Faddeev–Kulish Dressings in QED and Gravity

• Faddeev–Kulish dressings are coherent clouds of soft photons or gravitons appended to hard scattering states to cancel infrared divergences in theories with long-range interactions.
• They reorganize asymptotic dynamics by dressing bare particles with coherent-state operators and Wilson-line constructions, ensuring infrared-finite S-matrix elements.
• These dressings are linked to asymptotic symmetries, such as large gauge and BMS supertranslations, and play a key role in modern approaches like celestial holography.

Searching arXiv for the cited papers to ground the article and confirm metadata. Faddeev–Kulish dressings are coherent soft-boson clouds attached to asymptotic hard particles so that scattering states are not ordinary finite-particle Fock states but states already accompanied by their long-range gauge or gravitational field. In QED and perturbative gravity, the basic motivation is the infrared catastrophe: soft photons or gravitons are inevitably emitted and absorbed, virtual infrared divergences exponentiate, and ordinary $S$-matrix elements between bare Fock states become ill-defined or vanish in the infrared limit. The Faddeev–Kulish construction reorganizes this asymptotic dynamics by replacing bare hard particles with soft-dressed asymptotic states, yielding infrared-finite amplitudes in the cases discussed in the literature. Subsequent work has tied these dressings to large gauge symmetries, BMS supertranslations, Wilson and ’t Hooft line operators, celestial holography, worldline formulations, subleading soft theorems, and memory observables (Choi et al., 2017, 1803.02370, Choi et al., 2019, 2207.13719).

1. Infrared problem and the basic structure of the dressing

The standard setting is scattering in theories with long-range massless interactions. In QED and perturbative gravity, asymptotic charged or gravitating particles are never truly free; they are accompanied by infinitely many soft photons or gravitons. One formulation emphasizes that virtual infrared divergences exponentiate and make ordinary $S$-matrix elements between finite Fock states vanish, while real soft emission is unavoidable in any nontrivial scattering process (1803.02370). Another formulation states that perturbative amplitudes between bare charged Fock states are infrared divergent because charged particles interact via long-range fields, so one must redefine the asymptotic states themselves (Duary, 2022).

In this framework, a dressed state is obtained by acting on a hard Fock state with a coherent-state operator. A standard expression in QED is

$$\|\alpha\rangle\rangle \equiv W_\alpha |\alpha\rangle,$$

with dressed matrix elements

$$\mathbb S_{\beta\alpha} \equiv \langle\langle \beta\|S\|\alpha\rangle\rangle = \langle \beta|W_\beta^\dagger S W_\alpha|\alpha\rangle$$

that are infrared finite as $\lambda\to 0$ (1803.02370). In perturbative gravity, the analogous dressing operator can be written as

$$R_f(p_i) = \frac{\kappa}{2}\int d^3k \left[ \frac{p_i^\mu p_i^\nu}{p_i\cdot k} + c^{\mu\nu}(p_i,k) \right] \bigl(a^\dagger_{\mu\nu}(k)-a_{\mu\nu}(k)\bigr),$$

leading to the standard Faddeev–Kulish amplitude

$$M_{\text{FK}} = \bra{}\exp\!\left[-\sum_{i\in\text{out}}R_f(p_i)\right] S \exp\!\left[\sum_{i\in\text{in}}R_f(p_i)\right]\ket{}$$

(Choi et al., 2017).

The characteristic soft kernel is the same one that appears in the leading soft theorem. In QED,

$$F_l(\mathbf k , \alpha) = \sum_{\mathbf p \in \alpha}\frac{\epsilon_l \cdot p}{k \cdot p}\, \phi(\mathbf k, \mathbf p),$$

while in perturbative gravity

$$F^{gr}_l(\mathbf k , \alpha) = \sum_{\mathbf p \in \alpha} \frac{p_\mu \epsilon_l^{\mu\nu} p_\nu}{k \cdot p}\, \phi(\mathbf k, \mathbf p)$$

(1803.02370). The coherent cloud therefore reproduces the long-range Coulombic or gravitational field already at the level of asymptotic states.

A recurring claim in the literature is that the soft cloud is not merely a regulator. It is part of the physical asymptotic Hilbert space, and in several formulations it also organizes the asymptotic charge sector, the memory content of the state, or the Wilson-line structure associated with long-range interactions (1803.02370, Oertel et al., 21 Apr 2026, Feal et al., 2022).

2. Asymptotic symmetries, charge sectors, and the selection of dressed states

A major development is the identification of Faddeev–Kulish dressings with asymptotic symmetry requirements. In perturbative gravity, the BMS supertranslation charge is split as

$Q(f)=Q_S(f)+Q_H(f),$

with hard action

$S$0

and soft action

$S$1

(Choi et al., 2017). Conservation of supertranslation charge,

$S$2

then implies a relation between the soft zero mode and the hard external momenta. For eigenstates of the zero-mode operator $S$3, one obtains

$S$4

equivalently

$S$5

This gives the symmetry-based origin of the soft cloud: the in- and out-vacua differ by precisely the soft-graviton configuration required by the hard scattering (Choi et al., 2017).

The central result of that analysis is that amplitudes constructed from eigenstates of the BMS supertranslation charge coincide with Faddeev–Kulish amplitudes: $S$6 so charge-conserving amplitudes built from supertranslation eigenstates are exactly the Faddeev–Kulish amplitudes and are therefore infrared finite (Choi et al., 2017).

In QED, the same logic is described in terms of large $S$7 gauge symmetry. The asymptotic charges split into hard and soft pieces,

$S$8

and are conserved in scattering,

$S$9

(1803.02370). This leads to a selection-sector interpretation: bare momentum eigenstates generally lie in different charge sectors and do not interfere appropriately once soft radiation is traced out, whereas Faddeev–Kulish dressed states are eigenstates of the large-gauge charges with the same eigenvalue and can therefore interfere (1803.02370).

This suggests a common structural statement across gauge theory and gravity: the dressings are selected because asymptotic scattering is organized by conserved soft-plus-hard charges rather than by the naive Fock-space notion of isolated hard quanta. A plausible implication is that the infrared problem and the asymptotic symmetry problem are not separate issues but two descriptions of the same asymptotic mismatch.

3. Wilson lines, ’t Hooft lines, and geometric realizations

A second major theme is the reinterpretation of Faddeev–Kulish dressings as line operators. In QED, the Wilson-line form arises from Mandelstam’s gauge-invariant dressed field

$$\|\alpha\rangle\rangle \equiv W_\alpha |\alpha\rangle,$$0

For a timelike straight path corresponding to an asymptotic massive particle, this reduces to

$$\|\alpha\rangle\rangle \equiv W_\alpha |\alpha\rangle,$$1

which is identified with the Faddeev–Kulish operator (Choi et al., 2018). The same paper expresses the dressing directly in terms of boundary soft gauge fields,

$$\|\alpha\rangle\rangle \equiv W_\alpha |\alpha\rangle,$$2

and shows that it shifts the vacuum to a new vacuum with different soft charge (Choi et al., 2018).

This Wilson-line perspective is extended to boundaries other than null infinity. Timelike Wilson lines puncturing the future Rindler horizon or the Schwarzschild horizon create soft edge states carrying horizon soft charge, and are described as horizon analogs of flat-space Faddeev–Kulish dressings (Choi et al., 2018).

The line-operator picture also generalizes to electric–magnetic duality. In a theory with electric and magnetic charges, the magnetic analogue of the dressing is

$$\|\alpha\rangle\rangle \equiv W_\alpha |\alpha\rangle,$$3

and the paper shows that it can be written as a ’t Hooft line operator,

$$\|\alpha\rangle\rangle \equiv W_\alpha |\alpha\rangle,$$4

(Choi et al., 2019). The electric Wilson-line dressing is charged under electric large gauge transformations and neutral under magnetic ones, while the magnetic ’t Hooft-line dressing is charged under magnetic large gauge transformations and neutral under electric ones (Choi et al., 2019).

The gravitational extension follows the same pattern. The usual gravitational Wilson-line dressing is

$$\|\alpha\rangle\rangle \equiv W_\alpha |\alpha\rangle,$$5

while a gravitational ’t Hooft-line dressing

$$\|\alpha\rangle\rangle \equiv W_\alpha |\alpha\rangle,$$6

is charged under dual supertranslations and neutral under ordinary supertranslations (Choi et al., 2019).

These constructions show that Faddeev–Kulish dressings can be understood as coherent asymptotic line operators that carry definite asymptotic charges. This suggests a geometric interpretation of the dressing as a boundary puncture or asymptotic worldline rather than merely as an algebraic coherent-state factor.

4. Infrared finiteness, cloud relocation, and debates on completeness

The canonical cancellation mechanism is that virtual infrared divergences are canceled by the soft clouds attached to the external states. In the gravitational analysis tied to BMS symmetry, the one-loop infrared-divergent terms split into virtual graviton exchange, interactions between clouds and external lines, and cloud-to-cloud exchanges, with the cancellation schematically written as

$$\|\alpha\rangle\rangle \equiv W_\alpha |\alpha\rangle,$$7

(Choi et al., 2017). Since all charge-conserving amplitudes are equal to Faddeev–Kulish amplitudes, all such amplitudes are infrared finite (Choi et al., 2017).

An important structural result is that dressing clouds can be moved from the incoming side of the scattering operator to the outgoing side, or conversely, without changing the infrared finiteness properties of the amplitude. This is described as the clouds “weakly commuting” with $$\|\alpha\rangle\rangle \equiv W_\alpha |\alpha\rangle,$$8 (Choi et al., 2017). The significance assigned to this property is that the physical content lies in the total asymptotic soft charge balance, not in whether the cloud is written on the bra or ket side.

Other works sharpen or challenge the original prescription. One analysis argues that the original Kulish–Faddeev dressed states are necessary for the Gauss-law or BRST physical-state condition but are not sufficient to remove all infrared divergences in QED. In that treatment, asymptotic symmetry requires an additional transverse Chung-type dressing $$\|\alpha\rangle\rangle \equiv W_\alpha |\alpha\rangle,$$9, and the final states must include both the usual KF factor and the extra transverse cloud to obtain an infrared-finite $$\mathbb S_{\beta\alpha} \equiv \langle\langle \beta\|S\|\alpha\rangle\rangle = \langle \beta|W_\beta^\dagger S W_\alpha|\alpha\rangle$$0-matrix (Hirai et al., 2020). This is presented as a correction to the original prescription rather than a rejection of the dressed-state program.

A different critique comes from a low-energy approximation of massless Yukawa theory. There the authors derive a Faddeev–Kulish-type formula from an infrared-finite Dollard modifier and then decompose it into a cloud of virtual photons and a cloud of real photons. They argue that in the original Faddeev–Kulish construction the real-photon cloud is omitted, and consequently the scattering matrix is ill-defined on the Fock space of free electrons (Dybalski, 2017). In that account, the correct asymptotic state contains both the virtual cloud dressing the charged particle and the real cloud radiated away in the scattering process.

By contrast, the dressed-state analysis based on subleading soft dressings states that there is no real soft radiation in the asymptotic Hilbert space and that dressed amplitudes are equivalent to the infrared-finite part of traditional Fock amplitudes. In that framework, the dressed formalism gives the same cross sections as the Bloch–Nordsieck method (Choi et al., 2019). The contrast between these viewpoints is a genuine controversy in the literature. At minimum, it shows that the precise content of the asymptotic Hilbert space remains formulation-dependent, even when there is broad agreement that soft dressing is required.

A further reformulation appears in the worldline approach to QED. There the endpoint photon exchanges at $$\mathbb S_{\beta\alpha} \equiv \langle\langle \beta\|S\|\alpha\rangle\rangle = \langle \beta|W_\beta^\dagger S W_\alpha|\alpha\rangle$$1 and $$\mathbb S_{\beta\alpha} \equiv \langle\langle \beta\|S\|\alpha\rangle\rangle = \langle \beta|W_\beta^\dagger S W_\alpha|\alpha\rangle$$2 are kept explicitly, and these asymptotic endpoint contributions are claimed to be equivalent to the soft coherent dressings proposed by Faddeev and Kulish. In this picture, the ordinary Dyson $$\mathbb S_{\beta\alpha} \equiv \langle\langle \beta\|S\|\alpha\rangle\rangle = \langle \beta|W_\beta^\dagger S W_\alpha|\alpha\rangle$$3-matrix omits those asymptotic terms and therefore retains infrared divergences, while the worldline construction automatically includes them and yields an infrared-finite $$\mathbb S_{\beta\alpha} \equiv \langle\langle \beta\|S\|\alpha\rangle\rangle = \langle \beta|W_\beta^\dagger S W_\alpha|\alpha\rangle$$4-matrix to all loop orders (Feal et al., 2022).

5. Superpositions, decoherence, and the structure of the asymptotic Hilbert space

The need for dressing is especially sharp for incoming superpositions and wavepackets. In the inclusive formalism, after tracing over unobserved soft radiation, the reduced density matrix contains factors of the form

$$\mathbb S_{\beta\alpha} \equiv \langle\langle \beta\|S\|\alpha\rangle\rangle = \langle \beta|W_\beta^\dagger S W_\alpha|\alpha\rangle$$5

and for a generic superposition this suppresses off-diagonal terms unless the relevant hard currents match exactly (1803.02370). The result is that the scattering of an incoming superposition behaves as if the initial state were a classical ensemble: $$\mathbb S_{\beta\alpha} \equiv \langle\langle \beta\|S\|\alpha\rangle\rangle = \langle \beta|W_\beta^\dagger S W_\alpha|\alpha\rangle$$6 (1803.02370).

For wavepackets, the same suppression acts on a continuum and collapses the nontrivial scattering contribution onto a set of measure zero. The hard sector therefore appears unchanged,

$$\mathbb S_{\beta\alpha} \equiv \langle\langle \beta\|S\|\alpha\rangle\rangle = \langle \beta|W_\beta^\dagger S W_\alpha|\alpha\rangle$$7

leading to the statement that the inclusive formalism predicts no observable scattering for wavepackets (1803.02370).

With dressed asymptotic states, the reduced outgoing hard density matrix becomes

$$\mathbb S_{\beta\alpha} \equiv \langle\langle \beta\|S\|\alpha\rangle\rangle = \langle \beta|W_\beta^\dagger S W_\alpha|\alpha\rangle$$8

and for a definite outgoing hard state

$$\mathbb S_{\beta\alpha} \equiv \langle\langle \beta\|S\|\alpha\rangle\rangle = \langle \beta|W_\beta^\dagger S W_\alpha|\alpha\rangle$$9

(1803.02370). The interference terms are therefore restored. This is used to argue that dressed asymptotic states are the correct physical states for realistic coherent superpositions and localized particles.

The gravitational BMS-based analysis also discusses decoherence, but in a different sense. If one traces over soft gravitons, the overlap of distinct soft coherent clouds is

$\lambda\to 0$0

which vanishes if the hard momentum configurations differ (Choi et al., 2017). Thus the dressed-state formalism cures the infrared problem of the $\lambda\to 0$1-matrix, but reduced density matrices for hard particles can still decohere when the soft sector is traced out (Choi et al., 2017).

These results are compatible rather than contradictory. They indicate that Faddeev–Kulish dressing preserves the coherent scattering structure at the level of the full asymptotic state, while tracing over the soft sector can still destroy off-diagonal hard-sector coherence.

6. Celestial, holographic, AdS, and finite-time developments

In celestial holography, Faddeev–Kulish dressings are reformulated as conformal operators built from soft Goldstone fields. One construction introduces free bosons $\lambda\to 0$2 for large $\lambda\to 0$3 symmetry and scalar Goldstone fields $\lambda\to 0$4 for supertranslations, with single-particle dressings

$\lambda\to 0$5

(2207.13719). The corresponding dressed celestial operators are

$\lambda\to 0$6

in QED, and

$\lambda\to 0$7

in gravity (2207.13719). In that setting, the dressings remove all infrared divergent terms to all orders in perturbation theory in celestial two-point functions on particle-like Kerr–Schild backgrounds (2207.13719).

A related celestial construction organizes the conformally soft sector into “celestial diamonds.” There the top corners of the Goldstone diamonds are identified with conformal Faddeev–Kulish dressings, with leading Goldstone fields

$\lambda\to 0$8

and subleading conformal dressing fields $\lambda\to 0$9, $$R_f(p_i) = \frac{\kappa}{2}\int d^3k \left[ \frac{p_i^\mu p_i^\nu}{p_i\cdot k} + c^{\mu\nu}(p_i,k) \right] \bigl(a^\dagger_{\mu\nu}(k)-a_{\mu\nu}(k)\bigr),$$0 that encode the subleading soft sector (Pasterski et al., 2021). The claim is that dressing hard operators with $$R_f(p_i) = \frac{\kappa}{2}\int d^3k \left[ \frac{p_i^\mu p_i^\nu}{p_i\cdot k} + c^{\mu\nu}(p_i,k) \right] \bigl(a^\dagger_{\mu\nu}(k)-a_{\mu\nu}(k)\bigr),$$1 and $$R_f(p_i) = \frac{\kappa}{2}\int d^3k \left[ \frac{p_i^\mu p_i^\nu}{p_i\cdot k} + c^{\mu\nu}(p_i,k) \right] \bigl(a^\dagger_{\mu\nu}(k)-a_{\mu\nu}(k)\bigr),$$2 cancels the infrared divergent soft factor, leaving an infrared-finite amplitude (Pasterski et al., 2021).

The AdS/CFT setting provides another reformulation. There the Wilson-line-dressed scalar field

$$R_f(p_i) = \frac{\kappa}{2}\int d^3k \left[ \frac{p_i^\mu p_i^\nu}{p_i\cdot k} + c^{\mu\nu}(p_i,k) \right] \bigl(a^\dagger_{\mu\nu}(k)-a_{\mu\nu}(k)\bigr),$$3

is used as the starting point, with AdS acting as an infrared regulator and the flat-space Faddeev–Kulish state emerging only after the large-radius zoomed-in limit (Duary, 2022). The dressed scalar can then be reconstructed by “vanilla HKLL reconstruction” because only soft photon modes are used in the Wilson-line dressing (Duary, 2022).

Recent work has also emphasized the full time dependence of the dressing. In these finite-time treatments, the dressing kernels are

$$R_f(p_i) = \frac{\kappa}{2}\int d^3k \left[ \frac{p_i^\mu p_i^\nu}{p_i\cdot k} + c^{\mu\nu}(p_i,k) \right] \bigl(a^\dagger_{\mu\nu}(k)-a_{\mu\nu}(k)\bigr),$$4

for QED, and

$$R_f(p_i) = \frac{\kappa}{2}\int d^3k \left[ \frac{p_i^\mu p_i^\nu}{p_i\cdot k} + c^{\mu\nu}(p_i,k) \right] \bigl(a^\dagger_{\mu\nu}(k)-a_{\mu\nu}(k)\bigr),$$5

for gravity (Oertel, 7 May 2026). These papers argue that the dressing is fully constrained by symmetry, rotational invariance, and the requirement of reproducing classical memory, leading to the unique rotationally invariant choice

$$R_f(p_i) = \frac{\kappa}{2}\int d^3k \left[ \frac{p_i^\mu p_i^\nu}{p_i\cdot k} + c^{\mu\nu}(p_i,k) \right] \bigl(a^\dagger_{\mu\nu}(k)-a_{\mu\nu}(k)\bigr),$$6

for both QED and gravity (Oertel, 7 May 2026). Related detector-based analyses state that the full $$R_f(p_i) = \frac{\kappa}{2}\int d^3k \left[ \frac{p_i^\mu p_i^\nu}{p_i\cdot k} + c^{\mu\nu}(p_i,k) \right] \bigl(a^\dagger_{\mu\nu}(k)-a_{\mu\nu}(k)\bigr),$$7-dependent dressing correctly encodes the memory effect in the in- and out-scattering Fock spaces and contributes physical terms to the memory eigenvalues (Oertel et al., 21 Apr 2026).

7. Subleading soft dressings and current directions

The leading Faddeev–Kulish cloud captures the universal $$R_f(p_i) = \frac{\kappa}{2}\int d^3k \left[ \frac{p_i^\mu p_i^\nu}{p_i\cdot k} + c^{\mu\nu}(p_i,k) \right] \bigl(a^\dagger_{\mu\nu}(k)-a_{\mu\nu}(k)\bigr),$$8 soft pole, but several works extend the construction to the next order in the soft expansion. One approach derives subleading dressings from asymptotic symmetry charge conservation associated with the tree-level subleading soft theorems. In gravity the result is

$$R_f(p_i) = \frac{\kappa}{2}\int d^3k \left[ \frac{p_i^\mu p_i^\nu}{p_i\cdot k} + c^{\mu\nu}(p_i,k) \right] \bigl(a^\dagger_{\mu\nu}(k)-a_{\mu\nu}(k)\bigr),$$9

and in QED

$$M_{\text{FK}} = \bra{}\exp\!\left[-\sum_{i\in\text{out}}R_f(p_i)\right] S \exp\!\left[\sum_{i\in\text{in}}R_f(p_i)\right]\ket{}$$0

(Choi et al., 2019). In that construction, emission and absorption of extra soft photons or gravitons vanish in dressed amplitudes, which is taken to show that the soft and hard sectors are correlated rather than independent (Choi et al., 2019).

A more recent QED analysis introduces a dressing scale $$M_{\text{FK}} = \bra{}\exp\!\left[-\sum_{i\in\text{out}}R_f(p_i)\right] S \exp\!\left[\sum_{i\in\text{in}}R_f(p_i)\right]\ket{}$$1 and writes dressed states as

$$M_{\text{FK}} = \bra{}\exp\!\left[-\sum_{i\in\text{out}}R_f(p_i)\right] S \exp\!\left[\sum_{i\in\text{in}}R_f(p_i)\right]\ket{}$$2

with coherent clouds containing photons in the range

$$M_{\text{FK}} = \bra{}\exp\!\left[-\sum_{i\in\text{out}}R_f(p_i)\right] S \exp\!\left[\sum_{i\in\text{in}}R_f(p_i)\right]\ket{}$$3

The leading dressing function is

$$M_{\text{FK}} = \bra{}\exp\!\left[-\sum_{i\in\text{out}}R_f(p_i)\right] S \exp\!\left[\sum_{i\in\text{in}}R_f(p_i)\right]\ket{}$$4

and the dressed elastic amplitude becomes

$$M_{\text{FK}} = \bra{}\exp\!\left[-\sum_{i\in\text{out}}R_f(p_i)\right] S \exp\!\left[\sum_{i\in\text{in}}R_f(p_i)\right]\ket{}$$5

(Christodoulou et al., 19 Mar 2026). This is interpreted as replacing the usual infrared divergence by a finite dependence on the dressing scale $$M_{\text{FK}} = \bra{}\exp\!\left[-\sum_{i\in\text{out}}R_f(p_i)\right] S \exp\!\left[\sum_{i\in\text{in}}R_f(p_i)\right]\ket{}$$6, and if $$M_{\text{FK}} = \bra{}\exp\!\left[-\sum_{i\in\text{out}}R_f(p_i)\right] S \exp\!\left[\sum_{i\in\text{in}}R_f(p_i)\right]\ket{}$$7, the dressed amplitude is identified with the infrared-finite part of the corresponding Fock-basis amplitude (Christodoulou et al., 19 Mar 2026).

That same work includes a subleading cloud depending on total angular momentum,

$$M_{\text{FK}} = \bra{}\exp\!\left[-\sum_{i\in\text{out}}R_f(p_i)\right] S \exp\!\left[\sum_{i\in\text{in}}R_f(p_i)\right]\ket{}$$8

$$M_{\text{FK}} = \bra{}\exp\!\left[-\sum_{i\in\text{out}}R_f(p_i)\right] S \exp\!\left[\sum_{i\in\text{in}}R_f(p_i)\right]\ket{}$$9

and reports that tree-level soft-photon emission for $$F_l(\mathbf k , \alpha) = \sum_{\mathbf p \in \alpha}\frac{\epsilon_l \cdot p}{k \cdot p}\, \phi(\mathbf k, \mathbf p),$$0 is suppressed once the dressings are extended to subleading order,

$$F_l(\mathbf k , \alpha) = \sum_{\mathbf p \in \alpha}\frac{\epsilon_l \cdot p}{k \cdot p}\, \phi(\mathbf k, \mathbf p),$$1

(Christodoulou et al., 19 Mar 2026).

Across these developments, several points recur. First, the leading cloud is widely accepted as the asymptotic structure required by long-range interactions. Second, subleading dressings are increasingly understood in terms of asymptotic symmetries, angular momentum dependence, and conformally soft operators. Third, there remains active debate over the precise completeness of the original Kulish–Faddeev prescription, the role of extra transverse or real-radiation clouds, and the most appropriate definition of the asymptotic Hilbert space (Hirai et al., 2020, Dybalski, 2017, Choi et al., 2019).

Taken together, the literature presents Faddeev–Kulish dressings not as a single fixed operator formula, but as a general asymptotic principle: infrared-safe scattering in theories with massless long-range forces requires asymptotic states that already carry the soft structure dictated by the corresponding charges, line operators, and memory data.