On the Structure of Local Observables in String Field Theory

1. Introduction

In relativistic quantum field theory on Minkowski spacetime, locality is usually encoded by microcausality with operators associated with spacelike separated regions commute, in 1993 Sorkin demonstrated that this kinematic locality condition is not, by itself, compatible with the textbook measurement postulate when one attempts to implement ideal measurements of field observables localized to bounded spacetime regions [1,2]. The essential point is that the Lüders state-update map associated with an ideal measurement can act nonlocally on the state, and in a three-region protocol this nonlocal update can be converted into a superluminal signaling channel. The resulting obstruction is often summarized as the statement that most ideal measurements on quantum fields are impossible. It is important to distinguish the present claim from what is already known in algebraic and relativistic quantum field theory (AQFT), by AQFT we mean the Haag–Kastler framework in which to each suitable spacetime region $\mathcal{O}$ one assigns a local observable algebra $\mathcal{A}\left(\mathcal{O}\right)$, subject in particular to isotony and Einstein causality, that if ${\mathcal{O}}_1$ and ${\mathcal{O}}_2$ are spacelike separated, then every observable in $\mathcal{A}\left({\mathcal{O}}_1\right)$ commutes with every observable in $\mathcal{A}\left({\mathcal{O}}_2\right)$. In that framework, as well as in modern local-measurement theory for QFT, it has long been understood that localization, state preparation, and measurement are subtler than the textbook finite-dimensional picture suggested by nonrelativistic quantum mechanics. In particular, one should not expect the measurement theory of a relativistic field to be exhausted by sharply localized projector-valued idealizations and instantaneous Lüders updates [3,4,5].

Our point is therefore not just that relativistic QFT contains conceptual subtleties about localization but that it is already part of the standard lore of the theory. The new point of the present paper is that covariant open string field theory (SFT) supplies a simple ultraviolet mechanism for why the sharp-projector idealization fails operationally. Because the fundamental degrees of freedom are extended and the induced spacetime observables take the non-local form, the theory does not operationally realize strictly point-supported projectors in bounded regions. Instead, it satisfies an operational locality criterion that the change in any detector outcome probability induced by a spacelike separated intervention is bounded by a rapidly decaying function of the invariant separation and in the model studied below this bound is Gaussian. So SFT does not abandon relativistic locality but rather it replaces exact microcausality at arbitrarily sharp resolution by exponentially accurate locality at the string scale while recovering ordinary local QFT in the limit ${\ell }_s\to 0$ [3,4,5].

The broader program of replacing exact microcausality by an asymptotic or operational notion of locality in ultraviolet-complete deformations of local quantum field theory will be developed systematically elsewhere [2]; the present paper identifies covariant string field theory as a realization of that idea.

So the purpose of this paper is to show how the same logical tension is reformulated and resolved in covariant SFT, in SFT the fundamental object is the string field $\mathrm{\Psi }$ describing extended degrees of freedom and the induced spacetime description exhibits intrinsic nonlocality at the string length scale. This nonlocality is not a violation of relativistic causality in the operational sense relevant to communication but it does invalidate the assumptions needed to implement sharply localized projectors and exact Lüders state reductions. We show that when measurements are described dynamically by localized couplings of detectors to non-local SFT observables, any spacelike influence is bounded and exponentially suppressed beyond the string scale.

Throughout this paper we take $\square ={\eta }^{\mu \nu }{\partial }_{\mu }{\partial }_{\nu }$ as the d’Alembertian, $M_{*}$ is some UV mass scale, and ${\ell }_{*}=M_{*}^{-1}$ is the associated length scale. In the stringy setting we identify $M_{*}\sim {\ell }_s^{-1}$, where ${\ell }_s=\sqrt{{\alpha }^{\prime }}$ is the string length and ${\alpha }^{\prime }$ is the inverse string tension, in this paper $x,y\in {\mathbb{R}}^{1,3}$ are spacetime points with Minkowski metric ${\eta }_{\mu \nu }=\mathrm{diag}(1,-1,-1,-1)$, and ${(x-y)}^2={\eta }_{\mu \nu }{(x-y)}^{\mu }{(x-y)}^{\nu }$. The separation is spacelike when ${(x-y)}^2<0$. Microcausality asserts that for local observables $\mathcal{O}\left(x\right)$ and ${\mathcal{O}}^{\prime }\left(y\right)$ one has $[\mathcal{O}\left(x\right),{\mathcal{O}}^{\prime }\left(y\right)]=0$ whenever ${(x-y)}^2<0$, where $[X,Y]=XY-YX$ is the commutator.

2. Sorkin’s Impossible Measurement

We will start by reviewing the measurement structure that generates Sorkin’s paradox, $\mathcal{H}$ is the Hilbert space of the field theory we will use and ${\rho }_0$ is the initial density matrix meaning that ${\rho }_0$ is a positive trace-one operator on the Hilbert space $\mathcal{H}$. We consider three spacetime regions $O_1,O_2,O_3\subset {\mathbb{R}}^{1,3}$ such that $O_1$ lies to the past of $O_3$, we have that $O_2$ is a thickened spacelike hypersurface between them, and $O_1$ is spacelike separated from $O_3$[1,2]. If we let $A,B,C$ be self-adjoint observables associated with $O_1,O_2,O_3$ respectively and let $\left\{P_a\right\}$, $\left\{Q_b\right\}$, $\left\{R_c\right\}$ be their spectral projectors so that:

$$A=\sum _{a}a{P}_a,B=\sum _{b}b{Q}_b,\mathrm{and}C=\sum _{c}c{R}_c,$$

(1)

with $P_a{P}_{a^{\prime }}={\delta }_{a{a}^{\prime }}{P}_a$ and ${\sum }_a{P}_a=\mathbb{1}$, and analogously for $Q_b$ and $R_c$. Here $\mathbb{1}$ is the identity operator on our Hilbert space $\mathcal{H}$. If A is self-adjoint the spectral theorem says that A can be written in terms of its spectrum, its possible measurement values and the corresponding orthogonal projectors onto the associated eigenspaces. In the purely discrete case of finite or countable point spectrum this takes the simple sum over eigenvalues form:

$$A={\int }_{\mathbb{R}}\lambda dP\left(\lambda \right),$$

(2)

where $\lambda$ is just a generic spectral value of the observable. In operational terms $\lambda$ is the readout value on the measurement device such as position, momentum, energy, etc., while the projectors or $dP\left(\lambda \right)$ encode which subspace of states corresponds to that readout.

Definition 1 

(Ideal measurement instrument). Given a self-adjoint observable with spectral projectors $\left\{P_a\right\}$, the ideal, non-selective measurement map on states is

$${\mathcal{M}}_A\left(\rho \right)=\sum _a{P}_a\rho P_a,$$

(3)

where ρ is a density matrix. The corresponding selective update conditioned on outcome a is $\rho \mapsto P_a\rho P_a/\mathrm{Tr}\left(P_a\rho \right)$ when $\mathrm{Tr}\left(P_a\rho \right)\ne 0$.

Sorkin’s observation is that even when A and C are spacelike separated and thus commute in the microcausal sense the statistics of C after an intervening ideal measurement of B can depend on whether A was measured. Concretely, suppose B is certainly measured in $O_2$, then the non-selective state after measuring B but not A is ${\rho }_b={\mathcal{M}}_B\left({\rho }_0\right)$, and after measuring A and then B it is ${\rho }_{BA}={\mathcal{M}}_B\left({\mathcal{M}}_A\left({\rho }_0\right)\right)$. The probability of obtaining outcome c when measuring C is:

$$\begin{array}{cc}\hfill p\left(c\right|B)& =\mathrm{Tr}\left(R_c{\rho }_B\right),\hfill \end{array}$$

(4)

$$\begin{array}{cc}\hfill p\left(c\right|A,B)& =\mathrm{Tr}\left(R_c{\rho }_{BA}\right),\hfill \end{array}$$

(5)

these are just the Born rule applied to two different post-measurement states depending on what was done earlier. If we start with an initial state ${\rho }_0$ that is a positive, trace-one operator on the Hilbert space $\mathcal{H}$ then given that B was measured means we apply the ideal, non-selective Lüders instrument for B. For an observable with spectral projectors $\left\{P_a\right\}$, the non-selective ideal measurement map is given by (3). So if B is certainly measured but you don’t condition on which outcome b, the state becomes:

$${\rho }_B={\mathcal{M}}_B\left({\rho }_0\right).$$

(6)

If A is measured first and then B, the state becomes:

$${\rho }_{BA}={\mathcal{M}}_B\left({\mathcal{M}}_A\left({\rho }_0\right)\right).$$

(7)

Now we measure C, the outcome c corresponds to the projector $R_c$. The Born rule says that in state $\rho$ the probability of outcome c is:

$$p\left(c\right)=\mathrm{Tr}\left(R_c\rho \right).$$

(8)

We apply that once with $\rho ={\rho }_B$ and once with $\rho ={\rho }_{BA}$, giving (4). Here $\mathrm{Tr}$ denotes the trace on the Hilbert space $\mathcal{H}$. Sorkin shows us that for generic choices of ${\rho }_0$ and observables $A,B,C$, one finds $p\left(c\right|A,B)\ne p(c\left|B\right)$ even when $[A,C]=0$, meaning that the statistics in $O_3$ depend on an intervention in $O_1$ that is spacelike to $O_3$[1,2]. If one party in $O_1$ chooses whether to measure A and a second party in $O_3$ reads out C, then the certainty that B is measured in between allows the difference in (4) to function as a superluminal signaling channel. To state this simply Sorkin’s impossible measurement identifies a mismatch between microcausality and textbook ideal measurement postulates for sharply localized field observables [1,2].

The core assumption is not microcausality but the measurement postulate (3) applied to sharply localized projectors in bounded spacetime regions. The resolution strategy we pursue is to replace that postulate by a dynamical measurement model built from localized couplings to the degrees of freedom available in the UV-complete theory. In string field theory, this replacement is not optional as the theory does not furnish sharply localized projectors as operational observables.

It is important to clarify that the effect above is not the familiar EPR/Bell type nonlocality of entangled states in ordinary quantum mechanics. In EPR or Bell like scenarios one has spacelike separated choices of measurements on subsystems, but the no-signaling property holds as local interventions in $O_1$ cannot change the marginal outcome statistics in $O_3$ when the operations are implemented by genuinely local completely-positive instruments. The correlations can violate Bell inequalities but they do not allow superluminal control of $p\left(c\right)$. Sorkin’s setup is different as the spacetime separation is fully relativistic with $[A,C]=0$ by microcausality, and the putative signaling arises only after inserting an intermediate ideal measurement in $O_2$ and most importantly assuming the existence of sharply localized projectors and a Lüders-type projection postulate (3) for bounded regions. The paradox is therefore not mysterious action at a distance from entanglement but it is a clash between microcausal locality of observables and an instantaneous, nonselective, projective update rule for operators that are assumed to be strictly localized in spacetime. So what fails is not quantum mechanics per se, but the assumption that we can realize ideal von Neumann–Lüders measurements with sharply localized spectral projectors as operational procedures in a relativistic QFT.

3. Non-Locality from String Field Theory

We work in covariant open bosonic SFT for definiteness, although the present analysis is carried out in covariant open bosonic SFT, it is useful to say that closely related covariant constructions can also exist for closed strings where the interaction structure is organized by a Batalin–Vilkovisky master action rather than a purely cubic open-string vertex [6]. We will not need the closed-string formalism in detail here but it makes clear that the replacement of strictly local spacetime fields by extended string degrees of freedom is not special to the open-string sector. A useful point of contact is the older Hamiltonian often light–cone or proper–time formulation of string field theory due to Kaku and Kikkawa [7] where one starts from first–quantized open–string evolution in a worldsheet time parameter $\tau$:

$$i{\partial }_{\tau }|\mathrm{\Phi }\left(\tau \right)\rangle =H|\mathrm{\Phi }\left(\tau \right)\rangle ,$$

(9)

where $\left|\mathrm{\Phi }\right(\tau )\rangle$ is the second–quantized string field, a functional of the string embedding $X\left(\sigma \right)$, and in a covariant treatment, also the ghost degrees of freedom and H is the first–quantized string Hamiltonian or equivalently a Virasoro constraint operator where for the open bosonic string one may think of $H\sim L_0-1$ in appropriate units. Second quantization promotes $\left|\mathrm{\Phi }\right(\tau )\rangle$ to a field and packages the free evolution into the quadratic action:

$$S_0\left[\mathrm{\Phi }\right]={\displaystyle \frac{1}{2}}\int d\tau \langle \mathrm{\Phi }\left(\tau \right)|\left(i{\partial }_{\tau }-H\right)|\mathrm{\Phi }\left(\tau \right)\rangle .$$

(10)

Interactions are introduced by allowing strings to join and split, this is encoded by an associative string star products * implementing gluing of half–strings, and a cubic interaction term1:

$$S_{\mathrm{int}}\left[\mathrm{\Phi }\right]={\displaystyle \frac{g}{3}}\int d\tau \langle \mathrm{\Phi }\left(\tau \right)|\mathrm{\Phi }\left(\tau \right)*\mathrm{\Phi }\left(\tau \right)\rangle ,$$

(11)

where g is the string-field coupling constant, it sets the overall strength of the cubic vertex that glues or splits strings through the star product. In perturbation theory each insertion of the cubic interaction contributes one power of the string-field coupling g, so transition amplitudes are organized as an expansion in g, up to conventional rescalings of the string field. In the covariant Witten formulation we denote the same interaction-strength parameter by $g_o$, but the relation between g and $g_o$ is convention-dependent and can be absorbed into the normalization of $\mathrm{\Phi }$ versus $\mathrm{\Psi }$. The constant multiplying the cubic vertex is partly a matter of convention, because it can be traded against the normalization of the string field. For example a common convention writes Witten’s open string field theory action as an overall $g_o^{-2}$ times a universal Chern–Simons-like functional:

$$S\left[\mathrm{\Phi }\right]=-{\displaystyle \frac{1}{g_o^2}}\left[{\displaystyle \frac{1}{2}}\langle \mathrm{\Phi },Q_B\mathrm{\Phi }\rangle +{\displaystyle \frac{1}{3}}\langle \mathrm{\Phi },\mathrm{\Phi }*\mathrm{\Phi }\rangle \right],$$

(12)

where $g_o$ is the open-string coupling fixed operationally by matching an on-shell three-open-string amplitude. In contrast in a field-theory convention one often factors out the quadratic term and writes:

$$S\left[\mathrm{\Psi }\right]={\displaystyle \frac{1}{2}}\langle \mathrm{\Psi },Q_B\mathrm{\Psi }\rangle +{\displaystyle \frac{g}{3}}\langle \mathrm{\Psi },\mathrm{\Psi }*\mathrm{\Psi }\rangle .$$

(13)

These two forms are related by a field rescaling. Setting:

$$\mathrm{\Phi }=\alpha \mathrm{\Psi },$$

(14)

and inserting into (12) gives:

$$S\left[\mathrm{\Psi }\right]=-{\displaystyle \frac{1}{g_o^2}}\left[{\displaystyle \frac{1}{2}}{\alpha }^2\langle \mathrm{\Psi },Q_B\mathrm{\Psi }\rangle +{\displaystyle \frac{1}{3}}{\alpha }^3\langle \mathrm{\Psi },\mathrm{\Psi }*\mathrm{\Psi }\rangle \right].$$

(15)

Choosing $\alpha =g_o$ and absorbing the overall sign into the Euclidean/Minkowski convention for S if desired yields a canonically normalized kinetic term and identifies the cubic coefficient as $g=g_o$ in this particular normalization. More generally any change of convention in the BPZ inner product, the star product, or the overall prefactor of the action can be compensated by a rescaling $\mathrm{\Phi }\mapsto \alpha \mathrm{\Phi }$, which shifts the apparent relation between g and $g_o$. So the relation between g and $g_o$ is convention-dependent and may be absorbed into the normalization of $\mathrm{\Phi }$ versus $\mathrm{\Psi }$. The classical equation of motion then becomes:

$$\left(i{\partial }_{\tau }-H\right)|\mathrm{\Phi }\rangle +g|\mathrm{\Phi }*\mathrm{\Phi }\rangle =0.$$

(16)

This is the precise meaning of the popular schematic formula [7]:

$$L={\mathrm{\Phi }}^{†}[i{\partial }_{\tau }-H]\mathrm{\Phi }+{\mathrm{\Phi }}^{†}*\mathrm{\Phi }*\mathrm{\Phi },$$

(17)

where the first term is the second–quantized Schrödinger evolution of a string, and the second term is the join/split vertex.

Witten’s covariant open string field theory can be understood as the BRST– and gauge–invariant completion of this structure, in a covariant description the worldsheet theory has gauge redundancies, the reparametrizations and Weyl symmetry, this is so physical states are BRST cohomology classes. Accordingly, the kinetic operator is not a Hamiltonian H but the BRST charge $Q_B$, and the natural pairing is the BPZ inner product $\langle ·|·\rangle$ on the first–quantized state space. The unique cubic, associative, gauge–invariant action with these ingredients is Witten’s:

$$S_{\mathrm{SFT}}\left[\mathrm{\Psi }\right]=-{\displaystyle \frac{1}{g_o^2}}\left({\displaystyle \frac{1}{2}}\langle \mathrm{\Psi }|Q_B\mathrm{\Psi }\rangle +{\displaystyle \frac{1}{3}}\langle \mathrm{\Psi }|\mathrm{\Psi }*\mathrm{\Psi }\rangle \right),$$

(18)

where $g_o$ is the open-string coupling constant, $Q_B$ is the BRST operator encoding worldsheet gauge invariances, $〈·|·〉$ is the BPZ inner product on the string state space, and * is the associative star product implementing string interaction by gluing half-strings. The action (18) is gauge invariant under $\delta \mathrm{\Psi }=Q_B\Lambda +\mathrm{\Psi }*\Lambda -\Lambda *\mathrm{\Psi }$ with gauge parameter $\Lambda$ of ghost number zero. The match to the Hamiltonian form is seen after gauge fixing. In Siegel gauge $b_0\mathrm{\Psi }=0$ one has $Q_B\sim c_0(L_0-1)+\left(\mathrm{nonzero}\mathrm{modes}\right)$, so the free equation $Q_B\mathrm{\Psi }=0$ reduces to the familiar mass–shell constraint $(L_0-1)\mathrm{\Psi }=0$, the same operator content as the Hamiltonian H used above. The propagator in covariant SFT is generated by the Schwinger parameter representation:

$${\displaystyle \frac{1}{L_0-1}}={\int }_0^{\infty }ds{e}^{-s(L_0-1)},$$

(19)

so the string time/proper–time evolution parameter in the Hamiltonian picture is the worldsheet length modulus that appears in the covariant gluing construction. Thus Kaku–Kikkawa SFT is recovered as a noncovariant gauge–fixed presentation of the same joining/splitting algebra that Witten packages into a manifestly covariant BRST–invariant cubic action. The fundamental field is the string field $\mathrm{\Psi }$, an element of the first-quantized open-string state space with ghost number one. The classical action is the cubic Witten action [8].

To touch on Lorentz algebra closure and Poincaré invariance we note that in light-cone quantization one must explicitly check that the quantum Lorentz generators close as this requirement fixes the critical dimension and intercept for the bosonic string, eventually yielding $D=26$ with a massless vector at the first excited level, and it does not by itself remove the tachyonic ground state [9]. In the present paper we use the covariant BRST-invariant SFT formulation, in which Poincaré covariance is manifest at the level of the action and the gauge symmetry is encoded by the BRST operator.

We feel it is worth saying what would change if the Lorentz algebra did not close; in that case the would-be Poincaré generators fail to furnish a representation of the Lorentz group boosts become anomalous, transformation laws of states/operators become frame-dependent, and one loses a consistent Lorentz-covariant notion of asymptotic states and scattering. In light-cone language this appears as an anomalous commutator of the form $[M^{i-},M^{j-}]\ne 0$, while in covariant BRST language it is equivalent to a failure of BRST nilpotency, $Q_B^2\ne 0$, such as a breakdown of gauge invariance and the decoupling of negative-norm degrees of freedom. Such a theory is therefore not a consistent relativistic UV completion on Minkowski space so one must either restore consistency by anomaly cancellation/criticality, or by an appropriate noncritical completion or else interpret the setup as explicitly Lorentz-violating, in which case the operational locality bounds derived from Lorentz-invariant smearing $F(\square /M_{*}^2)$ would require a different, non-covariant reformulation.

Because covariant open SFT is defined by the BRST operator of an underlying gauge-fixed worldsheet theory, consistency requires that the would be gauge symmetry be anomaly free. The BRST charge must be nilpotent with $Q_B^2=2$ which in turn is equivalent to vanishing of the total worldsheet conformal anomaly, the vanishing total central charge. This is the precise sense in which conformal invariance (Weyl invariance) and the critical dimension enter covariant SFT.

In the holomorphic sector the stress tensor $T\left(z\right)$ admits the mode expansion [10], this is the mode expansion of the holomorphic stress–energy tensor in a 2D conformal field theory (CFT) in radial quantization. Where z is a complex coordinate on the Euclidean worldsheet. In radial quantization one may write $z=e^{\tau +i\sigma }$ with $\sigma$ as the spatial coordinate on the circle. The origin $z=0$ is the insertion point defining the vacuum state. The $L_n$ are the Virasoro generators. To make contact with the worldsheet conformal field theory underlying the BRST construction, it is useful to recall the radial-quantization formulation of the string worldsheet theory, the operator-state correspondence, and the Virasoro description of conformal symmetry [11]. They are defined by contour integrals:

$$L_n={\oint }_{\left|z\right|=r}{\displaystyle \frac{dz}{2\pi i}}{z}^{n+1}T\left(z\right),$$

(20)

They generate infinitesimal conformal transformations of $z⟼z^{\prime }=z+ϵz^{n+1}$. The imposrant special cases are $L_{-1}$ translations, $L_0$ dilatations or scale transformations, and $L_1$ special conformal transformations, meaning $L_0$ measures conformal dimension. The exponent is $z^{n+1}$ since the stress tensor has conformal weight 2. With Virasoro commutators [10]:

$$[L_m,L_n]=(m-n)L_{m+n}+{\displaystyle \frac{c}{12}}m(m^2-1){\delta }_{m+n,0},$$

(21)

where $L_m$ is the Virasoro generators or modes of the 2D conformal symmetry, $m,n$ are integer mode numbers where they label which Fourier/Laurent mode we are talking about (how many units of angular momentum around the circle/cylinder, in radial quantization language), and c is the central charge. The symbol ${\delta }_{m+n,0}$ is the Kronecker delta:

$${\delta }_{m+n,0}=\left\{\begin{array}{cc}1,\hfill & m+n=0,\hfill \\ 0,\hfill & m+n\ne 0,\hfill \end{array}\right.$$

(22)

this ensures the central term only appears when $m=-n$, such as when the commutator closes back onto the zero-mode sector in the required way For the bosonic string the matter CFT consists of D free scalars $X^{\mu }(\sigma ,\tau )$, giving:

$$c_{\mathrm{m}}=D,$$

(23)

where D is the target-space spacetime dimension of the bosonic string, and $c_m$ is the central charge of the matter sector of the worldsheet conformal field theory. Gauge-fixing reparametrizations/Weyl symmetry introduces the $(b,c)$ ghost system with:

$$c_{\mathrm{gh}}=-26,$$

(24)

in this c is the ghost field (Grassmann-odd) associated with an infinitesimal reparametrization parameter. It behaves like a worldsheet vector (more precisely, a conformal field of weight $-1$ in the holomorphic sector). b is the antighost field (also Grassmann-odd). It behaves like a worldsheet rank-2 tensor (a conformal field of weight 2 holomorphically). So the total central charge is:

$$c_{\mathrm{tot}}=c_{\mathrm{m}}+c_{\mathrm{gh}}=D-26.$$

(25)

The BRST charge holomorphic part may be written as:

$$\begin{array}{c}\hfill \begin{array}{c}\hfill Q_B=\oint {\displaystyle \frac{dz}{2\pi i}}\left[c\left(z\right)\left(T_{\mathrm{m}}\left(z\right)+\frac{1}{2}{T}_{\mathrm{gh}}\left(z\right)\right)+:b\left(z\right)c\left(z\right)\partial c\left(z\right):\right],\end{array}\end{array}$$

(26)

where $T_{\mathrm{m}}$ and $T_{\mathrm{gh}}$ are the matter and ghost stress tensors, $:\dots :$ denotes normal ordering, and $c\left(z\right)$ is the holomorphic reparametrization ghost field of the worldsheet $(b,c)$ Faddeev–Popov system, written as a function of the complex worldsheet coordinate z. A standard OPE computation shows that the potential anomaly in the BRST algebra is proportional to the total central charge:

$$Q_B^2\propto {\displaystyle \frac{c_{\mathrm{tot}}}{12}}\oint {\displaystyle \frac{dz}{2\pi i}}c\left(z\right){\partial }^{3}c\left(z\right).$$

(27)

Thus BRST nilpotency, hence gauge invariance and unitarity of the physical state space defined by BRST cohomology, this requires:

$$Q_B^2=0⟺c_{\mathrm{tot}}=0⟺D=26$$

(28)

for the bosonic string.

In light-cone quantization one can equivalently diagnose the same anomaly as a failure of the quantum Lorentz generators to close, in this the anomalous commutator takes the schematic form:

$$\begin{array}{c}\hfill \begin{array}{c}\hfill [M^{i-},M^{j-}]={\displaystyle \frac{2}{{\left(p^{+}\right)}^2}}\sum _{n>0}\left[\left({\displaystyle \frac{D-2}{24}}-1\right)n+{\displaystyle \frac{1}{n}}\left(a-{\displaystyle \frac{D-2}{24}}\right)\right]\left({\alpha }_{-n}^i{\alpha }_n^j-{\alpha }_{-n}^j{\alpha }_n^i\right)+(\alpha \leftrightarrow \tilde{\alpha }).\end{array}\end{array}$$

(29)

where $M^{i-}$ are the light–cone Lorentz generators mixing a transverse index $i=1,\dots ,D-2$ with the longitudinal − direction. D is the spacetime dimension. $p^{+}$ is the conserved light–cone momentum. The oscillators ${\alpha }_n^i$ are the transverse string modes satisfying $[{\alpha }_m^i,{\alpha }_n^j]=m{\delta }_{m+n,0}{\delta }^{ij}$. These ${\alpha }_n^i$ arise as the Fourier-mode coefficients in the mode expansion of the transverse string embedding $X^i(\sigma ,\tau )$ in radial/light-cone quantization. For $n>0$, ${\alpha }_{-n}^i$ and ${\alpha }_n^i$ play the role of creation and annihilation operators for the $n^{\mathrm{th}}$ vibrational mode, respectively. It is often convenient to introduce unit-normalized oscillators $a_n^i:={\alpha }_n^i/\sqrt{n}$ and $a_n^{i†}:={\alpha }_{-n}^i/\sqrt{n}$ ($n>0$), so that $[a_m^i,a_n^{j†}]={\delta }_{mn}{\delta }^{ij}$. The integer $n>0$ labels positive Fourier modes. The constant a is the normal–ordering constant. The term $(\alpha \leftrightarrow \tilde{\alpha })$ means add the same contribution with right–moving oscillators ${\tilde{\alpha }}_n^i$; for closed string only.

Intuitively, the ${\alpha }_n^i$ are nothing more than the quantized Fourier harmonics of a vibrating string, analogous to the normal modes of a guitar string. Each positive integer n labels a standing-wave pattern along the string, and quantization promotes the amplitude of each mode to a harmonic-oscillator degree of freedom. For $n>0$, ${\alpha }_{-n}^i$ creates one quantum of the nth vibrational mode, with transverse polarization i, while ${\alpha }_n^i$ annihilates it, the level operator N then counts the total vibrational excitation, weighted by mode number, so higher-n excitations cost more in $L_0$.

So closure enforces the same critical dimension [9]. In the present work, we use the covariant BRST-invariant SFT formulation, where Poincaré covariance is manifest at the level of the action, and the gauge symmetry is encoded by $Q_B$. We should note that for an open string, drop the right-moving piece $(\alpha \leftrightarrow \tilde{\alpha })$, and for for a closed string, keep it. One may also chose to rewrite the coefficients to make the $D=26$ structure manifest using ${\displaystyle \frac{D-2}{24}}-1={\displaystyle \frac{D-26}{24}}$:

$$\begin{array}{c}\hfill \begin{array}{c}\hfill [M^{i-},M^{j-}]={\displaystyle \frac{2}{{\left(p^{+}\right)}^2}}\sum _{n>0}\left[{\displaystyle \frac{D-26}{24}}n+{\displaystyle \frac{1}{n}}\left(a-{\displaystyle \frac{D-2}{24}}\right)\right]\left({\alpha }_{-n}^i{\alpha }_n^j-{\alpha }_{-n}^j{\alpha }_n^i\right)+(\alpha \leftrightarrow \tilde{\alpha }),\end{array}\end{array}$$

(30)

in this form the anomaly structure is explicit. The coefficient ${\displaystyle \frac{D-26}{24}}$ arises from the transverse zero–point energy and vanishes only in the critical dimension $D=26$. The second term involving $a-{\displaystyle \frac{D-2}{24}}$ reflects the normal–ordering shift in $L_0$. Lorentz algebra closure requires both coefficients to vanish, yielding $D=26$ and for the bosonic open string $a=1$ [9].

We work above with covariant open bosonic SFT only as a concrete and widely studied UV-complete framework in which the fundamental degrees of freedom are extended and the induced spacetime description is intrinsically non-local at the string scale. These are precisely the structural inputs needed for the operational locality analysis we present, such as sharp projectors localized to bounded spacetime regions are not operational observables and physical measurements are implemented by localized detector couplings to non-local operators, yielding completely positive instruments rather than an exact Lüders reduction.

We would also like to say that the perturbative vacuum of the open bosonic string contains a tachyonic mode with $m_T^2<0$, for the open string ground state one has $m_T^2=-1/{\alpha }^{\prime }$, this is signaling that the space-filling D-brane background around which one expands is an unstable stationary point, not a stable vacuum. This does not represent a violation of relativistic causality, but it is a statement about vacuum (in)stability. The physically relevant endpoint is the nonperturbative tachyon vacuum associated with D-brane annihilation, as formulated by Sen and quantitatively verified in open string field theory by level truncation and later established analytically by exact solutions [12,13,14,15]. If one wishes to avoid the tachyon already at the level of perturbative spectrum then we may equally formulate the discussion in a tachyon-free superstring field theory background but none of the operational-locality steps depend on the presence of a tachyonic scalar.

The use of covariant open bosonic string field theory above is for definiteness only as the operational-locality argument does not depend on the presence of a tachyonic scalar, and it extends to perturbatively tachyon-free superstring field theory backgrounds. Covariant open superstring field theories of Wess–Zumino–Witten type are known for the Neveu–Schwarz sector, and complete covariant actions including both the Neveu–Schwarz and Ramond sectors are also available [16,17,18].

Definition 2 

(Perturbatively tachyon-free string-field background). A covariant string-field-theory background will be called perturbatively tachyon-free if the physical BRST cohomology about that background contains no state with negative spacetime mass-squared. Equivalently, after gauge fixing and expansion about the chosen background, every physical component mode ${\varphi }_a$ in the perturbative spectrum satisfies

$$m_a^2\ge 0.$$

(31)

This condition concerns the stability of the perturbative vacuum but it is logically distinct from the question of locality or causality.

To see why the locality analysis is independent of the tachyon, it is useful to isolate the only structural ingredients used in the derivation. Let $L_0$ denote the worldsheet Virasoro zero mode, let a denote the intercept appropriate to the sector under consideration, and let $ϵ>0$ be the lower cutoff on the Schwinger or modulus parameter. Then the regulated propagator has the generic form:

$${\Delta }_{ϵ}={\int }_{ϵ}^{\infty }ds{e}^{-s(L_0-a)}={\displaystyle \frac{e^{-ϵ(L_0-a)}}{L_0-a}}.$$

(32)

After projection onto a fixed component level N, the matter contribution to $L_0$ takes the schematic form:

$$L_0={\alpha }^{\prime }{p}^2+N+\dots ,$$

(33)

where ${\alpha }^{\prime }$ is the inverse string tension, $p_{\mu }$ is the spacetime momentum, and the ellipsis denotes sector-dependent ghost and internal contributions. Accordingly, at fixed level one obtains a momentum-space kernel of the form:

$${\Delta }_{ϵ,N}\left(p\right)={\displaystyle \frac{e^{-ϵ({\alpha }^{\prime }{p}^2+N-a)}}{{\alpha }^{\prime }{p}^2+N-a}}.$$

(34)

The denominator determines the location of poles and therefore the perturbative mass spectrum. By contrast, the entire factor:

$$e^{-ϵ{\alpha }^{\prime }{p}^2}$$

(35)

is the ultraviolet ingredient responsible for non-locality. The key point is that the presence or absence of a tachyon changes the low-energy pole structure of (34), but it does not remove the entire-function factor (35) generated by the modulus cutoff.

Passing to Euclidean momentum $p_E$ makes the ultraviolet suppression manifest:

$${\Delta }_{ϵ,N}^E\left(p_E\right)={\displaystyle \frac{e^{-ϵ({\alpha }^{\prime }{p}_E^2+N-a)}}{{\alpha }^{\prime }{p}_E^2+N-a}},p_E^2\ge 0.$$

(36)

Because the exponential of a polynomial is entire, the regulated kernel admits analytic continuation back to Minkowski signature and induces a covariant entire functional calculus in spacetime. So exactly as in the bosonic discussion above, projection to component fields yields non-local operators of the form:

$$O_F\left(x\right)=F(\square /M_{*}^2)O\left(x\right),M_{*}^2\sim {\alpha }^{\prime -1},$$

(37)

with F entire and normalized by $F\left(0\right)=1$.

We can now state the precise relation between vacuum stability and operational locality, we consider a covariant SFT background satisfying the following conditions, (i) the theory admits a gauge-invariant perturbative expansion with a BRST-type kinetic operator and well-defined interaction vertices; (ii) a modulus cutoff $ϵ>0$ produces regulated propagators of the form (32); (iii) after projection to spacetime component fields, the induced observables are non-local and can be written as $O_F\left(x\right)=F(\square /M_{*}^2)O\left(x\right)$ with F entire; (iv) physical measurements are implemented by localized detector couplings to such non-local observables, so that outcome probabilities are governed by completely positive instruments rather than by an exact Lüders projection postulate.

Then the derivation of the operational-locality bound depends only on the entire-function smearing encoded in F and not on whether the perturbative spectrum contains a tachyon. Once the commutator bound:

$$\parallel [O_F\left(x\right),O_F\left(y\right)]\parallel \le CB\left({\displaystyle \frac{\sqrt{-{(x-y)}^2}}{{\ell }_{*}}}\right),$$

(38)

for spacelike ${(x-y)}^2<0$ has been established the corresponding detector-level signaling bound follows exactly as in the bosonic case, irrespective of the sign of the lowest physical mass-squared.

The proof is by inspection of the later argument, the detector-level probability bound is derived from Dyson expansion of localized interaction unitaries, followed by norm bounds on nested commutators of non-local observables. The only input from the SFT side is the existence of the entire-function smearing inherited from the regulated worldsheet propagator and vertex. The sign of the lowest mass-squared affects the stability of the background and the infrared pole structure of the propagator, but it does not alter the entire ultraviolet factor that produces the spacelike suppression of commutators. Therefore the operational-locality estimate is insensitive to the presence or absence of a tachyon.

A tachyon in the bosonic open-string vacuum is therefore a statement about perturbative vacuum instability not a structural ingredient of the locality mechanism discussed here. In a perturbatively tachyon-free superstring background, the same non-local entire-function smearing and the same detector-based replacement of ideal projectors remain available, so the operational resolution of Sorkin’s paradox goes through unchanged. For the purposes of the present paper, the bosonic theory should thus be regarded as a concrete and technically simple model in which the non-local mechanism can be displayed explicitly, rather than as an essential restriction of the argument.

Witten’s covariant open SFT has a cubic interaction and, when expressed in terms of component fields, induces an infinite-derivative non-polynomial effective action [8]. In particular, in truncated component descriptions the effective potential can appear unbounded along some directions but this is not an obstruction to the present analysis for two reasons, the first is that the superluminal signaling paradox concerns the measurement postulate applied to sharply localized projectors, not the global boundedness properties of an off-shell functional. The second reason is that the operational locality bound is derived from the algebra of non-local observables and detector dynamics, it requires controlled string-scale smearing and the resulting suppression of spacelike commutators that re independent of whether the underlying classical functional is globally bounded. When expanding about the tachyon vacuum, the nonperturbative minimum reproduces the expected vacuum energy shift, the cancellation of the D-brane tension, providing a consistent background for the effective low-energy description used in detector models.

The cutoff $ϵ$ on the worldsheet cylinder length in (19) can be rewritten in radial-quantized CFT language as a scale-smearing by the dilatation generator $L_0$ [7,8]. We introduce a radial modulus $z=e^{-s}$ so that $ds=-dz/z$. This radial-quantized interpretation of the cutoff in terms of the dilatation generator $L_0$ is standard in the worldsheet CFT description of strings [11]. Then the regulated evolution factor $e^{-s(L_0-a)}$, becomes $z^{L_0-a}$, and a natural scale-smeared operator is defined by the Mellin transform:

$${\mathrm{\Phi }}_{ϵ}:=\mathcal{M}\left\{\varphi \right\}(L_0-a)={\int }_{ϵ}^{\infty }dz{z}^{L_0-a-1}\varphi \left(z\right),$$

(39)

where $z>0$ is the radial scale, $L_0$ is the Virasoro zero-mode generating dilatations, a is the intercept, and $\varphi \left(z\right)$ is a world-sheet operator insertion at radius z (or equivalently a component operator appearing in the string-field expansion). The role of $ϵ$ is to exclude arbitrarily short scales, the UV moduli from the smearing.

In the open string, the matter contribution to $L_0$ takes the form:

$$L_0={\alpha }^{\prime }{p}^2+N,$$

(40)

where ${\alpha }^{\prime }$ is the inverse string tension, $p^{\mu }$ is the spacetime momentum, $p^2={\eta }_{\mu \nu }{p}^{\mu }{p}^{\nu }$, and $N\in {\mathbb{Z}}_{\ge 0}$ is the oscillator level. So imposing a minimal modulus $ϵ$ produces exponential damping factors such as $e^{-ϵ(L_0-a)}$, which in a fixed-level component truncation reduce to entire functions of □ at the scale $M_{*}^2\sim 1/{\alpha }^{\prime }$. This is the worldsheet/CFT origin of the non-local functional calculus used below.

To make explicit how the entire-function smearing used in (64) arises from the covariant open-string field theory propagator and interaction vertex after projection to spacetime component fields the string field $\mathrm{\Psi }$ is a ghost-number-one state in the first-quantized open-string Hilbert space. In Siegel gauge $b_0\mathrm{\Psi }=0$, we expand it in a momentum basis as:

$$\begin{array}{c}\hfill \begin{array}{c}\hfill |\mathrm{\Psi }\rangle =\int {\displaystyle \frac{d^{d}p}{{\left(2\pi \right)}^d}}\left(T\left(p\right)c_1|0;p\rangle +A_{\mu }\left(p\right){\alpha }_{-1}^{\mu }{c}_1|0;p\rangle +B\left(p\right)b_{-1}{c}_0{c}_1|0;p\rangle +\dots \right),\end{array}\end{array}$$

(41)

where $|0;p\rangle$ is the SL$(2,\mathbb{R})$-invariant vacuum of momentum $p_{\mu }$, $c_1$ is the usual ghost oscillator, ${\alpha }_{-1}^{\mu }$ is the first matter oscillator, and the ellipsis denotes higher oscillator levels. The coefficient $T\left(p\right)$ is the level-$N=0$ scalar mode (the open bosonic tachyon), while $A_{\mu }\left(p\right)$ is the level-$N=1$ vector mode. More generally, for a component field ${\varphi }_N$ at fixed oscillator level N, the matter part of the Virasoro zero mode contributes:

$$L_0={\alpha }^{\prime }{p}^2+N,p^2:={\eta }_{\mu \nu }{p}^{\mu }{p}^{\nu },$$

(42)

so that the corresponding mass is:

$$M_N^2:={\displaystyle \frac{N-1}{{\alpha }^{\prime }}}.$$

(43)

Here ${\alpha }^{\prime }$ is the inverse string tension and $d=26$ for the bosonic string. In Siegel gauge the free propagator is generated by the Schwinger parameter representation:

$${\displaystyle \frac{1}{L_0-1}}={\int }_0^{\infty }ds{e}^{-s(L_0-1)}.$$

(44)

Imposing a minimal worldsheet modulus $ϵ>0$ gives the regulated propagator:

$${\Delta }_{ϵ}={\int }_{ϵ}^{\infty }ds{e}^{-s(L_0-1)}={\displaystyle \frac{e^{-ϵ(L_0-1)}}{L_0-1}}.$$

(45)

Projecting (45) onto a fixed oscillator level N and using (42) yields:

$${\Delta }_{ϵ,N}\left(p\right)={\displaystyle \frac{e^{-ϵ({\alpha }^{\prime }{p}^2+N-1)}}{{\alpha }^{\prime }(p^2+M_N^2)}}.$$

(46)

Since N is fixed in a level truncation, the factor $e^{-ϵ(N-1)}$ is an irrelevant overall normalization. The momentum dependence responsible for non-locality is therefore the entire factor:

$$e^{-ϵ{\alpha }^{\prime }{p}^2}.$$

(47)

To interpret (47) as ultraviolet damping in a strictly unambiguous way, it is convenient first to Wick rotate to Euclidean momentum $p_E$, for which $p_E^2\ge 0$. Then (46) becomes:

$${\Delta }_{ϵ,N}^E\left(p_E\right)={\displaystyle \frac{e^{-ϵ({\alpha }^{\prime }{p}_E^2+N-1)}}{{\alpha }^{\prime }(p_E^2+M_N^2)}},$$

(48)

which manifestly suppresses large Euclidean momenta. Because the exponential of a polynomial is an entire function, (48) has a unique analytic continuation back to Minkowski signature. This is the precise sense in which the regulator is an entire function of the d’Alembertian.

At the level of the quadratic action, the entire factor in (46) may be moved from the propagator into the definition of the spacetime component field. We define the regulated fixed-level field in momentum space by:

$${\tilde{\varphi }}_{N,ϵ}\left(p_E\right):=e^{-{\displaystyle \frac{ϵ{\alpha }^{\prime }}{2}}{p}_E^2}{\tilde{\varphi }}_N\left(p_E\right),$$

(49)

where the level-dependent constant $e^{-ϵ(N-1)/2}$ has been absorbed into the normalization of ${\varphi }_N$. In coordinate space this is:

$${\varphi }_{N,ϵ}\left(x_E\right)=e^{{\displaystyle \frac{ϵ{\alpha }^{\prime }}{2}}{\Delta }_E}{\varphi }_N\left(x_E\right),$$

(50)

with ${\Delta }_E$ the Euclidean Laplacian. After analytic continuation back to Minkowski signature, (50) is written covariantly as:

$${\varphi }_{N,ϵ}\left(x\right)=F_{ϵ}(\square /M_{*}^2){\varphi }_N\left(x\right),F_{ϵ}\left(z\right)=e^{-{\beta }_{ϵ}z},$$

(51)

where:

$$M_{*}^2\sim {\displaystyle \frac{1}{{\alpha }^{\prime }}},{\beta }_{ϵ}=O\left(ϵ\right).$$

(52)

More precisely, if we choose:

$$M_{*}^2:={\displaystyle \frac{1}{{\alpha }^{\prime }}},$$

(53)

then $F_{ϵ}(-p^2/M_{*}^2)$ reproduces the same entire momentum kernel as (47) up to the dimensionless constant $ϵ$. Thus the functional calculus in (64) is not an ad hoc ansatz, it is the spacetime image of the regulated worldsheet propagator after projection to component fields.

The same conclusion follows from the interaction term. The cubic action is:

$$S_3\left[\mathrm{\Psi }\right]=-{\displaystyle \frac{1}{3g_o^2}}\langle \mathrm{\Psi },\mathrm{\Psi }*\mathrm{\Psi }\rangle .$$

(54)

In oscillator language, the three-string vertex has the standard form:

$$\begin{array}{c}\hfill \begin{array}{c}\hfill \langle V_3|={\mathcal{N}}_3{\left(2\pi \right)}^d{\delta }^{\left(d\right)}\left(\sum _{r=1}^3{p}_r\right)exp\left[-{\displaystyle \frac{1}{2}}\sum _{r,s=1}^3\sum _{m,n\ge 0}{\alpha }_m^{\left(r\right)}·V_{mn}^{rs}{\alpha }_n^{\left(s\right)}+\dots \right],\end{array}\end{array}$$

(55)

where ${\mathcal{N}}_3$ is a normalization constant, $V_{mn}^{rs}$ are the Neumann coefficients, $r,s\in \{1,2,3\}$ label the three external strings, and the ellipsis denotes the ghost sector. The zero modes satisfy:

$${\alpha }_0^{\left(r\right)\mu }=\sqrt{2{\alpha }^{\prime }}{p}_r^{\mu },$$

(56)

so the zero-mode part of the vertex contributes a Gaussian factor:

$$exp\left[-{\displaystyle \frac{{\alpha }^{\prime }}{2}}\sum _{r,s=1}^3{V}_{00}^{rs}{p}_r·p_s\right].$$

(57)

By momentum conservation ${\sum }_r{p}_r=0$ and cyclic symmetry of the Witten vertex, the quadratic form in (57) reduces to:

$$exp\left[-\gamma {\alpha }^{\prime }\sum _{r=1}^3{p}_r^2\right]$$

(58)

for some positive dimensionless constant $\gamma$ determined by the Neumann coefficients of the three-string overlap. The exact numerical value of $\gamma$ depends on conventions but what matters structurally is that (58) is entire in each external momentum.

Projecting onto the lowest scalar mode $T\left(p\right)$ gives a cubic term of the form:

$$\begin{array}{c}\hfill \begin{array}{c}\hfill S_{3,T}=-{\displaystyle \frac{{\kappa }_T}{3!}}\int \prod _{r=1}^3{\displaystyle \frac{d^d{p}_r}{{\left(2\pi \right)}^d}}{\left(2\pi \right)}^d{\delta }^{\left(d\right)}\left(\sum _{r=1}^3{p}_r\right)e^{-\gamma {\alpha }^{\prime }{\sum }_{r=1}^3{p}_r^2}T\left(p_1\right)T\left(p_2\right)T\left(p_3\right),\end{array}\end{array}$$

(59)

where ${\kappa }_T$ is an effective cubic coupling obtained after contracting the non-zero-mode oscillators and ghosts. Equation (59) is the desired component-level statement, that the cubic kernel itself carries an entire dependence on the spacetime momenta. Defining the smeared tachyon field by:

$${\tilde{T}}_F\left(p\right):=e^{-\gamma {\alpha }^{\prime }{p}^2}\tilde{T}\left(p\right),$$

(60)

the vertex (59) becomes local in the smeared field:

$$S_{3,T}=-{\displaystyle \frac{{\kappa }_T}{3!}}\int d^{d}x{T}_F{\left(x\right)}^3.$$

(61)

If we choose to work with the unsmeared field $T\left(x\right)$, then the interaction is non-local and contains infinitely many derivatives. These two descriptions are related by an entire field redefinition and are physically equivalent at the level relevant here.

The free propagator (46) and the cubic vertex (59) both show that once a minimal worldsheet modulus is imposed, projection to spacetime component fields produces kernels that are entire functions of momentum. After Fourier transform, this is precisely the statement that the induced spacetime description is governed by entire functions of the d’Alembertian at the scale $M_{*}^2\sim {\alpha }^{\prime -1}$. Therefore the regulated non-local observable:

$$O_F\left(x\right)=F(\square /M_{*}^2)O\left(x\right)$$

(62)

used below is not merely motivated by analogy but it is the natural covariant spacetime representation of the component-level kernels inherited from string field theory.

Finally, although we used the lowest scalar mode for notational simplicity the same reasoning applies to any fixed oscillator level. In particular, in a tachyon-free superstring background one may repeat the argument with the first physical level, so the emergence of entire-function non-locality is not tied to the presence of the bosonic tachyon.

This component-level derivation motivates the following operational definition, that the non-local regulator is a Lorentz scalar functional calculus of $\square ={\eta }^{\mu \nu }{\partial }_{\mu }{\partial }_{\nu }$, so the operational locality bounds we derive preserve Lorentz symmetry while softening strict microcausality at separations of order the string length.

In this paper, microcausality always means target-space microcausality, that for local spacetime observables $O\left(x\right)$ and $O^{\prime }\left(y\right)$, we require:

$$[O\left(x\right),O^{\prime }\left(y\right)]=0\forall {(x-y)}^2<0,$$

(63)

with $x,y\in {\mathbb{R}}^{1,3}$ spacelike separated. We use this notion, rather than a notion of worldsheet locality because the issue under study is Sorkin’s signaling paradox, which is formulated in terms of operations performed in separated spacetime regions $O_1,O_2,O_3$. By contrast, worldsheet locality refers to locality in the underlying two-dimensional worldsheet field theory, local operator insertions on the worldsheet have the usual short-distance OPE structure, and the worldsheet dynamics is organized by conformal symmetry and BRST consistency of the underlying two-dimensional worldsheet theory. This is conceptually distinct from target-space microcausality, which is the vanishing of commutators of spacetime observables at spacelike separation. So it is not the causal condition whose possible failure would correspond to superluminal signaling between spacetime observers. For us the worldsheet is the origin of the string-induced smearing that after projection to spacetime component fields produces non-local operators of the form $F(\square /M_{*}^2)O\left(x\right)$. So the relevant statement is that exact target-space microcausality is recovered in the local limit ${\ell }_s\to 0$, while at finite string length it is replaced by operational locality where spacelike commutators and detector-level influences are not identically zero, but are suppressed by a rapidly decaying bound at separations large compared with ${\ell }_s$. This asserts that for local observables $O\left(x\right)$ and $O^{\prime }\left(y\right)$ we have $[O\left(x\right),O^{\prime }\left(y\right)]=0$ whenever ${(x-y)}^2<0$, where $[X,Y]=XY-YX$ is the commutator.

Two features of (18) matter for measurement theory, the first is that $\mathrm{\Psi }$ describes extended configurations and cannot be localized to a spacetime point in the same way as a point-particle field. More generally this suggests that relativistic localization should be understood in an asymptotic or non-local sense rather than in terms of strictly sharp position operators, a point that has been developed further in the context of Newton–Wigner and Foldy–Wouthuysen localization in Ref. [19]. The second is that when one expands $\mathrm{\Psi }$ in a basis of string oscillators and projects onto spacetime component fields the interaction term generates infinitely many derivatives. In effective descriptions, these derivatives resum into entire functions of □ at a scale set by the string length ${\ell }_s$. This motivates the following sharpened operational definition, which separates the construction of non-local observables from the locality criterion they satisfy.

Definition 3 

(Regulated non-local observables and operational locality). Let $O\left(x\right)$ be a local composite operator in a low-energy field-theory truncation of string field theory, with $x\in {\mathbb{R}}^{1,3}$. Let

$$O_F\left(x\right)=F(\square /M_{*}^2)O\left(x\right),$$

(64)

where $\square ={\eta }^{\mu \nu }{\partial }_{\mu }{\partial }_{\nu }$ is the d’Alembertian, $M_{*}$ is a ultraviolet mass scale, ${\ell }_{*}:=M_{*}^{-1}$ is the corresponding length scale, and F is an entire function normalized by $F\left(0\right)=1$.

The operator $O_F\left(x\right)$ is called a regulated non-local observable if F is chosen so that $O_F\left(x\right)$ agrees with $O\left(x\right)$ at long wavelengths, while suppressing ultraviolet support at scales of order ${\ell }_{*}$. In momentum space, if $\tilde{O}\left(k\right)$ denotes the Fourier transform of $O\left(x\right)$, then

$$\widetilde{O_F}\left(k\right)=F(-k^2/M_{*}^2)\tilde{O}\left(k\right),$$

(65)

with $k^2={\eta }_{\mu \nu }{k}^{\mu }{k}^{\nu }$. In the string-field-theory setting one may take $M_{*}^2\sim {\alpha }^{\prime -1}$, so that ${\ell }_{*}\sim {\ell }_s=\sqrt{{\alpha }^{\prime }}$, and a canonical model is

$$F\left(z\right)=e^{-z/2}.$$

(66)

A relativistic quantum theory is said to satisfy operational locality at scale ${\ell }_{*}$ if for every pair of detector instruments localized in spacetime regions ${\mathcal{O}}_A,{\mathcal{O}}_C\subset {\mathbb{R}}^{1,3}$ with spacelike separation

$$r_{AC}:=\underset{x\in {\mathcal{O}}_A,y\in {\mathcal{O}}_C}{inf}\sqrt{-{(x-y)}^2}>0,$$

(67)

and for every detector outcome m at the C-side, the difference in outcome probabilities induced by toggling the A-side interaction obeys a bound

$$|p_m^{\left(A\mathrm{on}\right)}-p_m^{\left(A\mathrm{off}\right)}|\le {\mathcal{C}}_{A,C,m}B(r_{AC}/{\ell }_{*}),$$

(68)

where ${\mathcal{C}}_{A,C,m}$ is a finite constant depending on the detector couplings, switching functions, and initial state, and where $B:[0,\infty )\to [0,\infty )$ is a monotonically decreasing function with

$$B\left(\rho \right)\to 0(\rho \to \infty ).$$

(69)

If $B\left(\rho \right)\equiv 0$ for all $\rho >0$, then operational locality reduces to exact microcausality in the idealized sharp-localization limit.

Equation (64) should be read as a Lorentz-invariant functional calculus of the wave operator, the point is that $O_F\left(x\right)$ is not strictly point-supported but it is localized only up to tails controlled by the scale ${\ell }_{*}$. So the relevant physical question is no longer whether the commutator vanishes exactly outside the lightcone but whether any spacelike influence on realizable detector statistics is uniformly suppressed by a rapidly decaying bound. This is the sense in which exact microcausality is replaced by an operational sense of locality.

For the stringy entire regulator (66), the smeared observable is local on macroscopic scales but acquires controlled string-scale tails. Accordingly we expect B to decay rapidly once $r_{AC}\gg {\ell }_s$. The following theorem gives the corresponding quantitative commutator bound in the free-field model that captures this non-local behavior.

Theorem 1 

(Exponential suppression of spacelike commutators). Let $\varphi \left(x\right)$ be a free real scalar field of mass $m\ge 0$ on ${\mathbb{R}}^{1,3}$, and define the regulated field

$${\varphi }_F\left(x\right)=F(\square /M_{*}^2)\varphi \left(x\right)$$

(70)

with $F\left(z\right)=e^{-z/2}$ and $M_{*}={\ell }_{*}^{-1}$. For spacelike separated points $x,y$ with ${(x-y)}^2<0$, define the invariant distance

$$r:=\sqrt{-{(x-y)}^2}.$$

(71)

Then the commutator is not exactly zero, but satisfies the bound

$$\parallel [{\varphi }_F\left(x\right),{\varphi }_F\left(y\right)]\parallel \le Cexp\left(-{\displaystyle \frac{r^2}{4{\ell }_{*}^2}}\right),$$

(72)

where $\parallel ·\parallel$ denotes the operator norm on the regulated Hilbert-space domain and C is a constant depending on the field normalization and ultraviolet regularization.

Equation (72) identifies the decay function in (68) as:

$$B\left(\rho \right)=e^{-{\rho }^2/4}.$$

(73)

Thus the lightcone support of the local Pauli–Jordan commutator is replaced by a Gaussian tail of width ${\ell }_{*}$, in the string-field-theory interpretation we set ${\ell }_{*}={\ell }_s$, so spacelike noncommutativity is not absent but is uniformly and exponentially suppressed for separations $r\gg {\ell }_s$. This is the quantitative content of operational locality in the present framework.

This expresses that the lightcone support of the Pauli–Jordan commutator for $\varphi$ is smeared by the entire operator $F(\square /M_{*}^2)$ into a Gaussian tail whose width is ${\ell }_{*}$. The inequality is the quantitative form of operational locality that outside the lightcone the commutator is exponentially suppressed at separations $r\gg {\ell }_{*}$. In the SFT interpretation we set ${\ell }_{*}={\ell }_s$, this is to be read distributionally or after smearing with detector profiles/test functions as pointlike field insertions are shorthand.

In covariant open SFT, gauge fixing relates the BRST kinetic operator to the Virasoro zero mode via $Q_B\sim c_0(L_0-1)+\dots$ in Siegel gauge $b_0\mathrm{\Psi }=0$, so the free propagator is ${(L_0-1)}^{-1}$ and admits the Schwinger parameter representation ((19)) [7,8]. Excluding worldsheet cylinders shorter than a minimal length $ϵ>0$ yields the regulated propagator:

$${\Delta }_{ϵ}={\int }_{ϵ}^{\infty }ds{e}^{-s(L_0-1)}={\displaystyle \frac{e^{-ϵ(L_0-1)}}{L_0-1}}.$$

(74)

This is the precise sense in which a worldsheet cutoff induces non-locality in spacetime as the factor $e^{-ϵ(L_0-1)}$ is an entire function of $L_0$ that suppresses high worldsheet eigenvalues, the short-distance or UV excitations. The same mechanism can be phrased in radial-quantized CFT language through the scale-smeared operator:

$${\mathrm{\Phi }}_{ϵ}:={\int }_{ϵ}^{\infty }{\displaystyle \frac{dz}{z}}{z}^{L_0-a}\varphi \left(z\right)$$

(75)

where $z>0$ is the radial coordinate or the worldsheet modulus playing the role of a scale parameter, $L_0$ is the holomorphic Virasoro zero-mode generating dilatations in radial quantization, a is the intercept, and $\varphi \left(z\right)$ is a worldsheet operator or, in the gauge-fixed string field expansion, a component operator inserted at radius z.

To exhibit the entire-function structure we change variables $z=e^{-s}$ so that $dz/z=-ds$, then (75) becomes:

$${\mathrm{\Phi }}_{ϵ}={\int }_0^{ln(1/ϵ)}ds{e}^{-s(L_0-a)}\varphi \left(e^{-s}\right),$$

(76)

and the same exponential $e^{-s(L_0-a)}$ that appears in the Schwinger form (19) controls the UV. If one focuses on the induced spacetime description after expanding the string field in oscillator eigenstates of $L_0$, then for the open string:

$$L_0={\alpha }^{\prime }{p}^2+N,p^2:={\eta }_{\mu \nu }{p}^{\mu }{p}^{\nu },N\in {\mathbb{Z}}_{\ge 0},$$

(77)

where ${\alpha }^{\prime }$ is the inverse string tension, $p^{\mu }$ is the spacetime momentum, ${\eta }_{\mu \nu }$ is the Minkowski metric, and N is the oscillator level. With $a=1$ as the open bosonic intercept, the regulated propagator (74) in momentum space becomes:

$${\Delta }_{ϵ}\left(p\right)={\displaystyle \frac{e^{-ϵ({\alpha }^{\prime }{p}^2+N-1)}}{{\alpha }^{\prime }(p^2+M^2)}},M^2:={\displaystyle \frac{N-1}{{\alpha }^{\prime }}}.$$

(78)

In a low-energy truncation to a fixed level N and then fixed $M^2$, the only nonlocality is the entire factor $e^{-ϵ{\alpha }^{\prime }{p}^2}$. Identifying the UV scale by:

$$M_{*}^2\sim {\displaystyle \frac{1}{ϵ{\alpha }^{\prime }}}\sim {\displaystyle \frac{1}{{\alpha }^{\prime }}}={\ell }_s^{-2},{\ell }_{*}:=M_{*}^{-1},$$

(79)

the induced spacetime regulator acting on a momentum-space mode $\tilde{O}\left(k\right)$ takes the form:

$${\tilde{O}}_F\left(k\right)=F(-k^2/M_{*}^2)\tilde{O}\left(k\right),F\left(z\right)=e^{-z/2},$$

(80)

which is exactly the entire functional calculus used in Definition .2.

The proof of Theorem .3 is trivial when we let $\phi \left(x\right)$ be a free real scalar field of mass $m\ge 0$ on ${\mathbb{R}}^{1,3}$, acting on a Hilbert space $\mathcal{H}$, and define its regulated version by:

$$\begin{array}{cc}\hfill {\phi }_F\left(x\right)& :=F(\square /M_{*}^2)\phi \left(x\right),\hfill \end{array}$$

(81)

$$\begin{array}{cc}\hfill F\left(z\right)& =e^{-z/2},\hfill \end{array}$$

(82)

$$\begin{array}{cc}\hfill \square & :={\eta }^{\mu \nu }{\partial }_{\mu }{\partial }_{\nu }.\hfill \end{array}$$

(83)

For two points $x,y\in {\mathbb{R}}^{1,3}$ with spacelike separation ${(x-y)}^2<0$, define the invariant spacelike distance:

$$r:=\sqrt{-{(x-y)}^2}.$$

(84)

We prove that there exists a constant $C>0$, depending on m and on the chosen ultraviolet realization of $\phi$ as an operator-valued distribution such that:

$$\parallel [{\phi }_F\left(x\right),{\phi }_F\left(y\right)]\parallel \le Cexp\left(-{\displaystyle \frac{r^2}{4{\ell }_{*}^2}}\right),{\ell }_{*}:=M_{*}^{-1}.$$

(85)

For the free scalar field, the commutator is a c-number distribution:

$$\left[\phi \right(x),\phi (y\left)\right]=i\Delta (x-y)\mathbf{1},$$

(86)

where $\Delta$ is the Pauli–Jordan causal commutator function. Since $F(\square /M_{*}^2)$ is a Lorentz-invariant differential functional calculus commuting with translations, we have:

$$\begin{array}{cc}\hfill [{\phi }_F\left(x\right),{\phi }_F\left(y\right)]& =F({\square }_x/M_{*}^2)F({\square }_y/M_{*}^2)[\phi \left(x\right),\phi \left(y\right)]\hfill \end{array}$$

$$\begin{array}{c}\hfill =i{\Delta }_F(x-y)\mathbf{1}.\end{array}$$

(87)

With $F\left(z\right)=e^{-z/2}$ we can write:

$${\Delta }_F=e^{-\square /\left(2M_{*}^2\right)}\Delta .$$

(88)

For spacelike separation we can go to the Lorentz frame in which $x-y=(0,\mathbf{r})$ with $\left|\mathbf{r}\right|=r$. In this frame □ acting on functions of $\mathbf{r}$ reduces to $\square =-{\Delta }_{\mathbf{r}}$ (minus the Euclidean Laplacian in ${\mathbb{R}}^3$) when acting on time-independent kernels. On the spacelike slice relevant for bounding the commutator, (88) is the standard heat semigroup:

$$e^{-\square /\left(2M_{*}^2\right)}⇝e^{{\Delta }_{\mathbf{r}}/\left(2M_{*}^2\right)}.$$

(89)

Consequently ${\Delta }_F$ is the convolution of $\Delta$ with a Gaussian kernel of width ${\ell }_{*}$ in the spatial variable:

$$\begin{array}{cc}\hfill {\Delta }_F(0,\mathbf{r})& ={\int }_{{\mathbb{R}}^3}{d}^3\mathbf{u}{K}_{{\ell }_{*}}(\mathbf{r}-\mathbf{u})\Delta (0,\mathbf{u}),\hfill \end{array}$$

(90)

$$\begin{array}{cc}\hfill K_{{\ell }_{*}}\left(\mathbf{v}\right):& ={\displaystyle \frac{1}{{\left(4\pi {\ell }_{*}^2\right)}^{3/2}}}exp\left(-{\displaystyle \frac{{\left|\mathbf{v}\right|}^2}{4{\ell }_{*}^2}}\right).\hfill \end{array}$$

(91)

For the free scalar field, the Pauli–Jordan distribution $\Delta \left(\xi \right)$ is supported on and inside the lightcone, so in particular $\Delta (0,\mathbf{u})=0$ for $\mathbf{u}\ne 0$, equal-time commutator vanishes. More generally, for spacelike separated $(0,\mathbf{r})$, the only contribution to the smeared kernel (90) comes from values of $\Delta$ near the lightcone after smearing as the Gaussian factor dominates the tail. Using that $\Delta$ is a tempered distribution and the Gaussian is a Schwartz function we then obtain the pointwise estimate:

$$|{\Delta }_F(0,\mathbf{r})|\le C_{0}exp\left(-{\displaystyle \frac{r^2}{4{\ell }_{*}^2}}\right),$$

(92)

for some constant $C_0>0$ depending on m and on the chosen operator realization/ultraviolet regularization of $\phi$ (this is the only place where the distributional nature of $\phi$ enters).

Since (87) is proportional to the identity, its operator norm equals the absolute value of the scalar coefficient:

$$\parallel [{\phi }_F\left(x\right),{\phi }_F\left(y\right)]\parallel =\left|{\Delta }_F(x-y)\right|.$$

(93)

Combining (92) with (93) yields (85) with $C=C_0$. The entire functional calculus $F(\square /M_{*}^2)$ replaces exact lightcone support of $\Delta$ by a Gaussian tail of width ${\ell }_{*}=M_{*}^{-1}$. In the SFT interpretation one sets ${\ell }_{*}\sim {\ell }_s$, so the commutator is exponentially suppressed for $r\gg {\ell }_s$, which is the quantitative form of operational locality used in the measurement analysis.

The proof of Theorem .3 is explicit when we let $\phi \left(x\right)$ be a free real scalar field of mass $m\ge 0$ on ${\mathbb{R}}^{1,3}$, acting on a Hilbert space $\mathcal{H}$, and define its regulated version by (81). For two points $x,y\in {\mathbb{R}}^{1,3}$ with spacelike separation ${(x-y)}^2<0$, define the invariant spacelike distance of (84) we prove that there exists a constant $C>0$, depending on m and on the chosen ultraviolet realization of $\phi$ as an operator-valued distribution of (85).

For the free scalar field, the commutator is a c-number distribution as seen in (86) where $\Delta$ is the Pauli–Jordan causal commutator function. Since $F(\square /M_{*}^2)$ is a Lorentz-invariant differential functional calculus commuting with translations, we have (87) with:

$${\Delta }_F:=F(\square /M_{*}^2)\Delta ,$$

(94)

where □ acts on the difference variable. With $F\left(z\right)=e^{-z/2}$ we can write (88). For spacelike separation one may go to the Lorentz frame in which $x-y=(0,\mathbf{r})$ with $\left|\mathbf{r}\right|=r$, in this frame, □ acting on functions of $\mathbf{r}$ reduces to $\square =-{\Delta }_{\mathbf{r}}$ (minus the Euclidean Laplacian in ${\mathbb{R}}^3$) when acting on time-independent kernels. On the spacelike slice relevant for bounding the commutator, (88) is the standard heat semigroup of (89) Consequently, ${\Delta }_F$ is the convolution of $\Delta$ with a Gaussian kernel of width ${\ell }_{*}$ in the spatial variable we have (90).

For the free scalar field, the Pauli–Jordan distribution $\Delta \left(\xi \right)$ is supported on and inside the lightcone, so in particular $\Delta (0,\mathbf{u})=0$ for $\mathbf{u}\ne 0$, equal-time commutator vanishes. More generally for spacelike separated $(0,\mathbf{r})$, the only contribution to the smeared kernel (90) comes from values of $\Delta$ near the lightcone after smearing as the Gaussian factor dominates the tail. Using that $\Delta$ is a tempered distribution and the Gaussian is a Schwartz function, one obtains the pointwise estimate or (92) for some constant $C_0>0$ depending on m and on the chosen operator realization/ultraviolet regularization of $\phi$, this is the only place where the distributional nature of $\phi$ enters.

Since (87) is proportional to the identity, its operator norm equals the absolute value of the scalar coefficient as seen in (93). Combining (92) with (93) yields (85) with $C=C_0$ completing the proof.

A convenient way to estimate the spacelike commutator is to interpret $\phi$ and hence ${\phi }_F$ as an operator-valued distribution acting on test functions and to make explicit the UV regularization implicit in the constant C. For any UV-regularized realization of $\phi$ the commutator can be written as:

$$[{\phi }_F\left(x\right),{\phi }_F\left(y\right)]=i{\Delta }_F(x-y)\mathbf{1},$$

(95)

where ${\Delta }_F$ is the regulated Pauli–Jordan commutator kernel, a tempered distribution and $\mathbf{1}$ is the identity.

The entire operator $F(\square /M_{*}^2)=e^{-\square /\left(2M_{*}^2\right)}$ admits a kernel representation on tempered distributions as there exists a regularization-dependent Schwartz kernel $K_{{\ell }_{*}}\left(\xi \right)$ of width ${\ell }_{*}$ such that:

$${\Delta }_F\left(\xi \right)=(K_{{\ell }_{*}}*\Delta )\left(\xi \right)={\int }_{{\mathbb{R}}^{1,3}}{d}^{4}u{K}_{{\ell }_{*}}(\xi -u)\Delta \left(u\right),$$

(96)

where $\Delta$ is the usual Pauli–Jordan commutator distribution for the free scalar, * denotes convolution, and $\xi :=x-y$. Intuitively the entire functional calculus replaces sharp lightcone support by a controlled ${\ell }_{*}$-scale smearing; the precise $K_{{\ell }_{*}}$ depends on the chosen ultraviolet realization, which is why C is not universal.

Since $\Delta$ is a tempered distribution and $K_{{\ell }_{*}}$ is Schwartz, one has a general estimate of the form:

$$|{\Delta }_F\left(\xi \right)|\le C_0\underset{u\in \mathrm{supp}\left(\Delta \right)}{sup}\left|K_{{\ell }_{*}}(\xi -u)\right|,$$

(97)

for some constant $C_0>0$ depending on m and on the chosen UV regularization equivalently, on the choice of norms used to bound the action of $\Delta$ on Schwartz functions.

Now use that $\Delta \left(u\right)$ has lightcone support in the sense that it vanishes for spacelike u. For spacelike $\xi$ with invariant distance $r=\sqrt{-{\xi }^2}$, the closest points u on the lightcone to $\xi$ are at Minkowski distance of order r from $\xi$. Because $K_{{\ell }_{*}}$ has width ${\ell }_{*}$, its tail is dominated by a Gaussian at spacelike separation:

$$|K_{{\ell }_{*}}(\xi -u)|\le {\displaystyle \frac{C_1}{{\ell }_{*}^4}}exp\left(-{\displaystyle \frac{r^2}{4{\ell }_{*}^2}}\right),$$

(98)

uniformly for u on the lightcone with some constant $C_1>0$ depending only on the chosen UV kernel normalization. Combining (97) and (98) yields:

$$|{\Delta }_F\left(\xi \right)|\le Cexp\left(-{\displaystyle \frac{r^2}{4{\ell }_{*}^2}}\right),$$

(99)

where $C:=C_0{C}_1/{\ell }_{*}^4$, absorbing fixed factors into C as in the statement of the theorem.

Finally since (95) is proportional to the identity, its operator norm equals the absolute value of the coefficient:

$$\parallel [{\phi }_F\left(x\right),{\phi }_F\left(y\right)]\parallel =\left|{\Delta }_F(x-y)\right|\le Cexp\left(-{\displaystyle \frac{r^2}{4{\ell }_{*}^2}}\right).$$

(100)

The entire functional calculus $F(\square /M_{*}^2)$ replaces exact lightcone support of $\Delta$ by a Gaussian tail of width ${\ell }_{*}=M_{*}^{-1}$. In the SFT interpretation one sets ${\ell }_{*}\sim {\ell }_s$, so the commutator is exponentially suppressed for $r\gg {\ell }_s$, which is the quantitative form of operational locality used in the measurement analysis.

The next statement explains why the commutator estimate above is the correct notion of locality for physically realizable measurements, we let ${\mathcal{H}}_S$ be the system Hilbert space and let ${\mathcal{H}}_{D_A}$, ${\mathcal{H}}_{D_C}$ be detector Hilbert spaces for two localized detectors A and C. Suppose detector $X\in \{A,C\}$ couples to the system through an interaction Hamiltonian of the form:

$$H_{\mathrm{int}}^{\left(X\right)}\left(t\right)={\lambda }_X{\chi }_X\left(t\right)M_X\otimes O_{F,X}\left(x_X\left(t\right)\right),$$

(101)

where ${\lambda }_X\in \mathbb{R}$ is a coupling constant, ${\chi }_X$ is a switching function of compact support, $M_X$ is a bounded self-adjoint detector operator, and $O_{F,X}$ is a regulated non-local observable. Let $U_A$ and $U_C$ be the corresponding interaction unitaries, and let ${\Pi }_m^{\left(C\right)}$ be a projector associated with outcome m of the C-detector readout.

We assume that for all $x\in {\mathcal{O}}_A$ and $y\in {\mathcal{O}}_C$ one has the commutator bound:

$$\parallel [O_{F,A}\left(x\right),O_{F,C}\left(y\right)]\parallel \le K_{AC}B(r_{AC}/{\ell }_{*}),$$

(102)

where $r_{AC}$ is the minimal invariant spacelike separation defined in (67). Then, to leading nontrivial order in the couplings, the difference between the C-side outcome probabilities with A switched on and with A switched off obeys:

$$|p_m^{\left(A\mathrm{on}\right)}-p_m^{\left(A\mathrm{off}\right)}|\le K_m^{\prime }\left|{\lambda }_A{\lambda }_C\right|B(r_{AC}/{\ell }_{*})+O\left({\lambda }^3\right),$$

(103)

where $K_m^{\prime }$ depends on the detector operators, switching functions, and initial state, but not on $r_{AC}$ once those are fixed.

We expand the interaction unitaries in a Dyson series then the difference between the two experimental arrangements first appears in the cross terms involving both the A- and C-side couplings. Those terms are controlled by nested commutators of the form:

$$[O_{F,A}\left(x\right),O_{F,C}\left(y\right)]$$

(104)

after tracing over detector degrees of freedom and inserting the readout projector. Bounding the detector operators and switching functions in norm and integrating over their compact supports yields (103). Hence the ability of the A-side intervention to modify the C-side readout is controlled precisely by the same decay function B that governs non-local commutators.

4. Measurement as Localized Detector Dynamics

The signaling in (4) arises because (3) is imposed as an exact postulate for projectors localized to bounded spacetime regions. In a dynamical description a measurement is implemented by a localized coupling between the system and a detector, and the associated state-update map is derived by tracing out the detector. This produces a completely positive instrument without requiring an exact projector-valued idealization at finite spacetime resolution.

We let ${\mathcal{H}}_S$ be the system Hilbert space and ${\mathcal{H}}_D$ a detector Hilbert space, and let ${\rho }_0$ be the initial state on ${\mathcal{H}}_S$ and ${\sigma }_D$ the initial detector state on ${\mathcal{H}}_D$, we then consider an interaction Hamiltonian:

$$H_{\mathrm{int}}\left(t\right)=\lambda \chi \left(t\right)M\otimes {\mathcal{O}}_F\left(x\left(t\right)\right),$$

(105)

where $\lambda \in \mathbb{R}$ is a coupling constant, $\chi \left(t\right)$ is a switching function with compact support in a finite time interval, M is a self-adjoint detector operator on ${\mathcal{H}}_D$, and ${\mathcal{O}}_F\left(x\left(t\right)\right)$ is a regulated system observable evaluated along the detector worldline $x\left(t\right)\in {\mathbb{R}}^{1,3}$. The unitary evolution is:

$$U=\mathcal{T}exp\left(-i\int dt{H}_{\mathrm{int}}\left(t\right)\right),$$

(106)

where $\mathcal{T}$ denotes time-ordering. A measurement outcome m is associated with a detector projector ${\Pi }_m$ on ${\mathcal{H}}_D$. The induced selective operation on the system is:

$${\mathcal{I}}_m\left({\rho }_0\right)={\mathrm{Tr}}_D\left[({\Pi }_m\otimes \mathbb{1})U({\rho }_0\otimes {\sigma }_D)U^{†}\right],$$

(107)

where ${\mathrm{Tr}}_D$ is the partial trace over ${\mathcal{H}}_D$ and $U^{†}$ is the adjoint of U. The non-selective map is $\mathcal{I}\left({\rho }_0\right)={\sum }_m{\mathcal{I}}_m\left({\rho }_0\right)$.

Now consider two parties coupled to the field in two spacetime regions $O_A$ and $O_C$ that are spacelike separated, with interaction unitaries $U_A$ and $U_C$ of the form (106) generated by couplings (105) to regulated observables ${\mathcal{O}}_{F,A}$ and ${\mathcal{O}}_{F,C}$. The influence of the A-side interaction on C-side outcome probabilities is controlled by commutators of the coupled observables. Expanding perturbatively in the couplings ${\lambda }_A$ and ${\lambda }_C$, the leading change in a C-side probability is bounded by an integral of $∥[{\mathcal{O}}_{F,A}\left(x\right),{\mathcal{O}}_{F,C}\left(y\right)]∥$ over the supports of the switching functions. When the separation between $O_A$ and $O_C$ is r in the invariant sense and $r\gg {\ell }_s$, the bound makes this influence exponentially small.

This directly addresses the mechanism in (4). Sorkin’s protocol assumes an intermediate ideal measurement of B in a thickened spacelike hypersurface, implemented as an exact projector-valued reduction that can correlate spacelike separated interventions into a signaling channel. In SFT, a physically realizable intermediate measurement is described by a detector coupling (105) to a non-local operator ${\mathcal{O}}_F$ derived from the string field. The resulting map is an instrument of the form (107). The dependence of late statistics on an earlier spacelike intervention is then bounded and controlled by string-scale tails, rather than being an unbounded logical conflict between microcausality and ideal reduction.

5. Resolution in String Field Theory

The assumptions required to derive Sorkin’s superluminal signaling are that there exist sharply localized projectors in bounded spacetime regions and that measurements enact the exact Lüders map (3) on the system state. Covariant SFT replaces pointlike degrees of freedom by extended strings described by $\mathrm{\Psi }$ and induces non-local spacetime observables with intrinsic smearing at ${\ell }_s$. Therefore, the sharp projectors needed for an ideal measurement in a bounded region are not operationally available so any attempt to localize a measurement interaction to a region O necessarily couples to observables with tails at scale ${\ell }_s$, and the resulting state-update is given by (107), not by (3).

The crucial quantitative statement is that any spacelike influence is bounded by regulated commutators, which are exponentially suppressed outside the lightcone at distances large compared to ${\ell }_s$. This ensures that communication at spacelike separation cannot be achieved to arbitrary reliability. In the limit ${\ell }_s\to 0$ at fixed macroscopic separation, one has $F(\square /M_{*}^2)\to 1$, and the non-local tails vanish. In that limit the EFT approaches local QFT and the tension with ideal measurement resurfaces, consistent with Sorkin’s conclusion that the obstruction is a UV-sensitive statement about measurement idealizations.

6. Conclusions

Sorkin’s impossible measurement identifies a mismatch between microcausality and textbook ideal measurement postulates for sharply localized field observables. Covariant string field theory resolves this mismatch by altering what counts as a physically realizable observable and a physically realizable measurement. The fundamental string field $\mathrm{\Psi }$ leads to non-local spacetime observables regulated by entire functions of $\square /M_{*}^2$ at a scale $M_{*}\sim {\ell }_s^{-1}$. When measurements are derived from localized detector couplings to such observables, the associated state-update maps are completely positive instruments and any spacelike influence is bounded by exponentially suppressed commutators. This yields an operational notion of locality compatible with relativistic causality while retaining the long-distance local limit. A parallel use of entire-function ultraviolet regulation to construct a finite completion of gauge theory, in particular complete quantum electrodynamics, is developed in Ref. [20]. The corresponding construction of a regulated scattering framework and the passage from asymptotic microcausality to a macrocausal S-matrix description are developed in Ref. [21].