Past exam of the mathematics course of the University of Cambridge / 2001 / iii / Paper 64 (String Theory)

Use flat target spacetime with metric ${\eta }_{\mu \nu }=\mathrm{diag}(-1,1,\dots ,1)$, and let $h_{ab}$ be an independent Lorentzian worldsheet metric. The Polyakov action isHere $X$ is the string embedding map and $T$ the string tension. Variation of $X$, followed by integration by parts, givesThus the embedding equation isThe boundary term vanishes by periodicity for a closed string. At an open-string boundary it requires, for example, vanishing normal derivative for a Neumann boundary condition, or $\delta X=0$ for a Dirichlet boundary condition. Mixed choices give corresponding endpoint conditions.

Define the induced worldsheet metric ${\gamma }_{ab}={\partial }_{a}X\cdot {\partial }_{b}X$. Using $\delta \sqrt{-h}=-\frac{1}{2}\sqrt{-h}{h}_{ab}\delta h^{ab}$, metric variation givesThe second set of equations is therefore the vanishing worldsheet stress tensor:For a nondegenerate induced metric this says $h_{ab}=e^{2\omega }{\gamma }_{ab}$. Substituting into the action gives $S_P=-T\int \sqrt{-\det \gamma }{d}^2\sigma$, proving the classical equivalence of Polyakov and Nambu–Goto actions. The auxiliary metric has not added physical metric degrees of freedom.

There are three important symmetry statements. Worldsheet diffeomorphisms change the parameterization, with $X$ transforming as a scalar and $h$ as a tensor; the integral is invariant. Weyl transformations $h_{ab}\mapsto e^{2\omega (\sigma )}{h}_{ab}$ leave $\sqrt{-h}{h}^{ab}$ invariant specifically in two dimensions. These are gauge redundancies. Locally, the two diffeomorphism functions and one Weyl function fix the three independent components of $h$ to conformal gauge, $h_{ab}=e^{2\omega }{\eta }_{ab}$. Global worldsheet moduli may remain and cannot be gauged away. Separately the flat-target Poincaré group acts globally on $X$ and supplies conserved target momentum and angular momentum, rather than another worldsheet gauge constraint.

In conformal gauge the embedding equations and metric constraints becomeor equivalently ${\dot{X}}^2+X^{\prime 2}=0$ and $\dot{X}\cdot X^{\prime }=0$. The Virasoro constraints must still be imposed: fixing the metric in the action before varying it would miss them. Residual conformal changes of the two null worldsheet coordinates are generated by these constraints. Classically these statements hold in any target dimension; cancellation of the quantum worldsheet Weyl anomaly is a further condition, not part of the classical variation.

Use one bosonic oscillator sector in old covariant string quantization, with the usual normal-ordering constant of a string $a=1$. Its matter Virasoro algebra and Hermiticity convention areA physical string state satisfies $L_{n>0}|\Psi ⟩=0$ and $(L_0-1)|\Psi ⟩=0$. Let $|k⟩$ be the oscillator vacuum with $L_{n>0}|k⟩=0$ and $L_0|k⟩=h|k⟩$. For a level-two descendant, the zero-mode condition gives $h+2=1$, hence $h=-1$. The vacuum carrying this momentum is not itself a physical level-zero tachyon state; its descendant is the candidate physical state. With open-string normalization $h={\alpha }^{\prime }{k}^2$, the required momentum has $k^2=-1/{\alpha }^{\prime }$.

The positive-mode conditions can be computed without assuming the answer. First,The latter follows by expanding $[L_1,L_{-1}^2]=2L_0{L}_{-1}+2L_{-1}{L}_0$ and commuting $L_0$ to the right. Therefore $L_1(L_{-2}+\mu L_{-1}^2)|k⟩=0$ gives $3-2\mu =0$, or $\mu =3/2$. Similarly,At $h=-1$ and $\mu =3/2$ the second condition is $(d/2-13)|k⟩=0$. Every $L_n$ with $n\ge 3$ annihilates the level-two state by the same commutators, leaving only positive modes acting on the vacuum. ThusThis is the level-two scalar Virasoro null state.

For the norm, take the vacuum norm to be one, suppressing the continuum momentum delta function. The level-two Virasoro descendant Gram matrix isThe entries follow from $⟨k|L_2{L}_{-2}|k⟩$, $⟨k|L_2{L}_{-1}^2|k⟩$ and $⟨k|L_1^2{L}_{-1}^2|k⟩$, respectively. At $h=-1$,Consequently the physical state at $d=26$ has norm squared zero. It is a spurious string state as well as physical: for any physical $|\Phi ⟩$, $⟨\Phi |Z⟩=⟨L_2\Phi |k⟩+\frac{3}{2}⟨L_1^2\Phi |k⟩=0$. The null-state quotient of a string therefore identifies it with zero. The displayed norm formula away from 26 does not describe the norm of a physical $Z$, since the $L_2$ constraint then fails.

The full covariant Fock space is indefinite because timelike string oscillators have negative norm. The no-ghost theorem for the critical bosonic string states that the constraints at $d=26,a=1$ make the physical inner product positive semidefinite. Every physical state can be represented by a positive transverse state plus a physical null state. Modding out the latter leaves a positive physical space, rather than retaining negative-norm states and simply ignoring them.

One outline of the proof uses DDF operators. Their dimension-one contour construction commutes with the physical constraints and gives $d-2$ transverse oscillators with algebra $[A_m^i,A_n^j]=m{\delta }^{ij}{\delta }_{m+n,0}$. Their oscillator states have manifestly positive norm. The covariant constraints and spurious descendants can be organized triangularly in the remaining timelike and longitudinal oscillators. At the critical dimension the completeness step shows that their remaining physical contribution is null, so the 24 transverse DDF oscillators represent the whole quotient.

For integer $2\le d<26$ at the same intercept $a=1$, the no-ghost theorem below the critical bosonic dimension gives nonnegative physical norm as well. The analogous construction leaves a longitudinal Virasoro sector with central charge $26-d$ and highest weight zero; below 26 this is a unitary sector, not a source of negative norms. At 26 it becomes null. Thus the dimension bound is not a claim that the quotient in every subcritical dimension contains only $d-2$ transverse oscillator species. The argument assumes the ordinary nonzero-momentum sector where the reference null direction for the DDF construction can be chosen. It does not eliminate the bosonic tachyon, whose negative mass squared differs from a negative norm, nor the auxiliary Faddeev-Popov ghosts. Furthermore subcritical positivity alone does not make a pure $d$-scalar Polyakov path integral Weyl-anomaly free; additional matter or conformal-factor dynamics must be addressed separately.

The Polyakov path integral integrates over both the string embedding map and the worldsheet metric. After the usual analytic continuation, a formal closed-string vacuum functional has the schematic formThe genus weights organize string perturbation theory; the coupling $g_s$ is related to the constant dilaton. External states are inserted as string vertex operators, with the usual additional external coupling normalization. Division by the gauge group prevents counting different parameterizations and Weyl representatives as distinct configurations.

Gauge-fix $h=e^{2\varphi }\hat{h}(m)$, where $m$ labels inequivalent conformal structures. Conformal gauge fixes the local metric redundancies, not the worldsheet moduli. Unmarked compact surfaces have no complex moduli at genus zero, one at genus one and $3g-3$ at genus $g\ge 2$. Integrating over the appropriate moduli quotient includes large diffeomorphisms; at genus one it requires a modular fundamental domain, as in torus modular invariance.

The infinitesimal traceless diffeomorphism operator isIts Faddeev-Popov determinant is represented by anticommuting worldsheet ghost fields $c^a$ and traceless symmetric $b^{ab}$, with action proportional to $\int \sqrt{h}{b}^{ab}(P_{1}c{)}_{ab}$. In conformal complex coordinates this is the pair of chiral bc systems of weights $(2,-1)$. Ghost zero modes must be handled rather than included in an ordinary nonzero determinant: conformal Killing vectors correspond to residual gauge transformations, while antighost zero modes pair with surviving moduli variations. At genus zero three insertion positions can be fixed by the Möbius transformations; corresponding $c\tilde{c}$ factors supply the ghost zero modes. At higher genus the measure includes antighost insertions paired with the moduli.

In a flat target, integrating the free coordinate fields gives a Gaussian determinant $[{\det }^{\prime }(-{\Delta }_{\hat{h}}){]}^{-d/2}$, with the zero-mode integral supplying target volume or momentum conservation. Vertex insertions give the corresponding free-field correlators. For example, on a sphere, exponential operators have the factorThe remaining unfixed insertion positions and moduli are integrated with the ghost measure. Physical closed-string operators have matter conformal weights $(1,1)$; an unintegrated insertion is $c\tilde{c}V$, while an integrated insertion uses $\int d^{2}zV$. This explains how on-shell scattering amplitudes arise from the metric-and-embedding functional integral.

The quantum measure can break the classical Weyl symmetry. Each free coordinate contributes central charge one, while each chiral reparameterization ghost system contributes $-26$. Hence $c_{\mathrm{total}}=d-26$, and the worldsheet Weyl anomaly vanishes for the ordinary flat bosonic string exactly at $d=26$. Only then can the conformal factor be removed as a gauge degree of freedom without adding further dynamics. For a subcritical matter system it generally produces a Liouville-type theory, unless an appropriate additional conformal sector restores the central charge balance.

Equivalently the gauge-fixed construction uses a nilpotent string BRST operator. Physical states are its closed states modulo exact states, expressing the BRST cohomology of the gauge constraints. In the critical theory the ghost and matter contributions make this operator nilpotent and the measure gauge independent. Degenerations of moduli describe long-tube propagation and factorization onto intermediate string states. The bosonic tachyon can still cause infrared divergences; anomaly cancellation is not a claim that every vacuum integral is finite or that the bosonic vacuum is stable. The path integral sums embeddings and conformal structures, with ghosts removing local gauge overcounting and moduli retaining the genuine geometry.

Consider the closed bosonic string on a rectangular product of spatial circles with no background mixing between them. Let $Y^j$ have period $2\pi R_j$, take $0\le \sigma <2\pi$, and choose $Y_L(\tau +\sigma )$ to be the tilded oscillator sector. Closed-string topology permitsSingle-valued center-of-mass wavefunctions give quantized momentum $p^j=n_j/R_j$, $n_j\in \mathbb{Z}$. These are the momentum and winding modes. The oscillator modes remain integer moded; the zero modes are split asThe two intercept-one Virasoro constraints, with noncompact $k^2=-M^2$, giveThus the mass formula and compact-circle closed-string level matching condition areHere $N$ labels right movers and $\tilde{N}$ left movers; exchanging the labels changes the displayed sign convention, not the spectrum. Noncompact oscillator polarizations are still restricted by positive-mode Virasoro conditions and null-state identifications. T-duality exchanges $n_j,w_j$ and $R_j,{\alpha }^{\prime }/R_j$, preserving masses and the physical state conditions.

At generic radius, neutral $(N,\tilde{N})=(1,1)$ states with one noncompact oscillator and one compact oscillator give two massless vectors per circle. They are the two Abelian gauge fields associated with the compact metric and two-form components. Additional charged vectors can occur at oscillator level sum one. For one circle, set all other compact charges to zero. Take $n=w=\pm 1$ and $(N,\tilde{N})=(1,0)$, with a right-moving noncompact oscillator ${\alpha }_{-1}^{\mu }$; or take $n=-w=\pm 1$ and $(N,\tilde{N})=(0,1)$, with a left-moving noncompact oscillator ${\tilde{\alpha }}_{-1}^{\mu }$. These obey level matching. Their common squared mass isTherefore four extra charged vector particles become massless at $R=\sqrt{{\alpha }^{\prime }}$, the self-dual circle. Away from that radius they are massive. They are genuine vector states: the noncompact level-one oscillator carries a vector polarization, and the Virasoro constraint plus its null identification impose transversality and remove the gauge polarization at zero mass.

The resulting self-dual circle current algebra makes the gauge enhancement explicit. At the special radius, the left-moving charges above have $p_L=\pm 2/\sqrt{{\alpha }^{\prime }}$, $p_R=0$. With $Y_L(z)Y_L(w)\sim -{\alpha }^{\prime }\log (z-w)/2$, the currentshave conformal weight one. The operator products give $J^3{J}^3\sim 1/[2(z-w{)}^2]$, $J^3{J}^{\pm }\sim \pm J^{\pm }/(z-w)$ and $J^{+}{J}^{-}\sim 1/(z-w{)}^2+2J^3/(z-w)$. They form a level-one SU(2) current algebra; the right-moving copy supplies another SU(2). Pairing each chiral current with the opposite-sector noncompact derivative yields a weight-$(1,1)$ vector vertex. Thus the generic $U(1{)}_L\times U(1{)}_R$ becomes $SU(2{)}_L\times SU(2{)}_R$, with four charged generators in addition to the two Cartan vectors. Several self-dual circles give the corresponding product enhancement.

Compactification also creates scalar states and moduli; not every additional massless state is a vector. The mechanism just shown uses the intercept-one bosonic spectrum. A superstring projection changes the allowed oscillator states, so the same mass formula cannot simply be applied to a GSO-projected superstring without modification.

Introduce anticommuting matter fields ${\psi }^{\mu }$ as worldsheet Majorana fermions, transforming as target-spacetime vectors and two-dimensional spinors. The RNS string pairs them with $X^{\mu }$ through worldsheet supersymmetry. In flat superconformal gauge the matter action is, up to the conventional fermion normalization,and the field equations split $\psi$ into left- and right-moving components. The supersymmetry constraints include both the stress tensor and the supercurrent, so unphysical polarizations are removed rather than interpreting every fermionic worldsheet component as a physical particle.

The possible spin structures produce two sectors. On the closed chiral circle, or the doubled open string, the Neveu–Schwarz sector has $\psi (\sigma +2\pi )=-\psi (\sigma )$ and half-integer modes $b_r$. The Ramond sector has periodic $\psi$ and integer modes $d_n$, including $d_0$. Canonical quantization replaces the classical Grassmann variables by operators satisfyingIn particular ${\Gamma }^{\mu }=\sqrt{2}{d}_0^{\mu }$ obeys $\{{\Gamma }^{\mu },{\Gamma }^{\nu }\}=2{\eta }^{\mu \nu }$. This is the Ramond zero-mode Clifford algebra in covariant form. Its ground-state space must carry a representation of the Clifford algebra of the target metric, and hence transforms as a spinor, not a scalar vacuum. Indeed the zero-mode Lorentz generators proportional to $[{\Gamma }^{\mu },{\Gamma }^{\nu }]/4$ act on it as the spin representation. This is the essential mechanism converting anticommuting worldsheet matter into spacetime spinorial states.

The zero-mode supercurrent constraint on a Ramond ground state is proportional to $k\cdot d_0$. It becomesthe momentum-space Dirac equation. The Ramond level operator has zero intercept, so the ground state is massless. Exciting bosonic or nonzero fermionic oscillators gives tensor-spinor excitations; the Clifford ground-state factor is retained. In the Neveu–Schwarz sector there is no fermionic zero-mode degeneracy. Its ground state is a scalar and its excited states carry integer-spin tensor representations, despite the anticommuting oscillator construction. Thus worldsheet fermionic oscillators do not by themselves make every target-space state a fermion.

Consistent projection also matters. The GSO projection removes the NS tachyon and selects a definite Ramond chirality in the standard supersymmetric spectrum. At the critical dimension ten, the projected massless NS vector has eight transverse polarizations, while the projected Ramond ground state has eight physical spinor polarizations after the massless Dirac constraint and the Majorana-Weyl spinor conditions. The projection permits their organization into spacetime supersymmetry multiplets. The assignment of spacetime fermionic statistics is compatible with spinorial transformation, vertex-operator locality and the Spin-statistics theorem; it is not just a relabeling of the Grassmann variables in the classical action.

For a closed string the left and right sectors can be chosen independently subject to level matching and the consistent spin-structure sum. The NS-NS sector and Ramond–Ramond sector have bosonic spacetime representations; a pair of Ramond spinors produces integer-spin fields. The mixed NS-R sector and R-NS sector instead carry one spinor factor and yield spacetime fermions, including gravitino and dilatino states. Opposite versus equal chiralities of the two Ramond sectors distinguish the type IIA superstring theory and type IIB superstring theory choices.

Quantum consistency of this RNS matter-plus-gauge system requires $d=10$: its matter central charge is $d+d/2$, the reparameterization ghosts contribute $-26$ and the commuting superconformal ghosts contribute $11$, so the total is $3d/2-15$. Spacetime fermions arise through Ramond matter zero modes, their Clifford/spinor representations and the physical-state/projection conditions. The anticommuting $b,c$ gauge-fixing ghosts already present in the bosonic path integral are not such matter fields and do not create physical spacetime fermions.