to \mathbb{C} be continuous. The following three statements are equivalent. 1. ff has a primitive FF on DD. 2. For every contour γ\gamma in DD, ∫γf dz\int_\gamma f\,dz depends only on the initial and terminal points of γ\gamma (path independence). 3. For every closed contour γ\gamma in DD, ∮γf(z) dz=0\displaystyle\oint_\gamma f(z)\,dz = 0. Moreover, when these hold, for any contour γ\gamma in DD from pp to qq,

∫γf(z) dz=F(q)−F(p),\int_\gamma f(z)\, dz = F(q) - F(p),

(Fundamental Theorem for Contour Integrals).

Proof

We prove (1)⇒(2)⇒(3)⇒(1)(1)\Rightarrow(2)\Rightarrow(3)\Rightarrow(1), and along the way establish the displayed formula. The technique throughout is direct computation via the chain rule and the single-variable Fundamental Theorem of Calculus, reducing a complex-variable statement to real-variable calculus applied to real and imaginary parts.

Step 1: (1)⇒(2)(1)\Rightarrow(2), with the explicit formula. Suppose FF is a primitive of ff on DD: FF is holomorphic on DD and F′=fF'=f everywhere on DD. Let γ:[a,b]→D\gamma:[a,b]\to D be a contour, first assumed C1C^1 (no corners); we handle corners below. Define g(t)=F(γ(t))g(t) = F(\gamma(t)) for t∈[a,b]t\in[a,b]. Since FF is holomorphic and γ\gamma is a real-differentiable path with values in DD, the complex chain rule applies: FF is complex-differentiable at each γ(t)\gamma(t), and γ\gamma is differentiable at tt, so

g′(t)=F′(γ(t)) γ′(t)=f(γ(t)) γ′(t).g'(t) = F'(\gamma(t))\, \gamma'(t) = f(\gamma(t))\,\gamma'(t).

(This chain rule is justified exactly as in the real case: write F(γ(t+h))−F(γ(t))=F′(γ(t))(γ(t+h)−γ(t))+o(∣γ(t+h)−γ(t)∣)F(\gamma(t+h)) - F(\gamma(t)) = F'(\gamma(t))(\gamma(t+h)-\gamma(t)) + o(|\gamma(t+h)-\gamma(t)|) using complex differentiability of FF, divide by hh, and let h→0h \to 0, using continuity and differentiability of γ\gamma.) The function g:[a,b]→Cg:[a,b]\to \mathbb C is C1C^1 since g′=f(γ(⋅))γ′(⋅)g' = f(\gamma(\cdot))\gamma'(\cdot) is continuous (ff continuous, γ′\gamma' continuous). Write g(t)=u(t)+iv(t)g(t) = u(t) + iv(t) with u,vu,v real-valued C1C^1 functions on [a,b][a,b]. By the real Fundamental Theorem of Calculus applied separately to uu and vv,

∫abg′(t) dt=∫abu′(t) dt+i∫abv′(t) dt=(u(b)−u(a))+i(v(b)−v(a))=g(b)−g(a).\int_a^b g'(t)\,dt = \int_a^b u'(t)\,dt + i\int_a^b v'(t)\,dt = \big(u(b)-u(a)\big) + i\big(v(b)-v(a)\big) = g(b)-g(a).

Hence

∫γf(z) dz=∫abf(γ(t))γ′(t) dt=∫abg′(t) dt=g(b)−g(a)=F(γ(b))−F(γ(a))=F(q)−F(p),\int_\gamma f(z)\,dz = \int_a^b f(\gamma(t))\gamma'(t)\,dt = \int_a^b g'(t)\,dt = g(b) - g(a) = F(\gamma(b)) - F(\gamma(a)) = F(q) - F(p),

where p=γ(a)p = \gamma(a), q=γ(b)q=\gamma(b). If γ\gamma is only piecewise C1C^1, with corners at a=t0<t1<⋯<tn=ba = t_0 < t_1 < \cdots < t_n = b, apply the computation above on each subinterval [tk−1,tk][t_{k-1}, t_k] (on which γ\gamma restricts to a C1C^1 contour) to get

∫γf dz=∑k=1n(F(γ(tk))−F(γ(tk−1))),\int_\gamma f\,dz = \sum_{k=1}^n \big(F(\gamma(t_k)) - F(\gamma(t_{k-1}))\big),

a telescoping sum equal to F(γ(tn))−F(γ(t0))=F(q)−F(p)F(\gamma(t_n)) - F(\gamma(t_0)) = F(q)-F(p). This proves the displayed formula, and in particular the right-hand side depends only on p,qp,q (through FF), not on the specific contour γ\gamma joining them: this is exactly path independence, statement (2).

Step 2: (2)⇒(3)(2)\Rightarrow(3). Let γ\gamma be any closed contour in DD, say γ:[a,b]→D\gamma:[a,b]\to D with γ(a)=γ(b)=p\gamma(a) = \gamma(b) = p. Let γ0\gamma_0 denote the constant contour at pp (i.e. γ0(t)≡p\gamma_0(t) \equiv p on, say, [a,b][a,b]); this is a trivial contour from pp to pp lying in DD, and clearly ∫γ0f dz=0\int_{\gamma_0} f\,dz = 0 since γ0′≡0\gamma_0' \equiv 0. Both γ\gamma and γ0\gamma_0 are contours in DD from pp to pp. By path independence (2), applied with initial point pp and terminal point pp,

∮γf dz=∫γ0f dz=0.\oint_\gamma f\,dz = \int_{\gamma_0} f\,dz = 0.

This is exactly statement (3).

Step 3: (3)⇒(1)(3)\Rightarrow(1). Assume every closed contour integral of ff in DD vanishes. We construct a primitive directly, mimicking the construction of an antiderivative via F(x)=∫x0xfF(x) = \int_{x_0}^x f from single-variable calculus. Fix a basepoint. Since DD is open and connected, and open connected subsets of C\mathbb{C} are path-connected by piecewise-linear (in particular piecewise C1C^1) paths — a standard fact: the set of points reachable from a fixed z0∈Dz_0\in D by a polygonal path in DD is open (since DD is open, any small disk around a reachable point is still reachable by extending the polygonal path) and its complement in DD is also open by the same argument, so by connectedness of DD this set is all of DD — we may fix z0∈Dz_0 \in D and, for each z∈Dz \in D, choose a contour γz\gamma_z in DD from z0z_0 to zz (a finite polygonal path suffices). Define

F(z):=∫γzf(w) dw.F(z) := \int_{\gamma_z} f(w)\,dw.

Well-definedness. If γz,γ~z\gamma_z, \tilde\gamma_z are two contours in DD from z0z_0 to zz, form the closed contour Γ\Gamma obtained by traversing γz\gamma_z then γ~z\tilde\gamma_z reversed (concatenation of piecewise-C1C^1 paths is again piecewise C1C^1, and reversal negates the integral: ∫−σf dz=−∫σf dz\int_{-\sigma} f\,dz = -\int_\sigma f\,dz, directly from the substitution t↦a+b−tt \mapsto a+b-t in the definition of the contour integral). Then

0=∮Γf dz=∫γzf dz−∫γ~zf dz0 = \oint_\Gamma f\,dz = \int_{\gamma_z} f\,dz - \int_{\tilde\gamma_z} f\, dz

by hypothesis (3), so ∫γzf dz=∫γ~zf dz\int_{\gamma_z} f\,dz = \int_{\tilde\gamma_z} f\,dz. Hence F(z)F(z) does not depend on the choice of contour, and FF is well-defined on DD. Differentiability and F′=fF' = f. Fix z1∈Dz_1 \in D. Since DD is open, choose r>0r>0 with the disk B(z1,r)⊂DB(z_1, r) \subset D. For h∈Ch \in \mathbb{C} with 0<∣h∣<r0 < |h| < r, let σh\sigma_h be the straight segment from z1z_1 to z1+hz_1 + h; this lies inside B(z1,r)⊂DB(z_1,r) \subset D. Concatenating γz1\gamma_{z_1} (from z0z_0 to z1z_1) with σh\sigma_h gives a contour in DD from z0z_0 to z1+hz_1+h, so by well-definedness of FF,

F(z1+h)−F(z1)=∫σhf(w) dw.F(z_1 + h) - F(z_1) = \int_{\sigma_h} f(w)\, dw.

Now write, using ∫σhdw=h\int_{\sigma_h} dw = h (since σh(t)=z1+th\sigma_h(t) = z_1 + th, t∈[0,1]t\in[0,1], gives ∫01h dt=h\int_0^1 h\,dt = h),

F(z1+h)−F(z1)−f(z1)h=∫σh(f(w)−f(z1)) dw.F(z_1+h) - F(z_1) - f(z_1) h = \int_{\sigma_h} \big(f(w) - f(z_1)\big)\, dw.

Since ff is continuous at z1z_1, given ε>0\varepsilon>0 there is δ∈(0,r)\delta \in (0,r) such that ∣f(w)−f(z1)∣<ε|f(w) - f(z_1)| < \varepsilon whenever ∣w−z1∣<δ|w - z_1| < \delta. For ∣h∣<δ|h| < \delta, every point ww on σh\sigma_h satisfies ∣w−z1∣≤∣h∣<δ|w - z_1| \le |h| < \delta, so by the standard estimate ∣∫σg∣≤(sup⁡σ∣g∣)⋅length(σ)\left|\int_\sigma g\right| \le (\sup_\sigma |g|)\cdot \mathrm{length}(\sigma),

∣∫σh(f(w)−f(z1)) dw∣≤ε⋅∣h∣.\left| \int_{\sigma_h} \big(f(w)-f(z_1)\big)\,dw \right| \le \varepsilon \cdot |h|.

Hence

∣F(z1+h)−F(z1)h−f(z1)∣≤εwhenever 0<∣h∣<δ.\left| \frac{F(z_1+h)-F(z_1)}{h} - f(z_1) \right| \le \varepsilon \quad \text{whenever } 0<|h|<\delta.

Since ε>0\varepsilon>0 was arbitrary, lim⁡h→0F(z1+h)−F(z1)h=f(z1)\displaystyle\lim_{h\to 0} \frac{F(z_1+h)-F(z_1)}{h} = f(z_1), i.e. FF is complex-differentiable at z1z_1 with F′(z1)=f(z1)F'(z_1) = f(z_1). Since z1∈Dz_1 \in D was arbitrary, FF is holomorphic on all of DD with F′=fF' = f, i.e. FF is a primitive of ff. This proves (1).

Conclusion. We have shown (1)⇒(2)⇒(3)⇒(1)(1)\Rightarrow(2)\Rightarrow(3)\Rightarrow(1), so all three statements are equivalent, and Step 1 additionally established the fundamental-theorem formula whenever a primitive exists. ■\blacksquare

∎
CorollarySimple connectivity guarantees a primitive

Let DD be a simply connected domain and ff holomorphic on DD. Then ff has a primitive on DD, and consequently ∫γf dz\int_\gamma f \,dz is path-independent in DD and ∮γf dz=0\oint_\gamma f\,dz = 0 for every closed contour γ\gamma in DD.

Proof

By Cauchy's Theorem (in its homotopy or homology form for simply connected domains), if DD is simply connected and ff is holomorphic on DD, then ∮γf dz=0\oint_\gamma f\,dz = 0 for every closed contour γ\gamma in DD. This is exactly hypothesis (3) of the Path Independence ⇔ Existence of a Primitive theorem, whose proof (Step 3, not depending on this corollary) already shows (3)⇒(1)(3)\Rightarrow(1). Hence ff has a primitive on DD, and the remaining conclusions follow from the same theorem. ■\blacksquare

∎

Worked examples

Example 1: A primitive that is not an elementary formula lookup, on an asymmetric contour. Let D=CD = \mathbb{C} and f(z)=zez2f(z) = z e^{z^2}, which is entire (holomorphic on all of C\mathbb C). We seek ∫γf(z) dz\int_\gamma f(z)\,dz where γ\gamma is the contour that goes from z=1z=1 straight to z=iz = i, then along the parabola Im(w)=(Re(w)−1)2⋅1(i−1)-scaling\mathrm{Im}(w) = (\mathrm{Re}(w)-1)^2 \cdot \frac{1}{ (i - 1)\text{-scaling}}... — rather than track that curve explicitly, we exploit path independence: since C\mathbb C is simply connected and ff is entire, Corollary Simple connectivity guarantees a primitive guarantees a primitive exists, so we only need endpoints. Find the primitive. Guess F(z)=12ez2F(z) = \tfrac12 e^{z^2}. Check: F′(z)=12⋅2z⋅ez2=zez2=f(z)F'(z) = \tfrac12 \cdot 2z \cdot e^{z^2} = z e^{z^2} = f(z). So FF is a genuine primitive on all of C\mathbb C. Apply the Fundamental Theorem. Let γ\gamma be any contour from p=1p = 1 to q=3+2iq = 3+2i (an intentionally asymmetric, arbitrary endpoint) lying in C\mathbb{C} — the formula does not care which contour, by path independence:

∫γzez2 dz=F(3+2i)−F(1)=12e(3+2i)2−12e1.\int_\gamma z e^{z^2}\, dz = F(3+2i) - F(1) = \tfrac12 e^{(3+2i)^2} - \tfrac12 e^{1}.

Compute (3+2i)2=9+12i+4i2=9−4+12i=5+12i(3+2i)^2 = 9 + 12i + 4i^2 = 9 - 4 + 12i = 5 + 12i. So the answer is

∫γzez2 dz=12e5+12i−12e=12e5(cos⁡12+isin⁡12)−12e.\int_\gamma z e^{z^2}\, dz = \tfrac12 e^{5+12i} - \tfrac12 e = \tfrac12 e^5\big(\cos 12 + i \sin 12\big) - \tfrac12 e.

This is an exact closed form, obtained with zero knowledge of the actual shape of γ\gamma — only its endpoints 11 and 3+2i3+2i mattered, exactly because ff is entire (hence has a primitive on the simply connected domain C\mathbb{C}).

Example 2: A domain that is not simply connected, where path-independence must be checked directly (and can still hold for a special ff). Let D=C∖{0}D = \mathbb{C} \setminus \{0\} (not simply connected — it has a "hole") and consider f(z)=1/z2f(z) = 1/z^2 (note: not 1/z1/z). Is ∫γf dz\int_\gamma f\,dz path-independent in DD? DD is not simply connected, so the Corollary does not automatically apply — we must check directly whether a primitive exists. Candidate primitive. Try F(z)=−1/zF(z) = -1/z, holomorphic on DD (since z≠0z\ne 0 there). Then F′(z)=1/z2=f(z)F'(z) = 1/z^2 = f(z). So FF is a genuine primitive of ff on DD, even though DD is not simply connected! By the Theorem, (1)⇒(2)⇒(3)(1)\Rightarrow(2)\Rightarrow(3), so path independence and vanishing of all closed contour integrals hold for this particular ff, despite the topological "hole." Verify directly on a closed loop, as a sanity check. Let γ(t)=eit\gamma(t) = e^{it}, t∈[0,2π]t \in [0,2\pi] (unit circle, traversed once counterclockwise). Then γ′(t)=ieit\gamma'(t) = ie^{it}, and

∮γ1z2 dz=∫02π1e2it⋅ieit dt=i∫02πe−it dt=i[e−it−i]02π=−(e−2πi−e0)=−(1−1)=0,\oint_\gamma \frac{1}{z^2}\,dz = \int_0^{2\pi} \frac{1}{e^{2it}}\cdot i e^{it}\, dt = i\int_0^{2\pi} e^{-it}\,dt = i \left[\frac{e^{-it}}{-i}\right]_0^{2\pi} = -\big(e^{-2\pi i} - e^{0}\big) = -(1-1) = 0,

confirming ∮γz−2 dz=0\oint_\gamma z^{-2}\,dz = 0, consistent with the primitive's existence. Contrast this immediately with f(z)=1/zf(z) = 1/z on the same domain DD: there, ∮γ1z dz=∫02πieiteit dt=∫02πi dt=2πi≠0\displaystyle\oint_\gamma \frac1z\,dz = \int_0^{2\pi} \frac{ie^{it}}{e^{it}}\,dt = \int_0^{2\pi} i\,dt = 2\pi i \ne 0, so 1/z1/z has no primitive on DD — the topological hole matters for 1/z1/z but is irrelevant for 1/z21/z^2, illustrating that simple connectivity is sufficient, not necessary, for primitive existence.

CounterexampleWhy simple connectivity cannot be dropped from the Corollary

The Corollary requires DD to be simply connected. Drop only that hypothesis, keeping ff holomorphic on DD, and the conclusion can fail. Take D=C∖{0}D = \mathbb{C}\setminus\{0\}, which is a domain (open, connected) but not simply connected (the loop ∣z∣=1|z|=1 cannot be contracted to a point within DD, since any contraction would have to sweep across the removed point 00). Take f(z)=1/zf(z) = 1/z, holomorphic on DD. Claim: ff has no primitive on DD. Suppose, for contradiction, that FF were a primitive of ff on DD. By the Theorem ((1)⇒(3)(1) \Rightarrow (3), which uses no simple-connectivity hypothesis and is unconditionally true), we would need

∮γ1z dz=0\oint_\gamma \frac{1}{z}\, dz = 0

for every closed contour γ\gamma in DD. But taking γ(t)=eit\gamma(t) = e^{it}, t∈[0,2π]t\in[0,2\pi], direct computation (as in Example 2 above) gives

∮γ1z dz=2πi≠0.\oint_\gamma \frac{1}{z}\, dz = 2\pi i \ne 0.

This contradicts the requirement. Hence no primitive of 1/z1/z exists on C∖{0}\mathbb{C}\setminus\{0\}. This shows the Corollary's simple-connectivity hypothesis is doing real work: it is exactly what rules out this failure by forcing all closed-loop integrals to vanish via Cauchy's Theorem. Note the Theorem itself (the three-way equivalence) is not violated — it holds unconditionally for any domain; what fails here is only that hypothesis (3) of the Theorem is false on this particular non-simply-connected DD for this particular ff, which is precisely the scenario the Corollary is designed to prevent by adding the topological hypothesis.

Scope, edge cases & common traps

SituationWhat is trueCommon error
DD simply connected, ff holomorphic on DDPrimitive exists on all of DD; path independence holdsAssuming this works for domains with holes too
DD has holes, but ff happens to have vanishing residues / is itself a derivativePrimitive can still exist (e.g. 1/z21/z^2)Assuming any hole automatically kills every primitive
ff merely continuous, not holomorphicThe Theorem's equivalence (1)⇔(2)⇔(3) still holds (proof never used holomorphy of ff beyond continuity) but Cauchy's theorem (which produces case (3) automatically) needs holomorphyConflating "has antiderivative" with "is holomorphic"; these are different roles
Multi-valued "functions" like log⁡z\log z, z1/2z^{1/2} on C∖{0}\mathbb C \setminus \{0\}No single-valued holomorphic branch exists on all of C∖{0}\mathbb C\setminus\{0\}; branches exist on slit simply-connected subdomainsWriting "F(z)=log⁡zF(z) = \log z" as if it were single-valued globally
Piecewise C1C^1 vs merely continuous contoursTheorem stated and proved for piecewise C1C^1 contours (needed for γ′\gamma' to make sense)Trying to apply the chain-rule argument to a merely continuous, nowhere-differentiable path

The mathematician behind it

Origins · 1 minAugustin-Louis Cauchy and how this was discovered
AC
Augustin-Louis Cauchy
1789–1857 · Founder of rigorous complex analysis, École Polytechnique and Collège de France, Paris

Cauchy built the theory of contour integration and the theorem bearing his name that makes path independence provable rather than merely observed. His insistence on rigor — epsilon-delta arguments, careful definitions of continuity and convergence — transformed calculus and analysis from a computational art into a deductive science, and his 1825 memoir on integrals between imaginary limits is the direct ancestor of the theorem proved in this article.

History
- 1814 — Cauchy submits his first major memoir on definite integrals, containing early contour-integral ideas.
- 1825 — Cauchy publishes Mémoire sur les intégrales définies, prises entre des limites imaginaires, essentially proving Cauchy's theorem.
- 1831 — Cauchy, in exile in Turin, develops the integral formula bearing his name, deepening the primitive/path-independence circle of ideas.
- 1851 — Riemann's dissertation reframes these results geometrically via conformal mapping and Riemann surfaces, clarifying the role of multi-valuedness (e.g. log⁡z\log z) that this article's counterexample exhibits.
- 1900s — The homotopy and homology formulations of Cauchy's theorem (Goursat, and later textbook treatments) isolate exactly the simple-connectivity hypothesis used in the Corollary above.
Does every holomorphic function have a primitive on its domain of definition?

No. Existence of a primitive depends on the domain's topology, not just on holomorphy. A function like f(z)=1/zf(z) = 1/z is holomorphic on C∖{0}\mathbb{C}\setminus\{0\} but has no primitive there, because its integral around a loop encircling the origin is 2πi≠02\pi i \ne 0, violating the necessary condition from the Theorem. However, restricted to any simply connected subdomain not containing 00 (for instance the plane slit along the negative real axis), 1/z1/z does have a primitive, namely a branch of log⁡z\log z. So the answer depends essentially on whether the domain is simply connected, per the Corollary.

Is simple connectivity necessary for a primitive to exist, or just sufficient?

Only sufficient, and only in the "for every holomorphic ff on DD" sense. As Example 2 shows, f(z)=1/z2f(z)=1/z^2 has a primitive −1/z-1/z on the punctured plane C∖{0}\mathbb{C}\setminus\{0\}, which is not simply connected. So a specific holomorphic function can have a primitive on a non-simply-connected domain. What simple connectivity guarantees is uniform: every holomorphic function on a simply connected domain has a primitive there, with no exceptions and no case-by-case checking, via Cauchy's theorem forcing all closed-loop integrals to vanish.

How is this related to Cauchy's Theorem exactly?

Cauchy's Theorem states that if DD is simply connected and ff is holomorphic on DD, then ∮γf dz=0\oint_\gamma f\,dz = 0 for every closed contour γ\gamma in DD. That is precisely condition (3) in the Path Independence Theorem here. The Theorem's proof of (3)⇒(1)(3)\Rightarrow(1) then manufactures an actual primitive from that vanishing condition. So Cauchy's Theorem supplies the topological input (vanishing loop integrals), and the machinery in this article converts that input into the existence of an antiderivative and the fundamental theorem for contour integrals.

Why does the proof of (3)⇒(1)(3)\Rightarrow(1) bother with a "fixed basepoint" and specific contours to define FF?

This mirrors the real-variable construction F(x)=∫x0xf(t) dtF(x) = \int_{x_0}^x f(t)\,dt, which is well-defined because there is essentially only one way to get from x0x_0 to xx on the real line. In the plane there are infinitely many paths from z0z_0 to zz, so we must prove the integral is the same along all of them before F(z)F(z) even makes sense as a function; that is exactly the role of hypothesis (3), applied to the closed loop formed by two candidate paths traversed in opposite directions. Without this well-definedness check, "F(z)F(z)" would not be a legitimate function at all.

Can F′=fF' = f hold with FF merely continuous rather than holomorphic?

No — by definition, a primitive FF of a holomorphic (or continuous) ff on DD must itself be holomorphic on DD, since complex differentiability, "F′(z)F'(z) exists as a complex limit at every point," is precisely the definition of holomorphic. In fact once F′=fF' = f holds with FF holomorphic, ff itself is automatically holomorphic too (a derivative of a holomorphic function is holomorphic, provable later via Cauchy's integral formula and power series expansions), so the whole framework is self-consistently a theory of holomorphic functions throughout, even though the Theorem's proof only used continuity of ff.

What is the relationship between this and the ordinary Fundamental Theorem of Calculus from single-variable calculus?

It is a direct generalization. If DD is an interval on the real line and γ(t)=t\gamma(t) = t traces it, the contour integral ∫γf dz\int_\gamma f\,dz reduces exactly to the ordinary Riemann integral ∫abf(t) dt\int_a^b f(t)\,dt, and the formula F(b)−F(a)F(b)-F(a) recovers the classical Fundamental Theorem of Calculus. The complex version is more subtle only because there are infinitely many paths between two points in the plane, so "path independence" must be established as a nontrivial theorem rather than being automatic, as it trivially is on the real line where only one path (up to reparametrization) connects two points.

Does path independence require ff to be holomorphic, or just continuous?

The three-way equivalence in the Theorem (primitive exists ⇔ path independent ⇔ all loop integrals vanish) holds for any continuous ff on a domain DD — holomorphy of ff is never used in that proof. Holomorphy enters only through the Corollary, where Cauchy's Theorem is invoked to guarantee automatically that condition (3) holds for holomorphic ff on simply connected DD. For a merely continuous (non-holomorphic) ff, you would have to verify condition (3) by hand, since Cauchy's Theorem does not apply.

Before this

Where this leads