Superconducting metamaterials for waveguide quantum electrodynamics

Mohammad Mirhosseini,1, 2 Eunjong Kim,1, 2 Vinicius S. Ferreira,1, 2

Mahmoud Kalaee,1, 2 Alp Sipahigil,1, 2 Andrew J. Keller,1, 2 and Oskar Painter1, 2, ∗

1Kavli Nanoscience Institute and Thomas J. Watson, Sr., Laboratory of Applied Physics,California Institute of Technology, Pasadena, California 91125, USA.

2Institute for Quantum Information and Matter,California Institute of Technology, Pasadena, California 91125, USA.

(Dated: February 7, 2018)

The embedding of tunable quantum emitters in a photonic bandgap structure enables the controlof dissipative and dispersive interactions between emitters and their photonic bath. Operation inthe transmission band, outside the gap, allows for studying waveguide quantum electrodynamicsin the slow-light regime. Alternatively, tuning the emitter into the bandgap results in finite rangeemitter-emitter interactions via bound photonic states. Here we couple a transmon qubit to a su-perconducting metamaterial with a deep sub-wavelength lattice constant (λ/60). The metamaterialis formed by periodically loading a transmission line with compact, low loss, low disorder lumpedelement microwave resonators. We probe the coherent and dissipative dynamics of the system bymeasuring the Lamb shift and the change in the lifetime of the transmon qubit. Tuning the qubitfrequency in the vicinity of a band-edge with a group index of ng = 450, we observe an anomalousLamb shift of 10 MHz accompanied by a 24-fold enhancement in the qubit lifetime. In addition,we demonstrate selective enhancement and inhibition of spontaneous emission of different trans-mon transitions, which provide simultaneous access to long-lived metastable qubit states and statesstrongly coupled to propagating waveguide modes.

Cavity quantum electrodynamics (QED) studies theinteraction of an atom with a single electromagneticmode of a high-finesse cavity with a discrete spectrum[1, 2]. In this canonical setting, a large photon-atomcoupling is achieved by repeated interaction of the atomwith a single photon bouncing many times between thecavity mirrors. Recently, there has been much interestin achieving strong light-matter interaction in a cavity-free system such as a waveguide. Waveguide QED refersto a system where a chain of atoms are coupled to acommon optical channel with a continuum of electromag-netic modes over a large bandwidth. Slow-light photoniccrystal waveguides are of particular interest in waveguideQED because the reduced group velocity near a bandgappreferentially amplifies the desired radiation of the atomsinto the waveguide modes [3–5]. Moreover, in this con-figuration an interesting paradigm can be achieved byplacing the resonance frequency of the atom inside thebandgap of the waveguide [6–10]. In this case the atomcannot radiate into the waveguide but the evanescentfield surrounding it gives rise to a photonic bound state[8]. The interaction of such localized bound states hasbeen proposed for realizing tunable spin-exchange inter-action between atoms in a chain [11, 12], and also forrealizing effective non-local interactions between photons[13, 14].

While achieving efficient waveguide coupling in the op-tical regime requires the challenging task of interfacingatoms or atomic-like systems with nanoscale dielectricstructures [15–18], superconducting circuits provide anentirely different platform for studying the physics of

∗ opainter@caltech.edu; http://copilot.caltech.edu

light-matter interaction in the microwave regime [19].Development of the field of circuit QED has enabledfabrication of fast and tunable qubits with long co-herence times [20–22]. Moreover, strong coupling isreadily achieved in this platform due to the deep sub-wavelength transverse confinement of photons attainablein microwave waveguides and the large electrical dipole ofsuperconducting qubits [23]. Microwave waveguides withstrong dispersion, even “bandgaps” in frequency, can alsobe simply realized by periodically modulating the geome-try of a coplanar transmission line [24]. Such an approachwas recently demonstrated in a pioneering experiment byLiu and Houck [25], whereby a qubit was coupled to thelocalized photonic state within the bandgap of a mod-ulated coplanar waveguide (CPW). Satisfying the Braggcondition in a periodically modulated waveguide requiresa lattice constant on the order of the wavelength [26],however, which translates to a device size of approxi-mately a few centimeters for complete confinement of theevanescent fields in the frequency range suitable for mi-crowave qubits. Such a restriction significantly limits thescaling in this approach, both in qubit number and qubitconnectivity.

An alternative approach for tailoring dispersion in themicrowave domain is to take advantage of the meta-material concept. Metamaterials are composite struc-tures with sub-wavelength components which are de-signed to provide an effective electromagnetic response[27, 28]. Since the early microwave work, the electro-magnetic metamaterial concept has been expanded andextensively studied across a broad range of classical op-tical sciences [29–32]; however, their role in quantum op-tics has remained relatively unexplored, at least in partdue to the lossy nature of many sub-wavelength compo-

arX

iv:1

802.

0170

8v1

[qu

ant-

ph]

5 F

eb 2

018

2

4 4.5 5 5.5 6 6.5 7 7.5 80

0.2

0.4

0.6

0.8

1

Tran

smis

sion

0 1kx (π/d)

1

2

3

4

5

6

7

8

9

10Fr

eque

ncy

(GH

z)

bare waveguide dispersion

bandgap region(imaginary k-vector)

ω+,k

ω-,k

a b

c

sym

met

ry a

xis

x

20 μm

1/2 Lr

1/2 Lr

Cr

Ck

Frequency (GHz)

Band Gap

FIG. 1. Microwave metamaterial waveguide. a, Dis-persion relation of a CPW loaded with a periodic array ofmicrowave resonators. The dashed line shows the dispersionrelation of the waveguide without the resonators. Inset: cir-cuit diagram for a unit cell of the periodic structure. b, Scan-ning electron microscope (SEM) image of the fabricated ca-pacitively coupled microwave resonator with a wire width of500 nm. The resonator region is false-colored in purple, thewaveguide central conductor and the ground plane are coloredgreen, and the coupling capacitor is shown in orange. We haveused pairs of identical resonators symmetrically placed on thetwo sides of the transmission line to preserve the symmetry ofthe structure. c, Transmission measurement for the realizedmetamaterial waveguide made from 9 unit cells of resonatorpairs with a wire width of 1 µm, repeated with a lattice con-stant of d = 350µm. The blue curve depicts the experimentaldata and the red curve shows the lumped-element model fitto the data.

nents. Improvements in design and fabrication of low-loss superconducting circuit components in circuit QEDoffer a new prospect for utilizing microwave metamateri-als in quantum applications. Indeed, high quality-factorsuperconducting components such as resonators can bereadily fabricated on a chip [33, 34], and such elementshave been used as a tool for achieving phase-matchingin near quantum-limited traveling wave amplifiers [35–37] and for tailoring qubit interactions in a multimodecavity QED architecture [38].

In this paper, we utilize an array of coupled lumped-

element microwave resonators to form a compactbandgap waveguide with a deep sub-wavelength latticeconstant (λ/60) based on the metameterial concept. Inaddition to a compact footprint, these sort of structurescan exhibit highly nonlinear band dispersion surroundingthe bandgap, leading to exceptionally strong confinementof localized intra-gap photon states. We present the de-sign and fabrication of such a metamaterial waveguide,and characterize the resulting waveguide dispersion andbandgap properties via interaction with a tunable super-conducting transmon qubit. We measure the Lamb shiftand lifetime of the qubit in the bandgap and its vicinity,demonstrating the anomalous Lamb shift of the funda-mental qubit transition as well as selective inhibition andenhancement of spontaneous emission for the first twoexcited states of the transmon qubit.

We begin by considering the circuit model of a CPWthat is periodically loaded with microwave resonators asshown in the inset to Fig. 1a. The Lagrangian for thissystem can be constructed as a function of the node fluxesof the resonator and waveguide sections Φbn and Φan [40].Assuming periodic boundary conditions and applying therotating wave approximation, we derive the Hamiltonianfor this system and find the eigenstates and energies tobe (see App. A),

ω±,k =1

2

[(Ωk + ω0)±

√(Ωk − ω0)

2+ 4g2

k

], (1)

α±,k =(ω±,k − ω0)√

(ω±,k − ω0)2

+ g2k

ak +gk√

(ω±,k − ω0)2

+ g2k

bk,

(2)

where ak, and bk describe the momentum-space anni-hilation operators for the bare waveguide and bare res-onator sections, the index k denotes the wavevector, andthe parameters Ωk, ω0, and gk quantify the frequencyof traveling modes of the bare waveguide, the resonancefrequency of the microwave resonators, and coupling ratebetween resonator and waveguide modes, respectively.The operators α±,k represent quasi-particle solutions ofthe composite waveguide, where far from the bandgap thequasi-particle is primarily composed of the bare waveg-uide mode, while in the vicinity of ω0 most of its energyis confined in the microwave resonators.

Figure 1a depicts the numerically calculated energybands ω±,k as a function of the wavevector k. It is evidentthat the dispersion has the form of an avoided crossingbetween the energy bands of the bare waveguide and theuncoupled resonators. For small gap sizes, the midgapfrequency is close to the resonance frequency of uncou-pled resonators ω0, and unlike the case of a periodicallymodulated waveguide, there is no fundamental relationtying the midgap frequency to the lattice constant in thiscase. The form of the band structure near the higher cut-off frequency ωc+ can be approximated as a quadraticfunction (ω−ωc+) ∝ k2, whereas the band structure nearthe lower band-edge ωc− is inversely proportional to the

3

metamaterial waveguide

resonator

qubit100 µm

CPW input

XYZ

SQUID loop

25 µm Z

b

100 µm

100

10-5

10-10LDO

S (G

Hz -1

μm-1

)10

8

104

100

83 4 5 76

Frequency (GHz)83 4 5 76

Band Gap

a

c

Loca

lizat

ion

leng

th, l=d

FIG. 2. Disorder effects and qubit-waveguide coupling. a, Calculated localization length for a metamaterial waveguidewith structural disorder and resonator loss are shown as blue dots. The waveguide parameters are determined from the fit toa lumped element model with resonator loss to the transmission data in Fig. 1. Numerical simulation has been performed forN = 100 unit cells, averaged over 105 randomly realized values of the resonance frequency ω0, with the standard deviationδω0/ω0 = 0.5%. The red curve outside the gap is an analytic model based on Ref. [39]. b, SEM image of the fabricatedqubit-waveguide system. The metamaterial waveguide (gray) consists of 9 periods of the resonator unit cell. The waveguide iscapacitively coupled to an external CPW (red) for reflective read-out. Bottom left inset: The transmon qubit is capacitivelycoupled to the resonator at the end of the array. The Z drive is used to tune the qubit resonance frequency by controllingthe external flux bias in the superconducting quantum interference device (SQUID) loop. The XY drive is used to coherentlyexcite the qubit. Top right inset: capacitively coupled microwave resonator. c, Calculated local density of states (LDOS) atthe qubit position for a metamaterial waveguide with a length of 9 unit cells and open boundary conditions. The band-edgesfor an infinite structure are marked with dashed red lines.

square of the wavenumber (ω − ωc−) ∝ 1/k2. The anal-ysis above has been presented for resonators which arecapacitively coupled to a waveguide in a parallel geome-try; a similar band structure can also be achieved usingseries inductive coupling of resonators (see App. A).

A coplanar microwave resonator is often realized byterminating a short segment of a coplanar transmissionline with a length set to an integer multiple of λ/4, whereλ is the wavelength corresponding to the fundamentalresonance frequency [24, 33, 41]. However, it is possi-ble to significantly reduce the footprint of a resonator byusing components that mimic the behavior of lumped el-ements. We have used the design presented in Ref. [42]to realize resonators in the frequency range of 6-10 GHz.This design provides compact resonators by placing in-terdigital capacitors at the anti-nodes of the charge wavesand double spiral coils near the peak of the current wavesat the fundamental frequency (see Fig. 1b). Further, thesymmetry of the geometry results in the suppression ofthe second harmonic frequency and thus the eliminationof an undesired bandgap at twice the fundamental reso-nance frequency of the band-gap waveguide.

We fabricate individual resonator pairs using anelectron-beam deposited 120 nm Al film, patterned vialift-off, on a high resistivity silicon wafer substrate ofthickness 500 µm (see Ref. [43] for further details of fab-

rication techniques). In this work we have made a pe-riodic array of 9 resonator pairs with a wire width of 1µm and coupled them to a CPW in a periodic fashionwith a lattice constant of 350 µm to realize a metama-terial waveguide. The resonators are arranged in identi-cal pairs placed on the opposite sides across the centralwaveguide conductor to preserve the symmetry of thewaveguide. Figure 1c shows the measured power trans-mission through such a finite-length metamaterial waveg-uide. Here 50-Ω CPW segments, galvanically coupled tothe metamaterial waveguide, are used at the input andoutput ports. We find the midgap frequency of 5.83 GHzfor the structure, and a gap frequency span of 1.82 GHz.Using the simulated value of effective refractive index of2.54, the midgap frequency gives a lattice constant-to-wavelength ratio of d/λ ≈ 1/60.

Propagation of electromagnetic fields in the frequencyrange within the bandgap is exponentially attenuatedwith a localization length set by the imaginary part ofthe wavenumber. In addition, statistical variations inthe electromagnetic properties of the periodic structureresult in random scattering of the traveling waves in thetransmission band. Such random scatterings can lead tocomplete trapping of propagating photons in the pres-ence of strong disorder and an exponential extinction forweak disorder; a phenomenon known as the Anderson

4

5 5.5 6 6.5Frequency (GHz)

-40

-20

0

20

40

60La

mb

shift

(MH

z)

5 5.5 6 6.5Frequency (GHz)

0

5T 1 li

fetim

e (μ

s)

15

10

a b

FIG. 3. Measured dispersive and dissipative qubit dy-namics. a, Qubit frequency Lamb shift versus frequency.b, Qubit lifetime versus frequency. The open circles showexperimental data and the solid line presents a theory fit.The dashed red lines mark the estimated position of theband-edge corresponding to an infinite-length structure andthe shaded grey regions correspond to anti-crossing with thefirst individual resonances near the band-edges of the fi-nite structure. For calculating the Lamb shift, the barequbit frequency is calculated as a function of flux bias Φ as~ωge =

√8ECEJ(Φ)− EC using the extracted values of EC ,

EJ , and assuming the symmetrical SQUID flux bias relationEJ(Φ) = EJ,max cos(2πΦ/Φ0). The lifetime characterizationis performed in the time domain where the qubit is initially ex-cited with a π pulse through the XY drive. The excited statepopulation, determined from the state-dependent dispersiveshift of a close-by band-edge waveguide mode, is measuredsubsequent to a delay time during which the qubit freely de-cays.

localization of light [44]. We have measured a randomstandard deviation of 0.3% in the resonance frequencyof the fabricated lumped-element resonators. Figure 2ashows the calculated localization length as a functionof frequency from numerical simulation of the indepen-dently measured disorder and loss of the resonators inthe metamaterial waveguide (see App. C and D for fur-ther details). Near the edges of the bandgap the local-ization length from disorder dominates that from loss,rapidly approaching zero at the lower band-edge wherethe group index is largest and maintaining a large value(6×103 periods) at the higher band-edge where the groupindex is smaller. Similarly, the localization length in-side the gap is inversely proportional to the curvatureof the energy bands [12]. Owing to the divergence (inthe loss-less case) of the lower band curvature for thewaveguide studied here, the localization length inside thegap approaches zero near the lower band-edge frequencyas well. These results indicate that, even with practi-cal limitations on disorder and loss in such metamaterialwaveguides, a range of photon length scales of nearly fourorders of magnitude should be accessible for frequencieswithin a few hundred MHz of the band-edges.

To further probe the electromagnetic properties of themetamaterial waveguide we couple it to a superconduct-ing qubit. In this work we use a transmon qubit [20, 21]with the fundamental resonance frequency νge = 7.9 GHz

and Josephson energy to single electron charging energyratio of EJ/EC ≈ 100 at zero flux bias (details of ourqubit fabrication methods can also be found in Ref. [43]).Figure 2b shows the geometry of the device where thequbit is capacitively coupled to one end of the waveg-uide and the other end is capacitively coupled to a 50-ΩCPW transmission line. This geometry allows for form-ing narrow individual modes in the transmission band ofthe metamaterial, which can be used for dispersive qubitstate read-out [45] from reflection measurements at the50-Ω CPW input port (see Fig. 2b). Within the bandgapthe qubit is self-dressed by virtual photons which areemitted and re-absorbed due to the lack of escape chan-nels for the radiation. Near the band-edges surroundingthe bandgap, where the LDOS is rapidly varying with fre-quency, this can result in a large anomalous Lamb shiftof the dressed qubit frequency [9, 46]. To observe thiseffect, we tune the qubit frequency using a flux bias [21]and find the frequency shift by subtracting the measuredfrequency from the expected frequency of the qubit as afunction of flux bias. Figure 3a shows the measured fre-quency shift as a function of tuning. It is evident that thequbit frequency is repelled from the band-edges on thetwo sides, as a result of the asymmetric density of statesnear the cut-off frequencies. The measured frequencyshift is approximately 10 MHz at the band-edges (0.2%of the qubit frequency), in excellent agreement with thecircuit theory model (see App. E).

Another signature of the qubit-waveguide interactionis the change in the rate of spontaneous emission of thequbit. Tuning the qubit into the bandgap changes thelocalization length of the waveguide photonic state thatdresses the qubit. Since the finite waveguide is connectedto an external port which acts as a dissipative environ-ment, the change in localization length `(ω) is accompa-nied by a change in the radiative lifetime of the qubitTrad(ω) ∝ e2x/`(ω), where x is the total length of thewaveguide. Figure 3b shows the measured qubit lifetime(T1) as a function of its frequency in the bandgap. It isevident that the qubit lifetime drastically increases in-side the bandgap, where spontaneous emission into theoutput port is greatly suppressed due to the reduced lo-calization length of the photon bound state. Deep withinthe bandgap one observes the appearance of multiple nar-row spectral features in the measured frequency depen-dence of the qubit lifetime. These features, attributableto parasitic “box” modes of our chip packaging, high-light the ability of the metamaterial waveguide to enableeffectively-dissipation-free probing of the qubit’s environ-ment over a broad spectral range (> 1 GHz). As thequbit frequency approaches the band-edges, the lifetimeis sharply reduced because of the increase in the localiza-tion length of the waveguide modes. The slope of the life-time curve at the band-edge can be shown to be directlyproportional to the group delay, |∂Trad/∂ω| = Tradτdelay

(see App. E). We observe a 24-fold enhancement in thelifetime of the qubit near the upper band-edge, corre-sponding to a maximum group index of ng = 450 right

5

Delay, τ (μs)

Exci

ted

stat

e po

pula

tion

LDOS

LDOS

τ = 11 μs

τ = 1.1 μs

τ = 2.7 μs

τ = 5.7 μs

100

10-1

100

10-1

0 10 20 30

0 2 4 6 8 10 12

a

b

FIG. 4. State-selective enhancement and inhibition ofradiative decay. a, Measurement with the g-e transitiontuned into the bandgap, with the f -e transition in the lowertransmission band. b, Measurement with the g-e transitiontuned into the upper transmission band, with the f -e tran-sition in the bandgap. For measuring the f -e lifetime, weinitially excite the third energy level |f〉 via a two-photonπ pulse at the frequency of ωgf/2. Following the popula-tion decay in a selected time interval, the population in |f〉 ismapped to the ground state using a second π pulse. Finallythe ground state population is read using the dispersive shiftof a close-by band-edge resonance of the waveguide. g-e (f -e)transition data shown as red squares (blue circles)

at the band-edge.In addition to radiative decay into the output channel,

losses in the resonators in the waveguide also contributeto the qubit’s excited state decay. Using a low powerprobe in the single-photon regime we have measured in-trinsicQ-factors of 7.2±0.4×104 for the individual waveg-uide modes between 4.6-7.4 GHz. The solid line in Fig. 3bshows a fitted theoretical curve which takes into accountthe loss in the waveguide along with a phenomenologicalintrinsic lifetime of the qubit. While the measured life-time near the upper band is in excellent agreement withthe theoretical model, the data near the lower band showssignificant departure from the model. We attribute thisdeparture in the lower band to the presence of a spuriousresonance or resonances near the lower band-edge. Possi-ble candidates for such spurious modes include the asym-metric “slotline” modes of the metamaterial waveguide,which are weakly coupled to our symmetrically groundedCPW line but may couple to the qubit. Further studyof the spectrum of these modes and possible methods forsuppressing them using cross-over connections [47] willbe a topic of future studies.

The sharp variation in the photonic LDOS near themetamaterial waveguide band-edges may also be used to

engineer the multi-level dynamics of the qubit. A trans-mon qubit, by construct, is a nonlinear quantum oscilla-tor and thus it has a multilevel energy spectrum. In par-ticular, a third energy level (|f〉) exists at the frequencyωgf = 2ωge−EC/~. Although the transition g-f is not al-lowed because of the selection rules, the f -e transition isallowed and has a dipole moment that is

√2 larger than

the fundamental transition [20]. This is reminiscent ofthe scaling of transition amplitudes in a harmonic oscil-lator and results in a second transition lifetime that is halfof the fundamental transition lifetime for a uniform den-sity of states in the electromagnetic bath. Nonetheless,the sharply varying density of states in the metamaterialcan lead to strong suppression or enhancement of thespontaneous emission for each transition. Figure 4 showsthe measured lifetimes of the two transitions for two dif-ferent spectral configurations. In the first scenario, weenhance the ratio of the lifetimes Teg/Tfe by situating thefundamental transition frequency inside in the bandgapwhile having the second transition positioned inside thelower transmission band. The situation is reversed in thesecond configuration, where the fundamental frequencyis tuned to be within the upper energy band while thesecond transition lies inside the gap. In our fabricatedqubit, the second transition is 290 MHz lower than thefundamental transition frequency at zero flux bias, whichallows for achieving large lifetime contrast in both con-figurations.

Compact, low loss, low disorder superconducting meta-materials, as presented here, can help realize more scal-able superconducting quantum circuits with higher levelsof complexity and functionality in several regards. Theyoffer a method for densely packing qubits – both in spa-tial and frequency dimensions – with isolation from theenvironment by operation in forbidden bandgaps, and yetwith controllable connectivity achieved via bound qubit-waveguide polaritons [5, 12]. Moreover, the ability to se-lectively modify the transition lifetimes provides simul-taneous access to long-lived metastable qubit states aswell as short-lived states strongly coupled to waveguidemodes. This approach realizes an effective Λ-type levelstructure for the transmon, and can be used to createstate-dependent bound state localization lengths, quan-tum nonlinear media for propagating microwave pho-tons [14, 48, 49], or as recently demonstrated, to realizespin-photon entanglement and high-bandwidth itinerantsingle microwave photon detection [50, 51]. Combined,these attributes provide a unique platform for studyingthe many-body physics of quantum photonic matter [52–55].

ACKNOWLEDGMENTS

We would like to thank Paul Dieterle, Ana AsenjoGarcia and Darrick Chang for fruitful discussions re-garding waveguide QED. This work was supported bythe AFOSR MURI Quantum Photo4nic Matter (grant

6

16RT0696), the AFOSR MURI Wiring Quantum Net-works with Mechanical Transducers (grant FA9550-15-1-0015), the Institute for Quantum Information andMatter, an NSF Physics Frontiers Center (grant PHY-

1125565) with support of the Gordon and Betty MooreFoundation, and the Kavli Nanoscience Institute at Cal-tech. M.M. (A.J.K., A.S.) gratefully acknowledges sup-port from a KNI (IQIM) Postdoctoral Fellowship.

[1] J. M. Raimond and S. Haroche, Exploring the quantum(Oxford University Press, Oxford, 2006).

[2] A. Reiserer and G. Rempe, Reviews of Modern Physics87, 1379 (2015).

[3] P. Yao, C. Van Vlack, A. Reza, M. Patterson, M. M.Dignam, and S. Hughes, Phys. Rev. B 80 (2009).

[4] A. Goban, C. L. Hung, J. D. Hood, S. P. Yu, J. A. Muniz,O. Painter, and H. J. Kimble, Phys. Rev. Lett. 115,063601 (2015).

[5] G. Calajo, F. Ciccarello, D. Chang, and P. Rabl, Phys.Rev. A 93, 033833 (2016).

[6] V. P. Bykov, Soviet Journal of Quantum Electronics 4,861 (1975).

[7] E. Yablonovitch, Phys. Rev. Lett. 58, 2059 (1987).[8] S. John and J. Wang, Phys. Rev. Lett. 64, 2418 (1990).[9] A. G. Kofman, G. Kurizki, and B. Sherman, Journal of

Modern Optics 41, 353 (1994).[10] J. D. Hood, A. Goban, A. Asenjo-Garcia, M. Lu, S.-P.

Yu, D. E. Chang, and H. J. Kimble, Proceedings of theNational Academy of Sciences 113, 10507 (2016).

[11] E. Munro, L. C. Kwek, and D. E. Chang, New Journalof Physics 19, 083018 (2017).

[12] J. S. Douglas, H. Habibian, C. L. Hung, A. V. Gorshkov,H. J. Kimble, and D. E. Chang, Nature Photonics 9, 326(2015).

[13] E. Shahmoon, G. Kurizki, H. P. Stimming, I. Mazets,and P. Grisins, Optica 3, 725 (2016).

[14] J. S. Douglas, T. Caneva, and D. E. Chang, Phys. Rev.X 6, 031017 (2016).

[15] E. Vetsch, D. Reitz, G. Sague, R. Schmidt, S. T.Dawkins, and A. Rauschenbeutel, Phys. Rev. Lett. 104,203603 (2010).

[16] S.-P. Yu, J. D. Hood, J. A. Muniz, M. J. Martin,R. Norte, C.-L. Hung, S. M. Meenehan, J. D. Cohen,O. Painter, and H. J. Kimble, App. Phys. Lett. 104,111103 (2014).

[17] A. Javadi, I. Sollner, M. Arcari, S. L. Hansen, L. Midolo,S. Mahmoodian, G. Kirsanske, T. Pregnolato, E. Lee,J. Song, S. Stobbe1, and P. Lodahl, Nature Comm. 6,8655 (2015).

[18] M. K. Bhaskar, D. D. Sukachev, A. Sipahigil, R. E.Evans, M. J. Burek, C. T. Nguyen, L. J. Rogers, P. Siyu-shev, M. H. Metsch, H. Park, F. Jelezko, M. Loncar, andM. D. Lukin, Phys. Rev. Lett. 118, 223603 (2017).

[19] A. Blais, R.-S. Huang, A. Wallraff, S. M. Girvin, andR. J. Schoelkopf, Phys. Rev. A 69, 062320 (2004).

[20] J. Koch, T. M. Yu, J. Gambetta, A. A. Houck, D. I.Schuster, J. Majer, A. Blais, M. H. Devoret, S. M. Girvin,and R. J. Schoelkopf, Phys. Rev. A 76, 042319 (2007).

[21] R. Barends, J. Kelly, A. Megrant, D. Sank, E. Jef-frey, Y. Chen, Y. Yin, B. Chiaro, J. Mutus, C. Neill,P. O’Malley, P. Roushan, J. Wenner, T. C. White, A. N.Cleland, and J. M. Martinis, Phys. Rev. Lett. 111,080502 (2013).

[22] Y. Chen, C. Neill, P. Roushan, N. Leung, M. Fang,R. Barends, J. Kelly, B. Campbell, Z. Chen, B. Chiaro,A. Dunsworth, E. Jeffrey, A. Megrant, J. Y. Mu-tus, P. J. J. O’Malley, C. M. Quintana, D. Sank,A. Vainsencher, J. Wenner, T. C. White, M. R. Geller,A. N. Cleland, and J. M. Martinis, Phys. Rev. Lett. 113,220502 (2014).

[23] A. Wallraff, D. I. Schuster, A. Blais, L. Frunzio, R. S.Huang, J. Majer, S. Kumar, S. M. Girvin, and R. J.Schoelkopf, Nature 431, 162 (2004).

[24] D. M. Pozar, Microwave Engineering, 4th (Edition-JohnWiley, 1998).

[25] Y. Liu, Y. Liu, A. A. Houck, and A. A. Houck, NaturePhysics 13, 48 (2017).

[26] J. D. Joannopoulos, S. G. Johnson, J. N. Winn, andR. D. Meade, “Photonic crystals: molding the flow oflight,” (2011).

[27] D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, and S. Schultz, Phys. Rev. Lett. 84, 4184 (2000).

[28] T. Itoh, Electromagnetic metamaterials: transmissionline theory and microwave applications (the engineeringapproach) (New Jersey: A John Wiley & Sons Inc, 2006).

[29] T. Koschny, C. M. Soukoulis, and M. Wegener, Journalof Optics 19, 084005 (2017).

[30] A. Alu and N. Engheta, Journal of Optics 19, 084008(2017).

[31] H.-T. Chen, A. J. Taylor, and N. Yu, Rep. Prog. Phys.79, 076401 (2016).

[32] P. Genevet, F. Capasso, F. Aieta, M. Khorasaninejad,and R. Devlin, Optica 4, 139 (2017).

[33] M. Goppl, A. Fragner, M. Baur, R. Bianchetti, S. Fil-ipp, J. M. Fink, P. J. Leek, G. Puebla, L. Steffen, andA. Wallraff, Journal of Applied Physics 104, 113904(2008).

[34] A. Megrant, C. Neill, R. Barends, and B. Chiaro, App.Phys. Lett. 100, 113510 (2012).

[35] K. O’Brien, C. Macklin, I. Siddiqi, and X. Zhang, Phys.Rev. Lett. 113, 157001 (2014).

[36] C. Macklin, K. O’Brien, D. Hover, M. E. Schwartz,V. Bolkhovsky, X. Zhang, W. D. Oliver, and I. Siddiqi,Science 350, 307 (2015).

[37] T. C. White, J. Y. Mutus, I. C. Hoi, R. Barends,B. Campbell, Y. Chen, Z. Chen, B. Chiaro,A. Dunsworth, E. Jeffrey, J. Kelly, A. Megrant, C. Neill,P. J. J. O’Malley, P. Roushan, D. Sank, A. Vainsencher,J. Wenner, S. Chaudhuri, J. Gao, and J. M. Martinis,App. Phys. Lett. 106 (2015).

[38] D. C. McKay, R. Naik, P. Reinhold, L. S. Bishop, andD. I. Schuster, Phys. Rev. Lett. 114, 080501 (2015).

[39] J. C. Hernandez-Herrejon, F. M. Izrailev, and L. Tessieri,Journal of Physics A: Mathematical and Theoretical 43,425004 (2010).

[40] M. H. Devoret, Les Houches Lectures (1995).[41] J. Gao, The physics of superconducting microwave res-

onators, Ph.D. thesis (2008).

7

[42] J. Zhou, M. J. Lancaster, and F. Huang, in MicrowaveSymposium Digest, 2003 IEEE MTT-S International.(2003).

[43] A. J. Keller, P. B. Dieterle, M. Fang, B. Berger, J. M.Fink, and O. Painter, App. Phys. Lett. 111, 042603(2017).

[44] D. S. Wiersma, P. Bartolini, A. Lagendijk, and R. Righ-ini, Nature 390, 671 (1997).

[45] A. Wallraff, D. I. Schuster, A. Blais, L. Frunzio, J. Majer,M. H. Devoret, S. M. Girvin, and R. J. Schoelkopf, Phys.Rev. Lett. 95, 060501 (2005).

[46] S. John and J. Wang, Phys. Rev. B 43, 12772 (1991).[47] Z. Chen, A. Megrant, J. Kelly, R. Barends, J. Bochmann,

Y. Chen, B. Chiaro, A. Dunsworth, E. Jeffrey, J. Y. Mu-tus, P. J. J. O’Malley, C. Neill, P. Roushan, D. Sank,A. Vainsencher, J. Wenner, T. C. White, A. N. Cleland,and J. M. Martinis, App. Phys. Lett. 104 (2014).

[48] G. Nikoghosyan and M. Fleischhauer, Phys. Rev. Lett.105, 013601 (2010).

[49] A. Albrecht, T. Caneva, and D. E. Chang, New Journalof Physics 19, 115002 (2017).

[50] K. Inomata, Z. Lin, K. Koshino, W. D. Oliver, J.-S. Tsai,T. Yamamoto, and Y. Nakamura, Nature Communica-tions 7, 12303 (2016).

[51] J.-C. Besse, S. Gasparinetti, M. C. Collodo, T. Walter,P. Kurpiers, M. Pechal, C. Eichler, and A. Wallraff,arXiv:1711.11569 (2017).

[52] A. D. Greentree, C. Tahan, J. H. Cole, and L. C. L.Hollenberg, Nature Physics 2, 856 (2006).

[53] M. J. Hartmann, F. G. S. L. Brandao, and M. B. Plenio,Nature Physics 2, 849 (2006).

[54] A. A. Houck, H. E. Tureci1, and J. Koch, Nature Physics8, 292 (2012).

[55] C. Noh and D. G. Angelakis, Rep. Prog. Phys. 80, 016401(2017).

[56] B. M. Karyamapudi and J.-S. Hong, IEEE MTT-S In-ternational Microwave Symposium - IMS 2003 3, 1619(2003).

[57] V. Dossetti-Romero, F. M. Izrailev, and A. A. Krokhin,Physica E: Low-dimensional Systems and Nanostructures25, 13 (2004).

[58] F. M. Izrailev, T. Kottos, and G. P. Tsironis, Phys. Rev.B 52, 3274 (1995).

[59] P. B. Dieterle, M. Kalaee, J. M. Fink, and O. Painter,Phys. Rev. Applied 6, 014013 (2016).

[60] D. L. Underwood, W. E. Shanks, J. Koch, and A. A.Houck, Phys. Rev. A 86, 023837 (2012).

Appendix A: Band structure calculation

1. Quantization of a periodic resonator-loadedwaveguide

We consider the case of a waveguide that is periodi-cally loaded with microwave resonators. Fig. 5 depicts aunit cell for this configuration. The Lagrangian for this

system can be readily written as [40]

L =∑n

[1

2C0[Φan]

2 −[Φan − Φan−1]2

2L0

+1

2Cr[Φ

bn]

2+

1

2Ck[Φan − Φbn]

2 − [Φbn]2

2Lr

]. (A1)

In order to find solutions in form of traveling waves, it iseasier to work with the Fourier transform of node fluxes.We use the following convention for defining the (dis-crete) Fourier transformation

Φa,bκ =1√M

N∑n=−N

e−i2π(κ/M)nΦa,bn , (A2)

where M = 2N + 1 is the total number of periods inthe waveguide. Using the Fourier relation we find theLagrangian in k-space as

L =∑κ

[1

2(C0 + Ck)|Φaκ|

2 −∣∣∣1− e−i2π(κ/M)

∣∣∣2 |Φaκ|22L0

1

2(Ck + Cr)|Φbκ|

2 − |Φbκ|

2

2Lr− Ck

ΦbκΦa−κ + Φb−κΦaκ2

].

(A3)

To proceed further, we need to find the canonical nodecharges which are defined as Qa,bκ = ∂L

∂Φa,bκ, and subse-

quently derive the Hamiltonian of the system by using aLegendre transformation. Doing so we find

H =∑κ

[QaκQ

a−κ

2C ′0+∣∣∣1− e−i2π(κ/M)

∣∣∣2 ΦaκΦa−κ2L0

+

QbκQb−κ

2C ′r+

ΦbκΦb−κ2Lr

+QaκQ

b−κ +Qa−κQ

bκ

2C ′k

]. (A4)

Here, we have defined the following quantities

C ′0 =CkCr + CkC0 + C0Cr

Ck + Cr, (A5)

C ′r =CkCr + CkC0 + C0Cr

Ck + C0, (A6)

C ′k =CkCr + CkC0 + C0Cr

Ck. (A7)

The canonical commutation relation[Φiκ, Q

j−κ′ ] = i~δi,jδκ,κ′ allows us to define the fol-

lowing annihilation operators as a function of chargeand flux operators

aκ =

√C ′0Ωk

2~

(Φaκ +

i

C ′0ΩkQaκ

), (A8)

bκ =

√C ′rω0

2~

(Φbκ +

i

C ′rω0Qbκ

). (A9)

8

· · ·Φa

n−1

L0

Φan

L0

Φan+1 · · ·

C0

Ck

Φbn+1

CrLr

Ck

Φbn

CrLr

C0

· · · · · ·

· · ·L0

Lr Cr

LkC0

L0

Lr Cr

Lk

· · ·

C0

· · · · · ·

FIG. 5. Circuit diagram of metamaterial waveguides made from periodic arrays of transmission line sections loaded withcapacitively coupled resonators (top), and inductively loaded resonators (bottom).

Here, we have defined the resonance frequency for eachmode as

Ωk =

√4sin2(kd/2)

L0C ′0, (A10)

ω0 =1√LrC ′r

, (A11)

where k = (2πκ)/(Md) is the wavenumber. It is evidentthat Ωk has the expected dispersion relation of a discreteperiodic transmission line and ω0 is the resonance fre-quency of the loaded microwave resonators. Using the

above definitions for aκ, bκ

H =~2

∑k

[Ωk

(a†kak + a−ka

†−k

)+ ω0

(b†k bk + b−k b

†−k

)− gk

(b−k − b†k

)(ak − a†−k

)− gk

(a†k − a−k

)(b†−k − bk

)],

(A12)

along with the coupling coefficient

gk =

√C ′0C

′r

2C ′k

√ω0Ωk =

Ck√ω0Ωk

2√

(C0 + Ck)(Cr + Ck).

(A13)

An alternative structure for coupling microwave res-onators is depicted in the bottom panel of Fig. 5. Inthis geometry, the coupling is controlled by the inductiveelement Lk. Repeating the analysis above for this case,we find

Ωk =

√4sin2(kd/2)

C0L′0, (A14)

ω0 =1√CrL′r

, (A15)

gk =

√L′0L

′r

2L′k

√ω0Ωk. (A16)

We have defined the modified inductance values as

L′0 =LkLr + LkL0 + L0Lr

Lk + Lr, (A17)

L′r =LkLr + LkL0 + L0Lr

Lk + L0, (A18)

L′k =LkLr + LkL0 + L0Lr

Lk. (A19)

2. Band structure calculation with RWA

Using the rotating wave approximation, the Hamilto-nian in Eq. (A12) can be simplified to

H = ~∑k

[Ωka

†kak + ω0b

†k bk + gk

(b†kak + a†k bk

)],

(A20)

The simplified Hamiltonian can be written in the com-pact form

H = ~∑k

x†kHkxk, (A21)

9

where

Hk =

[Ωk gkgk ω0

],xk =

[akbk

]. (A22)

We desire to transform the Hamiltonian to a diagonalizedform

Hk =

[ω+,k 0

0 ω−,k

]. (A23)

It is straightforward to use the eigenvalue decompositionto find ω±,k as

ω±,k =1

2

[(Ωk + ω0)±

√(Ωk − ω0)

2+ 4g2

k

], (A24)

along with the corresponding eigenstates

α±,k =(ω±,k − ω0)√

(ω±,k − ω0)2

+ g2k

ak +gk√

(ω±,k − ω0)2

+ g2k

bk.

(A25)

3. Band structure calculation beyond RWA

The exact Hamiltonian in Eq. (A12) can be written inthe compact form

H =~2

∑k

x†kHkxk, (A26)

where

Hk =

Ωk 0 gk −gk0 Ωk −gk gkgk −gk ω0 0−gk gk 0 ω0

,xk =

aka†−kbkb†−k

. (A27)

To find the eigenstates of the system, we can use a lin-ear transform to map the state vector xk = Skxk such

that x†kHkxk = x†kHkxk with the transformed diagonalHamiltonian matrix

Hk =

ω+,k 0 0 00 ω+,k 0 00 0 ω−,k 00 0 0 ω−,k.

(A28)

In order to preserve the canonical commutation relations,

the matrix Sk has to be symplectic, i. e. Sk = SkJS†k,

with the matrix J defined as

J =

1 0 0 00 −1 0 00 0 1 00 0 0 −1.

(A29)

A linear transformation (such as Sk) that diagonalizes aset of quadratically coupled boson fields while preserving

their canonical commutation relations is often referred toas a Bogoliubov-Valatin transformation. While it is gen-erally difficult to find the transform matrix Sk, it is easyto find the eigenvalues of the diagonalized Hamiltonianby exploiting some of the properties of Sk. Note that

since Sk = SkJS†k, the matrices JHk and JHk share the

same set of eigenvalues. The eigenvalues of JHk are thetwo frequencies ω±,k, and thus we have

ω2±,k =

1

2

[(Ω2k + ω2

0

)±√

(Ω2k − ω2

0)2

+ 16ω0Ωkg2k

].

(A30)

4. Circuit theory derivation of the band structure

Consider the pair of equations that describe the prop-agation of a monochromatic electromagnetic wave of theform v(x, t) = V (x)e−ikxeiωt (along with the correspond-ing current relation) inside a transmission line

d

dxV (x) = −Z(ω)I(x),

d

dxI(x) = −Y (ω)V (x). (A31)

Here, Z(ω) and Y (ω) are frequency dependent impedanceand admittance functions that model the linear responseof the series and parallel portions of a transmissionline with length d. It is straightforward to check thatthe solutions to these equation satisfy k(ω) = nω/c =√−Z(ω)Y (ω)/d. For a loss-less waveguide and in the

absence of dispersion we have Z(ω) = iωL0 and Y (ω) =iωC0, and thus we find the familiar dispersion relationk(ω) = ω

√L0C0/d. Nevertheless, the pair of equations

above remain valid for arbitrary impedance and admit-tance functions Z(ω) and Y (ω), provided that the di-mension of the model circuit remains much smaller thanthe wavelength under consideration. In this model, a realand negative quantity for the product ZY results in animaginary wavenumber and subsequently creates a stopband in the dispersion relation. This situation can beachieved by periodically loading a transmission line withan array of resonators [35, 56]. Assuming a unit lengthof d we find

k2 =(ωc

)2

n2

[1 +

2cγend

1

ω20 − ω2

]. (A32)

Here, ω0 is the resonance frequency, and γe is the exter-nal coupling decay rate of an individual resonator in thearray. For moderate values of gap-midgap ratio (∆/ωm),the frequency gap can be found as

∆ =c

nd

(γeω0

), (A33)

and ωm = ω0 + ∆/2. We have defined the gap as therange of frequencies where the wavenumber is imaginary.

10

Although a microwave resonator can be realized by us-ing a two-elements LC-circuit, the three-element circuitsin Fig. 5 provide an additional degree of freedom whichenables setting the coupling γe independent of the reso-nance frequency ω0. Using circuit theory, it is straight-forward to show

ω0 =1√

Lr(Cr + Ck), (A34)

γe =Z0

2Lr

(Ck

Cr + Ck

)2

. (A35)

Here, Z0 is the characteristic impedance of the unloadedwaveguide. It is easy to check that for small values ofCk/Cr, the resonance frequency is only a weak functionof Ck. As a result, it is possible to adjust the couplingrate γe by setting the capacitor Ck while keeping the res-onance frequency almost constant. Figure 5 also depictsan alternative strategy for coupling microwave resonatorsto the waveguide. In this design, the inductive elementLk is used to set the coupling in a “current divider” ge-ometry. We provide experimental results for implemen-tation of bandgap waveguide based on both designs inthe next section.

While the “continuum” model described above pro-vides a heuristic explanation for formation of bandgap ina waveguide loaded with resonators, its results remainsvalid as far as k 2π/d. To avoid this approximation,we can use the transfer matrix method to find the exactdispersion relation for a system with discrete periodicsymmetry [24]. In this case Eq. (A32) is modified to

cos (kd) = 1−(ωc

)2n2d2

2− ndγe

c

ω2

ω20 − ω2

. (A36)

Note that this relation still requires d to be much smallerthan the wavelength of the unloaded waveguide λ =2πc/(nω).

5. Dispersion and group index near the band-edges

Equation (A30) can be reversed to find the wavenum-ber k as a function of frequency. Assuming, a lineardispersion relation of the form k = nΩk/c for the barewaveguide we find

k =nω

c

√ω2 − ω2

c+

ω2 − ω2c−. (A37)

Here, ωc+ = ω0 and ωc− = ω0

√1− 4g2

k/(Ωkω0) are theupper and lower cut-off frequencies, respectively. Thequantity g2

k/(Ωkω0) is a unit-less parameter quantify-ing the size of the bandgap and is independent of thewavenumber k.

The dispersion relation can be written in simpler formsby expanding the wavenumber in the vicinity of the two

band-edges

k =

nωc−c

√∆−δ− for ω ≈ ωc−,

nωc+c

√δ+∆ for ω ≈ ωc+.

(A38)

Here, ∆ = ωc+−ωc− is the frequency span of the bandgapand δ± = ω−ωc± are the detunings from the band-edges.

The form of the dispersion relation Eq. (A30) suggeststhat the maxima of the group index happens near theband-edges. Having the wavenumber, we can readilyevaluate the group velocity vg = ∂ω/∂k and find thegroup index ng = c/vg as

ng =

nωc−

√∆√

−4(δ−−iγi)3for ω ≈ ωc−,

nωc+√4∆(δ+−iγi)

for ω ≈ ωc+.(A39)

Note that we have replaced δ± with δ± − iγi to accountfor finite internal quality factor of the resonators in thestructure.

Appendix B: Coupling a Josephson junction qubit toa metamaterial waveguide

We consider the coupling of a Josephson junction qubitto the metamaterial waveguide. Assuming rotating waveapproximation, the Hamiltonian of this system can bewritten as

H = ~∑k

[ωka

†kak +

ωq2σz + fk

(a†kσ

− + akσ+).

](B1)

Here fk is the coupling factor of the qubit to the waveg-uide photons, and ωk = ω±,k, where the plus or minussign is chosen such that the qubit frequency ωq lies withinthe band. Without loss of generality, we assume fk to bea real number. The Heisenberg equations of motions forthe qubit and the photon operators can be written as

∂

∂tak = −iωkak − ifkσ− (B2)

∂

∂tσ− = −iωqσ− − i

∑k

fkak (B3)

The equation for ak can be formally integrated and sub-stituted in the equation for σ− to find

∂

∂tσ− = −iωqσ− − i

∑k

fke−iωk(t−t0)ak(t0) (B4)

−∑k

f2k

∫ t

t0

e−i(ωk)(t−τ)σ−(τ)dτ.

11

We now use the Markov approximation to write σ−(τ) ≈σ−(t)e−i(ωq)(τ−t), and thus

∂

∂tσ− = −iωqσ− − i

∑k

fke−iωk(t−t0)ak(t0) (B5)

−∑k

f2k

(∫ t

t0

e−i(ωk−ωq)(t−τ)dτ

)σ−(t).

Considering the generic equation of motion for a linearlydecaying qubit, (∂/∂t)σ− = −iωqσ− − (γ/2)σ−, we canidentify real part of the last term in the equation aboveas the decay rate due to radiation of the qubit into thewaveguide. We can extend the integral’s bound to ap-proximately evaluate this term as

γ = 2 Re

[∑k

f2k

∫ t

t0

e−i(ωk−ωq)(t−τ)dτ

]

≈ 2 Re

[∑k

f2k

∫ ∞t0

e−i(ωk−ωq)(t−τ)dτ

]= 2π

∑k

f2kδ(ωk − ωq). (B6)

Assuming the coupling rate fk is a smooth function ofthe k-vector, we can evaluate this some in the continuumlimit as

γ = 2π∑k

f2kδ(ωk − ωq) (B7)

≈Md

∫dkf2

kδ(ωk − ωq) (B8)

= L

∫dω

(∂k

∂ω

)f2kδ(ωk − ωq) (B9)

=L

cf(ωq)

2ng(ωq). (B10)

It is evident that reducing the group velocity increasesthe radiation decay rate of the qubit.A similar analysiscan be applied to find the decay rate of a linear cavitywith resonance frequency of ω0 (i.e. a harmonic oscilla-tor) that has been coupled to the waveguide with cou-pling constant g(ω). In this case we find

γ =L

cg(ω0)2ng(ω0),

Qe = ω0/γ =ω0c

L

1

g(ω0)2ng(ω0). (B11)

Appendix C: Disorder and Anderson localization

Propagation of electron waves in a one dimensionalquasi-periodic potential is described by[− ∂2

∂x2 +∑n

(U + Un)δ(x− an)

]ψq(x) = q2ψq(x).

(C1)

Here, q is the quasi momentum and Un is the randomvariable that models compositional disorder at positionx = na. Disorder leads to localization of waves with acharacteristic length defined as

`−1 = limN→∞

⟨1

N

N−1∑n=0

ln

∣∣∣∣ψn+1

ψn

∣∣∣∣⟩. (C2)

Here, the brackets represent averaging over different real-ization of the disorder, whereas the summation accountsfor spatial/temporal averaging for traveling waves. Forthis model, previous authors have found the localizationlength to be [39, 57, 58]

`

d=

2Γ(1/6)

61/3√πσ−2/3 ≈ 3.45σ−2/3. (C3)

In this model σ2 = 〈U2n〉 sin2 (q0a)/q2

0 is a parameter thatquantifies the strength of disorder, and q0 is the value ofquasi-momentum at the band-edge.

Now, we consider the propagation of current waves ina one dimensional waveguide that has been periodicallyloaded with resonators (a similar analysis can be appliedto the voltage waves for the case of inductively coupledresonators). Starting from Eq. (A32), it is straightfor-ward to find

∂2I(x)

∂x2 + I(x)(ωc

)2

n2

[1 +

∑n

d∆δ(x− an)

ω0,n − ω + iγi

]= 0.

(C4)

By comparing this equation with the Schrodinger equa-tion for the Kronig-Penny model Eq. (C1) we find

q2 →(ωc

)2

n2

U + Un → −(ωc

)2

n2

[d∆

ω0,n − ω + iγi

]. (C5)

For small variation in resonance frequencies, δω0, we canexpand the resonance potential term to find

Un = −(ω0

c

)2

n2 ∂

∂ω0,n

(d∆

ω0,n − ω + iγi

)δω0 (C6)

By evaluating the expression for Un and substituting itin the relation above for σ2, we find

σ2low =

(γeγi

)4(δω0

∆

)2

,

σ2high =

(γe∆

)4(δω0

∆

)2

. (C7)

The analysis above gives us the localization lengthfrom disorder, `diss. In addition to disorder, the loss inthe waveguide leads to an exponential extinction of thewave’s amplitude. The localization length from loss, `loss,

12

5.88 5.89 5.9 5.91 5.92

-50

-40

-30

-20

-10

Q e =325

Q i = 485562

5.88 5.89 5.9 5.91 5.92-1.5

-1

-0.5

0

0.5

1

1.5

Frequency (GHz)Frequency (GHz)

6.098 6.1 6.102 6.104 6.106

-40

-30

-20

-10

0

Q e =1098

Q i = 256332

6.098 6.1 6.102 6.104 6.106

-1

-0.5

0

0.5

1

)zHG( ycneuqerF)zHG( ycneuqerF

ba

dc

5.2 5.4 5.6 5.8 6 6.2 6.4 6.6 6.8 7Frequency (GHz)

-0.01

-0.005

0

0.005

0.01

arg(S 21

)

S 21 (d

B)S 21

(dB)

arg(S 21

)

FIG. 6. a, Optical and SEM images of microwave resonator array chip. Middle: optical image of the chip with two arraysof coupled resonators on a 1 × 1 cm silicon chip. Left and Right: SEM image (false-color) of the fabricated inductively (left)and capacitively (right) coupled microwave resonator pairs. The resonator region is colored red and the waveguide centralconductor is colored blue. b-c, Amplitude and phase response of two capacitively-coupled microwave resonator pairs measuredat the fridge temperature Tf ≈ 7 mK. The legends show the intrinsic (Qi = ω0/γi) and extrinsic (Qe = ω0/γe) quality factorsextracted from a Fano line shape fit. c, Statistical variations in the resonance frequency of 9 resonators with a wire width of500 nm. The dashed lines mark the standard deviation of the normalized error equal to σ = 0.3%.

can be found by solving for the complex band structureand setting `loss = 1/Im(k). Finally, the total localiza-tion length can be found by adding the effect of bothcontributions as

1

`total=

1

`diss+

1

`loss. (C8)

Appendix D: Characterization of lumped-elementmicrowave resonators

We have achieved a characteristic size of λ0/150(130µm by 76µm for ω0/2π ≈ 6 GHz) and λ0/76 (155µmby 92µm for ω0/2π ≈ 10 GHz), using a wire width of 500nm and 1 µm, respectively.

Figure 6 shows the typical amplitude and phase of mea-sured for a waveguide coupled to a pair of identical res-onators. Microwave spectroscopy of the fabricated res-onators is performed in a dilution refrigerator cooled-down to a temperature of Tf ≈ 7 mK. The input mi-crowave is launched onto the chip via a 50-Ω CPW. Theoutput microwave signal is subsequently amplified andanalyzed using a network analyzer (for more details re-garding the measurement setup, refer to Ref. [59]). Wehave extracted the intrinsic and extrinsic decay rates ofthe cavity by fitting the transmission data to a Fano line

shape of the form

S21(ω) = 1− γeeiφ0

γi + γe + 2i(ω − ω0). (D1)

Here γe and γi are the extrinsic and intrinsic decay ratesof the resonator, respectively. The phase φ0 is a param-eter that sets the asymmetry of the Fano line shape [41].The data demonstrates that it is possible to adjust theexternal coupling to the resonator in a wide range with-out much degradation in the internal quality factor (itis straightforward to convert the extrinsic quality factorQe to the coupling constants gk used in our theoreticalanalysis above). We have compared the measured reso-nance frequency with the resonance frequency found fromnumerical simulations in Fig. 6d. We find that the mea-sured resonance frequencies are in agreement with thesimulated values, with a multiplicative scaling factor of0.85. Using this scale factor, we have measured a randomvariation 0.3% in the resonance frequency. It has beenpreviously suggested that the shift in the resonance fre-quency and its statistical variation can be attributed tothe kinetic inductance of the free charge carriers in thesuperconductor, and the variations can be mitigated byincreasing the wire width [60].

13

Appendix E: Qubit frequency shift and thePurcell-limited lifetime

The qubit frequency shift can be derived from circuittheory by modeling the qubit as a linear resonator. Con-sider the circuit diagram in Fig. 7. The load impedanceseen from the qubit port can be written as

ZL(ω) =1

iωCg+ Zline(ω), (E1)

and

YL(ω) =iωCg

1 + Zline(ω)iωCg. (E2)

For weak coupling, the decay rate can be found using thereal part of the load impedance as

κ ' ω2qLJ Re [YL(ωq)]. (E3)

Here, ωq is the resonance frequency of the qubit. Simi-larly, the shift in qubit frequency is found as

∆ωq ' −ω2qLJ

2Im [YL(ωq)] . (E4)

For a transmon qubit, we have the following relationthat approximate its behavior in the linearized regime

LJ =

(Φ0

2π

)2EJ

, (E5)

ωq =1√LJCq

. (E6)

We first use the simplified continuum model to find theinput impedance Zline

Zline(ω) = ZB(ω)RL + ZB(ω) tanh [Im (k)x]

ZB(ω) +RL tanh [Im (k)x]. (E7)

Here, Im (k)(ω) is the attenuation constant (we are as-suming Re (k)(ω) = 0, i.e. valid when the value of ω iswithin the bandgap), ZB(ω) is the Bloch impedance ofthe periodic structure, and x is the length of the waveg-uide. Assuming Im (k)x 1, this expression can besimplified as

Zline(ω) ≈ZB(ω) +4RL|ZB(ω)|2

RL2 + |ZB(ω)|2

e−2 Im (k)x

≈ZB(ω) + 4RLe−2 Im (k)x. (E8)

Note that we have assumed RL |ZB(ω)| to makethe last approximation. For weak coupling, the qubitcoupling capacitance, Cg, should be chosen such thatthe (magnitude of ) impedance Zg = 1/(iωCg) is muchlarger than |Zline|. In this situation, we use Eq. (E4) andEq. (E8) to find

∆ωqωq

= −1

2(LJωq)(Cgωq)−

1

2(LJωq)(Cgωq)

2Im[ZB(ωq)]

= − Cg2Cq− Cg

2CqIm[ZB(ωq)]Cgωq. (E9)

Cg

Cq

Vq

EJ

metamaterial waveguide

Zline

RL

FIG. 7. Circuit diagram for qubit that is capacitively coupledto a metamaterial waveguide with a resistive termination.

Note that the first term in the frequency shift is merelycaused by addition of the coupling capacitor to the overallqubit capacitance.

We find the qubit’s radiation decay rate by substitut-ing Eq. (E8) in Eq. (E3)

κ =4ω2

qCg2

CqRLe

−2 Im (k)(ω)x. (E10)

Subsequently, the radiative lifetime of the qubit can bewritten as

Trad =Cq

4ω2qCg

2RLe2x/`(ωq), (E11)

where ` = 1/ Im (k) is the localization length in thebandgap. We note that the analysis from circuit theoryis only valid for weak qubit-waveguide coupling rates,where the Markov approximation can be applied. Inthe strong coupling regime, the qubit frequency and life-time can be found by numerically finding the zeros ofthe circuit’s admittance function Y = YL + Yq, whereYq = iωqCq + 1/(iωqLJ).

1. Group delay and the qubit lifetime profile

Equation (E11) demonstrates the relation between thequbit lifetime and the localization length. Moving thequbit frequency beyond the gap, results in a drastic in-crease in the localization length and subsequently reducesthe qubit lifetime. The normalized slope of the lifetimeprofile in the vicinity of the band-edge can be written as∣∣∣∣ 1

Trad

∂Trad

∂ω

∣∣∣∣ =

∣∣∣∣x∂ Im (k)

∂ω

∣∣∣∣ = |x Im (ng)/c| . (E12)

14

We now evaluate Eq. (A39) to find the group index at theupper and lower band-edges δ± = 0

|Re(ng)| = | Im(ng)| =

nωc−√

∆8γi3

for ω = ωc−,

nωc+1√

8∆γifor ω = ωc+.

(E13)

Consequently, we can write the normalized slope of thelifetime profile at the band-edge as(

1

Trad

∣∣∣∣∂Trad

∂ω

∣∣∣∣)∣∣∣∣ω=ωc±

= |x Im [ng(ωc±)]/c|

= |xRe [ng(ωc±)]/c|=τdelay. (E14)

This result has a simple description: the normalized slopeof the lifetime profile at the band-edge is equal to the(maximum) group delay.