Arxiv:1910.05828V3 [Quant-Ph] 7 May 2020 6] and Microwave Frequencies [7–9]

Dynamics of many-body photon bound states in chiral waveguide QED

Sahand Mahmoodian,1 Giuseppe Calajó,2 Darrick E. Chang,2, 3 Klemens Hammerer,1 and Anders S. Sørensen4 1Institute for Theoretical Physics, Institute for Gravitational Physics (Albert Einstein Institute), Leibniz University Hannover, Appelstraße 2, 30167 Hannover, Germany 2ICFO-Institut de Ciencies Fotoniques, The Barcelona Institute of Science and Technology, 08860 Castelldefels (Barcelona), Spain 3ICREA-InstitucióCatalana de Recerca i Estudis Avan¸cats,08015 Barcelona, Spain 4Center for Hybrid Quantum Networks (Hy-Q), Niels Bohr Institute, University of Copenhagen, Blegdamsvej 17, DK-2100 Copenhagen, Denmark (Dated: May 8, 2020) We theoretically study the few- and many-body dynamics of photons in chiral waveguides. In particular, we examine pulse propagation through an ensemble of N two-level systems chirally coupled to a waveguide. We show that the system supports correlated multi-photon bound states, which have a well-defined photon number n and propagate through the system with a group delay scaling as 1/n2. This has the interesting consequence that, during propagation, an incident coherent- state pulse breaks up into different bound-state components that can become spatially separated at the output in a sufficiently long system. For sufficiently many photons and sufficiently short systems, we show that linear combinations of n-body bound states recover the well-known phenomenon of mean-field solitons in self-induced transparency. Our work thus covers the entire spectrum from few- photon quantum propagation, to genuine quantum many-body (atom and photon) phenomena, and ultimately the quantum-to-classical transition. Finally, we demonstrate that the bound states can undergo elastic scattering with additional photons. Together, our results demonstrate that photon bound states are truly distinct physical objects emerging from the most elementary light-matter interaction between photons and two-level emitters. Our work opens the door to studying quantum many-body physics and soliton physics with photons in chiral waveguide QED.

I. INTRODUCTION In particular we theoretically consider the propagation of pulses of coherent and Fock states of light through a waveguide to which N TLSs are chirally coupled. We Generating quantum many-body states of light re- show that photons undergo a Wigner delay [21, 22] — a mains one of the outstanding challenges of modern quan- delay of the pulse center due the optical excitation trans- tum optics [1]. Such many-body states of light are of ferring to the atom and back to the waveguide — which is fundamental physical interest as they arise from non- dependent on the number of photons. Therefore, incident equilibrium systems with strong interactions between pulses break up into a state that is temporally ordered light and matter. On the other hand, they also promise to by its photon correlations. For example, as shown in form novel resources for quantum technologies, for exam- Fig.1(a), an incident coherent field can produce a pulse ple, in quantum-enhanced metrology [2]. The main ob- with three-photon correlations followed by two-photon stacle in the pursuit of generating such many-body states correlations states followed by uncorrelated photons. The has been the difficulty in developing one-dimensional sys- underlying physics that causes this time delay is exam- tems with a sufficiently strong nonlinear response at ined by considering the many-body photonic scattering the few-photon scale [3]. Recently however, significant eigenstates of TLSs chirally coupled to a waveguide. progress has been made in creating an ideal light–matter Of central importance are the class of bound eigen- interface between atoms or artificial emitters coupled to states, where two or more photons propagate together. a one-dimensional continuum of photons at optical [4– We show that the photon bound states propagate past arXiv:1910.05828v3 [quant-ph] 7 May 2020 6] and microwave frequencies [7–9]. Such an interface each atom with a photon-number-dependent delay of creates a highly nonlinear medium as photons propagat- 2 τn = 4/(n Γ), where Γ is the decay rate into the waveg- ing in a waveguide interact deterministically with atoms. uide, as can be understood in terms of absorption and Systems of this kind have thus far been used to propose stimulated emission of a photon as shown in Fig.1(b). or demonstrate the generation of states of photons with Since the interaction is chiral, the delay is also propor- strong two- or three-body correlations [10–20]. tional to the number of emitters N in the waveguide. Studies of photon correlations in these systems typi- We also show that the pulse distortion encountered by 6 cally consider steady-state driving and photon correla- an n-photon bound state scales as n− . Pulses of higher- tions are subsequently measured in relative coordinates. number bound states therefore propagate with negligible On the other hand, here we show that pulse propagation distortion, i.e. much like a soliton. Moving beyond the through two-level systems (TLSs) strongly coupled to a few-photon–few-atom limit we obtain a simple descrip- waveguide leads to very distinct temporal features which tion of photon propagation with a mesoscopic number of reveal the underlying dynamics and has the potential to photons n but a large number of atoms N 1. Here, the generate temporally ordered many-body states of light. number-dependent group delay breaks the input pulse 2

(a) TLS light-matter interaction between photons and two-level input field emitters. output field This manuscript is arranged as follows: in SectionII we introduce the model for chiral wQED and outline var- ious theoretical approaches for computing the dynamics throughIIA mean-ﬁeld theory,IIB the photon scattering eigenstates, andIIC the matrix-product states (MPS). In chiral waveguide SectionIII we compute how the input pulse propagates (b) BS propagation through the medium and compute the representation in terms of photon bound states. This is followed by Sec- tionIV which shows that the many-body photon bound states can be used to construct the mean-ﬁeld soliton solutions obtained in self-induced transparency. In Section FIG. 1. (a) N two-level atoms (blue circles) are chirally cou- V we show that photon bound states can undergo elastic pled to a waveguide with decay rate Γ and driven by an input scattering with individual photons modifying the delay of Gaussian pulse, which can be a coherent or Fock state. The light pulse propagates with a correlation-number-dependent the bound state but otherwise leaving it unaltered, much group velocity leading to an output state where one-, two- like classical solitons. In SectionVI, we show few-photon and three-photon bound states are spatially separated. (b) bound-state propagation can potentially be observed in Schematic of the bound states propagation. When an n- state-of-the-art experiments with a few emitters, and we photon bound state is scattered by an atom, it re-emits the discuss potential future applications. Finally, we con- absorbed photon with a stimulated emission rate Γn, coincid- clude in SectionVII. ing with the inverse of its width. Since only a single photon out of n is delayed by an amount 4/(Γn), the pulse preserves 2 its shape but is delayed by τn = 4/(Γn ). II. MODEL

We consider a system of N two-level systems (TLS) apart and produces a many-body ordered state of light. chirally coupled coupled to a linearly dispersive one- Finally, we show that the system approaches the classi- dimensional bath of photons. Here, chiral coupling means cal limit for n 1 and n N, where our full quantum that the TLSs couple only to right-propagating photons. description captures the quantum-to-classical transition The Hamiltonian for this system (~ = 1) is and reproduces the mean-field results of soliton propaga- Z N tion in self-induced transparency. X + Hˆ = i dx aˆ†(x)∂ aˆ(x)+√Γ σˆ−aˆ†(x ) +σ ˆ aˆ(x ) , The effect discussed here has features similar to − x i i i i i=1 vacuum-induced transparency (VIT) where photons un- (1) dergo a photon-number-dependent delay after interacting where all integrals are over ,σ ˆ± are Pauli operators for < i with atoms coupled to a cavity [23, 24]. A key difference the ith TLS,a ˆ(x) (â†(x)) is a photon annihilation (cre- is that the delay in VIT mainly depends on the total ation) operator at position x, the group velocity is set to photon number inside the entire system, and not on the unity vg = 1, and the energy has been renormalized to details of the pulse shape. The nonlinear effect we con- that of the TLS. In the limit of ideal chiral coupling, the sider is spatially localized to the individual atoms and system dynamics are not influenced by the positions of occurs for the simplest possible configuration of TLSs. the TLSs xi. In this Hamiltonian the first term captures This leads to a different spatiotemporal behaviour which the free-propagation of photons in the waveguide with we evaluate in a full multimode theory of the dynamics. linear dispersion, while the next terms take into account Our theory features a full quantum many-body treatment the interaction between the photons and the emitters. In of the system, where the photon time delay is examined the absence of the interaction terms the photonic eigen- by considering the many-body photonic scattering eigenstates of the Hamiltonian are plane waves with wavevec- states. tor k and frequency ω = k. Equation (1) constitutes the We also point out that in mean-field theories solitons typical scenario for chiral wQED [25]. In this manuscript, are known to be highly stable objects which are unaf- we are interested in describing the propagation of a multi- fected by external perturbations. Here we show that photon input field through the system. This goal can be similar properties exist for few-photon bound states. We achieved by exactly computing the scattering eigenstates outline how one can conduct scattering experiments be- of Hamiltonian (1), as illustrated in SubsectionIIB. tween photon bound states and individual photons. In In addition to computing the eigenstates of (1), we also the considered scattering experiments the bound state introduce an equivalent formulation that is well-suited for is deflected by the interaction but is otherwise unper- using the MPS technique that we are going to introduce turbed by it. Together the results obtained here demon- later in SubsectionIIC. This approach is based on the strate that the bound states should be considered as truly observation that the transmitted field depends directly distinct physical entities emerging from the underlying on the emitters’ evolution. This can be seen by formally 3 integrating the Heisenberg equation for the field opera- as a spin continuum under the mean-field approximation tora ˆ(x, t), that provides, within Born-Markov approxi- where quantum correlations between the atoms and the mation, the following generalized input-output relation light field can be neglected. Under these approximations, for the transmitted field [26, 27] the equations of motion give the SIT equations X aˆ (t) = (t) + i √Γˆσ−(t), (2) ∂ ∂ out Ein j + a(x, t) = i√Γσ (x, t), j ∂t ∂x − − ∂ + √ (5) where we have defineda ôut(t) =a ˆ(x , t) as the out- σ (x, t) = i Γσz(x, t)a(x, t), N ∂t − put field measured right after the last atom. Note that, ∂ within this approach, we assume the input field in(t) to σz(x, t) = 4√Γ Im a(x, t)σ∗ (x, t) . be a classical coherent field on resonance with theE atomic ∂t − transitions. With these assumptions the emitter dynam- /z Here, a(x, t) = aˆ(x, t) ,σ ˆ (x, t) = P σˆ− (t)δ(x ics driven by the input field is known to be described by /z i i x ), and σ (x, t)h = σˆ i (x,− t) , σ (x, t) = σˆ (x, t) are− a purely dissipative chiral master equation (ME) of the i z z the expectation− valuesh − of thei spin operatorsh in thei con- form [28, 29] (see Supplemental Material (SM) for more tinuum limit. These nonlinear equations of motion have information): SIT soliton solutions. Following the treatment in [34], X + for a resonant pulse the field can be taken to be real, ρ˙ = i H ρ ρH† + Γ σˆ−ρσˆ . (3) − eff − eff i j and one can map the equations of motion onto a nonlin- ij ear pendulum equation that can be solved exactly. This Here, gives the fundamental soliton solution for the field Γ X + X + √Γ¯n Γ¯n x Heff = i σˆ σˆ− + Hdrive iΓ σˆ σˆ−, (4) a(x, t) = sech t , (6) − 2 j j − l j 2 2 V − j l>j 0 is the effective Hamiltonian, which provides the non- wheren ¯ is the number of photons in the field, (or more Hermitian collective evolution of the emitters, while the precisely, the total energy in the original SIT work). The P + term Hdrive = √Γ in(t)ˆσ + H.c. gives the cou- pulse velocity within the medium in the laboratory frame j j 2 2 pling of the emitters to theE input field. is V 0 =n ¯ Γ/(¯n Γ + 4ν), where ν is the gas density (see The combination of Eqs. (2) and (3) provide a full de- SM for details). Transforming to a frame comoving with scription of the photon propagation through the chiral the pulse in the absence of emitters, i.e. at velocity vg = medium. In particular the spin dynamics can be effi- 1, gives the relative velocity in the backwards direction 2 ciently solved by making use of an MPS ansatz [30, 31] V = 4ν/(¯n Γ + 4ν). This corresponds to each emitter 2 as recently described in Ref. [27]. As will be shown in the imparting a delay of τn¯ = 4/(¯n Γ) on the pulse. following, this approach will allow us to fully explore the An important feature of the solitonic solution (6) is R limit of many photons and large atomic arrays, a scenario that the integrated Rabi frequency Ω = 2√Γ dta(x, t), that is challenging to simulate with standard numerical which is proportional to the area under the pulse, is techniques. fixed by the relationship between the pulse amplitude and pulse width, and always evaluates to be 2π. This corresponds to a full Rabi cycle of complete excitation A. Mean-field Theory and Self-induced and subsequent deexcitation. The SIT soliton therefore Transparency can be physically interpreted as a rapid excitation and de-excitation of the atoms, which suppresses spontaneous Before considering the full many-body dynamics of emission of the excited state and thus makes the medium the Hamiltonian (1), we consider the system within the transparent. mean-field limit. We present this mean-field limit to con- Inspired by the apparent dependence of velocity on trast its predictions with the full many-body theory pre- photon number, it is interesting to ask whether this prop- sented below. erty extends to the few-photon limit, thus enabling, e.g., The first treatment of (1) within mean-field theory photon-number separation at the output. A full quantum dates back to the work on self-induced transparency treatment is necessary to answer this question, which we (SIT) [32–34]. In these early experiments, gasses of two- turn to in the following sections. level atoms were excited by short intense laser pulses. Although the atoms are not ideally coupled to a single waveguide mode in such systems, the laser pulses are suf- B. Many-body Scattering Eigenstates ficiently short that decay channels to modes other than the laser mode can be neglected. Furthermore, both the In contrast to the mean-field treatment we now con- weak coupling of the atoms to this mode and the high in- sider the full many-body eigenstates. Since the Hamilto- tensity of the laser means that one can consider the atoms nian (1) preserves the combined number of atomic and 4 photonic excitations, eigenstates with different numbers been used to diagonalize the Hamiltonian in Eq. (1)[37]. of excitations decouple. By computing the eigenstates in Since we are interested in the state that emerges after the one- and two-excitation subspaces, one can general- interaction with the TLSs, we are interested in the scat- ize the result to an arbitrary number of excitations. This tering eigenstates. These are the photon states that in- technique is often referred to as Bethe’s ansatz [35] and teract with the TLSs and emerge unchanged apart from is used to diagonalize a class of one-dimensional many- an overall transmission coefficient. The n-body scatter- body Hamiltonians [36]. In particular, it has previously ing eigenstates have the form

Z n n 1 n Y Y ikj xj Sk = Ck d x aˆ†(x) 0 [k k + iΓ sgn (x x )] e + , (7) | in ,n,S √ | i l − m l − m ↔ n! l

where Ck,n,S is a normalization constant which varies p(n) is the number of partitions of n. For example, for with wavevector k, excitation number n, and the type three photons, there is a completely extended scattering of eigenstates S, where S labels different states as ex- eigenstate, a completely bound eigenstate, and a hybrid n plained below; aˆ†(x) =a ˆ†(x1)â†(x2) ... aˆ†(xn); d x = state with two bound photons and one extended photon. dx dx . . . dx ; and indicates summing over all n! per- The different string combinations give the different types 1 2 n ↔ mutations of xi to symmetrize the wavefunction. The of scattering eigenstates S by substituting the complex Pn energy of the eigenstates is E = i ki. Upon scattering wavevectors into (7). The transmission coefficient is also off all N emitters, the eigenstates are multiplied by the obtained by substituting the complex wavevectors into N Qn N eigenvalue t = t , where tk = (k iΓ/2)/(k + the expression for tk. k j=1 kj − iΓ/2). Since, by assumption, the system does not con- For n photons, the n-string state is a fully localized tain any dissipation and is chiral, all the transmission bound state with energy E, coefficients have t = 1. This means that the trans- k Z E P Γ P | | Cn,B n i xj xi xj mission coefficients simply multiply the eigenstate by a B = d x aˆ†(x) 0 e n j − 2 i 2 the m-body string (m n) has the wavevec- ertheless the form used here is particularly insightful as ≤ tor kj = K/m i [m + 1 2j]Γ/2 with j = 1, 2, . . . , m it gives direct access to the number-dependent transmis- where K is the− total energy− of the m-string. In total, the sion coefficient which, as we show, plays a central role in n-photon manifold has p(n) string combinations where understanding the many-body pulse propagation. 5

1 photon Fock state 0.24 2 photon Fock state output output

0.16

0.08

3 photon Fock state output

hybrid state

FIG. 2. Transmitted power P (x)√ (solid curves) for (a) one- (b) two-, and (c) three-photon transport through N = 20 TLSs for a resonant pulse with Γσ = 3 2 computed using the photon scattering theory. In (b) and (c) dashed lines show the contributions from the diﬀerent eigenstates. The input mode in all panels is the Gaussian curve shown in (a). (d) Transmitted power and correlation functions for a coherent state withn ¯ = 0.5. Solid lines show photon scattering results truncating at three photons while marker points show MPS calculations performed by solving the ME with a Runge-Kutta algorithm and (2) using maximum bond dimension Dmax = 150. (e) Second-order correlation function G (x1, x2) for the transmitted coherent state computed using photon scattering theory. The two- and three-photon bound state contributions appear as peaks that√ are localized in the relative coordinate. (f) MPS calculation for input coherent state withn ¯ = 8, N = 60 emitters and Γσ = 2 2. 2 Dashed vertical lines show N times the Wigner delay Tn = Nτn = 4N/(Γn ). With Pn(x) we indicate the contribution to the intensity coming from the emission of n photons within the space bin ∆x = 1. To perform this calculation we used a quantum trajectories MPS algorithm ﬁxing the maximum bond dimension to Dmax = 40.

C. MPS Ansatz ries algorithm where the state of the system evolves under the eﬀective Hamiltonian (4) and stochastically experi- In order to study the dynamics for stronger input ences quantum jumps [27] (method used for Fig.2(f)). pulses (n > 3) than the one computed with the S-matrix In both cases an MPS representation is applied either to formalism, we solve equations (2) and (3) using an MPS the quantum state or to a vectorized form of the density ansatz. Speciﬁcally, the system evolution can be solved matrix. Here for simplicity we limit the discussion to the either by directly solving the ME (3)[42] (method used former while the latter is discussed in the Supplemental for Fig.2(d) and Fig.3) or by using a quantum trajecto- Material. 6

The MPS ansatz consists of reshaping the generic faster velocity than the extended state. The bound state P quantum state φ = ψi ,i ,..i i1, i2, ..iN (with also undergoes significantly less broadening and distor- | i i1,..iN 1 2 N | i ij g, e ) into a matrix-product state of the form: tion. For the three-photon transport in Fig.2(c) again ∈ { } the extended state is distorted and delayed, while the X three-photon bound state has significantly less distortion φ = Ai Ai ...Ai i1, i2, . . . , iN , (11) | i 1 2 N | i and delay. In the three-photon manifold there is also a i1,...,iN string combination that forms a hybrid state where one where, for each specific set of physical indices photon is completely extended while the other two are i , i , . . . , i , the product of the A matrices gives bound. The evolution of this state is determined by the { 1 2 N } ij back the state coefficient ψi1,i2,...,iN . Each matrix Aij individual components of the state: the two bound pho- has dimension Dj 1 Dj known as the bond dimension tons propagate in a similar manner to the two-photon and finite edge boundary− × conditions are assumed by im- bound state in Fig.2(b), while the extended photon prop- posing D1 = 1 and DN = 1. The bond dimension reflects agates like a single photon. This separation of the prop- the entanglement entropy. For instance, if Dj = 1 for all agation of the bound and extended parts of this state j, the matrices A are scalars and the state (11) reduces can be shown explicitly in the long pulse limit σΓ 1 ij to a product state with no entanglement. For arbitrary (see SM). This is significant because it implies that in or- states the bond dimension grows exponentially with the der to understand the pulse evolution one does not have number of particles. The advantage of the MPS ansatz to understand the behaviour of all the different string is that, in many physical scenarios as the one consid- combinations S. Rather, one can focus on simply un- ered here, the entanglement grows slowly with the sys- derstanding the behaviour of the bound states. On the tem size allowing an efficient description of the state in other hand, for short pulses this separation is not com- terms of a smaller bond dimension [31]. An important pletely true because one needs to include the effect of figure of merit of the efficiency of the MPS ansatz is given interactions between the components, see Sec.V. by Dmax, the maximum bond dimension, that is needed to faithfully represent the system during the entire time evolution (see SM for more information). We will make A. Evolution of bound states use of this quantity in Sec.IV to quantify the amount of many-body correlations present in the system. To better quantify the difference in propagation observed above, let us compute the pulse evolution in the center-of-mass coordinate. This can be done by using the III. MANY-BODY PULSE PROPAGATION form of the n-photon bound state and its transmission coefficient given in Eqs. (8) and (9). Within the n-photon manifold the projection of an input Gaussian state on the We are interested in studying multi-photon propaga- 2 2 (E nk0) σ /(2n) tion through the chirally coupled array. Here, we consider bound state is n BE in = cne− − , where cn h | i coherent and Fock input states with mode creation oper- is a constant in E. Here it is convenient to use Jacobi co- Pn J Pi R ordinates xc = xj/n, x = xj/i xi+1, where atora în† = dx (x)â†(x), and we evaluate the transmit- j i j=1 − ted field with theE two methods described in the previous i 1, 2, . . . , n 1 . The resulting bound-state contri- ∈ { − } section. In particular, for the exact solution we com- bution is then Z pute the transmitted photon state using the eigenstates cn n Γ g(xJ ) out n,bound = d x aˆ†(x) 0 e− 2 for up to three photons while for higher excitations we | i √n! | i make use of the MPS ansatz. In Fig.2 we consider the (12) Z 2 2 N iExc E σ /(2n) propagation of a Gaussian photonic mode with ampli- dE t e − , 2 2 E,n ik0x x /(2σ ) 1/4 × tude (x) = e − /(√σπ ) where, throughout E J P the manuscript, resonant pulses are considered k0 = 0. where g(x ) is the exponent x x written in Ja- i

By taking higher-order derivatives we also compute We have shown that the hallmark of many-body pho- 2 2 2 2 ton propagation through an ensemble of quantum emit- the pulse broadening b(k0) = 32k0Γ/[n(4k0 + n Γ ) ], which is zero on resonance. The− third-order pulse distor- ters in chiral wQED is the number-dependent velocity tion term on resonance is d = 32/(n6Γ3). The pulse of the photon bound states. Here we analyse the influ- distortion is thus drastically reduced− for higher-order ence of imperfections such as losses, imperfect chirality, bound states. This indicates that many-photon bound and inhomogeneous broadening on this propagation. We states suffer negligible pulse distortion while propagat- first consider the influence of losses, where each emit- ing through the array of nonlinear and dispersive atoms. ter couples to an additional decay channel out of the waveguide with a rate Γ0. It is possible to obtain an an- alytic criterion for when these losses can be neglected. In order to verify that indeed the physics of the bound The form of the transmission coefficient of the bound states dominates the wave-packet evolution we also com- state can be obtained in the presence of loss by mapping pute the evolution of a coherent input pulse as shown in the total energy to a complex energy using the replace- Figure2(d). Here, the pulse width Γ σ = 3√2 is the same ment E E + inΓ /2. The reduction in probability of as in Figs.2(a)-(c), while the average photon number in 0 the output→ state then implies that the state can undergo the pulse isn ¯ = 0.5. We compute the output both by one or more quantum jumps. If the probability remains truncating the coherent state to three photons, and solv- close to unity the output state is only weakly affected. ing exactly using Eq. (10), or by solving Eqs. (2) and(3) Defining the efficiency or β-factor as β = Γ/Γ , where with the MPS algorithm. The evolution of the bound tot Γ = Γ + Γ , the transmission coefficient (9) in the states is seen in the position of the peaks of the power tot 0 presence of loss is distribution aˆ†(x)â(x) as well as in the difference be- h i tween the power and the m-th order correlation functions E + inΓtot[1 β(1 + n)]/2 (m) m m tE,n = − . (14) G (x) = aˆ†(x) [â(x)] . E + inΓ [1 β(1 n)]/2 h i tot − − After scattering off all N emitters the magnitude of the The localized nature of the bound states in the relative N resulting state is tE,n . For small imperfections 1 β coordinates is shown in Fig.2(e) where we show the two- | N | − 2 (2) 1, this gives tE,n = 1 2N(1 β)/n + (1 β) , point second-order correlation function G (x1, x2) = where the notation| | (M)− indicates− term ofO order−M and aˆ†(x )â†(x )â(x )â(x ) . Here the photons tend to lo- O h 2 1 1 2 i higher. This means that a sufficient condition for neglect- calize around the diagonal at small values of the rela- ing losses is N(1 β)/n 1. This implies that losses tive coordinate x x , while they are delocalized about − 1 − 2 have a reduced influence on higher-order bound states. the center-of-mass coordinate (x1 + x2)/2. This reflects If this condition is not met, there is a sizable probability that the photons are tightly bound together in the bound that one or more of the photons in a photon bound state state, but the bound state itself is free to propagate. is lost. While exact analytical calculations become infeasible for If a photon is lost at one point along the ensemble n¯ 1, the validity of our arguments and the impor- the remaining photons propagate through the rest of the tance of the bound states can still be seen in MPS sim- atoms with a different effective group velocity and disper- ulations. For example, in Fig.2(f), we calculate the sion. In addition to a reduced ampltitude, the fact that transmitted power for an input coherent staten ¯ = 8 the remaining transmitted photons have effectively prop- and N = 60. Here, for this particular system length, agated with a mix of velocities and dispersions causes the the photon-number-dependent Wigner delay clearly man- peaks associated with the different bound states in Fig.2 ifests itself as separate peaks for up to six-photon bound to broaden and eventually overlap (see SM). states. In addition to photon loss, we note that a non-zero value of Γ0 = 0 also affects the delay τn. This is also 6 3 2 The low-distortion propagation of the bound states computed analytically, τn = 4β / Γ[β(2 + β(n 1)) can be intuitively explained by returning to the sim- 1] = 4/(Γn2) 4(1 β)/(Γn2) +{ ((1 β)2).− Values− } − − O − ple schematic shown in Fig.1(b). When a multi-photon where τn diverges occur when tE,n approaches zero, i.e., bound state propagates trough the atomic array one of no light is transmitted. | | the photons in the wave packet can be absorbed and re- In the limit of large losses the photon bound states suf- emitted by the atom. This process occurs on a time scale fer exponential damping. For a steady-state input field, ruled by the inverse of the photon-number dependent this was considered for up to two photons in Ref. [19] stimulated emission rate that coincides with the bound- where it was shown that the output shows strong pho- state packet width, ∆tn 1/(Γn), allowing the pulse ton bunching. This bunching, however arises from the to preserve its shape. This∼ continuous absorption and extended states and does not reflect the bound-state dy- re-emission of photons during the bound-state propaga- namics. Even with finite pulse durations it is likely that tion leads to a time delay of one out of n photons by an the bound-state dynamics will be hard to discern in this amount 4/(Γn), leading to the group delay in Eq. (13). limit. 8

The introduction of imperfect chirality also influences width. Unlike the previous sections where we selected the nature of pulse propagation through the ensemble. an input pulse and propagated it through the medium, This occurs when, in addition to coupling to the forward- here we are simply considering a linear combination of propagating mode, each atom also couples to the back- bound eigenstates and compute observables for this state. ward propagating mode with a rate ΓL. The influence A localized function cn(E) ensures that the bound state of this cannot be considered analytically in a straightfor- ansatz is localized in the center-of-mass coordinate. ward manner. We have performed numerical MPS calcu- Such a state can be probed by measuring either the lations (see SM) for N = 20 emitters. When ΓL = 0.05Γ, field ψ aˆ(x) ψ or the m-th normally ordered observ- the shape of the output state remains qualitatively unable,h which| we| i consider in the center-of-mass coordinate m m changed. However, when ΓL = 0.2Γ the output state is ψ (â†(x)) (â(x)) ψ . We compute these observables completely distorted. These computations indicate that hin| the limit where| thei average photon number is large the effect observed in the ideal chiral case is robust to n¯ 1, the pulses are spectrally broad σ 1/Γ and imperfect chirality, provided that the imperfections are the order of the correlation function is much less than not too large. the photon number m n. These give (see SM for full We also consider the influence of inhomogeneous calculations), broadening in the two-level systems. The two-level atoms n¯√Γ n¯Γx are considered to have a normally distributed resonance ψ aˆ(x) ψ = sech , (17) frequency with dimensionless standard deviation ς/Γ. h | | i 2 2 This affects the transmission coefficient of the bound " #2m m m n¯√Γ n¯Γx states, which most notably affects the pulse delay. An ψ (â(x)†) (â(x)) ψ = sech , expression for the mean pulse delay can be computed h | | i 2 2 analytically, and for small broadening, gives to leading (18) order wheren ¯ is the average number of photons in the pulse. 4ς2 ς4 These observables reproduce the fundamental soliton so- τ = τ 1 + . (15) h ni n − n2Γ2 O Γ4 lution of SIT [33, 34, 43] given in Eq. (6). Self-induced transparency is thus the classical limit of the photon The reduction in the delay imparted by each emitter bound state when the photon number becomes large, or therefore scales quadratically in ς/(nΓ). This therefore conversely photon bound states are simply the quantum has a reduced impact on higher-order bound states pro- limit of a soliton, a quantum soliton. Just like the SIT vided that the broadening is limited Γ. We note that solitons, the integrated Rabi frequency of the pulse is . R the influence of inhomogeneous broadening can be com- Ω = 2√Γ dx aˆ(x) = 2π, i.e., the intense pulse of light h i pensated by introducing more emitters. rapidly excites and de-excites the emitters resulting in a 2π Rabi oscillation. We note that, the expressions in Eqs. (17)-(18) scale with Γ, which is the coupling to the IV. CONNECTION TO THE SIT SOLITON one-dimensional continuum. These expressions are therefore unchanged in the presence of coupling to an exter- In the previous sections we have shown how the Hamil- nal reservoir. This implies that, provided the ensemble tonian (1) leads to the SIT solitonic solutions in the does not act like a Bragg mirror, SIT is expected to oc- mean-field limit, and that the full quantum mechani- cur even in the presence of backscattering in accordance cal treatment of this Hamiltonian predicts correlation- with mean-field results. ordered photon propagation. In this section we aim to bridge the gap between these two regimes: first we show that indeed the many-body theory reduces to the mean- A. Beyond mean-field theory field result in the limit of large photon number. Secondly, we derive the quantum corrections to the mean-field re- As a lowest-order approximation, the variation in pho- sults which become relevant when both the number of ton number n making up the pulse (16) can be ig- photons n and the number of emitters N are large. Fi- nored, and the state will simply propagate at a re- nally, we push the numerical simulations to the many- duced speed dictated by the mean photon number. This photon limit to verify the analytical predictions. leads to the SIT soliton prediction, that simply maps Let us consider a wave packet composed of a linear x x + 4N/(¯n2Γ). However, for a coherent state, the → combination of many-body bound states. This is ex- uncertainty in the photon number scales as ∆n √n¯, ∼ pressed with the ansatz which leads to a gradual broadening of the pulse due to the different photon number components accumulating X Z ψ = dEc (E) B . (16) different time delays. This can become significant for a | i n | Ein sufficiently large number of emitters N. This broadening n is not captured by the mean-field theory. The breakdown We later show that this is the expected form of the state of the mean-field theory therefore occurs when the differ- for a high-power coherent input with a large spectral ence in the delay of then ¯ and then ¯ + ∆n photon bound 9

FIG. 3. (a) Output power for N = 20 atoms for coherent input pulses with a solitonic shape (see text) and different mean photon numbern ¯. The results are computed by directly solving the ME (3) (solid lines) with the MPS algorithm where we fixed Dmax = 150. In the second panel we include the results given in the many-body limit (19) (dashed lines), and in the third the mean-field solitonic ansatz (dotted line). We note that the x-axis is limited to focus on larger photon number components. −4 (b) Maximum bond dimensions Dmax of the MPS calculations required to obtain a tolerance less than 10 versus input power n¯ for different TLS number N. The blue shading highlights the regions for which panels (1)-(3) in (a) show the output power. states becomes on the order of the width of then ¯ photon simulation. In Fig.3(a) we plot the transmitted power of 3/2 n¯√Γ n¯Γx bound state. This gives ( n¯+∆n n)/∆tn N/n . input pulses with in(x) = sech for different T 3/2 − T ∼ 2 2 This means that when N/n & 1 the mean-field the- amplitude strengthE n ¯. This pulse shape is chosen such ory breaks down even for large input photon numbern ¯. that its electric field matches the SIT criterion. In the This provides the boundary between mean-field theory first box we see again the formation of the bound state and genuine quantum many-body dynamics. peaks which tend to reduce their time delay as the input Whenn ¯ 1 and N n3/2, one can consider a wave- & power is increased. For intermediate input pulses (sec- function composed of bound states with a number dis- ond box) the bound states with a large number of photons tribution given by a coherent state. In the limit where get more populated, and they accumulate toward a sin- mean-field theory breaks down, one must explicitly con- gle peak as the difference in delay times for large photon sider the Fock-state-dependent delay. In this limit the numbers become less distinguishable. In this regime the expression for the field and the power give (see SM), transmitted power starts to be well-described by Eq. (19). 2n For even higher input power (third box) the individual α 2 X α n√Γ nΓ 4N aˆ(x) = e−| | sech x + , bound states are no longer recognizable and a solitonic h i n! 2 2 n2Γ n pulse well-described by mean-field theory emerges. These 2n 2 2 results show how the SIT solitons emerge from a super- α 2 X α n Γ nΓ 4N aˆ†(x)â(x) = e−| | sech x + , position of photon bound states that, in the limit of few h i n! 4 2 n2Γ n photons, can be indeed interpreted as quantum solitons. (19) On the other hand it is important to emphasize the dif- with equivalent expressions for higher-order correlation ference in physical effects. While the SIT solitons can be functions. Here, α = √n¯ is the coherent field amplitude, fully described by a mean-field semi-classical treatment, which is assumed real. We note that similar expressions the formation of distinct bound-state peaks is character- exist for the bound state of the nonlinear Schrödinger ized by a highly-correlated state of light and represents equation with an attractive interaction [44, 45]. the break down of the mean-field solution due to quan- Equations (19) provide a simple description of the ob- tum effects. servables of a quantum many-body state of light. In Within the MPS ansatz, one natural way to character- order to investigate the full transition from the multi- ize the amount of correlations in the system is to allow the photon bound-state propagation to the formation of the maximum truncated bond dimension Dmax to vary, and th SIT solitons we make use again of the master equation to record the value Dmax(N, n¯) at which the truncation 10

20 20 20 0.03 state and an unbound photon. The collision between the bound state and the free photon is thus elastic and the 15 0.3 15 0.2 15 0.02 bound state is stable against external influence. 10 0.2 10 10 0.1 0.01 5 0.1 5 5 0 0 0 0 0 0 VI. OUTLOOK -50 0 50 -50 0 50 -50 0 50 Our theoretical and numerical predictions have shown FIG. 4. Evolution of the state |ini (see text) demonstrating that many-body photon bound state propagation can be the interaction between a two-photon bound state centered observed in chiral waveguide QED geometries with many at a2 = −10/Γ and a single√ photon centered at a1 = 15/Γ emitters and photons. While these predictions are in both with width σΓ = 3 2. The output observables (a) † (2) (3) the realm of quantum many-body physics, our work also haˆ (x)â(x)i, (b) G (x), and (c) G (x) are plotted versus predicts novel photon transport in the few-photon–few- the number of emitters N. emitter landscape. This is exemplified in Fig.5 where we consider coherent pulse propagation with average photon error exceeds some acceptable threshold value (see SM). numbern ¯ = 0.5 and N = 2 emitters. Figures5(a)-(b) In Fig.3(b), we show this quantity as a function of the show the the output power and correlation functions in solitonic input pulse strengthn ¯. We see that the break- the limit of ideal chiral coupling and no loss. Figure5(a) (2) down of the mean-field description occurs in the regime shows the two-time correlation function G (x1, x2) for where bound state formation occurs and the amount of an input pulse with width Γσ = 3√2. The width of this correlation in the system is high (large values of Dmax), input pulse is chosen such that, in the two-excitation sub- while in the limit of largen ¯ the bond dimension tends to space, it projects on both the two-photon bound state th shrink approaching the mean-field limit (Dmax = 1). subspace and the extended states with roughly equal probability. The distinct signatures of these two states can be seen: the bound state clearly propagates faster V. SOLITON INTERACTIONS and is seen as an antinode on the diagonal marked by the intersection of the dashed lines. The spread out tails So far we have characterized the propagation of light which propagate slower are the signature of the extended through ideal media and shown that it can be under- states. These are clearly not bunched in comparison to stood in terms of the photon bound states. To fully the bound state. Figure5(b) shows the equal-time corre- characterize and understand these objects it is impor- lation function for a narrower pulse width Γσ = √2. Here tant to also investigate their interactions and robustness a narrower pulsewidth is chosen so it dominantly projects to disturbances. To this end, we now make a preliminary on the two-photon bound state. The hallmark of the investigation of a scattering experiment between a single photon–photon interactions is observed in the difference photon and a two-photon bound state. We consider the between the power P (x) and the correlation functions R 3 (2) (3) input state in = C d x a†(x) 0 φ(x1, x2, x3) with C G (x), and G (x). Clearly, the left-most peak, cor- being a constant| i and | i responding to the single-photon component, undergoes

2 2 2 2 a larger time-delay than the bound states. The differ- (x1 a1) /(2σ ) (x2+x3 a2) /(4σ ) φ(x1, x2, x3) =e− − e− − ence in time-delay between two- and three-photon bound (20) Γ/2 x3 x2 states is also visible in the slight difference between the e− | − |+ , × ↔ peak centers of G(2)(x), and G(3)(x). We note that it is i.e., a product state composed of a two-photon bound also possible to observe a difference between P (x) and (2) state and a single photon which are centered at a2 and a1 G (x) for a single quantum emitter. respectively with a2 < a1. Figure4 shows the evolution We have also investigated the robustness of this effect of the pulse delay for this state as it propagates through when imperfections are introduced. Figures5(c)-(e) con- the ensemble, i.e. for different N. As in previous figures, sider additional coupling to the left-propagating mode at a frame comoving with the pulse in the absence of inter- a rate ΓL = 0.1Γ for different emitter spacings kd, where actions (i.e. N = 0) is assumed. Here since a2 < a1 and k is the propagation wavenumber, and d is the distance the Wigner delay is larger for single photons τ1 > τ2, the between each of the emitters. Note that unlike the fully bound-state photons catch up to and overtake the single chiral regime, when ΓL = 0 the distance between emit- photon. In this process the photons interact when the ters influences the dynamics6 of the system. Although two parts overlap. The interaction causes a change in the output field is slightly reduced in all three cases, the the Wigner delays which is seen as kinks in the lines in difference in the shape of the power and the correlation Figs.4(a)-(b). The region of interaction is highlighted functions is preserved as is the difference in the peak by the third-order correlations shown in Fig.4(c). After positions. Finally, Fig.5(f) considers ideal chiral cou- the interaction has ended (N & 10) the lines in Fig.4 pling, but with each emitter coupling to a loss reservoir (a) continue with the same slopes as before the inter- at a rate Γ0 = 0.1Γ. Here, the single-photon component action, signifying that there is still a two-photon bound suffers the largest loss, while the bound states suffer a 11

(a) -3 (b) (c) 10 x10 0.1 0.08

0 5 2 photon BS 0.05 0.04 -10

-20 0 0 0 -20 -10 0 10 -20 -10 0 10 -20 -10 100 (d) (e) 0.08 0.08 (f) 0.08

0.04 0.04 0.04

0 0 0 -20 -10 100 -20 -10 100 -20 -10 100

FIG. 5. Propagation of a coherent pulse with average√ photon numbern ¯ = 0.5 through N = 2 emitters. (a) Two-time (2) correlation function G (x1, x2) for pulse width Γσ = 3 2 for ideal chiral coupling computed using the three-photon theory. Dashed lines show N times the Wigner delay of the two-photon bound√ state τ2. (b) Power P (x) and second- and third-order correlation functions G(2)(x) and G(3)(x) for pulsewidth Γσ = 2 with ideal coupling. Power and correlations when also coupling to a backward mode with rate ΓL = 0.1Γ where the emitters are separated by (c) kd = π, (d) kd = π/2, and (e) kd = π/4. (f) Power and correlation functions for unidirectional coupling, but with each emitter also coupled to a loss mode with Γ0 = 0.1Γ. Panels (b)-(f) are computed using a full numerical treatment. reduced loss. The introduction of the the loss does not ters [20]. The effects we predict here are thus widely ap- spoil the difference in the shape of the power and the plicable and can be observed in many different physical correlation functions. systems spanning vastly different energy scales. The results here demonstrate that the propagation of Throughout this manuscript we have focused on un- few-photon bound states can be observed with as few as derstanding the fundamental physics of photonic bound two quantum emitters chirally coupled to a reservoir and state propagation. In addition to its fundamental inter- is robust to imperfections. This means that the phenom- est, the dynamics of this system is highly interesting from ena shown here can be realized by several platforms cur- an applied perspective. As a particular example we note rently under investigation. Optical quantum dots have that, for the output state shown in Fig.2(a)-(d), if one demonstrated chiral coupling between a single emitter selects a temporal window centered at Γx0 = N, a state and a waveguide [46, 47], while coupling between two most likely containing either zero or two photons− is pro- diamond impurity centers and a photonic nanostructure duced. Launching the output on a beamsplitter and con- has been achieved [48]. At microwave frequencies, circuit ditioning on the detection of a photon will thus produce QED platforms can also achieve strong coupling between a single-photon Fock state. For a small amplitude initial a superconducting qubit and a transmission line [8] and coherent state, the main contribution to this process will a scheme for unidirecitonal coupling has been proposed arise from the two-photon component of the incoming [49]. Multiple qubits have also been coupled to a single field. Since this is in a pure state, the outgoing single mode or propagating modes [9, 50–53]. Finally, gasses of photon will also be in a pure state, although the tempo- Rydberg atoms under the conditions of electromagnet- ral mode function will depend on the detection time. A ically induced transparency also exhibit strong nonlin- detailed investigation of using photon bound-state prop- earities [14] and possess bound eigenstates [54]. Under agation as a photon source is beyond the scope of this certain limits they can also be mapped on to a nonlinear work. We can however estimate that for the parame- Schrödingerequation with an attractive interaction [18]. ters in Fig.2(a)-(d), which are in no way optimized for Such a Hamiltonian possesses bound eigenstates [36, 45] the application, choosing a window with width Γxw = 8 and therefore sufficiently long samples of Rydberg gasses centered on Γx = N leads to a probability of obtain- 0 − can also potentially exhibit the bound state propagation ing two photons P2 = 0.061 and an error probability of 3 shown here. Alternatively, ensembles can be engineered obtaining one or three photons P1 + P3 = 1.5 10− . using Rydberg blockade to mimic chirally coupled emit- The efficiency and error of the source appear promising× 12 as they should be compared with the those of the input region of maximal quantum correlations, until finally re- coherent state which are P2 = 0.076 and P1 + P3 = 0.32 sulting in a state with weaker correlations. This work respectively. therefore paves the way for observing many-body quantum states of light in waveguide QED.

VII. CONCLUSION VIII. ACKNOWLEDGEMENTS

Our results show that chiral wQED platforms provide The authors would like to thank Mantas Cepulkovskis,ˇ a highly nonlinear medium suited for exploring nonlinear Peter Lodahl, Sebastian Hofferberth, Hanna le Jeannic, optics at the quantum level. From the most elementary Johannes Bjerlin, L. Henriet, and J. Douglas for fruitful light-matter interaction — the interaction between pho- discussions. S.M. and K.H. acknowledge funding from tons and two-level systems — emerges correlated pho- DFG through CRC 1227 DQ-mat, projects A05 and A06, tonic states. These bound states are truly distinct phys- and “Niedersächsisches Vorab” through the “Quantum- ical objects with their own dispersion relation and are and Nano-Metrology (QUANOMET)”. G.C. and D.E.C. stable against external influence. Our work provides a acknowledge support from ERC Starting Grant FOQAL, clear recipe for how these features can potentially be MINECO Severo Ochoa Grant SEV-2015-0522, CERCA observed in experiments. In the limit where the num- Programme/Generalitat de Catalunya, FundacióPrivada ber of photons is high, the bound states approach the Cellex, Fundació Mir-Puig, Fundación Ramón Areces well-known soliton solution of SIT. Our full quantum de- Project CODEC, European Quantum Flagship Project scription on the other hand covers the entire spectrum QIA, QuantumCAT (funded within the framework of from few-photon quantum propagation to genuine quan- the ERDF Operational Program of Catalonia), and Plan tum many-body (atom and photon) phenomena, and ul- Nacional Grant ALIQS, funded by MCIU, AEI, and timately the quantum-to-classical transition. In partic- FEDER. A.S.S. acknowledges support from the Danish ular, the analysis highlights how the mean-field solution National Research Foundation (Center of Excellence Hy- with weak quantum correlations breaks down through a Q).

[1] D. E. Chang, V. Vuletić,and M. D. Lukin, Quantum space, Phys. Rev. Lett. 108, 263601 (2012). nonlinear optics — photon by photon, Nat. Photonics 8, [9] M. Mirhosseini, E. Kim, X. Zhang, A. Sipahigil, P. B. 685 (2014). Dieterle, A. J. Keller, A. Asenjo-Garcia, D. E. Chang, [2] V. Paulisch, M. Perarnau-Llobet, A. González-Tudela, and O. Painter, Cavity quantum electrodynamics with and J. I. Cirac, Quantum metrology with one- atom-like mirrors, Nature 569, 692 (2019). dimensional superradiant photonic states, Phys. Rev. A [10] K. M. Birnbaum, A. Boca, R. Miller, A. D. Boozer, T. E. 99, 043807 (2019). Northup, and H. J. Kimble, Photon blockade in an optical [3] D. E. Chang, J. S. Douglas, A. González-Tudela, C.-L. cavity with one trapped atom, Nature 436, 87 (2005). Hung, and H. J. Kimble, Colloquium: Quantum matter [11] A. Faraon, I. Fushman, D. Englund, N. Stoltz, P. Petroff, built from nanoscopic lattices of atoms and photons, Rev. and J. Vuˇckovic, Coherent generation of non-classical Mod. Phys. 90, 031002 (2018). light on a chip via photon-induced tunnelling and block- [4] M. Arcari, I. Söllner,A. Javadi, S. L. Hansen, S. Mah- ade, Nat. Phys. 4, 859 (2008). moodian, J. Liu, H. Thyrrestrup, E. H. Lee, J. D. Song, [12] A. Reinhard, T. Volz, M. Winger, A. Badolato, K. J. S. Stobbe, and P. Lodahl, Near-unity coupling efficiency Hennessy, E. L. Hu, and A. Imamo˘glu,Strongly corre- of a quantum emitter to a photonic crystal waveguide, lated photons on a chip, Nat. Photonics 6, 93 (2012). Phys. Rev. Lett. 113, 093603 (2014). [13] T. Peyronel, O. Firstenberg, Q.-Y. Liang, S. Hofferberth, [5] A. Goban, C.-L. Hung, S.-P. Yu, J. D. Hood, J. A. Muniz, A. V. Gorshkov, T. Pohl, M. D. Lukin, and V. Vuletić, J. H. Lee, M. J. Martin, A. C. McClung, K. S. Choi, Quantum nonlinear optics with single photons enabled D. E. Chang, O. Painter, and H. J. Kimble, Atomlight by strongly interacting atoms, Nature 488, 57 (2012). interactions in photonic crystals, Nat. Commun. 5, 3808 [14] O. Firstenberg, T. Peyronel, Q.-Y. Liang, A. V. Gor- (2014). shkov, M. D. Lukin, and V. Vuletić,Attractive photons [6] P. Lodahl, S. Mahmoodian, and S. Stobbe, Interfacing in a quantum nonlinear medium, Nature 502, 71 (2013). single photons and single quantum dots with photonic [15] A. Javadi, I. Söllner,M. Arcari, S. Lindskov Hansen, nanostructures, Rev. Mod. Phys. 87, 347 (2015). L. Midolo, S. Mahmoodian, G. Kirˇsansk˙e, T. Pregno- [7] L. S. Bishop, J. M. Chow, J. Koch, A. A. Houck, lato, E. H. Lee, J. D. Song, S. Stobbe, and P. Lodahl, M. H. Devoret, E. Thuneberg, S. M. Girvin, and R. J. Single-photon non-linear optics with a quantum dot in a Schoelkopf, Nonlinear response of the vacuum rabi reso- waveguide, Nat. Commun. 6, 8655 (2015). nance, Nature Physics 5, 105 (2009). [16] C. Hamsen, K. N. Tolazzi, T. Wilk, and G. Rempe, Two- [8] I.-C. Hoi, T. Palomaki, J. Lindkvist, G. Johansson, photon blockade in an atom-driven cavity qed system, P. Delsing, and C. M. Wilson, Generation of nonclas- Phys. Rev. Lett. 118, 133604 (2017). sical microwave states using an artificial atom in 1d open 13

[17] L. De Santis, C. Antón,B. Reznychenko, N. Somaschi, [35] H. A. Bethe, On the theory of metals, Z. Phys. 71, 205 G. Coppola, J. Senellart, C. Gómez, A. Lemaˆıtre, (1931). I. Sagnes, A. G. White, L. Lanco, A. Aufféves, and [36] H. B. Thacker, Exact integrability in quantum field the- P. Senellart, A solid-state single-photon filter, Nat. Nan- ory and statistical systems, Rev. Mod. Phys. 53, 253 otechnol. 12, 663 (2017). (1981). [18] Q.-Y. Liang, A. V. Venkatramani, S. H. Cantu, T. L. [37] V. I. Rupasov and V. I. Yudson, Exact dicke superradi- Nicholson, M. J. Gullans, A. V. Gorshkov, J. D. Thomp- ance theory: Bethe wavefunctions in the discrete atom son, C. Chin, M. D. Lukin, and V. Vuletić,Observation model, Sov. Phys. JETP 59, 478 (1984). of three-photon bound states in a quantum nonlinear [38] J.-T. Shen and S. Fan, Strongly correlated multiparticle medium, Science 359, 783 (2018). transport in one dimension through a quantum impurity, [19] S. Mahmoodian, M. Cepulkovskis,ˇ S. Das, P. Lodahl, Phys. Rev. A 76, 062709 (2007). K. Hammerer, and A. S. Sørensen, Strongly correlated [39] H. Zheng, J. Gauthier, and H. U. Baranger, Cavity-free photon transport in waveguide quantum electrodynam- photon blockade induced by many-body bound states, ics with weakly coupled emitters, Phys. Rev. Lett. 121, Phys. Rev. Lett. 107, 223601 (2011). 143601 (2018). [40] Y. Shen and J.-T. Shen, Photonic-fock-state scattering in [20] N. Stiesdal, J. Kumlin, K. Kleinbeck, P. Lunt, C. Braun, a waveguide-qed system and their correlation functions, A. Paris-Mandoki, C. Tresp, H. P. Büchler, and S. Hoffer- Phys. Rev. A 92, 033803 (2015). berth, Observation of three-body correlations for photons [41] V. I. Yudson, Dynamics of integrable quantum systems, coupled to a rydberg superatom, Phys. Rev. Lett. 121, Sov. Phys. JETP 88, 1043 (1985). 103601 (2018). [42] P. Bienias, J. Douglas, A. Paris-Mandoki, P. Titum, [21] E. P. Wigner, Lower limit for the energy derivative of the I. Mirgorodskiy, C. Tresp, E. Zeuthen, M. J. Gullans, scattering phase shift, Phys. Rev. 98, 145 (1955). M. Manzoni, S. Hofferberth, D. Chang, and A. Gorshkov, [22] R. Bourgain, J. Pellegrino, S. Jennewein, Y. R. P. Sortais, Photon propagation through dissipative rydberg media and A. Browaeys, Direct measurement of the wigner time at large input rates, arXiv:1807.07586 (2018). delay for the scattering of light by a single atom, Opt. [43] A. V. Yulin, I. V. Iorsh, and I. A. Shelykh, Resonant Lett. 38, 1963 (2013). interaction of slow light solitons and dispersive waves [23] G. Nikoghosyan and M. Fleischhauer, Photon-number se- in nonlinear chiral photonic waveguide, New Journal of lective group delay in cavity induced transparency, Phys. Physics 20, 053065 (2018). Rev. Lett. 105, 013601 (2010). [44] Y. Lai and H. A. Haus, Quantum theory of solitons in [24] H. Tanji-Suzuki, W. Chen, R. Landig, J. Simon, and optical fibers. i. time-dependent hartree approximation, Vuletić, Vacuum-induced transparency, Science 333, Phys. Rev. A 40, 844 (1989). 1266 (2011). [45] Y. Lai and H. A. Haus, Quantum theory of solitons in [25] P. Lodahl, S. Mahmoodian, S. Stobbe, A. Rauschenbeu- optical fibers. ii. exact solution, Phys. Rev. A 40, 854 tel, P. Schneeweiss, J. Volz, H. Pichler, and P. Zoller, (1989). Chiral quantum optics, Nature 541, 473 (2017). [46] I. Söllner, S. Mahmoodian, S. L. Hansen, L. Midolo, [26] T. Caneva, M. T. Manzoni, T. Shi, J. S. Douglas, J. I. G. Kirsanske, T. Pregnolato, H. El-Ella, E. H. Lee, J. D. Cirac, and D. E. Chang, Quantum dynamics of prop- Song, S. Stobbe, and P. Lodahl, Deterministic photon– agating photons with strong interactions: a generalized emitter coupling in chiral photonic circuits, Nat. Nan- input-output formalism, New J. Phys. 17, 113001 (2015). otechnol. 10, 775 (2015). [27] M. T. Manzoni, D. E. Chang, and J. S. Douglas, Simu- [47] A. Javadi, D. Ding, M. H. Appel, S. Mahmoodian, M. C. lating quantum light propagation through atomic ensem- Löbl,I. Söllner,R. Schott, C. Papon, T. Pregnolato, bles using matrix product states, Nat. Commun. 8, 1743 S. Stobbe, L. Midolo, T. Schröder,A. D. Wieck, A. Lud- (2017). wig, R. J. Warburton, and P. Lodahl, Spinphoton inter- [28] K. Stannigel, P. Rabl, and P. Zoller, Driven-dissipative face and spin-controlled photon switching in a nanobeam preparation of entangled states in cascaded quantum- waveguide, Nat. Nanotechnol. 13, 398 (2018). optical networks, New J. Phys. 14, 063014 (2012). [48] A. Sipahigil, R. E. Evans, D. D. Sukachev, M. J. Bu- [29] H. Pichler, T. Ramos, A. J. Daley, and P. Zoller, Quan- rek, J. Borregaard, M. K. Bhaskar, C. T. Nguyen, J. L. tum optics of chiral spin networks, Phys. Rev. A 91, Pacheco, H. A. Atikian, C. Meuwly, R. M. Camacho, 042116 (2015). F. Jelezko, E. Bielejec, H. Park, M. Lonˇcar,and M. D. [30] F. Verstraete, V. Murg, and J. I. Cirac, Matrix prod- Lukin, An integrated diamond nanophotonics platform uct states, projected entangled pair states, and varia- for quantum-optical networks, Science 354, 847 (2016). tional renormalization group methods for quantum spin [49] P. O. Guimond, B. Vermersch, M. L. Juan, A. Sharafiev, systems, Adv. Phys. 143, 143 (2008). G. Kirchmair, and P. Zoller, A unidirectional on- [31] U. Schollwöck, The density-matrix renormalization group chip photonic interface for superconducting circuits, in the age of matrix product states, Ann. Phys. 326, 96 arXiv:1911.02460 (2019). (2011). [50] A. F. Van Loo, A. Fedorov, K. Lalumière,B. C. Sanders, [32] S. L. McCall and E. L. Hahn, Self-induced transparency A. Blais, and A. Wallraff, Photon-mediated interac- by pulsed coherent light, Phys. Rev. Lett. 18, 908 (1967). tions between distant artificial atoms, Science 342, 1494 [33] S. L. McCall and E. L. Hahn, Self-induced transparency, (2013). Phys. Rev. 183, 457 (1969). [51] P. Y. Wen, K. T. Lin, A. F. Kockum, B. Suri, H. Ian, J. C. [34] R. K. Bullough, P. J. Caudrey, J. C. Eilbeck, and J. D. Chen, S. Y. Mao, C. C. Chiu, P. Delsing, F. Nori, G. D, Gibbon, A general theory of self-induced transparency, Lin, and I. C. Hoi, Large collective lamb shift of two dis- Opto-electronics 6, 121 (1974). tant superconducting artificial atoms, arXiv:1904.12473 (2019). 14

[52] M. Fitzpatrick, N. M. Sundaresan, A. C. Y. Li, J. Koch, [55] H. Bateman, Tables of Integral Transforms, Vol. I and A. A. Houck, Observation of a dissipative phase tran- (McGaw-Hill Book Company, 1954) p. 30. sition in a one-dimensional circuit qed lattice, Phys. Rev. [56] D. E. Chang, L. Jiang, A. V. Gorshkov, and H. J. Kimble, X 7, 011016 (2017). Cavity qed with atomic mirrors, New J. Phys. 14, 063003 [53] N. M. Sundaresan, R. Lundgren, G. Zhu, A. V. Gorshkov, (2012). and A. A. Houck, Interacting qubit-photon bound states [57] A. Gonzalez-Tudela, D. Martin-Cano, E. Moreno, with superconducting circuits, Phys. Rev. X 9, 011021 L. Martin-Moreno, C. Tejedor, and F. J. Garcia-Vidal, (2019). Entanglement of two qubits mediated by one-dimensional [54] M. F. Maghrebi, M. J. Gullans, P. Bienias, S. Choi, plasmonic waveguides, Phys. Rev. Lett. 106, 020501 I. Martin, O. Firstenberg, M. D. Lukin, H. P. B¨uchler, (2011). and A. V. Gorshkov, Coulomb bound states of strongly interacting photons, Phys. Rev. Lett. 115, 123601 (2015). 15

Supplemental Material: Dynamics of many-body photon bound states in chiral waveguide QED

S1. SELF-INDUCED TRANSPARENCY

Here we derive the fundamental soliton solution of self-induced transparency. While this calculation is covered in detail in [34], we include it here for completeness. The calculations in the main text set vg = 1. Here we include the group velocity in the calculations to keep track of the units of each expression. The Hamiltonian (1) becomes

Z N ∂ p X + Hˆ = iv dx aˆ†(x) aˆ(x) + Γv σˆ−aˆ†(x ) +σ ˆ aˆ(x ) . (S1) − g ∂x g i i i i i=1

The two-level systems in SIT are modelled as a spin continuum. We do this via the mappingσ ˆ σˆ(x)/ν, where i → σˆi is any of the Pauli operators for the ith TLS, and ν is the linear density of TLSs. This mapping produces the commutation relation

[ˆσ (x), σˆ+(x0)] = σˆz(x) δ (x x0). (S2) − − − We note that the continuum spin operators now have units of linear density. Within the continuum limit the Hamil- tonian is Z Z ∂ p + Hˆ = iv dx aˆ†(x) aˆ(x) + Γv dx σˆ−(x)â†(x ) +σ ˆ (x)â(x ) . (S3) − g ∂x g i i Obtaining the equations of motion for the operators and taking their expectation values within a mean-field approximation gives the SIT equations ∂ ∂ p + a(x, t) = i Γvgσ (x, t), ∂t ∂x − − ∂ p σ (x, t) = i Γvgσz(x, t)a(x, t), (S4) ∂t − ∂ p σz(x, t) = 4 Γvg Im a(x, t)σ∗ (x, t) . ∂t − Since we are considering a resonant pulse, σ (x, t) is imaginary and we can thus set σ (x, t) = is(x, t). Examining the form of the last two equations of (S4), they− can both be satisfied by defining − ν s(x, t) = sin [2χ(x, t)], − 2 (S5) σ (x, t) = ν cos [2χ(x, t)], z − where Z t p χ(x, t) = Γvg dt0a(x, t0). (S6) −∞ Here s(x, t) and σ (x, t) are chosen such that in the limit of no field σ (x, t) ν, which is the desired physical limit. z p z → − Noting that a(x, t) = ∂χ/∂t/ Γvg, the first equation in (S4) becomes

∂2 ∂2 Γv ν + χ(x, t) = g sin [2χ(x, t)]. (S7) ∂t2 ∂x∂t 2

Since we are looking for soliton solutions, we look for a functional form that propagates without changing shape. Such functions therefore only depend on a single variable ξ = x V 0t, where V 0 is the propagation velocity of the soliton. One can then express (S7) as −

∂2 κ2 χ(ξ) = sin [2χ(ξ)], (S8) ∂ξ2 2 16 where Γv ν κ2 = g . (S9) V (v V ) 0 g − 0 Equation (S8) has the solution χ(ξ) = arcsin [tanh (κξ)]. This gives the solution

V 0κ a(x, t) = p sech [κ(x V 0t)]. (S10) Γvg − p R This solution fulﬁlls the criterion for the fundamental SIT soliton 2 Γvg dxa(x, t)/V 0 = 2π. An expression for the R 2 velocity can be obtained in terms of the average photon numbern ¯ = vg dxa(x, t) /V 0. This gives

2 n¯ vgΓ V 0 = 2 . (S11) n¯ Γ + 4vgν Substituting the velocity back into the expression for the ﬁeld gives

n¯√Γ n¯Γ x a(x, t) = sech t . (S12) 2√vg 2 V 0 −

In the main text we consider a frame comoving with the pulse in the absence of atoms, i.e., at velocity vg. Within 2 4vg ν 2 1 this frame the pulse has a velocity in the backwards direction V = vg V 0 = n¯2Γ (1 + 4vgν/n¯ Γ)− . The delay per emitter can be obtained by noting that the velocity can be expressed− in terms of the time it takes for the pulse to 1 1 2 pass an atom V 0 = ν− / (νvg)− + τn¯ . This gives an average delay per emitter of τn¯ = 4/(¯n Γ) which is equivalent to the delay of the n-photon bound state.

S2. ONE-, TWO- AND THREE-PHOTON TRANSPORT USING SCATTERING EIGENSTATES

In this section we detail how we compute the output photon states when N emitters are impinged upon by one, two, or three photons. We also combine these scattering calculations to consider scattering of a coherent state with up to three photons.

A. Single-Photon Transport

In the few-photon transport calculations we consider Gaussian input states. The normalized single-photon input state has the form

Z 2 2 1 ik0x x /(2σ ) in 1 = dx aˆ†(x) 0 1 e e− (S13) | i | i√σπ 4 with its Fourier representation

Z 2 2 √σ (k k0) σ /2 in = dk aˆ†(k) 0 e− − . (S14) | i1 | iπ1/4

The creation operatorsa ˆ†(k) are eigenstates of the single-photon scattering operator. After N scattering events, we N havea ˆ†(k) t aˆ†(k), with the transmission coeﬃcient → k k iΓ/2 t = − . (S15) k k + iΓ/2

Throughout we consider an ideal 1D reservoir without losses, and with vg = 1. The scattered state is thus

r Z Z 2 2 σ 1 N (k k0) σ /2 ikx out = dx aˆ†(x) 0 dk t e− − e . (S16) | i1 2π π1/4 | i k This is computed numerically. Most of the integrals that appear in the few-photon transport calculations are Fourier integrals, which can be eﬃciently computed using Fast Fourier Transforms. 17

B. Two-Photon Transport

The two-photon scattering matrix has two types of eigenstates, extended states WE,∆ and bound states BE [38]. The scattering matrix has the form | i | i Z Z ˆN 1 N N N S22 = dEd∆ t E +∆t E ∆ WE,∆ 2 WE,∆ + dE tE,2 BE 2 BE , (S17) 2 2 2 − | i h | | i h | where all integrals range from to . The two-photon scattering eigenstates can be obtained from Bethe’s ansatz in the main text (7). In a position-space−∞ ∞ representation they are 1 Z WE,∆ 2 = dx1dx2aˆ†(x1)ˆa†(x2) 0 WE,∆(xc, x) | i √2 | i (S18) 1 Z BE 2 = dx1dx2aˆ†(x1)ˆa†(x2) 0 BE(xc, x). | i √2 | i Here, √ 1 2 iExc WE,∆(xc, x) = e [2∆ cos (∆x) Γ sgn (x) sin (∆x)] , √4∆2 + Γ2 2π − r (S19) Γ Γ iExc x B (x , x) = e e− 2 | |, E c 4π and E 2iΓ t = − , (S20) E,2 E + 2iΓ

x1+x2 k p and xc = 2 , x = x1 x2, E = k + p is a two-photon detuning, and ∆ = −2 is a diﬀerence in photon energies, where k and p are photon− detunings. The input state for two photons is

Z 2 2 2 2 1 1 x /2σ x /2σ ik0(x1+x2) in 2 = dx1dx2 aˆ†(x1)ˆa†(x2) 0 e− 1 e− 2 e | i √2 | iσ√π (S21) Z 2 2 2 2 1 1 x /σ x /4σ 2ik0xc = dx1dx2 aˆ†(x1)ˆa†(x2) 0 e− c e− e . √2 | iσ√π Projecting this on the two-photon states gives

2 2 2 2 (E 2k0) σ /4 Γ σ /4 Γσ B in = √Γσe− − e Erfc (S22) 2h E| i2 2 and

2 2 h 2 2 i 4σ 1 (E 2k0) σ /4 ∆ σ 2 WE,∆ in 2 = e− − √π∆e− ΓDF (∆σ) , (S23) h | i π√2 √4∆2 + Γ2 − where Erfc = 1 Erf, where Erf is the error function, and DF is Dawson’s Function. The output state− is then written in terms of two integrals given by the bound-state and extended-state contributions 1 Z out 2 = dx1dx2 aˆ†(x1)ˆa†(x2) 0 [ψB(xc, x) + ψW (xc, x)] , (S24) | i √2 | i with r 2 2 Γ Z 2 2 Γ Γ σ /4 Γσ x N (E 2k0) σ /4 iExc ψ (x , x) = √Γσ e Erfc e− 2 | | dE t e− − e , (S25) B c 4π 2 E,2 and

Z 2 2 h 2 2 i σ N N (E 2k0) σ /4 iExc 1 ∆ σ E E √ ψW (xc, x) = 2 dEd∆ t +∆t ∆e− − e 2 2 π∆e− ΓDF (∆σ) π 2 2 − 4∆ + Γ − (S26) [2∆ cos (∆x) Γ sgn (x) sin (∆x)] . × − The integrals in these two functions are computed numerically. 18

C. Three-photon transport

Using the same rules rules we construct the three-photon scattering matrix for N emitters Z Z Z N 1 3 N 1 N N N Sˆ = d k t Wk Wk + dKdk t t H H + dE t B B , (S27) 33 3! k | i3h | 2 k K,2| K,ki3h K,k| E,3 | Ei3h E| where E 9iΓ/2 tE,3 = tE/3 iΓtE/3tE/3+iΓ = − . (S28) − E + 9iΓ/2 The eigenstates are 1 Z Wk 3 = dx1dx2dx3aˆ†(x1)â†(x2)â†(x3) 0 Wk,3(x1, x2, x3) | i √3! | i 1 Z HK,k 3 = dx1dx2dx3aˆ†(x1)â†(x2)â†(x3) 0 HK,k,3(x1, x2, x3) (S29) | i √3! | i 1 Z BE 3 = dx1dx2dx3aˆ†(x1)â†(x2)â†(x3) 0 BE,3(x1, x2, x3) | i √3! | i with the normalized real-space representations,    3 3  1 Y Y ikj xj Wk,3(x1, x2, x3) = q [km kn iΓ sgn (xn xm)] e + , Q3 − − − ↔ 3!(2π)3 [(k k )2 + Γ2] m

Γ iE(x1+x2+x3)/3 Γ/2( x1 x2 + x1 x3 + x2 x3 ) BE,3(x1, x2, x3) = e e− | − | | − | | − | . √3π The three-photon Gaussian input state is

Z 2 2 2 2 2 2 1 1 ik0(x1+x2+x3) x /(2σ ) x /(2σ ) x /(2σ ) in 3 = dx1dx2dx3aˆ†(x1)ˆa†(x2)ˆa†(x3) 0 e e− 1 e− 2 e− 3 . (S31) | i √3! | iπ3/4σ3/2 Both the projection of the input state on the scattering eigenstates and the subsequent integrals in (S27) are performed numerically. The total contribution of the bound state varies with the input pulse width. Figure S1 shows the overlap between the three-photon bound state and an input Gaussian versus pulsewidth. We note that the maximum overlap for the three-photon bound state occurs for shorter pulses than the two-photon bound state.

D. Output states and observables

We now have one-, two-, and three-photon output states which we express as Z out = dx aˆ†(x ) 0 ψ (x ), | i1 1 1 | i 1 1 1 Z out 2 = dx1dx2 aˆ†(x1)â†(x2) 0 ψ2(x1, x2), (S32) | i √2 | i 1 Z out 3 = dx1dx2dx3 aˆ†(x1)â†(x2)â†(x3) 0 ψ3(x1, x2, x3). | i √3! | i These are also used to compute the output for coherent states 2 3 α 2/2 α α αout = e−| | 0 + α out 1 + out 2 + out 3 + ... (S33) | i | i | i √2| i √3!| i 19

1.0 B in 2 |2h E| i| B in 2 |3h E| i| 0.8

0.6

0.4

0.2

0.0 1 0 1 10− 10 10 σΓ

FIG. S1. Overlap of a two- and three-photon Gaussian states with two- and three-photon bound states versus the width of the Gaussian state σΓ.

(2) We are interested in computing the normally ordered expectation values and correlations aˆ†(x)â(x) , G (xc, x) = (3) h i aˆ†(xc)â†(xc + x)â(xc + x)â(xc) , and G (xc, x, y) = aˆ†(xc)â†(xc + x)â†(xc + y)â(xc + y)â(xc + x)â(xc) . For a hcoherent state with up to three photons,i these give the expressionsh i Z 2 2 4 2 2 aˆ†(x)â(x) = α ψ (x) + α dx ψ (x, x ) ψ (x) h i | | | 1 | | | 1| 2 1 | − | 1 | Z Z 6 1 2 2 1 2 + α dx1dx2 ψ3(x, x1, x2) dx1 ψ2(x, x1) + ψ1(x) , | | 2 | | − | | 2| | (S34) Z G(2)(x , x) = α 4 ψ (x , x + x) 2 + α 6 dx ψ (x , x + x, x ) 2 ψ (x , x + x) 2 , c | | | 2 c c | | | 1| 3 c c 1 | − | 2 c c | G(3)(x , x, y) = α 6 ψ (x , x + x, x + y) 2. c | | | 3 c c c |

S3. SEPARABILITY OF THREE-PHOTON HYBRID STATES

The behaviour of the hybrid state in Fig.2(c) of the main text is signiﬁcant. Its shape indicates that the bound two-photon part of this state propagates with the same eﬀective velocity as the actual two-photon bound state, while the extended photon in the hybrid state evolves like a single photon. In this section we show that, for spectrally narrow input states, the evolution of the Hybrid state follows that of an independent two-photon bound state and an independent single photon. We consider the input state in given | i3 in (S31) that is on resonance k0 = 0. The hybrid state has the form given in (S30). The contribution of the hybrid state to the full three-photon output state is

1 Z out = dKdk tN tN H H in | ihybrid 2 K,2 k | K,ki3h K,k| i (S35) 1 Z = dx1dx2dx3ψhybrid(x1, x2, x3)â†(x1)â†(x2)â†(x3) 0 √3! | i

We now focus on the inner product HK,k in . Noting that the input state is separable and therefore already fully symmetric we can relabel the 3! termsh in the| Hybridi state to make them identical. We end up with

iΓ 3! Z 2 2 2 2 (x1+x2+x3)/(2σ ) 3 in HK,k 3 = q − 3/4 3/2 dx1dx2dx3 e− θ(x2 x1) h | i 2π 3Γ ( K k)2 + Γ2 ( K k)2 + 9Γ2 π σ − 2 − 4 2 − 4 (S36) K Γ K iΓ K iΓ i (x1+x2) (x2 x1) ikx3 k iΓ sgn (x x ) k + iΓ sgn (x x ) e 2 e− 2 − e . 2 − − 2 − 3 − 1 2 − 2 − 3 − 2 20

Using the two-body coordinates x = x + x /2 and x = x x allows writing the integral as c 1 2 1 − 2 Z (2x2+x2/2+x2)/(2σ2) K iΓ x I(K, k) = dx dx dx e− c 3 θ( x) k iΓ sgn x x c 3 − 2 − − 2 − 3 − c − 2 K iΓ x iKx Γ x ikx k + iΓ sgn x x + e c e 2 e 3 2 − 2 − 3 − c 2 (S37) Z 2 2 2 Z Γ 2 2 (2x +x )/(2σ ) iKxc ikx3 x x /(4σ ) K iΓ x = dx dx e− c 3 e e dx e 2 e− θ( x) k iΓ sgn x x c 3 − 2 − − 2 − 3 − c − 2 K iΓ x k + iΓ sgn x x + . 2 − 2 − 3 − c 2 Unfortunately all the integrals cannot be computed exactly. We now perform the integrals over the x coordinate by considering Γσ 1. The exponentially decaying functions decay with Γ, while the Gaussian functions have a width Γ/2x x2/(2σ2) Γ/2x 1/σ. When taking Γσ 1, we can make the approximation e e− e for x < 0. We can then write the∼ integral over x as ∼

Z " 2 2 2 Γ x K Γ Γ x K x dx e 2 θ( x) k + sgn a + iΓ k sgn + a − 2 − 4 − 2 2 − 2 − 2 (S38) K x Γ2 x x x iΓ k sgn a sgn + a + Γ2 sgn a sgn + a , − 2 − 2 − − 2 2 2 − 2 where a = x x . We then compute the integrals 3 − c Z 2 I = dx θ( x)eΓ/2x = 0 − Γ Z Γ/2x x 2 aΓ I = dx θ( x)e sgn + a = (1 2e− )θ(a) θ( a) 1 − 2 Γ − − − (S39) Z x 2 I = dx θ( x)eΓ/2x sgn a = (1 2eaΓ)θ( a) θ(a) 2 − 2 − Γ − − − Z Γ/2x x x 2 Γ a I = dx θ( x)e sgn a sgn + a = 2e− | | 1 . 3 − 2 − 2 Γ − The integral in (S37) then becomes

( 2 Z 2 2 2 2 2 (2x +x )/(2σ ) iKxc ikx3 K Γ 2 Γ a I(K, k) = dx dx e− c 3 e e k + + Γ e− | | Γ c 3 2 − 4 (S40) K Γ a 2 Γ a 2iΓ k sgn (a)(1 e− | |) + Γ (2e− | | 1) , − 2 − − −

Γ a Γ a where we have used I + I = 4/Γe− | | and I I = 4/Γ sgn (a)(1 e− | |). Unfortunately we are unable to 1 2 − 1 − 2 − compute I(K, k) exactly. Instead we compute the projection of this function on the K-axis, i.e. I˜(K) = R dkI(K, k). We show that this function is much narrower than the linewidth. Consequently, conservation of energy E = K + k will imply that its projection on the k-axis I¯(k) = R dKI(K, k) is also narrower than the linewidth. After computing the integrals and some manipulation, in the limit Γσ 1, we obtain 2 2 2 K2σ2/4 K σ 4 8πΓ 2K K2σ2/4 Kσ I(K) = π√πe− 3Γσ + + + 4πK √πσe− Erﬁ , (S41) Γσ − Γσ K2 + Γ2 K2 + Γ2 − 2 where Erﬁ is the imaginary error function (which is actually real-valued). In the limit K 0 and Γσ 1 the function is → 4 K2σ2/4 2 2 I(K) = π√π 3Γσ + e− + O(K σ ). (S42) − Γσ

x2/4 On the other hand, in the limit Kσ 1 and Γσ 1, using the large argument approximation e− Erﬁ (x/2) = 2/(√πx) + O(1/x3) we have 2 2 K2σ2/4 K σ 4 1 I(K) = π√πe− 3Γσ + + O . (S43) Γσ − Γσ K2σ2 21

The Gaussian part becomes negligible here and the function therefore scales as 1/K2σ2 which is much less than one when K Γ. Within the linewidth of the emitter, the function is largest when Kσ 1, where it is well-approximated by a Gaussian→ function. Since the tails of this function scales as 1/K2σ2, they only serve to slightly broaden the Gaussian function. This implies that (S42) is a reasonable approximation for the function. This means that the Gaussian part of the function is dominant over the other parts. This part of the function I(K, k) comes from the integrals I and I and can actually be computed analytically when Γσ 1. It is given by 0 3 " 2 2 # 2 K 3Γ 2 K2σ2/4 k2σ2/2 I(K, k) k √2πσ e− e− ∼ Γ 2 − − 4 (S44) 3Γ 2 K2σ2/4 k2σ2/2 √2πσ e− e− . ∼ 2 This then gives

iΓ 3! 3Γ 2 2 2 2 √ 2 K σ /4 k σ /2 in HK,k q − 3/4 3/2 2πσ e− e− . (S45) h | i ∼ 2π 3Γ ( K k)2 + Γ2 ( K k)2 + 9Γ2 π σ 2 2 − 4 2 − 4 This projection has two very useful features. The Gaussian part is separable in K and k and this part is highly localized about the origin. In fact the function multiplying the Gaussians can be approximated by a constant in K and k when Γσ 1. Substituting the projection H in into the (S35) gives the hybrid wavefunction h K,k| i Z 2 1 N N Γ 3! 3Γ 2 K2σ2/4 k2σ2/2 ψhybrid(x1, x2, x3) = dKdk t t 2 2 √2πσ e− e− 2 K,2 k 12π2Γ ( K k)2 + Γ ( K k)2 + 9Γ π3/4σ3/2 2 2 − 4 2 − 4 K Γ K iΓ K iΓ i (x1+x2) (x2 x1) ikx3 θ(x x ) k iΓ sgn (x x ) k + iΓ sgn (x x ) e 2 e− 2 − e + . 2 − 1 2 − − 2 − 3 − 1 2 − 2 − 3 − 2 ↔ 2 2 Z 4Γ σ √2π 1 1 N N K2σ2/4 k2σ2/2 dKdk t t e− e− ∼ Γ2 π3/4σ3/2 2 K,2 k

K Γ 1 1 i (x1+x2) (x2 x1) ikx3 θ(x x ) + sgn (x x ) sgn (x x ) e 2 e− 2 − e + . 2 − 1 2 3 − 1 2 − 3 − 2 ↔ (S46)

The integrals over K and k are separable and can be computed individually. They are simply inverse Fourier transforms of the Gaussian functions multiplied by their transmission coefficients. The two-photon bound part of the hybrid state therefore evolves according to the two-photon bound state transmission coefficient, while the single photon part evolves according to the single-photon transmission coefficient. One can therefore interpret the hybrid eigenstates as a two-photon bound state and a single photon which are weakly interacting. The effect of interactions then becomes negligible in the long pulse limit, since the two components are unlikely to be at the same spatial positions.

S4. ELECTRIC FIELD OF BOUND STATES AND RELATION TO SELF-INDUCED TRANSPARENCY

In this section we compute the electric ﬁeld of a linear combination of bound states. In particular, we show that in the classical limit, where the sum is composed of a sum of many-body bound states, the ﬁeld resembles that of a self-induced transparency soliton. We begin by considering an arbitrary state composed of a sum of bound states

X Z ψ = dEc (E) BE (S47) | i n | in n

The electric ﬁeld is then proportional to Z X E E aˆ(x) = ψ aˆ(x) ψ = dEdE0cn∗ 1(E0)cn(E)n 1 B aˆ(x) B n. (S48) h i h | | i − − h | | i n 22

The bulk of this calculation involves computing the matrix element in the integrand. Using the expression for the n-photon bound state in (8) of the main text, the matrix element is r r n 2 n 1 Z E E0 E E √n Γ − (n 2)! Γ − (n 1)! i n n−1 (x1+x2+...xn−1) n 1 B aˆ(x) B n = − − dx1dx2 . . . dxn 1e − − h | | i 2π n 1 n − Pn−1 − Γ P Γ xi xj iEx/n xi x e− i

nE0 where Kn = E n 1 . The bound-state− − wave functions are highly symmetric, which can be used to greatly restrict the domain of integration. This symmetry will be exploited to reduce the exponentials with negative absolute value arguments to exponentials without absolute values, but with arguments that are always negative. Noting that the integral is in- variant under permutation of any xi xj, we can restrict the domain of integration to x1 > x2 > . . . xn 1 and correspondingly multiply by (n 1)!. This↔ gives − − √ Z Γ P E E Γ n 2 iEx/n iKn(x1+x2+...xn−1)/n xi x n 1 B aˆ(x) B n = Γ − (n 2)!(n 1)! e dx1dx2 . . . dxn 1e e− 2 i | − | − h | | i 2π − − > − Γ(x1 x2) Γ(x1 x3) Γ(x1 x4) Γ(x1 xn−1) e− − e− − e− − ... e− − × × × Γ(x2 x3) Γ(x2 x4) Γ(x2 xn−1) e− − e− − ... e− − × × × (S50) . . . .

Γ(xn−3 xn−2) Γ(xn−3 xn−1) e− − e− − × Γ(xn−2 xn−1) e− − , × where the subscript > of the integral indicates the domain x1 > x2 > . . . > xn 1. All these exponents can be combined to give −

Γ(n 2)x1 Γ(n 4)x2 Γ(n 6)x3 Γ[n 2 2(i 1)]xi Γ(n 2)xn−1 e− − e− − e− − . . . e− − − − . . . e − . (S51) There is still an exponential factor with an absolute value in its exponent. This is n 1 Γ Pn−1 − h Γ Γ i xi x Y (xi x) (xi x) e− 2 i | − | = θ(x x)e− 2 − + θ(x x )e 2 − , (S52) i − − i i=1 n 1 where θ(x) is Heaviside’s step function. In general, this contains 2 − terms when fully expanded, but most of these terms integrate to zero because the integral is over x1 > x2 > . . . > xn 1. In fact, there are only n terms that contribute, − Γ Pn−1 Γ Γ Γ xi x (x x1) (x x2) (x x1) e− 2 i | − | θ(x > x1 > x2 > . . . > xn 1)e− 2 − e− 2 − . . . e− 2 − → − Γ (x x1) Γ (x x2) Γ (x x1) + θ(x1 > x > x2 > . . . > xn 1)e 2 − e− 2 − . . . e− 2 − − Γ (x x1) Γ (x x2) Γ (x x1) + θ(x1 > x2 > x > . . . > xn 1)e 2 − e 2 − . . . e− 2 − (S53) − + ... + Γ (x x1) Γ (x x2) Γ (x x1) + θ(x1 > x2 > . . . > xn 1 > x)e 2 − e 2 − . . . e 2 − , − where the notation θ(x > x1 > x2 > . . . > xn 1) = θ(x x1)θ(x1 x2)θ(x2 x3) . . . θ(xn 2 xn 1). This then gives n many-body integrals each− over n −1 variables,− − − − − − √ Z E E Γ n 2 iEx/n iKn(x1+x2+...+xn)/n n 1 B aˆ(x) B n = Γ − (n 2)!(n 1)!e dx1dx2 . . . dxn 1e − h | | i 2π − − > −

Γ(n 2)x1 Γ(n 4)x2 Γ[n 2n 2 2(k 1)]xk Γ(n 2)xn−1 e− − e− − . . . e− − − − − . . . e − × n Γ Γ Pn−1 Γ Γ Γ Pn−1 (n 1)x xi (n 1)x x1 xi θ(x x )e− 2 − e 2 i=1 + θ(x x)θ(x x )e− 3 − e− 2 e 2 i=2 (S54) − 1 1 − − 2 Γ Γ Pj−1 Γ Pn−1 [n 1 2(j 1)]x xi xi + ... + θ(xj 1 x)θ(x xj)e− 2 − − − e− 2 i=1 e 2 i=j − − − Γ Γ Pn−1 o (n 1)x xi + ... + θ(xn 1 x)e 2 − e− 2 i=1 . − − 23

The j-th integral is

Γ Z [n 1 2(j 1)] iKn(x1+x2+...+xn)/n Γ(n 2)x1 Γ(n 4)x2 Γ(n 6)x3 Ij(x) = e− 2 − − − dx1dx2 . . . dxn 1e e− − e− − e− − > − Γ Γ Γ[n 2 2(i 1)]xi Γ(n 2)xn−1 (x1+x2+...+xj−1) (xh+xj+1+...+xn−1) . . . e− − − − . . . e − θ(xj 1 x)θ(x xj)e− 2 e− 2 − − − x1 Γ Z Γ Z Γ [n 1 2(j 1)] ∞ Γ(n 2)x1 iKnx1/n x1 Γ(n 4)x2 iKnx2/n x2 = e− 2 − − − dx1e− − e e− 2 dx2e− − e e− 2 x x Z xj−2 (S55) Γ[n 2 2(j 2)]xj−1 iKnxj−1/n ... dxj 1e− − − − e × × x − Z x Z xj Γ[n 2 2(j 1)]xj iKnxj /n Γ[n 2 2j]xj+1 iKnxj+1/n dxje− − − − e dxj+1e− − − e −∞ −∞ xn−2 Z Γ Γ(n 2)xn−1 iKnxn−1/n xn−1 ... dxn 1e − e e 2 . × × − −∞ The n 1 integrals thus split into two sets of integrals: the ﬁrst j 1 integrals and the second n j. The ﬁrst j 1 integrals− have a form giving − − −

h Pj 1 i x1 xj−2 − Z ∞ Z Z exp i=1 αix α1x1 α2x2 αj−1xj−1 − K˜j 1(x) = dx1e− dx2e− ... dxj 1e− = , (S56) j 1 i − x x × × x − Q P i=1− k=1 αk where αk = Γ [n 2 2(j 1) + 1/2] iKn/n and the sums over αk can be easily evaluated. The latter n j integrals must be evaluated− in− reverse− order. They− give −

n j − Y 1 iKn(n j)x/n Γ(n j)(j 1/2)x I˜n j(x) = e − e − − . (S57) − Γm (n m 1/2) + imK /n m=1 − − n Combining the two integrals we have

Γ [n 1 2(j 1)]x Ij(x) = e− 2 − − − I˜n j(x)K˜j 1(x) − − n j j 1 − − iK0 (n 1)x Y 1 Y 1 = e n − (S58) Γm (n m 1/2) + imK Γk (n k 1/2) ikK m=1 − − n0 k=1 − − − n0 iK0 (n 1)x e n − I0 (K0 ), ≡ j n where we have deﬁned Kn0 = Kn/n. We note that in Ij(x) the exponents containing Γ all cancel each other. Putting everything together the matrix element can be written

n E E √Γ n 2 i(E E0)x X n 1 B aˆ(x) B n = Γ − (n 2)!(n 1)!e − Ij0 (Kn0 ). (S59) − h | | i 2π − − j=1

Using the properties of the complex Gamma function Γ (z) one can express

h 0 i n π Γ1 n sech Knπ X − Γ I0 (K0 ) = (S60) j n Γ n 1 iK /Γ Γ n 1 + iK /Γ j=1 − 2 − n0 − 2 n0 The entire expression becomes

h 0 i Z sech Knπ 1 X i(E E0)x Γ aˆ(x) = (n 2)!(n 1)! dE dE0cn∗ 1(E0)cn(E)e − 1 1 , (S61) h i 2√Γ − − − Γ n iK /Γ Γ n + iK /Γ n − 2 − n0 − 2 n0 which is thus far an exact result without approximation. 24

A. Recovering the result of self-induced transparency

We now consider Gaussian wavepackets of bound states. Here c (E) = BE in . We do not need to know the full n nh | i form of cn(E), but it is separable and can be written as

2 2 (E nk0) σ /(2n) cn(E) = e− − cn, (S62) where cn is not known but is constrained by the normalization of the state ψ ψ = 1. Using this expression in (S61) gives h | i

Z 0 2 2 2 2 0 1 X (E (n 1)k0) σ /[2(n 1)] (E nk0) σ /(2n) i(E E )x aˆ(x) = (n 2)!(n 1)!cncn∗ 1 dE dE0e− − − − e− − e − h i √ − − − 2 Γ n h 0 i (S63) Knπ sech Γ . × Γ n 1 iK /Γ Γ n 1 + iK /Γ − 2 − n0 − 2 n0 ¯ Again, as in the previous section, the integral can be recast over Kn0 = E/n E0/(n 1) and Kn = E/n + E0/(n 1) with Jacobian n(n 1)/2 giving − − − − Z 0 1 X ¯ 2 2 2 2 0 ¯ (Kn 2k0) nσ /4 Kn nσ /4 iKnnx aˆ(x) = n!(n 1)!cncn∗ 1 dKn0 dKne− − e− e h i √ − − 4 Γ n h 0 i (S64) Knπ sech Γ , × Γ n 1 iK /Γ Γ n 1 + iK /Γ − 2 − n0 − 2 n0 where we have used the approximation n/(n 1) 1 for large n to simplify the Gaussian functions. Integrating over − ∼ K¯n gives

1 2√π Z 02 2 0 X Kn nσ /4 iKnnx aˆ(x) = cncn∗ 1 dKn0 e− e h i √ − √ 2 4 Γ n nσ h 0 i (S65) Knπ n!(n 1)! sech Γ − , × Γ n 1 iK /Γ Γ n 1 + iK /Γ − 2 − n0 − 2 n0 Again, we make the approximation in the limit of large n, n!(n 1)! − = n2 + O(n), (S66) Γ n 1 iK /Γ Γ n 1 + iK /Γ − 2 − n0 − 2 n0 which leads to 2 Z 1 n √π 02 2 0 K0 π X Kn nσ /4 iKnnx n aˆ(x) = cncn∗ 1 dKn0 e− e sech . (S67) h i √ − √ 2 Γ 2 Γ n nσ

If the Gaussian function is slowly varying with respect to the sech function, one can expand it about Kn0 = 0 giving 1 + O(Kn0 ). Taking this limit and evaluating the Fourier integral gives

√Γ X n2√π nΓx aˆ(x) = cncn∗ 1 sech . (S68) h i 2 − √ 2 2 n nσ

Deﬁning the integration Rabi frequency as Ω = 2√Γ R dx aˆ(x) , one obtains h i X √nπ Ω = 2π cncn∗ 1 . (S69) − σ n

P R 2 P 2 √nπ The normalization of the wavefunction implies 1 = n dE cn(E) = n cn σ and therefore the integrated Rabi frequency in the large n limit approaches | | | |

Ω = 2π. (S70) 25

From (S68), in the limit of a large average photon number and a narrow photon distribution, the electric ﬁeld can also be approximated as

n¯√Γ n¯Γx aˆ(x) = sech , (S71) h i 2 2 wheren ¯ is the average photon number.

S5. NORMALLY-ORDERED CORRELATION FUNCTION OF BOUND STATES AND MEAN-FIELD THEORY

Here we compute the m-th-order normally ordered correlation function of the bound states (n m). Following the previous section one obtains an integral ≥

n 1 0 m Z 0 E m E Γ − (n 1)! n! i(E E )(x1+x2+...xn−m)/n n B aˆ†(x) [ˆa(x)] B n = − dx1dx2 . . . dxn me − h | | i 2πn (n m)! − − n m (S72) Pn−m − Γ xi xj iEmx/n Y mΓ xi x e− i

As before, one must reduce the integral to the domain x1 > x2 > . . . > xn m and then expand the product and reduce it to its n m + 1 non-zero terms. Computing all the integrals using− the techniques of the previous section, one obtains − m 1 K K E0 m m E Γ − n! (n 1)!(n + m 1)! Γ (1 n i nΓ ) Γ (1 n + i nΓ ) iKx n B aˆ†(x) [ˆa(x)] B n = − − − − − e , h | | i 2πn (n m)! (2m 1)! Γ (1 m i K ) Γ (1 m + i K ) − − − − nΓ − nΓ (S73) where now K = E E0. −

A. Reduction to mean-ﬁeld theory (n m)

We now show that if the particle number n is much large than the order of the correlation function m, the outcome of the measurement probes the mean ﬁelds and the correlation function factorizes. One can start by making the approximation K K π2n2 (n!)2 Γ 1 n i Γ 1 n + i = + O(n). (S74) nΓ nΓ 2 πK − − − sinh ( nΓ ) Additionally one can use the recurrence relation of the Gamma function Γ(z n) = Γ(z)/ [(z 1)(z 2) ... (z n)] to write − − − − m 1 K K iK iK Y− 1 Γ 1 m i Γ 1 m + i = Γ Γ − − nΓ − nΓ − nΓ nΓ 2 K2 k + n2Γ2 k=1 (S75) m 1 π Y− 1 = . K sinh πK 2 K2 nΓ nΓ k=1 k + n2Γ2 Combining these, the matrix element becomes

m 1 2m 1 m 1 2 E0 m m E Γ − n − K πK iKx Y− 2 K B aˆ†(x) [ˆa(x)] B = cosech e k + , (S76) nh | | in 2 (2m 1)! nΓ nΓ n2Γ2 − k=1 (n+m 1)! 2m 1 where we have used Stirling’s approximation to write − n − . (n m)! ∼ One can now consider an n-photon Fock state composed− of bound states

1/2 Z 2 2 σ (E nk0) σ /(2n) E ψ = dE e− − B . (S77) | in √nπ | n i 26

The correlation function of this state is given by

m Z 0 2 2 2 2 0 m m σ (E nk0) σ /(2n) (E nk0) σ /(2n) E m E ψ aˆ†(x) [ˆa(x)] ψ = dEdE0 e− − e− − B aˆ†(x) [ˆa(x)] B . (S78) nh | | in √nπ h n | | n i

Changing the integration variables to K = E E0 and K¯ = (E + E0)/2 and integrating out K¯ gives − m 1 m 1 2m 1 Z 2 2 2 m m Γ − n − K πK K σ iKx Y− 2 K ψ aˆ†(x) [ˆa(x)] ψ = dK cosech e− 4n e k + . (S79) nh | | in 2 (2m 1)! nΓ nΓ n2Γ2 − k=1 At this point one can make use of the tabulated Fourier Transform [55],

Z n 1 n 1 2 4 − y πy Y− y [sech (ax)]2n = dy eiyx cosech + r2 . (S80) (2n 1)!a 2a 2a 4a2 − r=1 In the limit σΓ 1 the Gaussian factor approaches unity and the tabulated integral can be used to give m 1 2m 1 2m m m Γ − n − nΓ 1 nΓx n ψ aˆ†(x) [ˆa(x)] ψ n = m 1 sech h | | i 2 2 4 − 2 " #2m (S81) n√Γ nΓx = sech . 2 2

As in the previous section, one can also consider a state composed of different Fock states n. Such a state returns the same result but with n replaced with the average photon numbern ¯. Comparing this result with (S71) indicates that the m-th order correlation function can be obtained using mean-field theory when n m. We note that the expression for the correlation function has a different form when n m. This is particularly clear from (S73) when n = m which causes the Gamma functions to evaluate to unity leaving∼ only the plane-wave factor varying with K.

S6. PROPAGATION IN THE LARGE n LIMIT

In this section we show how the observables change under propagation when n 1. We start with the electric field aˆ(x) which is given in (S61). For an incident coherent field in the limit of n 1, we expect that the field amplitudes ofh thei bound states approach those of a coherent state and therefore

1/2 2 n 2 2 α /2 α σ (E nk0) σ /(2n) cn(E) e−| | e− − (S82) ∼ √n! √nπ Since the bound states are eigenstates, scattering oﬀ N emitters simply maps c (E) tN c (E). As in the main n → E,n n N iNφn(E) text, one can write tE,n = e and Taylor expand φn(E). Since we have found the pulse distortion terms decrease rapidly as n increases we simply consider the ﬁrst order term in the expansion of φn(E). Following the calculations starting from (S63) one obtains for zero detuning (k0 = 0)

2n α 2 X α n√Γ nΓ 4N aˆ(x) = e−| | sech x + , (S83) h i n! 2 2 n2Γ n which gives (19) from the main text. A similar process can be followed to obtain the normally ordered correlation functions.

S7. EFFECTIVE SPIN MODEL AND MPS ANSATZ

In this appendix we present an alternative route to the S-matrix scattering formalism used in the main text to describe the light propagation through an array of chirally coupled emitters. The full system dynamics can be indeed described by the combination of a driven-dissipative master equation for the emitters and a generalized input-output relation that allowed to reconstruct the electric ﬁeld [26, 27]. The dynamics of the emitter can than be eﬃciently solved by using a matrix product states (MPS) ansatz [27, 30, 31, 42] which allow us to push the simulation to larger atomic arrays and higher input power. 27

0.025

0.02

0.015

0.01

0.005

0 -80 -60 -40 -20 0

FIG. S2. Comparison between the results obtained by directly solving the master equation (continuous line) and by using the quantum trajectories algorithm (markers points) for the same parameters used in Fig.2(d) of the main text.

A. Eﬀective spin model

Let us consider the most general case (extension of the cascade scenario described by Eq.(1) of the main text) of N TLA with equal frequency ω0 = ck0 coupled asymmetrically to a waveguide with decay rates ΓR and ΓL associated respectively to the emission of right and left propagating photons. The addition of the two decay rates gives the total emission rate of the TLA into the waveguide Γ = ΓL + ΓR. In a realistic implementation the emitters can also radiate into other external channels, the decay rate associated to this emission is indicated with Γ0. The right propagating input pulse is treated as a classical coherent field, E (t, x) = (t)eikinx and for the rest of the discussion we will in Ein assume it to be on resonance with the atomic transition, i.e. kin = k0. The coupling of the emitters to the input field is P + + given by the Hamiltonian Hdrive = j √ΓR Ein(t, xj)ˆσj + H.c. whereσ ˆj = g e j is the j-th emitter annihilation operator. | ih | Under Born-Markov approximation the emitters dynamics is known to be described by a chiral master equation(ME) of the general form [29]: ρ˙ = i H ρ ρH† + [ρ], (S84) − eff − eff J where X (Γ + Γ0) + X ik0 xl xj + ik0 xl xj + H = i σˆ σˆ− + H i Γ e | − |σˆ σˆ− + Γ e | − |σˆ σˆ− eff − 2 j j drive − L j l R l j (S85) j l>j is the effective Hamiltonian which provides the non-Hermitian collective evolution of the emitters. The recycling term h i X + X ik0 xl xj ik0 xl xj + [ρ] = (Γ + Γ ) σˆ−ρσˆ + Γ e | − | + Γ e− | − | σˆ−ρσˆ + H.c. J 0 j j R L j l (S86) j l>j assures the conservation of the density operator trace and it arises by the quantum jumps that the emitters experience after the emission of a photon into the waveguide or into free space. For ΓL = ΓR Eq. (S84) coincides with the standard waveguide QED master equation [56, 57] where both coherent and dissipative interactions between the emitters occur depending on the atoms position. In the limit of a perfect chiral waveguide considered in the main text, i.e. ΓL = Γ0 = 0 and Γ = ΓR, the master equation Eq. (S84) reduces to the well known cascade master equation [28, 29] characterized by an unidirectional purely dissipative interaction between the emitters. In this limit only the spatial order of the atoms and not the specific positions matters and the excitation is transferred only to the right without any back action. Once solved the emitters dynamics, the transmitted and reflected (for the semi-chiral case) electric field is obtained by the following generalized input-output relations [26, 27]:

X p ik0xj E (t) = (t) + i Γ e− σˆ−(t), (S87) R Ein R j j 28

X p ik0xj EL(t) = i ΓLe σˆj−(t). (S88) j

The combination of Eq. (S84), (S87), and (S88) allows to correctly describes the photon propagation trough the chiral medium.

B. MPS ansatz

In order to observe the main results obtained in the main text, e.g. photon number dependent bound states separation, it is necessary to consider substantial input pulse amplitudes and large atomic arrays, a scenario that is challenging to simulate with standard numerical techniques. To treat this many body problem we use a recently developed algorithm which involve an MPS representation to efficiently describe the effective spin dynamics [27]. The system evolution can be solved with two different approaches, either by directly solving the ME (S84)[42](method used for Fig. 3 of the main text) or by using a quantum trajectories algorithm where the state of the system evolves under the effective Hamiltonian (S85) and it stochastically experiences the occurring of quantum jumps [27] (method used for Fig. 2(f) of the main text). In both case an MPS representation is applied either to the quantum state or to the the density matrix once mapped to a vector. The MPS ansatz consists in reshaping the generic quantum state P φ/ρ = ψi ,i ,..i i1, i2, ..iN into a matrix product state of the form: | i i1,..iN 1 2 N | i X φ/ρ = A A ...A i , i , ..i (S89) | i i1 i2 iN | 1 2 N i i1,..iN where, for each specific set of physical indices ¯i ,¯i , ..¯i , the product of the A¯ matrices gives back the state { 1 2 N } ij coefficient ψ¯i ,¯i ,..¯i . Each matrix A¯i has dimension Dj 1 Dj known as bond dimension and finite-edge boundary 1 2 N j − × conditions are assumed by imposing D1 = 1 and DN = 1. The bond dimension reflects the entanglement entropy. For arbitrary states the bond dimensions grow exponentially with the system size. The advantage of the MPS ansatz it is that, in many physical scenarios, as the one considered here, the entanglement can grow slowly with the system size allowing an efficient description of the state in terms of a smaller bond dimension [31]. In order to compute the evolution of the system it is possible, in both methods, to derive a matrix product operator (MPO) representation for either the Liouvillian or the effective Hamiltonian and jump operators. The derivation of this representation for our semi-chiral case follows straightforwardly from the bi-directional case presented in [27, 42]. The evolution of the system is then computed by applying a linear expansion of the master equation (or of the time evolution operator) 1 + dt (Heff ) or a Runge-Kutta method. In addition a MPO representation can be derived also for the observables and it allowsL to evaluate the expectation values. After each application of an MPO to an MPS the MPS bond dimension increases. The bond dimension is than truncated after each step in order to keep an efficient description of the state. We indicate as Dmax the maximum bond dimension used to represent the system during an entire time evolution. In all the plots considered in this work we used Dmax = 150 for the ME simulations and Dmax = 40 for the quantum trajectories algorithm.

S8. EFFECT OF OTHER DECAY CHANNELS

In the main text we considered the ideal scenario of an array of atoms perfectly chirally coupled to a waveguide. In real implementations this is not always the case and the atoms can either emit into the other direction or into free space. Let us first assume that the atoms can emit also into the direction opposite to the pulse propagation, i.e. ΓL = 0, while the emission to other external channels is suppressed Γ0 = 0. In this case the master equation (S84) becomes6 dependent on the atomic positions. To better understand this point we consider the two paradigmatic cases of atoms equally spaced either by a distance k0d = k0 xi xj = (n + 1)π/2 or k0d = k0 xi xj = nπ with n = 1, 2, ... We focus on these two configurations because in the| bi-directional− | limit they are the ones| − providing| purely coherent or dissipative interactions respectively. In Fig.S3(a)-(b) we plot the transmitted intensity for the two cases and we compare it to the one obtained in the perfectly chiral scenario. We observe that the emission into the opposite channel affects the intensity in a totally different manner depending on the atomic distance. In the first configuration, Fig.S3(a), it leads to an overall spreading of the bound states peaks for increasing values of ΓL. In the second, Fig.S3(b), the bound state separation not only seems to be robust to the deviation from a perfect chiral emission, but it even leads to a better resolution of the bound state peaks at the price of a weaker pulse in transmission. This effect is caused by an additional time-delay of the BS peaks, induced by partial reflection, which affects mostly the few photon bound states making the bound states separation even more evident. The many-photons BS, on the contrary, 29

(a) 0.16 (b) 0.16

0.12 0.12

0.08 0.08

0.04 0.04

0 0 -50 -40-30 -20 0-10 10 -50 -40-30 -20 0-10 10

input (c) 0.1 (d) 0.16 reflected field 0.08 0.12 transmitted field 0.06 0.08 0.04

0.04 0.02

0 0 -50 -40-30 -20 0-10 10 -50 -40-30 -20 0-10 10

FIG. S3. (a)-(b) Transmitted power as function of the position for different values of the decay along the left direction. In (a) we fixed the atomic distance to kd = π/2 while in (b) to kd = π. (c) Transmitted and reflected power for ΓL/Γ = 0.2 and kd = π. (d) Transmitted power for different values of an external decay rate Γ0. In panels (a),(b) and (d) the dashed lines show the expected delay√ of the BS for the ΓL = Γ0 = 0 case while in panel (c) it shows the input field position. In all panels N = 20,n ¯ = 1.8, Γσ = 2 2 and Dmax = 170. not only experience less delay than their few-photons counterpart, but they are also more robust against distortion tending to keep their solitonic behaviour. These features are more evident in Fig. S3(c) where we plot the reflected field intensity. Here the majority of the reflection occurs at the interface and it causes the damping in transmission. On the other end, when the pulse enter in the medium, we observe a major reflection for the BS with few photons. This phenomena make sense in light of the connection to the SIT limit at large photon pulses. In this limit, the chirality is not an essential requirement to observe the formation of solitons. The role of chirality is however crucial to bring the solitonic behaviour to the level of few photons. In addition this observation suggests that the k0d = k0 xi xj = nπ spacing is the optimal configuration to observe the effect in an experiment. On the other hand, in many| realistic− | implementation, it is difficult to have control on the position of the emitters. For this reason it makes sense consider the effect of a semi-chiral emission by making an average on random positions. This scenario is similar to consider a decay of the atoms in an external channel with rate Γ0. The effect of this external decay is plotted in Fig. S3(d) and shows that in general the effect is visible for values Γ 0.2. Again we expect the overall effect to be robust for many-photon BS. 0 ≤