Bosonic topological insulator intermediate state in the superconductor-insulator transition

M. C. Diamantini,1 A. Yu. Mironov,2, 3 S. V. Postolova,2, 4 X. Liu,5 Z. Hao,5 D. M. Silevitch,6 Ya. Kopelevich,7 P. Kim,5 C. A. Trugenberger,8 and V. M. Vinokur9, 10 1NiPS Laboratory, INFN and Dipartimento di Fisica e Geologia, University of Perugia, via A. Pascoli, I-06100 Perugia, Italy 2A. V. Rzhanov Institute of Semiconductor Physics SB RAS, 13 Lavrentjev Avenue, Novosibirsk, 630090 Russia 3Novosibirsk State University, Pirogova str. 2, Novosibirsk 630090, Russia 4Institute for Physics of Microstructures RAS, GSP-105, Nizhny Novgorod 603950, Russia 5Department of Physics, Harvard University, Cambridge, Massachusetts 02138, USA 6Division of Physics, Mathematics, and Astronomy, California Institute of Technology, Pasadena, CA 91125, USA 7Universidade Estadual de Campinas-UNICAMP, Instituto de F¨ı¿œ¨ı¿œsica “Gleb Wataghin”/DFA Rua Sergio Buarque de Holanda, 777, Brasil 8SwissScientific Technologies SA, rue du Rhone 59, CH-1204 Geneva, Switzerland 9Materials Science Division, Argonne National Laboratory, 9700 S. Cass. Ave, Lemont, IL 60437, USA. 10Consortium for Advanced Science and Engineering (CASE) University of Chicago 5801 S Ellis Ave Chicago, IL 60637 A low-temperature intervening metallic regime arising in the two-dimensional superconductor- insulator transition challenges our understanding of electronic fluids. Here we develop a gauge theory revealing that this emergent anomalous metal is a bosonic topological insulator where bulk transport is suppressed by mutual statistics interactions between out-of-condensate Cooper pairs and vortices and the longitudinal conductivity is mediated by symmetry-protected gapless edge modes. We explore the magnetic-field-driven superconductor-insulator transition in a niobium titanium nitride device and find marked signatures of a bosonic topological insulator behavior of the intervening regime with the saturating resistance. The observed superconductor-anomalous metal and insulator- anomalous metal dual phase transitions exhibit quantum Berezinskii-Kosterlitz-Thouless criticality in accord with the gauge theory.

An anomalous metallic regime intervening between the superconductor and insulator has been reported in a wide a b c variety of two-dimensional electronic systems experienc- ing the superconductor-to-insulator transition (SIT) 1–15, � � � and is often referred to as “Bose metal” 16. Despite decades of dedicated studies 17,18, its nature remains un- � � � clear and poses a challenge to our understanding of elec- tron fluids. The very existence of a 2D metal is at odds with the 2D orthodoxy, as conventional theories expect a Figure 1. Mutual statistics interactions. a: A charge direct quantum SIT with no intermediate metallic phase. (blue ball) evolves along the Euclidean time axis and is en- circled by a vortex (red). The two trajectories are topo- Yet a gauge theory of Josephson junction arrays (JJA) logically linked since one cannot decouple the charge tra- at T =0 19 predicted a metallic phase intervening between jectory from the circle without breaking it. This linking the superconductor and superinsulator. Below, building encodes the Aharonov-Casher effect. b: A dual situation on our field theory of the SIT 20, we extend the approach in which a vortex (red ball) evolves along the Euclidean of 19 to finite temperatures and construct a gauge theory time direction and is encircled by a charge. This linking of the Bose metal. We unravel the fundamental role of represent the Aharonov-Bohm effect. c The linked charge- infrared-dominant Aharonov-Bohm-Casher (ABC) mu- anticharge and vortex-antivortex fluctuations representing coupled Aharonov-Bohm-Casher interactions which are en-

arXiv:1906.07969v2 [cond-mat.str-el] 31 Aug 2019 tual statistics interactions determining the SIT phase coded in the Chern-Simons term in the action. structure. We report transport measurements on nio- bium titanium nitride (NbTiN) and van der Waals het- erostructures of twisted double bilayer graphene (TDBG) films offering strong support to the proposed picture. We show that strong quantum fluctuations prevent Bose condensation of both vortices and Cooper pairs (CP). The mutual statistics interactions, see Fig. 1, between T T them induce a gap in their fluctuation spectrum, quanti- symmetry-protected U(1)oZ2 edge modes, where Z2 de- 21 fied by the Chern-Simons mass, mCS , preventing bulk notes time-reversal symmetry. Hence the Bose metal re- transport. The longitudinal conductance is mediated by alizes the long sought 22 bosonic topological insulator. 2

e g

f Dimensionless SIT-driving parameter, g

Figure 2. Phase structure of the SIT. a: Superconductor: Strings with electric quantum numbers condense. b: Superinsula- tor: Strings with magnetic quantum numbers condense. c: Coexistence of long electric and magnetic stringsnear the first-order direct transition from a superconductor to a superinsulator. d: Bosonic topological insulator/Bose metal: all strings are sup- pressed by their high self-energy. e: Phase transitions induced by an external magnetic field: Direct, magnetic-field-induced transition from a superconductor to a superinsulator for η < 1. f: Magnetic-field-driven transition from a superconductor to a topological insulator for η > 1. g: Schematic phase diagram for strong quantum fluctuations, η > 1 (not to scale), near the SIT. The Bose metal/topological insulator state (BM) is separated from the superconducting (SC) and superinsulating (SI) states by vortex- and charge-BKT transitions respectively (magenta and blue solid lines respectively). In the vicinity of the TVBKT line, RRQ, towards the TCBKT line. The yellow strip depicts the domain of thermally activated resistance, which crosses over to BKT critical behavior on approach to TCBKT. The SIT-driving values g1 and g2 that mark superinsulator-Bose metal and superinsulator-Bose metal quantum transitions satisfy the duality relation g2 = 1/g1. The parameter g can be either the dimensionless conductance of the film, or magnetic field, or gate voltage in the gate-driven SIT.

PHASE STRUCTURE OF THE SIT CP near the SIT is governed by the free energy which we derive from the mixed Chern-Simons action 20, explicitly accounting for the ABC effects 21,23. To that end we inte- We consider a two-dimensional superconducting film grate out the gauge fields and arrive at the free energy of in the vicinity of the SIT at temperatures T 6 Tc0, where a system of strings carrying electric and magnetic quan- Tc0 is the mean-field temperature of formation of Cooper tum numbers Q and M and representing the Euclidean pairs with the infinite lifetime. The film harbors inter- trajectories of charges and vortices on a lattice of spacing acting elementary excitations, Cooper pairs and vortices, 2 2  `, see Appendix: F = Q /g + gM − 1/η µeηN, where with characteristic energies e2 = 4e2/` and e2 = Φ2/λ , q v 0 ⊥ the string length L = N` and g=e /e =(π/e2)p`/λ is where e is the charge (we use natural units, v q ⊥ the dimensionless tuning parameter with g=g =1 cor- c = 1, = 1, restoring physical units when necesssary), c ~ responding to the SIT. In experiments on films expe- Φ = π/e is the flux quantum, the ultraviolet (UV) cut- 0 riencing the disorder-driven SIT, the tuning parameter off of the theory is defined as ` ' min{d, ξ}, with d and is the dimensionless conductance, g = R /R , where ξ being the thickness of the film and the superconduct- Q 0 R = h/4e2 is the quantum resistance for CP, h is the ing coherence length, respectively, and λ = λ2/d is the Q ⊥ L Planck constant, and R is the sheet resistance of the film Pearl length of the film, with λ being the London pen- 0 L measured at predefined standard conditions. The param- etration depth of the bulk material. Mapping to JJA is eter g describes thus the tuning of the SIT not only by achieved via the e → 4E , e → 2π2E replacement, q C v J regulating disorder, but implicitly also by varying dissi- where E and E are the charging and Josephson cou- C J pation 24 (in JJA g = p(π2/2)(E /E )). The dimen- pling energies of the array, respectively. Since E ∼ g∆, J C J sionless parameter η describes the strength of quantum where ∆ is the gap in a superconducting granule in the fluctuations. Near the SIT, where e ≈e , η acquires the JJA and g is the dimensionless tunneling conductance be- q v form (in physical units) tween the adjacent granules, the model implicitly takes into account dissipation effects. 1 π2`  π`  η = G , (1) The behavior of an ensemble of interacting vortices and α µeλ⊥ αλ⊥ 3 where α = e2/(~c) ≈ 1/137 is the fine structure constant, the relativistic case µe is the positional entropy (see SI), and G is the diago- 2 g  qe¯  3 2  nal element of the 3D Green function describing electro- Seff (Aµ) = d d x Aµ −δµν ∇ + ∂µ∂ν Aν , magnetic interactions screened by the CS mass, see SI. 2 2π ˆ Identifying the UV cutoff with the superconducting co- (4) herence length ξ, yields the geometric factor as d/(κλ ) where Aµ is the external electromagnetic potential, the L dimensionless charge unitq ¯ = 2 for CP and the renor- where κ = λL/ξ is the Landau-Ginzburg parameter. √ malized effective chargeqe ¯ =qe ¯ g. Accordingly, the Bose condensation of charges and/or vortices means eff charge current j is found as jµ = (δ/δA )S (A ) = proliferation of strings of an arbitrary size and occurs if µ ind µ eff ν 2 µν F is negative, i.e. if g (¯qe/2π) d ∂ν F , with Fµν = ∂µAν − ∂ν Aµ. One sees that only the derivatives of the external fields, but not Q2/g + gM 2 < 1/η . (2) their constant parts, induce a current. Therefore, in the bulk, both the longitudinal and the quantum Hall com- ponents of the linear conductance vanish at T = 0. The phase emerging at particular values of g and η is However, the Chern-Simons effective action is not in- determined by the geometric condition that the nods on variant under gauge transformations not vanishing on the a square lattice of integer electric and magnetic charges, sample boundaries. To restore the gauge invariance, one {Q, M}, fall within the interior of an ellipse with semi- has to add edge chiral bosons 27. This operation is an ex- axes r = (g/η)1/2 and r = 1/(gη)1/2, see Fig. 2(a–d). Q M act analogue to the description of the edge modes in the The Bose metal emerges if none of the condensates can quantum Hall effect 28. The resulting edge action gives form i.e. if simultaneously rise to the equation of motion, see Appendix: g/η > 1 and gη > 1 . (3) qe¯ v ρ = eff V, (5) b 2π This relation resolves the enigma of why some materials where v is the velocity of the edge modes and V is the exhibit a direct SIT while others go through the inter- b applied voltage. Using I =qe ¯ v ρ, one finds the longi- mediate Bose metal phase. The direct SIT at g = 1 eff b tudinal sheet resistance corresponds to η < 1; tuning g, one crosses the region η < g < 1/η, where both vortex and Cooper pair conden- V R R ≡ = Q , (6) sates coexist, i.e. the direct SIT is a first-order quantum  I g transition. The Bose metal phase opens up for η > 1 and is favored by thicker films and for materials with which is in a full concert with the early elegant charge- vortex duality arguments 24,29 leading to R = R 29,30 low κ. Its domain is delimited by the lines g=η and  Q at the SIT. In experiments, R may well deviate from g=1/η. As we show below, these lines represent quan-  25,26 tum Berezinskii-Kosterlitz-Thouless (BKT) transi- RQ upon departure from the SIT, but the standard SIT scaling analysis yields the convergence of R (T → 0) to tions, and g=1, η=1 is a quantum tri-critical point.  2,3,15 To describe the magnetic-field-driven SIT in systems RQ upon approach to the presumed quantum criti- with g ≈ 1 i.e. which are already on the brink of the SIT, cal point. At zero temperature and η>1, the Bose metal we introduce the frustration factor f = B/BΦ, where BΦ forms for between the points g = 1/η and g = η. Hence is the magnetic field corresponding to one flux quantum its sheet resistance is lower than the quantum resistance, R < R , on the superconducting side, g > 1 and larger Φ0 = π/e per unit cell. Then the external magnetic  Q than the quantum resistance, R > R , on the insulating field shifts M → M + f and modifies the condensation  Q conditions, see Fig. 2e,f. Setting g = 1 + , with   1, side, g < 1, with equality achieved at g = 1. Duality is 2 realized in the form R /R ↔R /R when g ↔ 1/g, gen- one finds for a direct SIT at η < 1, fc = (1/2)(g − Q   Q 1) ≈ . At η > 1, but still close to the tri-critical point, eralizing the universality arguments of 24,29,30 onto the one sets g = η +  and obtains for the superconductor- Bose metal. √ 3/2 1/2 3/2 BM transition fc = /η = (g − η) /η see SI for Bulk transport suppressed by the topological gap and details. ballistic symmetry-protected edge modes are the hall- mark of topological insulators. In our case, while the flux quantum is π/e, the charge is carried by bosonic ex- citations of charge 2e. The BM is thus an integer bosonic THE NATURE OF THE INTERMEDIATE BOSE topological insulator with edge modes protected by the METAL T U(1)oZ2 symmetry. This is one of the generic integer topological phases recently classified in 31,32. The quan- The response of a BM to an applied field is deter- tum fluctuations parameter η, the suppression of bulk mined by its effective electromagnetic action, obtained conductances by the topological CS mass mCS, and equa- by integrating out the effective gauge fields in the gen- tions (5,6) for the longitudinal resistivity mediated by the T eral Chern-Simons action (see Appendix), which (with- U(1)oZ2 -symmetry protected gapless edge states are the out any loss of generality) assumes the simplest form in central results of our work. 4

a b 6 c 1 0 0 ) ) 1 0 B = Ω Ω 0 . 2 5 T k k 5 ( (

1 0 0 . 1 2 T

0 . 0 7 T R R

4 , , 1 0 0 . 0 5 T e e c c 0 . 0 3 T T = 0 . 6 5 K n n 3

a 1 0 a 1 0 ( B I B , R I B ) t t 0 . 0 2 T s s i i

s ( B , R ) s 2 0 . 0 1 T S I S I e e 1 0 r r

0 T t t ( B S B , R S B ) e e 1 e e 1 0 h h S S T = 0 . 0 5 K 1 0 0 1 0 . 1 1 1 0 - 3 1 0 - 2 1 0 - 1 T e m p e r a t u r e , T ( K ) M a g n e t i c f i e l d , B ( T ) d 1 0 0 e f ) T = Ω T = B = 0 . 1 6 T

k I B

( 0 . 1 2 K 0 . 3 0 K

T = 2 K 0 . 1 5 K 0 . 3 5 K 0 R

, 0 . 2 0 K 0 . 4 0 K e R c 0 . 2 5 K I B n 1 0 T = a

R

t S I 0 . 4 5 K s i

s R 0 . 5 0 K

e S B r

0 . 5 5 K

t B = 0 . 0 1 1 T S B B = 0 . 0 3 9 T e S I 0 . 6 0 K e T = 2 K

h 0 T 0 = 1 K 0 . 6 5 K S 1 1 0 - 3 1 0 - 2 1 0 - 1 1 0 0 1 0 - 3 1 0 - 2 1 0 - 1 1 0 0 1 0 - 3 1 0 - 2 1 0 - 1 1 0 0 | B - B | ∗( l n ( T / T ) ) 2 , ( T ) | B - B | ∗( l n ( T / T ) ) 2 , ( T ) | B - B | ∗( l n ( T / T ) ) 2 , ( T ) S B 0 S I 0 I B 0

Figure 3. Temperature and magnetic field dependencies and quantum BKT scaling of R (T,B) in the vicinity of the magnetic field-driven SIT. a: The device scheme and a setup for the two-terminal resistance measurements sketch. b:

Sheet resistance, R, as function of temperature, T , spanning the the anomalous metallic regime in the magnetic field range from 0 T to 0.25 T. The residual resistance grows with decreasing g and becomes R = RQ at g ' 1. c: Sheet resistance, R, as function of B for different temperatures exhibits three crossing points (marked by arrows) at BSB = 0.011 T, BIB = 0.16 T, and

BSI = 0.039 T, corresponding to superconductor-BM, BM-superinsulator, and the SIT transitions, respectively. The crossing points satisfy the duality relation BSB/BSI = BSI/BIB with great accuracy. d: Quantum BKT scaling near the BSB critical point. e: The BKT scaling for the remnant of the SIT. f: Quantum BKT scaling near the BIB critical point.

Shown in Fig. 2g is a general phase diagram compris- we recall that the particle number operator Nq and ing the BM near the SIT, with the tuning parameter the U(1) phase ϕ, form a pair of canonically conju- g denoting either the conductance, or magnetic field or gate variables, since Nq is the generator of global U(1) gate voltage in a proximity array 15. The BM forms at charge transformations. Only two symmetry realiza- η > 1 and occupies the area between the charge- and tions are allowed in infinite systems. Either (i) Nq is vortex-BKT transition lines. Quantum transitions be- q 2 fixed, ∆N = hN − N¯  i = 0, and ϕ is undefined, tween the superinsulator and the BM and between the q q q ∆ϕ = ∞ so that the U(1) symmetry is linearly realized, superconductor and the BM occur at g = g1, and g = g2, or (ii) Nq does not annihilate the vacuum, ∆Nq=∞, ϕ respectively. Duality requires g2 = 1/g1. By self-duality, the SIT can still be identified as the line g = 1 on which is fixed, ∆ϕ=0 and the global U(1) symmetry is sponta- R (T ) = R (the dashed line in Fig 1g) although there neously broken. These two possibilities define the zero-  Q 33,34 is no phase transition anymore at this point. temperature superinsulators and superconductors . However, because of the topological Chern-Simons in- teractions, the charge and vortex numbers do not obey anymore the Gauss-law constraints generating the two STABILITY OF THE BOSONIC TOPOLOGICAL U(1) symmetries, and the third possibility, ∆Nq 6= 0, INSULATOR ∆Nv 6= 0, arises. We find that in the Bose metal, the equal-time quantum correlation functions in the ground To unravel the mechanism preventing the conden- sation of CP and ensuring stability of the Bose metal, 5 a b c ) N b T i N f i l m e x p [ C / ( B - B ) 1 / 2 ] Ω T ( B ) I B ( m i n 1 2 0 . 0 5 T 8 G r a p h e n e # 1

؏

) G r a p h e n e # 2

4 1 R - ) 1 0 , J J A , R e f . [ 4 ] 1 K - 9 ( e

0 . 0 3 T K

c 6 G a f i l m s , R e f . [ 2 ] n (

i - 1

n 1 / T ( K )

v m

a

0 . 0 1 5 T B K T f i t , E q . ( 7 ) e t T d 0 1 0 2 0 / s 6 i T 1

0 . 0 1 T / s 4

4 1

e 2 x 1 0 r )

t 0 . 0 0 5 T Ω k

e

3 (

e 4

0 1 ؏ h 0 T T ( B ) 2 R d e v S 1 0 3 0 0 5 1 0 1 5 2 0 2 5 1 1 0 0 . 0 0 0 . 0 5 0 . 1 0 - 1 1 / T ( K ) X / X c M a g n e t i c f i e l d , B ( T )

Figure 4. Sheet resistance vs 1/T plots and the BKT scaling for correlation lengths. a: Representative log R ∗ vs. 1/T plots. The dashed black line depicts R ∝ exp(−T /T ) dependence. The dotted line describes R(1/T ) behavior for ∗∗ two parallel resistors comprising Rb ∝ exp(T /T ) (bulk modes) insulating dependence and constant (edge ballistic modes) 1/2 behavior, see text. b: The BKT scaling of 1/Tmin ∼ ξcorr ∼ exp[const/(X/Xc − 1) ] of the BKT correlation length near the superconductor-BM quantum BKT transition in different materials driven by either magnetic field, disorder, or gate voltage, demonstrating the universality of the BKT superconductor-BM transition. The circles stand for Tmin(B) in NbTiN films,

X/Xc = B/BSB; the diamonds mark data for Van der Waals heterostructures of twisted double bilayer graphene (TDBG) where the SIT is tuned by the in-plane magnetic field, X/Xc = B||/Bc, Bc = 2.5 T is an adjustment parameter and Tmin is divided by 35; the triangles mark data for the SIT driven by the frustration factor f = Φ/Φ0, with Φ being the magnetic flux 4 2 per plaquette in JJA, fc = 0.1 , and squares mark the film thickness driven SIT in granular films , for which X/Xc = RN/RNc,

RNc = 32 kΩ, Tmin is divided by 16. c: The BKT scaling of the correlation length ∼ 1/Tdev near the insulator-Bose metal transition, detected by deviation of the resistance from the exponential insulating exp(T ∗∗/T ) behavior (the dashed line in the inset). state are given by EXPERIMENT

 |x − y|  hj0(x)j0(y)i ∝ exp − , ξ (g/g ) Transport measurements are taken on NbTiN 10 nm corr 1 thick films prepared by the atomic layer deposition  |x − y|  hφ0(x)φ0(y)i ∝ exp − , (7) (ALD) technique based on sequential surface reaction, ξcorr(g2/g) step-by-step film growth. The films were lithographically patterned into bars, see Fig. 3a, and resistivity measure- where j0 and φ0 are the charge and vortex densi- p ments were performed at sub-Kelvin temperatures in he- ties, respectively and ξcorr(x) ∝ exp(const/ |x − 1|), lium dilution refrigerators (see the details of the sample 25,26 is the BKT correlation length with x set by the preparation, geometry, measurement technique and char- quantum coupling constant g and its critical point gc. acterization in 35). All the resistance measurements were Accordingly, the two dual transitions, superconductor- carried out in the linear regime, with the proper elec- BM and superinsulator-BM are quantum BKT transi- tric line filtering. Since the expected frequency of typical tions. Another far-reaching implication of Eq. (7) is that quantum fluctuations responsible for BM behaviour ex- charges and vortices form an intertwined liquid com- ceeds by several orders of magnitude the system of filtra- prising fluctuating macroscopic islands with typical di- tion, this is effective for noise elimination but does not af- mensions ξ(g/g1) and ξ(g2/g), respectively. The emer- fect the relevant physics at higher frequencies. Shown in gent texture is referred to as the self-induced electronic Fig. 3a is the sketch of the two-terminal setup. Figure 3b 34 granularity . The associated characteristic frequency presents the log-log plot of the low-temperature part of of the BM quantum fluctuations is ω = v/ξcorr. While R (T ) across the magnetic field-driven SIT. At magnetic  the exact expression for the correlation length in the fields B 0.04 T, R saturates at lowest measured tem- .  ground state is not yet available, it must lie in the in- peratures to the magnetic field-dependent value spanning terval ξ < ξcorr < ~/(vmCS), where the upper bound about an order of magnitude in sheet resistance, from is the length scale associated with the CS gap to the R ' 1 kΩ to R ' 20 kΩ, suggesting metallic behaviour   first excited state. We obtain thus the lower bound across this range. At fields above B ' 0.011 T, the R (T ) ω > m v2/ . Using m = e e /πv and e ≈e at  CS ~ CS ~ q v v q dependence develops a minimum. Above B & 0.04 T, the center of the BM phase, we find ω>αv/d. This is the curve shows a trend to an insulating upturn, and, as the typical frequency associated with the electrostatic soon as B exceeds 0.16 T, R (T ) exhibits pronounced  energy ~αv/d of a . For the NbTiN param- insulating behaviour. Plotting R (B) isotherms, see √ √  eters 35 d = 10 nm and v/c = 1/ ε = 1/ 800, we find panel Fig. 3c, exposes three sequential crossing points ω > 7 THz. 6

BSB = 0.011 ± 0.001 T, BSI = 0.039 ± 0.001 T, and shown by dashed line. At fields, B < BSB, the resistance B = 0.16 ± 0.01 T. The corresponing resistances are R (T ) saturates at low temperatures, indicating the pos- IB  RSB = 3.63 ± 0.01 kΩ, RSI = 6.57 ± 0.01 kΩ, and RSI = sible crossover to quantum vortex creep. Our findings 11.9 ± 0.1 kΩ. are in accord with the recently reported dissipative state Temperature dependencies of the resistance R(T ) of with non-zero resistance in two-dimensional 2H-NbSe2 films 38. Above B = 0.011 T the R (1/T ) dependencies TDBG measured at optimal doping in the presence of SB  the magnetic field, B||, parallel to the film indicates the develop minima, which become less pronounced above 37 field-induced SIT, see , where also the details of fabri- the BSI field, where the insulating behavior becomes dom- cation of TDBG devices and measurements protocol can inant. These emergent minima signal that the bulk spec- be found. In the intermediate B|| region R(T ) develops trum of charge excitations acquires a gap, which prevents a minimum the position of which depends on B||. As we bulk electronic transport. Such minima are often viewed discuss below, these minima may indicate the possibil- as the hallmark of topological insulators, see, for exam- ity of the formation of the bosonic topological insulator ple, 39 and references therein. We now note that in the state. BM, i.e. topological insulator domain, BSB < B < BIB, one can neglect the contribution from moving vortices as DISCUSSION AND CONCLUSION it is seen from Fig. 4a. Then the sheet resistance of the BM results from the charge current contribution from two parallel channels, (i) the ballistic edge modes and (ii) the We identify the crossing points B and B as quan- SB IB thermally activated bulk modes over the CS gap. Accord- tum transitions between the superconductor and the BM ingly, below the minimum, T < T , the sheet resistance and the BM and the superinsulator, respectively, and min is perfectly fitted by the two parallel resistors formula the range of fields BSB < B < BIB as the domain of R (T,B) = RCS(T )Rbal(B)/[RCS(T ) + Rbal(B)], where the existence of the Bose metal, i.e. the bosonic topo-  R (T ) ∝ exp(T /T ) is the bulk thermally activated logical insulator. The intermediate crossing point at CS CS resistance corresponding to surmounting the insulating B = B is a remnant of the SIT tri-critical point. One SI gap T , while R (B) is the field dependent resistance immediately observes that the expected duality relations CS bal mediated by the edge modes. A fit for the field 0.05 T B /B = B /B and R /R = R /R , are satisfied SB SI SI IB SB SI SI IB is shown by the dashed line in Fig. 4a. Upon increas- with fantastic accuracy. This accuracy is an additional ing the magnetic field above B = 0.039 T (i.e. moving indication of the topological nature of the anomalous SI into the g < 1 region), R (B) increases and the edge metal appearing between the insulating and supercon- bal states incrementally mix with the bulk modes. As a re- ducting phases at the SIT. Generalizing the standard SIT sult, the minimum in R (T,B) becomes less pronounced scaling considerations of 30 onto the BKT transition, we  and the film crosses over continuously to insulating be- introduce the scaling variable |g − g |(ln(T /T ))2, where c 0 havior. Markedly, the R(T ) vs.1/T dependence in the T is the adjusting parameter to be determined by the 0 TDBG exhibits remarkable similarity to that in NbTiN, best fit. One expects BKT scaling behaviors near both see Appendix. quantum critical points BSB and BIB. Furthermore, since the observed SIT physics is dominated by the proxim- The correlation lengths ξSB∼1/Tmin and ξIB∼1/Tdev ity to the quantum tri-critical point at η = 1, one may associated with the quantum BKT superconductor-BM expect the remnant of the analogous behavior at BSI. and superinsulator-BM transitions are expected to dis- 30 This calls for revisiting the original scaling result of play the BKT criticality. Here Tmin is the position of the as well as the observed persistent scaling in the experi- minimum in R , heralding the emergence of the bulk CS 15  ment . The results of the BKT scaling analysis around insulating gap. The temperature Tdev, identified by the B , B , and B are presented in Fig. 3d,e,f respectively, deviation of R (T ) in Fig. 4a from the insulating expo- SB IB SI  and demonstrate an excellent critical fit, supporting the nential dependence exp(const/T ), marks the switching quantum BKT nature of the transitions to the Bose metal on of the edge modes and the start of shunting the in- phase. Note that, while for both the superconductor- and sulating bulk by metallic edge channels. The latter can insulator-to-Bose metal transitions the best fit yields the be viewed as quantum wires, hence their switching on characteristic temperature T0 = 2 K, the corresponding can be described as a BKT quantum phase transition 40 characteristic parameter for the SIT is T0 = 1 K, which in analogy to the seminal work . Scaling of 1/Tmin as reflects the different energy characteristics for the rem- function of the dimensionless tuning parameter B/BSI is nant quantum SIT occurring near the tri-critical point presented in Fig. 4b by green solid circles. Accordingly, g = 1, η = 1. the positions of the minima of R(T ) of TDBG at differ- Shown in Fig. 4a are the magnified R (T ) dependen- ent B are shown by violet diamonds. Also displayed  || cies which we re-plotted as functions of 1/T for the fields in Fig. 4b are similar dependencies for other materials, below 0.16 T. We present a few representative curves metallic granular films 2 and JJA 4. The data comply to avoid data crowding. At relatively high tempera- perfectly with the BKT exponential scaling illustrating tures one sees the resistance rapidly dropping as func- the universality of the transition into the BM in differ- tion of 1/T due to thermally activated vortex motion, ent systems. The scaling of 1/Tdev is shown in Fig. 4c, R (1/T ) ∝ exp(−T ∗/T ), the exponential behavior is although there the window of the magnetic fields is less  7

wide and the error-bar near B = BIB grows large. Note able contribution at the initial stage of work on NbTiN that the exponential BKT scaling can be viewed as a re- films. The work at Argonne (V.M.V.) was supported by sult of the formal ν → ∞ limit. An “infinite” critical the U.S. Department of Energy, Office of Science, Ba- exponent ν implies, by the Harris criterion 41, that disor- sic Energy Sciences, Materials Sciences and Engineer- der is irrelevant for the SIT in the renormalization group ing Division. The work on transport measurements at sense. The role of disorder in the large-scale properties Novosibirsk was supported by the grant of RF president of the system is only to tune the SIT and renormalize (MK-5455.2018.2). The work by Y.K. was supported by material parameters. FAPESP, CNPq and AFOSR Grant FA9550-17-1-0132. The presented gauge theory of the Bose metal re- The work at Caltech (D.S.) was supported by National veals that the long-debated SIT-intervening metallic Science Foundation Grant No. DMR-1606858. phase is a bosonic topological insulator. Its metallic conductance is mediated by bosonic edge modes and the superconductor-topological insulator and topological APPENDIX insulator-superinsulator transitions are quantum BKT phase transitions. The BKT scaling at the transition A. Free energy points, together with the high-precision duality rela- tions for the transition fields and the corresponding re- We start with the the action describing the inter- sistances, providing an unambiguous evidence for the twined vortex-CP dynamics derived in 20: bosonic topological insulator and the associated quantum  q¯ v 1 BKT transitions, are reported for the first time. Our ob- 2 c S = dtd x i aµµαν ∂αbν + 2 f0f0 + 2 fifi servation of a bosonic topological insulator in NbTiN and ˆ 2π 2ev 2evvc similar behavior in TDBG, stresses its universal charac- vc 1 √ √  + 2 g0g0 + 2 gigi +i qa¯ µQµ + i qb¯ µMµ (8), ter as a result of the Chern-Simons gap in the spectrum of 2eq 2eqvc relevant excitations. Note that the occurrence of bosonic √ where v = 1/ µε is the speed of light in the film ma- topological insulator in the double belayered graphene is c terial, expressed in terms of the magnetic permeability in a concert with the expectations of 22. Interestingly, µ and the electric permittivity ε (we use natural units our previous findings of 35 support also the percolation c = 1, = 1). This action (8) is a non-relativistic version network picture 36 of gapless bosonic channels near the ~ of the topologically massive gauge theory 21 describing BKT transition to TI. Finally, our results resolve the a (2+1)-dimensional vector particle with the CS mass, long-standing puzzle of the SIT, unraveling that the sys- m =qe ¯ e /2πv , arising without spontaneous symme- tem follows an indirect transition scenario through the in- CS q v c try. The emergent gauge fields a and b mediate the tervening metallic state in systems with strong quantum µ µ mutual statistics interactions between Cooper pairs, with fluctuations, while films with suppressed quantum fluc- word-lines Q and chargeq ¯ = 2 and vortices of flux 2π/q¯, tuations (more disordered and/or with the higher carrier µ with world-lines M . When appropriately regularized on densities) exhibit the direct SIT. µ 14,42 a lattice of spacing ` (see Supplementary Information), Recent transport studies reported that this these world-lines can be viewed as “strings” of typical anomalous Bose metal state possesses neither Hall re- length L = N`, carrying electric and magnetic quantum sistance nor cyclotron resonance, which are expected to numbers Q and M, respectively. To derive the free energy be pronounced in states that have longitudinal resistance of the interacting Cooper pair-vortex system the gauge much smaller than the normal resistance of the respec- fields in the above quadratic action are integrated out via tive materials. These findings comply with our predic- the standard Gaussian integration procedure, obtaining tion that the Bose metal is a bosonic topological insu- thus an effective action for the charge and vortex strings lator. More experimental research, however, is required alone. As usual in statistical field theory this has the for conclusive evidence. One can also propose an imag- interpretation of an energy for the “string gas”, which is ing study of the fluctuation pattern using local probes, proportional to the string length. To obtain a free en- like scanning squid, scanning NV microscopy, or scanning ergy associated with the strings, one has to include the impedance microscopy, as well as using a non-local mea- contribution from the positional string entropy, which is surement geometry to obtain indirect evidences of the also proportional to its length, with the proportionality edge modes. factor µe = ln(5) representing the 5 possible choices for string continuation at each lattice site. For the relevant case of Cooper pairs (i.e.q ¯ = 2), we find the free energy Acknowledgments in the main text.

We are delighted to thank Thomas Proslier for prepar- ing NbTiN samples used in the experiments and Tom B. Effective action for the topological insulator Rosenbaum for valuable discussions. M.C.D. thanks CERN, where she completed this work, for kind hospi- To determine the nature of the Bose metal we find its tality. S.V.P. is grateful to Tatyana Baturina for valu- electromagnetic response by coupling the charge current 8

(¯qe)jµ to an external electromagentic potential Aµ and gauge invariance is restored by adding the edge action we compute its effective action by integrating out gauge 1 2 fields aµ and bµ, S = d x (∂ ξ∂ ξ − ∂ η∂ η) + edge π ˆ 0 s 0 s √  2 q¯ 1 qe¯ eff d x A0 ∂sχ , (12) e−Seff (Aµ) = Da Db e−S(aµ,bµ)+i(¯qe)jµAµ , ˆ 2π Z ˆ µ µ including the electromagnetic√ coupling of the edge charge Z = Da Db e−S(aµ,bµ) . (9) density ρ = (¯qe )( q/¯ 2π)∂ χ in the A = 0 gauge,qe ¯ ˆ µ µ eff s s eff being the effective charge of the Cooper pairs in the Bose √ metal phase,qe ¯ eff =qe ¯ g. As in the case of the quantum This gives Hall effect, the non-universal dynamics of the edge modes is generated by boundary effects28, which result in the Hamiltonian 2   g  qe¯  3 2 1 2 Seff (Aµ) = d d x vcF0 + Fi , (10) 1 h 2 2i 4 2π ˆ vc H = ds −v (∂ ξ) − v (∂ η) , (13) π ˆ b s b s where Fµ=µαν ∂αAν is the dual field strength and we where vb is the velocity of propagation of the edge modes have identified the geometric lattice factor 4µeη` with along the boundary. Upon adding this term, the total the relevant thickness parameter d of the film, so as to edge action becomes maintain self-duality (see main text). This is the action 1 2 of a bulk insulator which becomes best evident in the Sedge = d x [(∂0 − vb∂s) ξ∂sξ − (∂0 + vb∂s) η∂sη] π ˆ relativistic case, vc = 1: √  q¯  +¯qe d2x A ∂ χ (14). eff ˆ 0 2π s g  qe¯ 2 S (A ) = d d3x A −δ ∇2 + ∂ ∂  A . The equation of motion generated by this action is eff µ 2 2π ˆ µ µν µ ν ν (11) qe¯ eff qe¯ eff vb∂sρ = E = ∂sA0 . (15) Varying this action with respect to the vector potential 2π 2π A gives the main text formula for the electric current. ν Integrating this equation gives eq. (5) in the main text, which represent ballistic charge conduction with the re- sistance R = RQ/g.

C. Conduction by edge modes D. Bose metal stability

The Chern-Simons effective action is not invariant un- To analyze an intervening phase harboring dilute der gauge transformations ai = ∂iλ and bi = ∂iχ at the topological excitations in the Hamiltonian formalism, we 27 edges. Two chiral bosons λ = ξ + η and χ = ξ − η set Qµ=Mµ=0 and decompose the original gauge fields i ij i ij have to be introduced to restore the full gauge invari- in Eq. (8) as a =∂iξ+ ∂jφ, b =∂iλ+ ∂jψ. Quantizing ance, exactly as it is done in the quantum Hall effect the action in (8) we arrive at the ground state wave func- framework 28 and for topological insulators 43. The full tional

i i h 2 2  2 1 2i Ψ[a , b ]=exp i(¯q/4π) d x (ψ∆ξ + φ∆λ) − (¯q/4π) d x g(∂iφ) + g (∂iψ) , ´ ´ generalizing the Schr¨odingerwave function to a system account also the corresponding topological excitations, with an infinite number of degrees of freedom. The fields vortices and point charges. These can be in highly en- φ and ψ represent vortex- and CP charge-density waves, tangled mixed states or in their pure free state. Quantum which are gapped due to the mutual statistics interac- operator expectation values in the entangled mixed state tions. When the two symmetries are compact, however, in which vortices are the non-observed environment, are the fields ξ and λ are angles and we have to take into given by

 2 q¯ 2 q¯ 2  hOi∝ DψDλ O(ψ, λ) exp − d x 2πg (∂iψ) + 2πg (∂iλ) − 2zcosλ , ´ ´ 9 see SI. The quantum fugacity z governs the degree of en- films.We used NbCl5, TiCl4, and NH3 as gaseous reac- tanglement. This is the classical partition function of the tants; the stoichiometry was tuned by varying the ratio 25,26 46 sine-Gordon model undergoing the BKT transition . of TiCl4/NbCl5 cycles during growth . The supercon- In a quantum case, it is a quantum BKT transition at ducting properties of these ultrathin NbTiN films were the “effective temperature” for vortex liberation set by optimized by utilizing AlN buffer layers grown on top 45 the quantum conductance parameter g. Correspondingly, of the Si substrate . Nb1−xTixN films of thicknesses charge liberation from the dual entangled state is set by d = 10 were grown. Films have a fine-dispersed polycrys- 1/g. The highly entangled states correspond to super- talline structure35. The average crystallite size is ≈ 5 nm. conductor and superinsulator at high and low values of Deposition temperature is 3500C. Ti fraction x is 0.3. g, respectively, where vortices and charges have algebraic correlation functions. The Bose topological insulator is The films were lithographically patterned into bridges the intervening state at g ' 1, where both charges and 50 µm wide, the distance between current-contacts was vortices are screened by strong quantum fluctuations, 2500 µm and distance between voltage-contacts was 450 leading to correlation functions (7). µm. Most resistive transport measurements are carried out using low-frequency ac techniques in a two-terminal configuration with V ≈ 100 µV, f ≈ 1 Hz. Addition- E. Samples and measurements ally we measured temperature dependence of resistance at zero magnetic field. From comparison two- and four- To grow NbTiN films, we employed the atomic layer terminal configuration we determined number of square deposition (ALD) technique based on sequential surface in a two-terminal configuration for obtaining resistance reaction step-by-step film growth. The fabrication tech- per square. For ac measurements we use SR830 Lock-ins nique is described in detail in the Supplemental Mate- and current preamplifiers SR570. All the resistance mea- rial. This highly controllable process provides superior surement are carried out in linear regime with using ad- thickness and stoichiometric uniformity and an atom- equately system of filtration. Resistivity measurements ically smooth surface 44as compared to chemical vapor at sub-Kelvin temperatures were performed in dilution deposition, the standard technique used to grow NbTiN refrigerators 3He/4He with superconducting magnet.

F. Low-temperature resistance in NbTiN and double belayer graphene

N b T i N 1 . 0 B = 0 . 0 5 T | G r a p h e n e

e

c B | | = 1 2 T n 0 . 8 a t s i s e r

t 0 . 6 e e h s

d 0 . 4 e z i l a m r

o 0 . 2

N T m i n

0 . 0 0 1 2 3 4 5 6 7 8 9 1 0 - 1 - 1 T / ( T m i n ) Figure 5. Sheet resistance vs 1/T plots in NbTiN and twisted double belayer graphene (TDBG) Representative

R vs. 1/T plots for fields B⊥ = 0.05 T for NbTiN and B|| = 0.05 for TDBG (sample No. 2, see main text). The temperature scales are normalized with respect to positions of the minima, resistances are normalized with respect to their saturation values. Importantly, normalized temperatures of the saturation coincide for both systems indicating a universal topological character of the quantum BKT transitions confining the domain of the existence of the bosonic topological insulators.

1 Haviland, D., Liu, Y., & Goldman, A. Onset of Supercon- 3 Hebard, A. & Paalanen, M. A. Magnetic-Field-Tuned ductivity in the Two-Dimensional Limit. Phys. Rev. Lett. Superconductor-Insulator Transition in Two-Dimensional 62, 2180 – 2183 (1989). Films. Phys. Rev. Lett. 65, 927 – 930 (1990). 2 Jaeger, H. M., Haviland, D. B., Orr, B. G., & Gold- 4 van der Zant, H. S. J., Fritschy, F. C., Elion, W. J., man, A. M. Onset of in ultrathin Geerligs, L. J., & Mooij,J. E. Field-induced granular metal films. Phys. Rev. B 40 183 – 196 (1989). superconductor-to-insulator transitions in Josephson- 10

junction arrays Phys. Rev. Lett. 69, 2971 (1992). tinuous symmetry group. I. Classical systems. Sov. Phys.– 5 Mason, N.& Kapitulnik, A. Dissipation effects on the su- JETP, 32, 493-500 (1970). perconductor? insulator transition in 2D superconductors. 26 Kosterlitz, J. M. & Thouless, D. J. Long range order and Phys. Rev. Lett. 82, 5341?5344 (1999). metastability in two dimensional solids and superfluids. 6 Markovi´c,N., Christiansen, C., Mack, A. M., Huber, W. H. (application of dislocation theory. J. Phys. C: Solid State & Goldman, A. M. Superconductor?insulator transition in Phys. 5, L124 (1972). two dimensions. Phys. Rev.B 60, 4320?4328 (1999). 27 Floreanini, R. & Jackiw, R. Self-dual fields as charge- 7 Steiner, M. A., Breznay, N. P. & Kapitulnik, A. Approach density solitons. Phys. Rev. Lett. 59 1873 – 1876 (1987). to a superconductor-to-Bose-insulator transition in disor- 28 Wen, X.-G. Theory of the Edge States in Fractional Quan- dered films. Phys. Rev.B 77, 212501 (2008). tum Hall Effects. Int. J. Mod. Phys. B6 1711-1762 (1992). 8 Bollinger, A. T. et al. Superconductor-insulator transition 29 Fisher, M. P. A., Grinstein, G. & Girvin, S. M. Presence of in La2−xSrxCuO4 at the pair quantum resistance. Nature Quantum Diffusion in Two Dimensions: Universal Resis- 472, 458?460 (2011). tance at the Superconductor-Insulator Transition. Phys. 9 Eley, S., Gopalakrishnan, S., Goldbart, P. M. & Mason, N. Rev. Lett. 64, 587 – 590 (1990). Approaching zero-temperature metallic states in meso- 30 Fisher, M. P. A. Quantum Phase Transitions in Disordered scopic superconductor?normal? superconductor arrays. Two-Dimensional Superconductors. Phys. Rev. Lett. 65, Nat. Phys. 8, 59?62 (2012). 923 – 926 (1990). 10 Allain, A., Han, Z. & Bouchiat, V. Electrical control of 31 Lu, Y.-M. & Vishwanath, A. Theory and Classification of the superconductingto- insulating transition in graphene- interacting integer topological phases in two dimensions: a metal hybrids. Nat. Mater. 11, 590?594 (2012). Chern-Simons approach. Phys. Rev B 86, 125119 (2012). 11 Han, Z. et al. Collapse of superconductivity in a hybrid tin- 32 Wang, C. & Senthil, T. Boson topological insulators: A graphene Josephson junction array. Nat. Phys. 10, 380?386 window into highly entangled quantum phases. Phys. Rev. (2014). B 87, 235122 (2013). 12 Cou¨edo,F. et al. Dissipative phases across the 33 Vinokur, V. M. et al. Superinsulator and quantum synchro- superconductor-to-insulator transition. Sci. Rep. 6, nization. Nature 452, 613 – 615 (2008). 35834 (2016). 34 Baturina, T. I. &Vinokur, V. M. Superinsulator- 13 Park, S., Shin, J. & Kim, E. Scaling analysis of field-tuned superconductor duality in two dimensions. Ann. Phys. superconductor? insulator transition in two-dimensional 331, 236 – 257 (2013). tantalum thin films. Sci. Rep. 7, 42969 (2017). 35 Mironov, A. Yu. et al. Charge Berezinskii-Kosterlitz- 14 Breznay, N. P. & Kapitulnik, A. Particle-hole symmetry re- Thouless transition in superconducting NbTiN films. Sci- veals failed superconductivity in the metallic phase of two- entific Rep. 8, 4082 (2018). dimensional superconducting films. Sci. Adv. 3, e1700612 36 Chalker J. T. & Coddington P. D. Percolation, quantum (2017). tunnelling and the integer Hall effect. J. Phys.C 21, 2665 15 Bottcher, C. G. L. et al. Superconducting, insulating and – 2679 (1988). anomalous metallic regimes in a gated two-dimensional 37 Liu, X. et al. Spin-polarized Correlated Insulator and semiconductor-superconductor array. Nature Physics 14, Superconductor in Twisted Double Bilayer Graphene. 1138-1144 (2018). arXiv:1903.08130v2 (2019). 16 Das, D. & Doniach, S. Existence of a Bose metal at T = 0. 38 A. Benyamini, et al. Fragility of the dissipationless state Phys. Rev.B 60, 1261 – 1275 (1999). in clean two-dimensional superconductors. Nature Physics 17 Goldman, A. M. Superconductor-Insulator Transitions. (2019) DOI:https://doi.org/10.1038/s41567-019-0571-z Int. J. Mod. Phys. B24, 4081 – 4101 (2010). 39 Wang, K., Graf, D., Li, L., Wang, L. & Petrovic, C. 18 Kapitulnik, A., Kivelson, S. A. & Spivak, B. Anomalous Anisotropic giant magnetoresistance in NbSb2. Scientific metals – failed superconductors. Rev. Mod. Phys. 91, Reports 4, 7328 (2014). 011002 (2019). 40 Bobbert, P., Fazio, R., Schon, G. & Zaikin, A. D. Phase 19 Diamantini, M. C., Sodano,P., & Trugenberger,C. A. transitions in dissipative Josephson chains: Monte Carlo Gauge theories of Josephson junction arrays. Nuclear results and response functions. Phys. Rev.B 45, 2294 – Physics B474, 641 – 677 (1996). 2304 (1992). 20 Diamantini, M. C., Trugenberger C. A. & Vinokur, V. M. 41 Harris, A. B. Upper bounds for the transition temperatures Confinement and Asymptotic Freedom with Cooper of generalized Ising models. J. Phys. C7 1671 (1974). Pairs. Communications Physics, 1, 77 (2018). DOI: 42 Wang, Y., Tamir, I., Shahar, D. & Armitage, N. P. Absence 10.1038/s42005-018-0073-9. of Cyclotron Resonance in the Anomalous Metallic Phase 21 Deser, S., Jackiw, R. & Templeton, S. Three-dimensional in InOx. Phys. Rev. Lett. 120, 167002 (2018). massive gauge theories. Phys. Rev. Lett. 48 975 (1982). 43 Cho, G. Y. & Moore, J. E. Topological BF Field Theory De- 22 Bi, Z et aL. Bilayer Graphene as a Platform for Bosonic scription of Topological Insulators. Ann. Phys. 326 1515- Symmetry-Protected Topological States. Phys. Rev. Lett. 1535 (2011). 118, 126801 (2017). 44 Lim, B. S., Rahtu, A., & Gordon, R. G. Atomic layer depo- 23 Wilczek, F., Disassembling Anyons. Phys. Rev. Lett. 69 sition of transition metals. Nature Materials 2, 749 (2003). 132-135 (1992). 45 Shiino, T. et al. Improvement of the critical temperature of 24 Fazio, R. & Sch¨on,G. Charge and Vortex Dynamics in superconducting NbTiN and NbN thin films using the AlN Arrays of Tunnel Junctions. Physical Review B 43, 5307 – buffer layer. Supercond. Sci. Technol. 23, 045004 (2010). 5320 (1991). 46 Proslier, T., Klug, J. A., Becker, N. C., Elam, J. W. & 25 Berezinskii, V. L. Destruction of long-range order in one- Pellin, M. J. Atomic Layer Deposition of Superconductors. dimensional and two-dimensional systems having a con- ECS Transactions 41, 237 (2011).