Reading the first Seiberg–Witten paper

What can an exact solution determine when the microscopic theory is strongly coupled? Here, the useful answer is a set of local low-energy variables, their couplings and their central charges. A curve organizes how these variables fit together. The curve is not a replacement for the quantum field theory, and its periods do not by themselves list every stable particle.

Read Nathan Seiberg and Edward Witten, Electric-Magnetic Duality, Monopole Condensation, And Confinement In N=2 Supersymmetric Yang-Mills Theory, arXiv v1, posted 15 July 1994. Page references below give the printed page followed by the PDF page. Start with what the solution solves and the classical Coulomb branch. The later reading blocks identify additional preparation where it becomes useful.

The journal erratum, pp.485–486 corrects a production interchange of Figures 1 and 2. Figure 1 is the vacuum $u$-plane with based monodromy loops; Figure 2 is a contour around branch points in the curve’s $x$-plane. The pinned arXiv v1 already has this pairing: Figure 1, p.28/PDF29, and Figure 2, p.38/PDF39. The erratum prints no equation correction. Keep the two planes distinct: moving a vacuum around a puncture is not the same operation as integrating a differential along a cycle of one fixed curve.

Keep the theory and the notation fixed

The microscopic model is pure four-dimensional $\mathcal N=2$ $SU(2)$ Yang–Mills. It has eight real Poincaré supercharges. Its $\mathcal N=1$ superspace presentation makes four manifest; it does not reduce the full theory to four. A generic Coulomb vacuum preserves all eight and leaves one Abelian vector multiplet light. The gauge group has rank one, although its fundamental matrices are two by two. No curved background or background-preserved subset is involved here.

The paper sometimes calls an $\mathcal N=2$ vector multiplet a “chiral multiplet.” This is not an $\mathcal N=1$ chiral multiplet. Its section 2.1 decomposition resolves the terminology: one $\mathcal N=1$ vector plus one adjoint chiral. Keep that distinction when reading its final discussion of the charged vector particles.

Original-paper notation	Convention used in the lessons
$\mathcal F$, $h=\partial\mathcal F/\partial a$	Local holomorphic prepotential; $h=a_D$. This is not a superpotential.
$u=\langle\operatorname{Tr}\phi^2\rangle$	$u=\langle\operatorname{Tr}_2\phi_{\rm SW}^2\rangle$, with $\operatorname{Tr}_2(t^at^b)=\delta^{ab}/2$. Classically $u=a_{\rm cl}^2/2$.
Dimensionless $x,u,y$ in section 6	$X=x/\Lambda^2$, $U=u/\Lambda^2$, $Y=y/\Lambda^3$. Thus finite singularities are $U=\pm1$.
$(a_D,a)$ in that dimensionless curve	The physical periods are $\Lambda(a_D^{\rm source},a^{\rm source})$ at fixed $U$.
$(n_m,n_e)$	A charge row, acting on the period column $\Pi=(a_D,a)^T$. The charged vector has $n_e=\pm1$.
$Z$ and the mass bound	$Z=n_ma_D+n_ea$, $M\geq\sqrt2\lvert Z\rvert$. Equality needs a BPS state that actually exists.
$\tau$	$\partial_a a_D=(da_D/du)/(da/du)$, not $a_D/a$.
$m$ in section 5.6	The lesson calls this adjoint deformation coefficient $\mu$, to distinguish it from a particle mass.

There is no electric-label doubling in this guide. That change belongs to the second-paper convention guide. Restoring $\Lambda$ also fixes a curve-scale convention; it does not establish equality to an arbitrary subtraction-scheme scale.

Read the low-energy argument

Read sections 2 and 3 to learn what must be reconstructed, section 4 to learn what its periods mean, and sections 5.1–5.5 to identify the local singularity data. Use the linked teaching derivations before the corresponding passage when the method is new.

Passage in v1	Question to answer while reading	Canonical preparation
Section 2.2, (2.6), p.6/PDF 7	Why are inequivalent vacua labeled by one complex invariant?	Solve the vacuum equation.
Section 2.3, (2.7)–(2.16), pp.6–9/PDF 7–10	Which couplings belong to the same prepotential, and where does the scale enter?	The local prepotential action and the running scale.
Section 3.2, (3.18)–(3.21), p.14/PDF 15	Why does the dual photon use $a_D$ and $\tau_D=-1/\tau$?	Dualize the whole multiplet. Section 3.3's gravity extension can wait.
Section 4, (4.3)–(4.5), p.17/PDF 18; decay discussion, pp.19–20/PDF 20–21	What is the exact central charge, and what additional question determines stability?	The physical BPS bound and the marginal threshold.
Section 5.1, (5.1)–(5.4), pp.21–22/PDF 22–23; sections 5.4–5.5, (5.7)–(5.15), pp.26–27/PDF 27–28	How do a weak-coupling logarithm and a light monopole constrain continuation around punctures?	The light-field threshold and based monodromy paths.

The low-energy action is local in a chosen electric–magnetic frame. At a singularity, a field that was integrated out becomes light and must be restored. A divergent photon-only coupling is therefore a reason to change the field content and frame, not to continue using the same truncated action indiscriminately.

Read the curve and the periods as one specification

Prepare the two-sheeted surface and its cycles. Then read section 6 with these landmarks open:

Landmark in v1	Object to identify	Teaching home
(6.6), p.35/PDF 36	The cubic family and its degenerations	Reconstruction inputs.
(6.7), (6.14), pp.35–36/PDF 36–37	Cycle ordering and period ordering	The sheet and oriented periods.
(6.15)–(6.20), pp.37–38/PDF 38–39	A meromorphic differential whose vacuum derivative is holomorphic	Find and normalize the differential.
(6.22)–(6.23), p.39/PDF 40	The two complete-cycle integrals	Evaluate the oriented periods.
(6.24)–(6.30), pp.39–40/PDF 40–41	Weak coupling and the monopole limit	The local period expansion.
“The Spectrum,” pp.40–41/PDF 41–42	Why monodromy is not a proof of an unchanged stable spectrum	Supply the spectral information.

In dimensionful variables the first two essential formulas are

$$y^2=(x^2-\Lambda^4)(x-u),\qquad \lambda_{\rm SW}=\frac{\sqrt2}{2\pi}\frac{(x-u)\,dx}{y}, \qquad [x]=[u]=2,\quad[y]=3,\quad[\lambda_{\rm SW}]=1. \tag{PC10.1}$$

They translate the source’s curve (6.6) and differential (6.19). The curve coordinate $x$ is not a spacetime coordinate.

For the initial real vacuum $u>s=\Lambda^2>0$, use the electric contour from $-s$ to $s$ on the bank where $y<0$, and the magnetic contour from $s$ to $u$ where $y$ is positive imaginary. Each returns on the other sheet. Then

$$\begin{aligned} a&=\frac{\sqrt2}{\pi}\int_{-s}^{s} \sqrt{\frac{u-x}{s^2-x^2}}\,dx>0,\\ a_D&=\frac{i\sqrt2}{\pi}\int_s^u \sqrt{\frac{u-x}{x^2-s^2}}\,dx\in i\mathbb R_{>0}. \end{aligned} \tag{PC10.2}$$

The factor of two relative to the coefficient of $\lambda$ is the return leg, not an optional normalization. These are v1 (6.22)–(6.23). The lesson’s usual geometric intersection is $\gamma_e\cdot\gamma_D=+1$; the paper writes its positive intersection symbol in the opposite order. Keep the actual contours and translate the intersection convention, rather than reversing a physical period.

For example $x=s\cos\theta$ makes the electric integral an ordinary integral of $\sqrt{u-s\cos\theta}$ on $0\leq\theta\leq\pi$. Its leading weak limit is $\sqrt{2u}$. Averaging the next even binomial term gives $a/\sqrt{2u}=1-\Lambda^4/(16u^2)+\cdots$. The normalization is classical only at leading order; $u=a^2/2$ is not an exact identity.

Restore the logarithm that produces the monodromy

The paper moves from the local integrals to their logarithmic behavior in (6.27)–(6.30). Here is a way to recover the crucial coefficient without assuming the monodromy matrix in advance. The period equation supplies a short alternative derivation; this is the course’s reconstruction, not an equation printed in v1.

Put $z=(u-s)/s$ and $p=a/\Lambda$ or $a_D/\Lambda$. The equation becomes

$$z(z+2)p''(z)+\frac14p(z)=0. \tag{PC10.3}$$

First fix the regular solution from the shrinking magnetic integral. With $x=s+(u-s)t$,

$$\frac{a_D}{\Lambda} =\frac{i\sqrt2\,z}{\pi}\int_0^1 \frac{\sqrt{(1-t)/t}}{\sqrt{2+zt}}\,dt =\frac i2d(z),\qquad d(z)=z-\frac{z^2}{16}+O(z^3). \tag{PC10.4}$$

The leading integral is $\pi/2$. The electric integral at $z=0$ separately gives $p_0=4/\pi$. A second solution of (PC10.3) can therefore be written $p_e(z)=h(z)+k\,d(z)\log z$, with $h(0)=p_0$ and $h,d$ holomorphic near zero. Why is the whole regular function $d$ multiplying the logarithm? Substitution into the differential equation forces its logarithmic coefficient to solve the same homogeneous equation; the regular solution vanishing at zero spans that one-dimensional space.

Since $d(z)=z+\cdots$, the nonlogarithmic constant term in (PC10.3) is $2k+p_0/4$. Its vanishing fixes

$$k=-\frac1{2\pi},\qquad a=\Lambda h(z)+\frac i\pi a_D\log z. \tag{PC10.5}$$

Now continue once counterclockwise around $z=0$. The holomorphic functions return, but the continuously followed logarithm gains $2\pi i$. Consequently

$$\Pi\longmapsto M_+\Pi,\qquad M_+=\begin{pmatrix}1&0\\-2&1\end{pmatrix}. \tag{PC10.6}$$

This reproduces v1 (5.11)–(5.12). Both the sign and the coefficient follow from the fixed periods. Resetting the logarithm to its principal value at the endpoint would erase the effect. The local argument specifies the small loop; transporting it to a distant base point requires the explicit connecting paths of the monodromy lesson.

Return to confinement after the curve

Now return to section 5.6, especially the superpotential and vacuum equations (5.17)–(5.21). Read the complete local vacuum calculation before interpreting the result.

The small deformation $W_{\rm UV}=\mu\operatorname{Tr}\Phi^2$ leaves four real supercharges in the action. Near the monopole point the appropriate fields are the dual photon multiplet and oppositely charged monopole chirals $M,\widetilde M$, with

$$W_{\rm eff}=\sqrt2 A_D M\widetilde M+\mu u(A_D). \tag{PC10.7}$$

The local solution is a supersymmetric vacuum of the remaining $\mathcal N=1$ theory. The magnetic Higgs mechanism gaps the dual photon and describes confinement of the corresponding original electric probes. Use $|\mu|/\Lambda\ll1$ for the controlled local dynamics. Holomorphy of a superpotential is not an exact determination of every kinetic term or of a finite-deformation string profile. Near the other singularity the light dyon needs its own local frame; there is no single weak local action making mutually nonlocal particles simultaneously elementary.

Exit calculations

1. Change the period frame, then follow the loop

Use $T=\left(\begin{smallmatrix}1&1\\0&1\end{smallmatrix}\right)$ and define $\Pi'=T\Pi$. Find the monopole row, local monodromy and coupling in the new frame. Does a clockwise circuit give the same matrix? Explain why changing the frame alone does not create a different massive particle.

Hint 1

Require $\gamma'\Pi'=\gamma\Pi$ for every period column. Transform both the entering and leaving columns in the continuation equation.

Hint 2

The formulas are $\gamma'=\gamma T^{-1}$ and $M_+'=TM_+T^{-1}$. Reversing a based loop gives its inverse, independently of the choice of frame.

Solution

The old monopole row $(1,0)$ becomes $(1,-1)$. Its central charge is still $a_D=a_D'-a'$, where primes here label the new frame, not derivatives. Matrix multiplication gives

$$M_+'=\begin{pmatrix}-1&2\\-2&3\end{pmatrix},\qquad \tau'=\tau+1,\qquad (M_+')^{-1}=\begin{pmatrix}3&-2\\2&-1\end{pmatrix}.$$

The last matrix belongs to the clockwise circuit. The first happens to have the numerical form of the paper’s other-singularity matrix in its original based convention; that coincidence does not move the monopole singularity. Its frame and the charge row have changed together. The preserved central charge and $\operatorname{Im}\tau$ check the result. Following an actual path across a stability wall is a separate physical question.

2. Change the scale without changing the dimensionless vacuum

Replace $\Lambda$ by $c\Lambda$ with $c>0$, holding $U=u/\Lambda^2>1$ fixed. How do $a,a_D,\tau$ and the BPS bound change? Contrast this with keeping physical $u$ fixed. Can the ratio $a_D/a$ be used for $\tau$ at $U=3$?

Hint 1

In the period integral use $x=\Lambda^2X$. The differential carries one overall power of $\Lambda$.

Hint 2

At $U=3$, the elliptic parameter of the period lesson is $2/(U+1)=1/2$. The two derivative integrals therefore have equal positive magnitudes.

Solution

At fixed $U$, both periods multiply by $c$, so $Z$ and the BPS mass bound do too. Their $u$ derivatives each multiply by $1/c$, and $\tau$ is unchanged. At fixed physical $u$, however, $U$ becomes $U/c^2$: this changes the dimensionless vacuum and can leave the original weak-coupling regime. These are different comparisons.

At $U=3$, the derivative ratio is $\tau=i$. The original contour integrals give $a/\Lambda\simeq2.43201345$ and $a_D/\Lambda\simeq0.906493920i$, so $a_D/a\simeq0.372734i$. Their ratio is a useful central-charge phase diagnostic but is not the gauge coupling. The derivative in the definition of $\tau$ cannot be dropped.

3. Rotate the adjoint deformation phase

Let $\mu=|\mu|e^{i\alpha}\ne0$, with $|\mu|\ll\Lambda$, and use the local slope $a_D=i(u-s)/(2\Lambda)+\cdots$. Solve the F- and D-flat equations near the monopole point, including the complex product $M\widetilde M$. Does a generic phase $\alpha$ break the remaining supersymmetry? What conclusion about a string tension is justified without a separate approximation?

Hint 1

Invert the local slope to obtain $u'(0)$, with the derivative now taken with respect to $a_D$. Do not replace a complex product by its absolute value in an F equation.

Hint 2

The F equations include $\sqrt2M\widetilde M+\mu u'(a_D)=0$ and $a_D M=a_D\widetilde M=0$. The D equation fixes equal magnitudes, while a gauge transformation changes the relative phase.

Solution

The slope gives $u'(0)=-2i\Lambda$. Thus

$$a_D=0,\qquad M\widetilde M=i\sqrt2\,\mu\Lambda,\qquad |M|=|\widetilde M|=(\sqrt2|\mu|\Lambda)^{1/2}.$$

One representative has both fields equal to the last magnitude times $e^{i(\alpha/2+\pi/4)}$. Their common sign is gauge equivalent. Every F and D term vanishes, so the phase does not spontaneously break the remaining four real supercharges. Setting the condensates to zero at nonzero $\mu$ would fail the first F equation.

The condensate establishes a dual Higgs mechanism in the controlled local theory. It does not by itself establish an exact string tension for finite $\mu$. A BPS tension formula requires the additional leading local approximation and its energy completion, taught in the flux-tube calculation. The local kinetic description and omitted higher terms must be controlled separately.

Return to the paper’s final spectrum discussion with two separate questions: which central charge can become small, and which particle actually occupies that charge sector? Answering the first from periods is a major part of the solution; it does not silently answer the second.