Schwarz Reflection Principle

Analytic continuation usually feels like archaeology: you are handed a function on a small patch and asked to reconstruct the larger function it belongs to, guessing at what lies beyond the boundary. The Schwarz reflection principle is the one case where the answer is not a guess but a mirror. If a holomorphic $f$ lives in a region whose edge includes a segment of the real axis, and $f$ takes real values on that segment, then $f$ extends across the axis by a formula you write down instantly — reflect both input and output by complex conjugation, $f(\bar z)=\overline{f(z)}$ . Reality on the axis acts as a perfect mirror: the function's behaviour just below is forced to be the conjugate image of its behaviour just above. Two facts do all the work. First, the conjugated map $g(z)=\overline{f(\bar z)}$ is genuinely holomorphic, not merely smooth: conjugation reverses orientation, and doing it once to the argument and once to the value composes two orientation reversals back into an orientation-preserving, angle-preserving map. Second, reality on the segment makes the boundary values of $f$ and $g$ agree exactly, so the two analytic pieces glue continuously — and continuity across an analytic arc, by Morera's theorem, upgrades for free to full analyticity. The seam vanishes, and the domain of definition has doubled. The principle reaches far past the real axis. Reflection is a conformal symmetry, so one can mirror across any line or, through inversion $z\mapsto R^2/\bar z$ , across any circle. This one idea tiles the plane in the Schwarz–Christoffel and modular functions, justifies the electrostatic method of images, forces the nontrivial zeros of the Riemann zeta function into conjugate pairs, and — as Euclidean reflection positivity — underlies the reconstruction of quantum field theory from its imaginary-time correlators.

Key takeaways

If $f$ is holomorphic on a domain touching a real-axis segment $I$ and is real-valued on $I$ , it continues analytically across $I$ by $f(\bar z)=\overline{f(z)}$ — the mirror obtained by conjugating both argument and value.
The reflected piece $g(z)=\overline{f(\bar z)}$ is holomorphic because a double conjugation preserves the Cauchy–Riemann equations; reality on $I$ makes its boundary values match those of $f$ .
The glued function is analytic, not just continuous: Morera's theorem certifies it because every contour integral straddling the axis vanishes.
Reflection generalizes to any line and, via inversion $z^{*}=R^2/\bar z$ , to any circle, and to harmonic potentials as odd reflection $u(x,-y)=-u(x,y)$ .
Consequences are everywhere: real power-series coefficients, conjugate-pair zeros and poles, the method of images, Schwarz–Christoffel tiling, and reflection positivity in quantum field theory.
Key idea: real boundary values are conjugation-invariant, so a function holomorphic on one side of a real segment and real on it has a unique mirror continuation $f(\bar z)=\overline{f(z)}$ to the reflected domain, doubling its domain of definition for free.

The core insight

If $f$ is analytic on a domain bordering a real-axis segment and is real-valued there, then $f(\bar z)=\overline{f(z)}$ defines a single-valued analytic continuation into the mirror domain; the two pieces match because real boundary values are conjugation-invariant.

Where this lives today

The reflection principle is load-bearing in three very different modern enterprises. In numerical conformal mapping, the Schwarz–Christoffel transformation — continued across polygon edges by exactly this principle — ships in production software (the SC Toolbox of Driscoll and Trefethen) and solves Laplace and potential-flow problems on complicated planar domains, from airfoil and microstrip design to the flattening of cortical surfaces in medical imaging. In quantum field theory, the Osterwalder–Schrader reconstruction theorem elevates reflection to an axiom: a Euclidean theory whose correlators are reflection positive under mirror reflection of imaginary time continues back to a genuine relativistic theory on Minkowski space with a positive Hilbert-space inner product. That Euclidean reflection is Schwarz's mirror wearing a physicist's hat, now standard in constructive field theory, lattice gauge theory, and the conformal bootstrap. And in the analytic theory of the Riemann zeta function, the symmetry $\zeta(\bar s)=\overline{\zeta(s)}$ pairs the nontrivial zeros into conjugates and, with the functional equation, arranges them symmetrically about the critical line — a structural fact behind every numerical test of the Riemann hypothesis. The same idea continues fluid potentials past cylinders (Milne–Thomson) and supplies boundary reflections in two-dimensional conformal field theory.

The physical picture

Think of the real segment as a silvered mirror laid along the axis. Above it, in the domain $\Omega^{+}$ , the function $f$ does whatever it does; the reality condition $f(x)\in\mathbb{R}$ means it approaches the mirror with no imaginary part, nothing sticking out of the mirror plane. There is then exactly one holomorphic continuation into the lower domain $\Omega^{-}$ : the conjugate image, so that a value $w$ attained at $z$ just above the axis is matched by $\bar w$ at the reflected point $\bar z$ just below. The first figure shows this with $f(z)=z^{2}$ : the point $z=1+i$ carries the value $2i$ , and its mirror $\bar z=1-i$ is forced to carry $-2i$ .

Figure

The mirror map with f(z) = z². The domain point z = 1 + i in Ω⁺ carries the value f(z) = 2i in the w-plane; its reflection z̄ = 1 − i in Ω⁻ is forced to carry the conjugate value f(z̄) = −2i. Real values on the segment I are the fixed axis of the mirror, so the whole lower half is the conjugate image of the upper half

The same picture in the language of potentials is the method of images. A holomorphic function is a pair of harmonic conjugate fields, and $\operatorname{Im}f=0$ on the axis says a potential is pinned to a constant on a flat conductor. The reflection principle then makes the field on the far side of the conductor the odd mirror image of the field in front of it — precisely the fictitious "image charge" that electrostatics places behind a grounded plane. The second figure shows this: a charge above a grounded plane, its image below, and the antisymmetric potential $u(x,-y)=-u(x,y)$ that satisfies Laplace's equation everywhere.

Figure

The potential-theory face: a charge +q at height h above a grounded plane at y = 0. Schwarz reflection places an image charge −q at −h, so the continued potential obeys the odd relation u(x, −y) = −u(x, y) and vanishes on the plane. For q = 1 nC and h = 0.10 m the plane attracts the charge with force F = q²/(16πε₀h²) = 2.25×10⁻⁷ N

The core equations

The whole theorem is a hypothesis, a formula, and a certificate that the formula is analytic. The hypothesis: $f$ is holomorphic on a domain leaning against a real segment and reaches real values as it approaches it. Reality on the axis is the load-bearing assumption; analyticity up to the boundary alone is not enough.

f \text{ holomorphic on } \Omega^{+},\qquad f(x)\in\mathbb{R}\ \text{ for } x\in I\subset\mathbb{R} \qquad (1)

where: $\Omega^{+}$ — a domain in the upper half-plane whose boundary contains the open real segment $I$ ; $I$ — an open interval of the real axis on which $f$ is continuous and real-valued; $f$ — holomorphic on $\Omega^{+}$ and continuous up to $I$ . Reading: all data live above the axis; the only thing known on it is that the values are real. The continuation is an explicit two-piece formula: the original $f$ on $\Omega^{+}\cup I$ , and the conjugate of $f$ at the conjugate point on $\Omega^{-}$ .

\tilde f(z)=\begin{cases} f(z), & z\in\Omega^{+}\cup I,\\[4pt] \overline{f(\bar z)}, & z\in\Omega^{-}. \end{cases} \qquad (2)

where: $\Omega^{-}=\{\bar z:z\in\Omega^{+}\}$ — the mirror image of $\Omega^{+}$ across the real axis; $\tilde f$ — the analytic continuation of $f$ to $\Omega^{+}\cup I\cup\Omega^{-}$ ; $\bar z$ — the complex conjugate of $z$ , its reflection across the real axis. Reading: to evaluate the continuation at a lower point $z$ , reflect up to $\bar z$ , apply $f$ , then conjugate the answer. The lower piece is not merely a copy — it is itself holomorphic, which is what makes the two halves join into one analytic object rather than a piecewise patch.

g(z)=\overline{f(\bar z)}\quad\text{is holomorphic on } \Omega^{-} \qquad (3)

where: $g$ — the reflected function defining $\tilde f$ on $\Omega^{-}$ ; conjugating both argument and value preserves the Cauchy–Riemann equations. Reading: one conjugation is anti-holomorphic, but two compose back to holomorphic — Derivation 1 verifies this. The glue holds because on the segment the two pieces take the same value; reality is exactly the condition that makes the upper and lower limits coincide.

\lim_{y\to 0^{+}} f(x+iy)=\lim_{y\to 0^{-}} g(x+iy)=f(x)\in\mathbb{R},\qquad x\in I \qquad (4)

where: $y\to 0^{\pm}$ — approach to $I$ from above or below; $f(x)$ — the common real boundary value, with $g(x)=\overline{f(\bar x)}=\overline{f(x)}=f(x)$ . Reading: were $f(x)$ not real, $g(x)\ne f(x)$ and the seam would not close. A structural payoff: expanding $\tilde f$ about any real point of $I$ gives real coefficients, since a function equal to its own conjugate-reflection is real with real derivatives on the axis.

\tilde f(z)=\sum_{n\ge 0} a_n\,(z-x_0)^n,\qquad a_n\in\mathbb{R}\ \text{ for } x_0\in I \qquad (5)

where: $a_n=\tilde f^{(n)}(x_0)/n!$ — Taylor coefficients about a real point $x_0\in I$ ; the symmetry $\tilde f(\bar z)=\overline{\tilde f(z)}$ forces every $a_n$ real. Reading: real power series and reflection symmetry are the same statement — why $e^{z}$ , $\cos z$ , and $\zeta(s)$ all obey $f(\bar z)=\overline{f(z)}$ . Reflection is not tied to the real axis. Across a circle of radius $R$ the conjugate is replaced by the inversive point, and boundary values on a line or circle continue by inversion in that curve.

f(z^{*})=\overline{f(z)},\qquad z^{*}=\frac{R^{2}}{\bar z}\quad(\text{circle } |z|=R) \qquad (6)

where: $z^{*}=R^{2}/\bar z$ — the point inverse to $z$ in the circle $|z|=R$ , same argument as $z$ , modulus $R^{2}/|z|$ ; $R$ — the circle across whose arc $f$ takes real (or circular) boundary values. Reading: inversion fixes $|z|=R$ pointwise, so it plays the role conjugation plays for a line. Finally, the harmonic version: a harmonic function vanishing on the axis continues by an odd reflection, and stays harmonic because the Laplacian is invariant under $y\to-y$ .

u(x,-y)=-\,u(x,y)\ \Longrightarrow\ \nabla^{2}u=0\ \text{ across } y=0 \qquad (7)

where: $u$ — a harmonic potential with $u=0$ on the line $y=0$ ; $\nabla^{2}=\partial_{xx}+\partial_{yy}$ — the two-dimensional Laplacian. Reading: odd reflection is the analytic-function statement of the image charge; a constant boundary value $c$ instead of $0$ reflects about $c$ , giving $u(x,-y)=2c-u(x,y)$ .

Derivation AWhy the mirror function is holomorphic

The goal is to prove the fact on which everything rests: that $g(z)=\overline{f(\bar z)}$ is genuinely holomorphic on the reflected domain $\Omega^{-}$ , not merely a smooth real-analytic map. This is not obvious, because complex conjugation by itself is the archetypal non-holomorphic operation — $z\mapsto\bar z$ fails the Cauchy–Riemann equations everywhere. The resolution is that $g$ conjugates twice, in the argument and in the value, and two orientation-reversing reflections compose into an orientation-preserving map. We make this precise by writing $f=u+iv$ , expressing $\operatorname{Re}g$ and $\operatorname{Im}g$ through $u,v$ , and checking the Cauchy–Riemann equations for $g$ .

Step 1. Write $z=x+iy$ , so $\bar z=x-iy$ , and split $f$ into real and imaginary parts $f=u+iv$ . Then $f(\bar z)=u(x,-y)+i\,v(x,-y)$ , and conjugating gives $g(z)=\overline{f(\bar z)}=u(x,-y)-i\,v(x,-y)$ . Why: laying $g$ out in components is the only way to test holomorphy directly through the Cauchy–Riemann equations, since holomorphy is a statement about the partial derivatives of the real and imaginary parts.

Step 2. Read off the real and imaginary parts of $g$ as $U(x,y)=u(x,-y)$ and $V(x,y)=-v(x,-y)$ . Differentiating by the chain rule, with all derivatives of $u,v$ at $(x,-y)$ , gives $U_x=u_x$ , $U_y=-u_y$ , $V_x=-v_x$ , $V_y=v_y$ . Why: the minus signs come from the $y\to-y$ in the argument and the conjugation in the value; tracking them is the whole proof.

Step 3. Impose holomorphy of $f$ at $(x,-y)$ , so $u_x=v_y$ and $u_y=-v_x$ there. Substitute: $U_x=u_x=v_y=V_y$ and $U_y=-u_y=v_x=-V_x$ . Hence $U_x=V_y$ and $U_y=-V_x$ — the Cauchy–Riemann equations for $g$ . Why: the two reflections conspire so that the CR equations for $f$ at $(x,-y)$ become those for $g$ at $(x,y)$ .

Step 4. With continuous first partials (inherited from $f$ ) and the CR equations satisfied throughout $\Omega^{-}$ , $g$ is holomorphic there. Geometrically $z\mapsto\bar z$ reverses orientation, $f$ preserves it, $w\mapsto\bar w$ reverses it again, so $g$ preserves orientation and angles. Why: it certifies the lower piece of $\tilde f$ is analytic, so gluing it to $f$ yields one analytic object, not a continuous patch.

Go deeper · Derivation B · 2 minGluing the two halves — the Morera argument

Derivation BGluing the two halves — the Morera argument

The goal is to upgrade "the two pieces agree continuously on $I$ " to "the glued function $\tilde f$ is analytic across $I$ ." We have $f$ holomorphic on $\Omega^{+}$ , $g=\overline{f(\bar z)}$ holomorphic on $\Omega^{-}$ by the previous derivation, and — because $f$ is real on $I$ — the two sharing the same continuous boundary value there. Continuity across a curve does not by itself imply analyticity. The bridge is Morera's theorem, the converse of Cauchy's: a continuous function whose integral around every triangle vanishes is holomorphic. We verify that hypothesis by splitting triangles that straddle the axis and cancelling the two boundary integrals in the limit.

Step 1. Establish continuity of $\tilde f$ . For $x\in I$ the lower limit is $\lim_{z\to x,\ z\in\Omega^{-}}g(z)=\overline{f(x)}$ , and since $f(x)$ is real this equals the upper limit $f(x)$ . So $\tilde f$ is continuous on $\Omega^{+}\cup I\cup\Omega^{-}$ . Why: Morera needs a continuous function to start from, and the reality condition is what supplies continuity across the seam.

Step 2. Dispose of triangles that avoid the axis. For any closed triangle $\Delta$ entirely in $\Omega^{+}$ or entirely in $\Omega^{-}$ , Cauchy's theorem gives $\oint_{\partial\Delta}\tilde f\,\mathrm{d}z=0$ , since $\tilde f$ is holomorphic there. Why: the subtlety is concentrated in triangles crossing $I$ , which the next step reduces to this case.

Step 3. Handle a triangle $\Delta$ straddling the axis. Cut it with $y=+\varepsilon$ and $y=-\varepsilon$ into an upper sub-polygon in $\Omega^{+}$ , a lower one in $\Omega^{-}$ , and a thin strip. Each off-axis piece integrates to zero by Step 2, so $\oint_{\partial\Delta}\tilde f\,\mathrm{d}z$ reduces to the integrals along $y=\pm\varepsilon$ ; as $\varepsilon\to0$ these approach the same points of $I$ with opposite orientation and cancel by uniform continuity. Why: this limiting cancellation is the technical heart — where "no jump on the seam" becomes "no obstruction to a contour integral."

Step 4. Conclude by Morera. Since $\oint_{\partial\Delta}\tilde f\,\mathrm{d}z=0$ for every triangle and $\tilde f$ is continuous, $\tilde f$ is holomorphic throughout $\Omega^{+}\cup I\cup\Omega^{-}$ — analytic on $I$ , so the real segment is now an interior curve of holomorphy, not a boundary. Why: the domain has doubled, and the extension is unique since holomorphic functions agreeing on $\Omega^{+}$ agree wherever both are defined.

Go deeper · Derivation C · 2 minReflecting a harmonic potential — the image charge

Derivation CReflecting a harmonic potential — the image charge

The goal is to derive the potential-theory form of the principle: a function $u$ harmonic in the upper half-plane and vanishing on the axis extends to a globally harmonic function by the odd reflection $u(x,-y)=-u(x,y)$ , which is exactly the electrostatic method of images. Harmonic functions are real parts of holomorphic ones, so we can bring the full analytic machinery to bear — but with a twist: the boundary condition here is $\operatorname{Re}f=0$ on $I$ , a reflection in the imaginary axis of the target rather than the real axis. Handling that correctly turns an even reflection into the physically required odd one, placing a charge of opposite sign behind the conductor.

Step 1. Complete $u$ to a holomorphic $f=u+iv$ by choosing a harmonic conjugate $v$ on the simply connected upper region. The condition $u=0$ on $I$ says $\operatorname{Re}f=0$ there, so $f$ maps $I$ into the imaginary axis of the $w$ -plane. Why: recasting the harmonic condition as "where a holomorphic map sends the boundary" is what lets the reflection principle apply.

Step 2. Reflect across the correct target line. Reflection of a value $w$ across the imaginary axis is $w\mapsto-\bar w$ , so the reflection principle continues $f$ into $\Omega^{-}$ by $\tilde f(z)=-\,\overline{f(\bar z)}$ . Why: reflecting in the imaginary axis, not the real axis, is the one decision that makes the potential odd; a real-axis reflection gives the wrong even sign and violates $u=0$ .

Step 3. Read off the potential: $\operatorname{Re}\tilde f=\operatorname{Re}\!\big(-\overline{f(\bar z)}\big)=-\,u(x,-y)$ . Extending $u$ as itself in $\Omega^{+}$ and this in $\Omega^{-}$ gives the odd reflection $u(x,-y)=-u(x,y)$ , with $u=0$ on $y=0$ by continuity. Why: this is the image-charge prescription written analytically — the field below the plane is the sign-flipped mirror of the field above.

Step 4. Confirm the physics. Under $y\to-y$ the Laplacian $\nabla^{2}=\partial_{xx}+\partial_{yy}$ is unchanged, so $u(x,-y)$ is harmonic wherever $u$ is. The configuration — a source at $+h$ plus an equal-and-opposite image at $-h$ — is exactly Kelvin's method of images, with the reflection principle as its uniqueness proof. Why: "add a fictitious mirror charge" is not a lucky guess but a theorem guaranteeing the combined field is the unique harmonic solution.

Limiting cases

The one principle takes different concrete forms depending on the geometry of the boundary curve and on which set the boundary values are pinned to. The table collects the standard specializations.

Case	Result	Meaning
Across the real axis (classical)	$f(\bar z)=\overline{f(z)}$	$f$ analytic and real on a real segment continues by conjugate symmetry; its Taylor coefficients about real points are real.
Across a general straight line	$f(z^{})=\overline{f(z)}+2ic,\quad z^{}=a+e^{2i\theta}\,\overline{(z-a)}$	For a domain line through $a$ at angle $\theta$ and boundary values on a target line $\operatorname{Im}w=c$ , conjugate both domain and target after rotating the line to the real axis.
Across a circle (inversive symmetry)	$f(z^{})=\overline{f(z)},\quad z^{}=R^{2}/\bar z$	If $f$ maps the arc $\|z\|=R$ into a line or circle, continue by inversion in the circle; this is the engine of the Milne–Thomson circle theorem.
Harmonic potential, value 0 on axis	$u(x,-y)=-u(x,y)$	A harmonic function vanishing on the axis continues by odd reflection — the grounded-plane image charge.
Harmonic potential, value $c$ on axis	$u(x,-y)=2c-u(x,y)$	Held at constant $c$ , the potential reflects about the level $c$ ; subtract $c$ to reduce to the vanishing case.

Common trapReflection works for any function that is merely analytic up to the boundary

It does not. The reality (or line/circle) condition on the boundary values is essential. The formula $f(\bar z)=\overline{f(z)}$ continues $f$ only when $f$ actually attains real values on the segment — more generally, only when the boundary values lie on the line or circle you are reflecting in. If $f$ is holomorphic up to $I$ but complex-valued there, then $g(x)=\overline{f(x)}\ne f(x)$ , the two pieces disagree on the seam, and $\tilde f$ is discontinuous — no continuation of this form exists. You also need genuine continuity of $f$ up to $I$ ; analyticity on the open domain alone does not guarantee boundary limits exist. This is not a technicality: it is the entire reason the mirror closes.

Worked examples

1 · Continuing a real polynomial across the axis (fully worked)

Suppose you are told only that a function $f$ is holomorphic in the upper half-plane, is real on the real axis, and equals $z^{3}-3z$ there and just above it. Confirm the reflection identity at $z=1+i$ , and use the principle to evaluate the continuation at the lower-half point $z=1-i$ without any new information.

Step 1. Evaluate $f$ at the upper point. With $z=1+i$ , first $(1+i)^{2}=2i$ , so $(1+i)^{3}=(1+i)(2i)=2i+2i^{2}=-2+2i$ . Then $f(1+i)=(-2+2i)-3(1+i)=-2+2i-3-3i=\mathbf{-5-i}$ . This is the value the function actually takes at $1+i$ .

Step 2. Apply the reflection formula to get the lower value. The continuation is $\tilde f(1-i)=\overline{f(\overline{1-i})}=\overline{f(1+i)}=\overline{-5-i}=\mathbf{-5+i}$ . Notice we obtained the value at a point where $f$ was not originally defined, using only the reflection rule.

Step 3. Cross-check by direct substitution, which is legitimate here because $z^{3}-3z$ is entire. Compute $(1-i)^{2}=-2i$ , so $(1-i)^{3}=(1-i)(-2i)=-2i+2i^{2}=-2-2i$ , and $f(1-i)=(-2-2i)-3(1-i)=-2-2i-3+3i=-5+i$ . This matches the reflected value exactly. The reflection identity gives $f(1+i)=-5-i$ and $f(1-i)=-5+i$ , a conjugate pair, with no independent lower-half data needed.

Step 4. Read the structural lesson. Because the coefficients $1,0,-3,0$ of $z^{3}-3z$ are real, the reflection symmetry $f(\bar z)=\overline{f(z)}$ is automatic, and any value in the lower half-plane is fixed the instant the upper half is known. Real coefficients and mirror symmetry are the same fact.

2 · The grounded plane and the image force (fully worked)

A point charge $q=1\ \mathrm{nC}$ sits at height $h=0.10\ \mathrm{m}$ above an infinite grounded conducting plane. Use the reflection (image) principle to find the force the plane exerts on the charge, and confirm the potential is odd across the plane.

Step 1. Replace the conductor by its image. Odd reflection of the potential across $y=0$ is equivalent to deleting the conductor and inserting an image charge $-q$ at $y=-h$ , so that $u(x,-y)=-u(x,y)$ vanishes on the plane. The physical field above the plane equals that of the pair $\{+q\ \text{at}\ +h,\ -q\ \text{at}\ -h\}$ .

Step 2. Compute the force from the image alone (the real charge feels only the image, not itself). The separation is $2h=0.20\ \mathrm{m}$ , so $F=\dfrac{1}{4\pi\varepsilon_0}\dfrac{q^{2}}{(2h)^{2}}=\dfrac{q^{2}}{16\pi\varepsilon_0 h^{2}}$ . Insert $q^{2}=10^{-18}\ \mathrm{C^{2}}$ , $h^{2}=10^{-2}\ \mathrm{m^{2}}$ , and $16\pi\varepsilon_0=4.451\times10^{-10}$ : $F=\dfrac{10^{-18}}{4.451\times10^{-10}\times10^{-2}}=\mathbf{2.25\times10^{-7}\ N}$ , directed toward the plane (attractive).

Step 3. Cross-check the energy. The interaction energy of the real charge with the plane is half the naive pair energy, $W=-\dfrac{q^{2}}{16\pi\varepsilon_0 h}=-2.25\times10^{-8}\ \mathrm{J}$ , and $F=-\partial W/\partial h$ reproduces the force above. F = 2.25×10⁻⁷ N attractive, W = −2.25×10⁻⁸ J — the reflection principle turns a boundary-value problem into a two-charge sum.

3 · Conjugate-pair zeros of a real function (structural)

A function $f$ is holomorphic on a domain symmetric about the real axis, real on the axis, and has a zero at $z_0=2+3i$ . What else can you conclude with no further computation?

Step 1. Apply reflection symmetry to the zero. Since $f(\bar z)=\overline{f(z)}$ , evaluate at $\bar z_0=2-3i$ : $f(2-3i)=\overline{f(2+3i)}=\overline{0}=0$ . So $2-3i$ is automatically a zero of the same multiplicity. Off-axis zeros of a real-analytic function occur in conjugate pairs.

Step 2. Generalize to poles and to famous cases. The identical argument on $1/f$ pairs poles, and applied to the Riemann zeta function — real for real $s>1$ — it forces the nontrivial zeros to come in conjugate pairs $\rho,\bar\rho$ , one of the basic symmetries of the critical strip. Reality on a segment is a global constraint on the entire singularity structure.

Physics of the portrait

The reflection principle is inseparable from the mathematician who found it while wrestling with a stubborn problem in conformal mapping — how to continue the map of a polygon past its straight edges. What began as a bookkeeping device for tiling the plane became one of the most reused symmetries in mathematical physics.

Origins · 1 minHermann Schwarz and how this was discovered

Hermann Schwarz

1843–1921 · conformal mapping, minimal surfaces, and the Dirichlet problem

Hermann Amandus Schwarz was the most geometric of the great Berlin analysts, a student of Kummer and Weierstrass who inherited Weierstrass's chair in 1892. His mathematics is a catalogue of tools still in daily use: the Cauchy–Schwarz inequality, the Schwarz lemma bounding holomorphic self-maps of the disc, the Schwarzian derivative, the Schwarz alternating method for the Dirichlet problem, and the Schwarz–Christoffel formula that maps the half-plane onto any polygon. The reflection principle arose in the late 1860s from exactly this last circle of ideas. Asking how the conformal image of a polygon could be continued past a straight boundary edge — where the mapping function is real on a segment of the axis — Schwarz saw that the continuation was forced: reflect the polygon, reflect the map, and the two fit together analytically. Repeated reflection then tiles the plane, and the same argument continues the modular elliptic functions and triangle-group maps central to hyperbolic geometry and automorphic forms. Schwarz worked slowly and published sparingly, prizing a single decisive construction, and his reflection principle has the character of his best work: a small, exact idea that turned out to be a master key — reappearing a century later, essentially unchanged, as the reflection-positivity axiom that reconstructs a quantum theory from its Euclidean correlations.

History

1848

— Kelvin introduces the method of images for a charge near a grounded plane

1851

— Riemann's dissertation founds conformal mapping and analytic continuation

1869

— Schwarz states the reflection principle while mapping polygons conformally

1886

— Morera proves the converse of Cauchy's theorem, the modern engine of the gluing proof

1940

— Milne-Thomson recasts circular reflection as the circle theorem of ideal flow

Going further

Texts. Ahlfors, Complex Analysis, 3rd ed., §4.6 (the reflection principle) · Stone and Goldbart, Mathematics for Physics, complex-analysis chapter (analytic continuation) · Arfken, Weber and Harris, Mathematical Methods for Physicists, 7th ed., complex-variable chapters · Whittaker and Watson, A Course of Modern Analysis, §5.5 · Nakahara, Geometry, Topology and Physics (conformal maps and complex structure) · Byron and Fuller, Mathematics of Classical and Quantum Physics (analytic functions and potential theory). The numeric image-force example uses $q=1\ \mathrm{nC}$ , $h=0.10\ \mathrm{m}$ , giving $F=2.25\times10^{-7}\ \mathrm{N}$ , computed from Coulomb's law with $16\pi\varepsilon_0=4.451\times10^{-10}$ . Key checkpoints in the article are: the Cauchy–Riemann verification that $g=\overline{f(\bar z)}$ is holomorphic; the Morera triangle-splitting argument; invariance of $\nabla^{2}$ under $y\to-y$ ; the polynomial values $f(1+i)=-5-i$ and $f(1-i)=-5+i$ ; and the conjugate-pair zero statement.

Prerequisites. Analytic Functions and the Cauchy–Riemann Equations · Consequences: Liouville, Maximum Modulus, Morera · Harmonic Functions and Conjugates · Cauchy's Integral Theorem. Leads to. Elementary Complex Functions · Zeros, Poles, and Isolated Singularities · Laurent Series.

Ask the tutor: Ask me to walk through the Morera gluing step by step, reflect a specific function of yours across a line or circle, or show how reflection positivity turns a Euclidean correlator back into a quantum theory.

Why must f be real on the segment, and not just analytic up to it?

Because continuity of the glued function on the seam demands it. Approaching a segment point $x$ from below, the continuation gives $\overline{f(\bar x)}=\overline{f(x)}$ ; approaching from above gives $f(x)$ . These agree only if $\overline{f(x)}=f(x)$ , i.e. $f(x)$ is real. If $f$ were complex-valued on the segment, the two limits would differ, the mirror image would fail to match the original, and $\tilde f$ would jump across the axis instead of continuing smoothly through it. Reality is therefore not a convenience but the precise algebraic condition that makes conjugation a symmetry of the boundary data. The generalization keeps the same logic: the boundary values must lie on a line or circle so that the corresponding reflection fixes them.

How can conjugation appear twice yet leave the function holomorphic?

A single conjugation, $z\mapsto\bar z$ or $w\mapsto\bar w$ , is anti-holomorphic: it reverses orientation and negates angles, and it violates the Cauchy–Riemann equations. But $g=\overline{f(\bar z)}$ conjugates twice — once on the input and once on the output — and two orientation reversals compose into an orientation preservation. Concretely, writing $g=U+iV$ with $U(x,y)=u(x,-y)$ and $V(x,y)=-v(x,-y)$ and substituting the Cauchy–Riemann equations for $f$ at $(x,-y)$ yields $U_x=V_y$ and $U_y=-V_x$ exactly. The minus signs from reflecting the argument and from conjugating the value cancel in just the right pattern. It is the same reason a mirror image of a mirror image is not mirrored.

What is the precise link to the electrostatic method of images?

A two-dimensional electrostatic potential $u$ is harmonic, and any harmonic function is the real part of a holomorphic $f$ . Pinning $u$ to zero on a grounded plane is the condition $\operatorname{Re}f=0$ on a boundary segment, which the reflection principle continues by an odd reflection $u(x,-y)=-u(x,y)$ . That odd extension is identical to deleting the conductor and placing an equal-and-opposite image charge on the far side. So the "trick" of image charges is a theorem: the reflection principle proves that the two-charge field is the unique harmonic solution satisfying the grounded boundary condition, rather than a fortunate guess that happens to work. The worked example computes the resulting attractive force explicitly.

How does reflection across a circle differ from reflection across a line?

The geometry changes but the logic does not. Reflection across a line is realized by complex conjugation (for the real axis) or a rotated conjugation for a general line. Reflection across a circle of radius $R$ is realized by inversion, $z\mapsto z^{*}=R^{2}/\bar z$ , the point with the same direction from the center but reciprocal-scaled distance, so that the circle itself is fixed pointwise. If $f$ carries the circular arc into a line or another circle, the continuation reads $f(R^{2}/\bar z)=$ (reflection of $\overline{f(z)}$ in the target curve). This is what powers the Milne–Thomson circle theorem in fluid flow, where inserting the inverse-point image makes a cylinder a streamline.

Does the principle apply to harmonic functions, or only to holomorphic ones?

Both — they are two faces of one statement. For holomorphic $f$ real on a segment, the extension is $\overline{f(\bar z)}$ . For harmonic $u$ vanishing on a segment, it is the odd reflection $-u(x,-y)$ ; for $u$ equal to a constant $c$ , it reflects about $c$ . The harmonic version follows from the holomorphic one by completing $u$ to $f=u+iv$ and reflecting in the appropriate target line, as the third derivation shows. Physicists usually meet the harmonic form first, as the method of images for potentials and steady temperature fields, without realizing it is Schwarz's theorem in disguise.

Why do the zeros of real-analytic functions come in conjugate pairs?

Directly from $f(\bar z)=\overline{f(z)}$ . If $f(z_0)=0$ then $f(\bar z_0)=\overline{f(z_0)}=\overline{0}=0$ , so $\bar z_0$ is a zero of the same multiplicity; a zero off the real axis drags its mirror partner along. The same argument on $1/f$ pairs the poles. For the Riemann zeta function, which is real for real arguments greater than one, this forces the nontrivial zeros into conjugate pairs $\rho,\bar\rho$ ; combined with the functional equation's reflection $s\leftrightarrow 1-s$ , the zeros arrange symmetrically about the critical line. The conjugate symmetry is thus a global constraint on singularity structure that costs nothing beyond reality on a single segment.