Why do dualities matter? In perturbative string theory we expand in a small parameter, either the string coupling \(g_s\) or the ratio \(\ell_s/R\). Dualities connect regimes where those expansions fail to regimes where they succeed. They tell us that geometry, coupling, and even the choice of fundamental objects are not invariant, only the physical observables are.
1. What is a duality?
A duality is an exact isomorphism of Hilbert spaces and observables between two theoretical descriptions. If theory A with parameters \(\lambda\) looks complicated, there is often a dual theory B with parameters \(\tilde\lambda = f(\lambda)\) that looks simple. The map preserves spectra, scattering amplitudes, and symmetries.
Three archetypes in string theory:
- T-duality: large radius \(\leftrightarrow\) small radius, momentum \(\leftrightarrow\) winding.
- S-duality: strong coupling \(\leftrightarrow\) weak coupling, electric \(\leftrightarrow\) magnetic.
- U-duality: the discrete combination of S and T, unifying the dualities of M-theory compactifications.
2. T-duality: circles and strings
Compactify one spatial direction on a circle of radius \(R\). A closed string has quantized momentum \(p = n/R\) with \(n\in\mathbb{Z}\), and it can wind \(m\) times around the circle. The winding contribution to the energy is proportional to the string tension \(1/(2\pi\alpha')\) times the length \(2\pi R m\), so \(w = mR/\alpha'\).
The closed string mass formula, ignoring constants, is
\[ M^2 = \left(\frac{n}{R}\right)^2 + \left(\frac{mR}{\alpha'}\right)^2 + \frac{2}{\alpha'}(N+\tilde N-2) \]This is invariant under the simultaneous exchange
\[ R \;\longleftrightarrow\; \tilde R = \frac{\alpha'}{R}, \qquad n \;\longleftrightarrow\; m \]which maps Type IIA to Type IIB, and Heterotic \(SO(32)\) to \(E_8\times E_8\). For open strings, T-duality exchanges Neumann and Dirichlet boundary conditions, turning Dp-branes into D(p\(\pm\)1)-branes.
Try it interactive
Left circle is the original radius, right is the T-dual. A string wound \(m\) times around a small circle looks like a particle with momentum \(n\) on a large circle.
Low-lying spectrum
\(M^2\) shown without oscillator terms. Swapping \((n,m)\) and replacing \(R\) by \(\tilde R\) leaves the table invariant.
3. S-duality: strong meets weak
Many theories have a complex coupling
\[ \tau = \frac{\theta}{2\pi} + \frac{4\pi i}{g^2} \]The transformation \(\tau \to -1/\tau\) inverts the coupling \(g \to 1/g\) when \(\theta=0\). It exchanges electrically charged perturbative states with magnetically charged solitons.
- Type IIB is self-dual under \(SL(2,\mathbb{Z})\). The fundamental string \((1,0)\) is exchanged with the D1-string \((0,1)\).
- Type I \(SO(32)\) is S-dual to Heterotic \(SO(32)\): \(g_{\text{het}} = 1/g_{\text{I}}\).
Try it interactive
With \(\theta=0\), \(\tau = 4\pi i/g^2\) lies on the imaginary axis. S-duality reflects it across the unit circle.
Numbers
| \(\tau\) | i·34.9 |
|---|---|
| \(-1/\tau\) | i·0.029 |
| Im \(\tau\) | 34.906 |
| Im \((-1/\tau)\) | 0.029 |
Large Im \(\tau\) means weak coupling. Its dual sits near the real axis, strong coupling.
4. U-duality and M-Theory
The five ten-dimensional superstring theories, Type IIA, Type IIB, Type I, Heterotic \(SO(32)\), and Heterotic \(E_8\times E_8\), are connected by chains of S and T dualities. They are all limits of a single eleven-dimensional theory, M-theory.
Type IIA at coupling \(g_s\) lifts to M-theory on a circle of radius
\[ R_{11} = g_s^{2/3}\, \ell_p, \qquad \ell_s^2 = \frac{\ell_p^3}{R_{11}} \]so strong IIA coupling decompactifies an extra dimension. The low energy limit of M-theory is eleven-dimensional supergravity.
Try it interactive
With \(\ell_p=1\), \(R_{11}=g_s^{2/3}\). When \(g_s \gg 1\), the circle is large and the theory is eleven-dimensional.
Duality web
| \(R_{11}/\ell_p\) | 1.000 |
|---|---|
| \(\ell_s/\ell_p\) | 1.000 |
| Regime | crossover |
5. Mathematics behind it
Buscher rules for T-duality
For a target space with an isometry along \(x^0\), the dual metric \(\tilde G\) and B-field \(\tilde B\) are
\[ \tilde G_{00} = \frac{1}{G_{00}},\quad \tilde G_{0i} = \frac{B_{0i}}{G_{00}},\quad \tilde B_{0i} = \frac{G_{0i}}{G_{00}},\quad \tilde G_{ij} = G_{ij} - \frac{G_{0i}G_{0j}-B_{0i}B_{0j}}{G_{00}} \]with dilaton shift \(e^{-2\tilde\phi} = e^{-2\phi} G_{00}\). This preserves conformal invariance of the worldsheet sigma model.
Narain lattices
On \(T^d\), left and right momenta \((p_L,p_R)\) form an even self-dual Lorentzian lattice \(\Gamma^{d,d}\). T-duality is the orthogonal group \(O(d,d;\mathbb{Z})\) acting on this lattice, preserving \(p_L^2-p_R^2\in 2\mathbb{Z}\).
Modular group
S-duality sits inside \(SL(2,\mathbb{Z})\) acting on \(\tau\):
\[ \tau \mapsto \frac{a\tau+b}{c\tau+d},\quad ad-bc=1,\quad a,b,c,d\in\mathbb{Z} \]The generators \(T:\tau\to\tau+1\) and \(S:\tau\to-1/\tau\) generate the full duality group of Type IIB.
U-duality groups for toroidal M-theory: in \(D=11-d\) dimensions you get \(E_{d(d)}(\mathbb{Z})\). For \(d=6\), that is \(E_{6(6)}(\mathbb{Z})\).
6. GNU Octave examples
Copy these into Octave to explore the maps numerically.
1) t_duality_spectrum.m
% computes M^2 for momentum n and winding m on R and dual Rd alpha = 1.0; R = 1.5; Rd = alpha / R; nmax = 3; mmax = 3; [M,N] = meshgrid(0:mmax, 0:nmax); % M=m, N=n M2_R = (N./R).^2 + (M.*R/alpha).^2; M2_Rd = (N./Rd).^2 + (M.*Rd/alpha).^2; disp(' n m M2(R) M2(Rd) '); disp([(N(:)) (M(:)) M2_R(:) M2_Rd(:)]); % check invariance under n<->m and R<->Rd err = max(abs(M2_R(:) - M2_Rd([1 4 2 5 3 6]))); % for small grid
2) s_duality_map.m
% plots tau and its S-dual -1/tau in upper half-plane g = logspace(-1, 1, 400); % 0.1 to 10 theta = 0; tau = theta/(2*pi) + 4*pi*1i./(g.^2); tau_dual = -1./tau; figure; hold on; grid on; axis equal; plot(real(tau), imag(tau), 'LineWidth',2); plot(real(tau_dual), imag(tau_dual), '--', 'LineWidth',2); xlabel('Re \tau'); ylabel('Im \tau'); title('S-duality: \tau \to -1/\tau'); legend('\tau(g)', '-1/\tau', 'location','northeast');
3) mtheory_scale.m
% relates IIA string coupling to M-theory circle lp = 1.0; gs = logspace(-1, 1, 200); R11 = gs.^(2/3) * lp; ls = sqrt(lp^3 ./ R11); % ls^2 = lp^3 / R11 loglog(gs, R11, 'LineWidth',2); hold on; loglog(gs, ls, '--', 'LineWidth',2); grid on; xlabel('g_s'); ylabel('length / l_p'); legend('R_{11}','\ell_s'); title('M-theory lift of IIA');