Abstract
One of the most interesting predictions resulting from quantum physics, is the violation of classical symmetries, collectively referred to as anomalies. A remarkable class of anomalies occurs when the continuous scale symmetry of a scalefree quantum system is broken into a discrete scale symmetry for a critical value of a control parameter. This is an example of a (zero temperature) quantum phase transition. Such an anomaly takes place for the quantum inverse square potential known to describe ‘Efimov physics’. Broken continuous scale symmetry into discrete scale symmetry also appears for a charged and massless Dirac fermion in an attractive 1/r Coulomb potential. The purpose of this article is to demonstrate the universality of this quantum phase transition and to present convincing experimental evidence of its existence for a charged and massless fermion in an attractive Coulomb potential as realized in graphene.
Introduction
Continuous scale symmetry (CS)—a common property of physical systems—expresses the invariance of a physical quantity f(x) (e.g., the mass) when changing a control parameter x (e.g., the length). This property is expressed by a simple scaling relation, f(ax) = bf(x), satisfied ∀a > 0 and corresponding b(a), whose general solution is the power law f(x) = Cx ^{α} with α = ln b/ln a. Other physical systems possess the weaker discrete scale symmetry (DS) expressed by the same aforementioned scaling relation but now satisfied for fixed values (a, b) and whose solution becomes f(x) = x ^{α} G(ln x/ln a), where G(u + 1) = G(u) is a periodic function. Physical systems having a DS are also known as selfsimilar fractals^{1} (Fig. 1a). It is possible to break CS into DS at the quantum level, a result which constitutes the basis of a special kind of scale anomaly^{2, 3}.
A wellstudied example is provided by the problem of a particle of mass μ in an attractive inverse square potential^{4, 5}, which plays a role in various systems^{6,7,8,9} and more importantly in Efimov physics^{10, 11}. Although well defined classically, the quantum mechanics of the scale—and conformal^{12}—invariant Hamiltonian H = −Δ/2μ − ξ/r ^{2} (with ħ = 1) is well posed, but for large enough values of ξ, H is no longer selfadjoint^{13, 14}. The corresponding Schrödinger equation for a normalisable wave function ψ(r) of energy k ^{2} = −2μE is,
where ζ ≡ 2μξ − l(l + d − 2) is a dimensionless parameter, d the space dimensionality and l the orbital angular momentum. Equation (1) is invariant under the transformation r → λr and k → k/λ, ∀λ (CS), namely to every normalisable wave function of energy k ^{2} corresponds a continuous family of states with energies (λk)^{2}, so that the bound spectrum is a continuum unbounded from below. Various ways exist to cure this problem, based on cutoff regularisation and renormalisation group^{15,16,17,18,19,20,21}, and all lead for the lowenergy spectrum to a quantum phase transition (QPT) monitored by ζ, between a single bound state for ζ < ζ _{c} to an infinite and discrete energy spectrum for ζ > ζ _{c}, independent of the regularisation procedure and given by
which clearly displays DS. The critical value ζ _{c} = (d − 2)^{2}/4 depends on the space dimensionality only, and \({\epsilon _{\rm{0}}}\) is a regularization dependent energy scale. In the overcritical phase ζ > ζ _{c}, the corresponding renormalization group solution provides a rare example of a limit cycle^{15, 16, 22}. Building on the previous example, it can be anticipated that the problem of a massless Dirac fermion in an attractive Coulomb potential^{23,24,25}, −Zα/r, is also scale invariant (CS) and that the spectrum of resonant quasibound states presents similar features and a corresponding QPT.
In this work, we demonstrate the existence of such a universal QPT for arbitrary space dimension d ≥ 2 and independently of the short distance regularisation. We obtain an explicit formula for the lowenergy fractal spectrum in the overcritical regime. In contrast to the Schrödinger case equation (1), the massless Dirac Hamiltonian displays an additional parity symmetry which may be broken by the regularisation. In that case, the degeneracy of the overcritical fractal spectrum is removed and two intertwined geometric ladders of quasibound states appear in the swave channel. All these features are experimentally demonstrated using a charged vacancy in graphene. We observe the overcritical spectrum and we obtain an experimental value for the universal geometric ladder factor in full agreement with the theoretical prediction. We also explain the observation of two intertwined ladders of quasibound states as resulting from the breaking of parity symmetry. Finally, we relate our findings to Efimov physics as measured in cold atomic gases.
Results
The Dirac model
The Dirac equation of a massless fermion in the presence of a −Zα/r potential is obtained from the Hamiltonian (with ħ = c = 1),
where (γ ^{0}, γ ^{j}) are Dirac matrices. Here the dimensionless parameter monitoring the transition is β = Zα, where Z is the Coulomb charge and α the fine structure constant. The QPT occurs at the critical value β _{c} = (d − 1)/2 (Supplementary Note 1) (A related anomalous behavior in the Dirac Coulomb problem has been identified long ago^{26} but its physical relevance was marginal since it required non existent heavynuclei Coulomb charges Z ≃ 1/α ≃ 137 to be observed. Moreover, the problem of a massive Dirac particle is different due to the existence of a finite gap which breaks CS.). For resonant quasibound states, we look for scattering solutions of the form ψ _{in} + e ^{2iη} ψ _{sc}, where η(E) is the energydependent scattering phase shift and ψ _{in,sc}(r, E) are two component objects representing the radial part of the Dirac spinor which behave asymptotically as,
for \(\left E \rightr \gg 1\) and, using \(\gamma \equiv \sqrt {{\beta ^2}  \beta _{\rm{c}}^2} \),
for \(\left E \rightr \ll 1\) and for the lowest angular momentum channels. The two component objects V _{in,sc} and \(U_{{\rm{in,sc}}}^ \pm \) in Eqs. (4) and (5) are constants. It is easy to infer from (5) that β = β _{c} plays a special role. Indeed for β > β _{c}, there exists a family of normalisable solutions that admit complex eigenvalues E = −i \(\epsilon \), hence the Hamiltonian (3) is not selfadjoint (H ≠ H ^{†}). To properly define this quantum problem, a regularisation is thus needed for the too strong potential at overcritical values of β = Zα. This is achieved by introducing a cutoff length L and a boundary condition at r = L, which is equivalent to replacing the Coulomb potential at short distances by a wellbehaved potential whose exact form is irrelevant in the lowenergy regime \(EL \ll 1\). The resulting mixed boundary condition can be written as \(h = {\left. {{\Psi _2}\left( {r,E} \right){\rm{/}}{\Psi _1}\left( {r,E} \right)} \right_{r \to {L^ + }}}\), where Ψ_{1,2} represent the two components of the aforementioned radial part of the Dirac spinor. The resulting scattering phase shift η(E, L, h), which contains all the information about the regularisation, thus becomes a function of L and of the parameter h. The quasibound states energy spectrum is obtained from the scattering phase shift by means of the Krein–Schwinger relation^{27, 28} which relates the change of density of states δρ to the energy derivative of η, (This is also related to the Wigner time delay^{29} and to the Friedel sum rule)
Theoretical structure of quasibound spectrum
From now on, and to compare to experimental results further discussed, we consider the case d = 2, for which there is a single orbital angular momentum quantum number \(m \in {\Bbb Z}\). The corresponding critical coupling becomes β _{c} = m + 1/2 ≥ 1/2, giving rise to the swave channels, m = 0, −1, for which β _{c} = 1/2. Depending on the choice of boundary condition h, δρ(E) can be degenerate or nondegenerate over these two swave channels. This degeneracy originates from the symmetry of the (2 + 1) Dirac Hamiltonian (3) under parity, (x, y) → (−x, y), and its existence is equivalent to whether or not the boundary condition breaks parity (Supplementary Note 2). In what follows, we will consider the generic case in which there is no degeneracy. In the undercritical, β < β _{c}, and lowenergy regime \(EL \ll 1\), we observe (Figs. 1b, 2a) a single quasibound state originating from only one of the swave channels and which broadens as β increases. In the overcritical regime β > β _{c}, this picture changes dramatically. (We emphasize that this picture remains valid for all values of β > β _{c} and not only in the vicinity of β _{c}). The lowenergy (\(EL \ll 1\)) scattering phase shift displays two intertwined, infinite geometric ladders of quasibound states (Figs. 1c, 3) at energies E _{ n } still given by (2) but with ζ − ζ _{c} now replaced by \({\beta ^2}  \beta _{\rm{c}}^2\). (Moreover, note that the energy scale \({\epsilon _0}\) for the Dirac case is different from the inverse square Schrödinger case defined in equation (1)). This sharp transition at β _{c} belongs to the same universality class as presented for the inverse square Schrödinger problem, namely CS of the quasibound states spectrum is broken for β > β _{c} into a DS phase characterized by a fractal distribution of quasibound states. The QPT thus reflects the lack of selfadjointness of the Hamiltonian equation (3) and the necessary regularisation procedure leads to a scale anomaly in which CS is broken into DS.
Experimental realization in graphene
A particularly interesting condensed matter system, where the previous considerations seem to be relevant is graphene in the presence of implanted Coulomb charges in conveniently created vacancies^{30}. It is indeed known that lowenergy excitations in graphene behave as a massless Dirac fermion field with a linear dispersion \(\epsilon \) = ±v _{F}p and a Fermi velocity v _{F} ≃ 10^{6} m/s^{31}. These characteristics have been extensively exploited to make graphene a very useful platform to emulate specific features of quantum field theory, topology and especially QED^{23}, since an effective fine structure constant α _{ G } ≡ e ^{2}/ħv _{F} of order unity is obtained by replacing the velocity of light c by v _{F}.
It has been recently shown that singleatom vacancies in graphene can stably host local charge^{30}. Density functional theory calculations have shown that, when a carbon atom is removed from the honeycomb lattice, the atoms around the vacancy site rearrange into a lower energy configuration^{32}. The resulting lattice reconstruction causes a charge redistribution, which in the ground state has an effective local charge of ≈+1. Recent Kelvin probe force microscopy measurements of the local charge at the vacancy sites are in good agreement with the Density functional theory predictions. Vacancies are generated by sputtering graphene with He^{+} ions^{33, 34}. Charge is modified and measured at the vacancy site by means of scanning tunnelling spectroscopy and Landau level spectroscopy as detailed in ref. ^{30}. Applying multiple pulses allows for a gradual increase in the vacancy charge, which in turn acts as an effective tunable Coulomb source. Moreover, the size of the source inside the vacancy is small (≈1 nm) as compared to the method of deposited metal clusters^{35}. Using this method, we are able to observe the transition expected to occur at β = 1/2 and to measure and analyze three resonances for a broad range of β values.
To establish a relation between the measured differential conductance and the spectrum of quasibound states, we recall that the tunnel current I(V) is proportional to both the density of states ρ _{ t }(\(\epsilon \)) of the STM tip and ρ(\(\epsilon \)) of massless electronic excitations in graphene at the vacancy location. We also assume that the tunnel matrix element t^{2} depends only weakly on energy and that both voltage and temperature are small compared to the Fermi energy and height of the tunnelling potential, so that the current I(V) = G _{t} V is linear with V thus defining the tunnel conductance G _{t} = 2π(e ^{2}/ħ)t^{2} ρ _{ t } ρ(\(\epsilon \)). Assuming that ρ _{ t } of the reference electrode (the tip) is energy independent, a variation δρ(\(\epsilon \)) of the local density of states at the vacancy leads to a variation δI(V) of the current and thus to a variation δG _{t}(V) of the tunnel conductance so that, at zero temperature, we obtain^{36}
where ρ _{0} is the density of states in the absence of vacancy. By considering the vacancy as a local perturbation, each quasiparticle state is characterized by its scattering phase shift taken to be the phase shift η(E) of the quasibound Dirac states previously calculated. Then, the change of density of states δρ(E) is obtained from equation (6) and combining together with equation (7) leads to the relation,
between the differential tunnel conductance and the scattering phase shift.
The measurements and the data analysis presented here were carried out as follows: positive charges are gradually injected into an initially prepared single atom vacancy and the differential conductance δG _{t}(V) is measured at each step as a function of voltage. Since we are looking at the positions of resonant quasibound states, both quantities displayed in Figs. 2, 4 give the same set of resonant energies, independently of the energyindependent factor G _{t}/πρ _{0}. For low enough values of the charge, the differential conductance displayed in Fig. 2b, shows the existence of a single quasibound state resonance. The behavior close to the Dirac point, namely in the lowenergy regime independent on the short distance regularization, is very similar to the theoretical prediction of Fig. 2a. When the build up charge exceeds a certain value, we note the appearance of three resonances, emerging out of the Dirac point. We interpret these resonances as the lowest overcritical (β > 1/2) resonances, which we denote E _{1}, \(E_1^\prime \), E _{2,} respectively. The corresponding theoretical and experimental behaviors displayed in Figs. 3, 4, show a very good qualitative agreement. To achieve a quantitative comparison solely based on the previous Dirac Hamiltonian equation (3), we fix L and the boundary condition h and deduce the theoretical β values corresponding to the respective positions of the lowest overcritical resonance E _{1} (as demonstrated in Fig. 4). This allows to determine the lowest branch E _{1}(β) for n = 1 represented in Fig. 5. Then, the experimental points \(E_1^\prime \), E _{2} are directly compared to their corresponding theoretical branch as seen in Fig. 5. We determine L and h, according to the ansatz h = a(m + 1), and obtain the best correspondence for L ≃ 0.2 nm, a ≃ −0.85. We compare the experimental E _{2}/E _{1} ratio with the universal prediction \({E_{n + 1}}{\rm{/}}{E_n} = {e^{  \pi /\sqrt {{\beta ^2}  1/4} }}\) as seen in Fig. 6. A trendline of the form \({e^{  b/\sqrt {{\beta ^2}  1/4} }}\) is fitted to the ratios E _{2}/E _{1}, yielding a statistical value of b = 3.145 with standard error of Δb = 0.06 consistent with the predicted value π. An error of ±1 mV is assumed for the position of the energy resonances.
A few comments are appropriate: (i) The points on the E _{2}(β) curve follow very closely the theoretical prediction \({E_{n + 1}}{\rm{/}}{E_n} = {e^{  \pi /\sqrt {{\beta ^2}  1/4} }}\). This result is insensitive to the choice of h, thus manifesting the universality of the ratio E _{ n+1}/E _{ n }. (ii) In contrast, the correspondence between the \(E_1^\prime \) points and the theoretical branch is sensitive to the choice of h. This reflects the fact that while each geometric ladder is of the form equation (2) (with the appropriate ζ → β change), the energy scale \({\epsilon _{\rm{0}}}\) is different between the two thus leading to a shifted relative position of the two geometric ladders in Fig. 3. The ansatz taken for h is phenomenological (Supplementary Note 2), however, we find that in order to get reasonable correspondence to theory, the explicit dependence on m is needed. More importantly, it is necessary to use a degeneracy breaking boundary condition to describe the \(E_1^\prime \)(β) points. For instance, if the Coulomb potential is regularized by a constant potential for r ≤ L ^{37}, then both angular momentum channels (i.e., the \(E_1^\prime \) and E _{1} points) become degenerate. The existence of the experimental \(E_1^\prime \) branch is therefore a distinct signal that parity symmetry in the corresponding Dirac description equation (3) is broken. In graphene, exchanging the triangular sublattices is equivalent to a parity transformation. Creating a vacancy breaks the symmetry between the two sublattices and is therefore at the origin of broken parity in the Dirac model. (iii) The value L ≃ 0.2 nm is fully consistent with the lowenergy requirement \({E_1}L{\rm{/}}\hbar {v_{\rm{F}}} \simeq 0.03 \ll 1\) necessary to be in the regime relevant to observe the βdriven QPT.
Discussion
A further argument in support of the universality of this QPT is achieved by comparing the experimental results obtained in graphene with those deduced from a completely different physical problem. To that purpose, we dwell for a short while recalling the basics underlying Efimov physics^{38}. Back to 1970, Efimov^{10} studied the quantum problem of three identical nucleons of mass m interacting through a short range (r _{0}) potential. He pointed out that when the scattering length a of the twobody interaction becomes very large, \(a \gg {r_0}\), there exists a scalefree regime for the lowenergy spectrum, \({\hbar ^2}{\rm{/}}m{a^2} \ll E \ll {\hbar ^2}{\rm{/}}mr_0^2\), where the corresponding boundstates energies follow the geometric series \(\left( {\sqrt {  {E_n}} =  {{\tilde \epsilon }_0}{e^{  \pi n/{s_0}}}} \right),\) where \({s_0} \simeq 1.00624\) is a dimensionless number and \({\tilde \epsilon _0}\) a problemdependent energy scale. Efimov deduced these results from an effective Schrödinger equation in d = 3 with the radial (l = 0) attractive potential \(V(r) =  \left( {s_0^2 + 1{\rm{/}}4} \right){\rm{/}}{r^2}\). Using Eqs. (1) and (2) and the critical value ζ _{c} = (d − 2)^{2}/4 = 1/4 for this Schrödinger problem, we deduce the ζ value for the Efimov effect to be \(s_0^2 + 1{\rm{/}}4 \, >\, {\zeta _{\rm{c}}}\) corresponding to the overcritical regime of the QPT. The value of β matching to the Efimov geometric series factor \({e^{\pi /{s_0}}}\) is \({\beta _{\rm{E}}} = \sqrt {s_0^2 + 1{\rm{/}}4} = 1.1236\), referred to as the fixed Efimov value. Despite being initially controversial, Efimov physics has turned into an active field especially in atomic and molecular physics where the universal spectrum has been studied experimentally^{39,40,41,42,43,44,45,46} and theoretically^{38}. The first two Efimov states E _{ n } (n = 1, 2) have been recently determined using an ultracold gas of caesium atoms^{47}. Although the Efimov spectrum always lies at a fixed and overcritical value of the coupling, unlike the case of graphene where β can be tuned, the universal character of the overcritical regime allows nevertheless for a direct comparison of these two extremely remote physical systems. To that purpose, we include the Efimov value β _{E} in the expression obtained for the massless Dirac fermion in a Coulomb potential and insert the corresponding data points obtained for cold atomic caesium in the graphene plot (Fig. 5) up to an appropriate scaling of \({\tilde \epsilon _0}\). The results are fully consistent thus showing in another way the universality presented.
There are other remote examples of systems displaying this universal QPT, e.g., flavoured QED3^{48}, and the XY model (KosterlitzThouless^{8} and roughening transitions^{22}). Our results provide a useful and original probe of characteristic features of this universal QPT and motivate a more thorough study of this transition.
Methods
Our sample is stacked two layers of graphene on top of a thin BN flake (see Fig. 1f). The standard dry transfer procedure is followed to get this heterostructure. A large twisted angle between the two layers graphene is selected in order to weaken the coupling. The freestanding like feature for the top layer graphene is checked by the Landau levels spectroscopy. To achieve the diluted single vacancies, the sample is exposed to the helium ion beam for short time (100 eV for 5 s) followed by the high temperature annealing. The experiment is performed at 4.2 K with a homebuilt STM. The dI/dV (I is the current, V is the bias) is recorded by the standard lockin technique, with a small AC modulation 2 mV at 473.1 Hz added on the DC bias. To tune the effective charge on the vacancy, we apply the voltage pulse (−2 V, 100 ms) with the STM tip directly locating on top of the vacancy.
Data availability
The data that support the findings of this study are available from the corresponding author upon request.
Additional information
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
 1.
Akkermans, E. in Fractal Geometry and Dynamical Systems in Pure and Applied Mathematics II: Fractals in Applied Mathematics vol. 601, 1–21 (eds Carf, D., Lapidus, M. L., Pearse, E. P. J. & van Frankenhuijsen, M.) (American Mathematical Society (AMS), 2013).
 2.
Adler, S. L. Axialvector vertex in spinor electrodynamics. Phys. Rev. 177, 2426–2438 (1969).
 3.
Bell, J. S. & Jackiw, R. A PCAC puzzle: π ^{0} → γγ in the σmodel. Il Nuovo Cimento A 60, 47–61 (1969).
 4.
Case, K. M. Singular potentials. Phys. Rev. 80, 797–806 (1950).
 5.
Landau, L. D. Quantum Mechanics: NonRelativistic Theory (ButterworthHeinemann, 1991).
 6.
LévyLeblond, J.M. Electron capture by polar molecules. Phys. Rev. 153, 1–4 (1967).
 7.
Camblong, H. E., Epele, L. N., Fanchiotti, H. & Garca Canal, C. A. Quantum anomaly in molecular physics. Phys. Rev. Lett. 87, 220402 (2001).
 8.
Kaplan, D. B., Lee, J.W., Son, D. T. & Stephanov, M. A. Conformality lost. Phys. Rev. D 80, 125005 (2009).
 9.
Nisoli, C. & Bishop, A. R. Attractive inverse square potential, U(1) gauge, and winding transitions. Phys. Rev. Lett. 112, 070401 (2014).
 10.
Efimov, V. Energy levels arising from resonant twobody forces in a threebody system. Phys. Lett. B 33, 563–564 (1970).
 11.
Efimov, V. Weaklybound states of three resonantlyinteracting particles. Sov. J. Nucl. Phys 12, 589–595 (1971).
 12.
Jackiw, R. W. Diverse Topics in Theoretical and Mathematical Physics (World Scientific, 1995).
 13.
Meetz, K. Singular potentials in nonrelativistic quantum mechanics. Il Nuovo Cimento 34, 690–708 (1964).
 14.
Gitman, D. M., Tyutin, I. & Voronov, B. L. Selfadjoint Extensions in Quantum Mechanics: General Theory and Applications to Schrödinger and Dirac Equations with Singular Potentials vol. 62 (Springer, 2012).
 15.
Albeverio, S., HøeghKrohn, R. & Wu, T. T. A class of exactly solvable threebody quantum mechanical problems and the universal low energy behavior. Phys. Lett. A 83, 105–109 (1981).
 16.
Beane, S. R. et al. Singular potentials and limit cycles. Phys. Rev. A 64, 042103 (2001).
 17.
Mueller, E. J. & Ho, T.L. Renormalization group limit cycles in quantum mechanical problems. Preprint at arXiv:condmat/0403283 (2004).
 18.
Braaten, E. & Phillips, D. Renormalizationgroup limit cycle for the 1/r ^{2} potential. Phys. Rev. A 70, 052111 (2004).
 19.
Hammer, H.W. & Swingle, B. G. On the limit cycle for the 1/r ^{2} potential in momentum space. Ann. Phys. 321, 306–317 (2006).
 20.
Moroz, S. & Schmidt, R. Nonrelativistic inverse square potential, scale anomaly, and complex extension. Ann. Phys. 325, 491–513 (2010).
 21.
De Martino, A., Klöpfer, D., Matrasulov, D. & Egger, R. Electricdipoleinduced universality for Dirac fermions in graphene. Phys. Rev. Lett. 112, 186603 (2014).
 22.
Kolomeisky, E. B. & Straley, J. P. Universality classes for linedepinning transitions. Phys. Rev. B 46, 12664–12674 (1992).
 23.
Shytov, A. V., Katsnelson, M. I. & Levitov, L. S. Atomic collapse and quasirydberg states in graphene. Phys. Rev. Lett. 99, 246802 (2007).
 24.
Pereira, V. M., Nilsson, J. & Castro Neto, A. H. Coulomb impurity problem in graphene. Phys. Rev. Lett. 99, 166802 (2007).
 25.
Nishida, Y. Vacuum polarization of graphene with a supercritical Coulomb impurity: lowenergy universality and discrete scale invariance. Phys. Rev. B 90, 165414 (2014).
 26.
Pomeranchuk, I. & Smorodinsky, J. On the energy levels of systems with Z > 137. J. Phys. 9, 100 (1945).
 27.
Akkermans, E. Twisted boundary conditions and transport in disordered systems. J. Math. Phys. 38, 1781–1793 (1997).
 28.
Akkermans, E., Dunne, G. & Levy, E. in Optics of Aperiodic Structures: Fundamentals and Device Applications (ed. Negro, L. D.) (Pan Stanford Publishing, 2013).
 29.
Smith, F. T. Lifetime matrix in collision theory. Phys. Rev. 118, 349–356 (1960).
 30.
Mao, J. et al. Realization of a tunable artificial atom at a supercritically charged vacancy in graphene. Nat. Phys. 12, 545–549 (2016).
 31.
Andrei, E. Y., Li, G. & Du, X. Electronic properties of graphene: a perspective from scanning tunneling microscopy and magnetotransport. Rep. Prog. Phys. 75, 056501 (2012).
 32.
Liu, Y., Weinert, M. & Li, L. Determining charge state of graphene vacancy by noncontact atomic force microscopy and firstprinciples calculations. Nanotechnology 26, 035702 (2015).
 33.
Lehtinen, O. et al. Effects of ion bombardment on a twodimensional target: atomistic simulations of graphene irradiation. Phys. Rev. B 81, 153401 (2010).
 34.
Chen, J.H., Li, L., Cullen, W. G., Williams, E. D. & Fuhrer, M. S. Tunable Kondo effect in graphene with defects. Nat. Phys. 7, 535–538 (2011).
 35.
Wang, Y. et al. Observing atomic collapse resonances in artificial nuclei on graphene. Science 340, 734–737 (2013).
 36.
Akkermans, E. & Montambaux, G. In Mesoscopic Physics of Electrons and Photons, ch. 7 (Cambridge University Press, 2007).
 37.
Pereira, V. M., Kotov, V. N. & Castro Neto, A. H. Supercritical Coulomb impurities in gapped graphene. Phys. Rev. B 78, 085101 (2008).
 38.
Braaten, E. & Hammer, H.W. Universality in fewbody systems with large scattering length. Phys. Rep. 428, 259–390 (2006).
 39.
Kraemer, T. et al. Evidence for Efimov quantum states in an ultracold gas of caesium atoms. Nature 440, 315–318 (2006).
 40.
Tung, S.K., JiménezGarca, K., Johansen, J., Parker, C. V. & Chin, C. Geometric scaling of Efimov states in a ^{6}Li–^{133}Cs mixture. Phys. Rev. Lett. 113, 240402 (2014).
 41.
Pires, R. et al. Observation of Efimov resonances in a mixture with extreme mass imbalance. Phys. Rev. Lett. 112, 250404 (2014).
 42.
Pollack, S. E., Dries, D. & Hulet, R. G. Universality in three and fourbody bound states of ultracold atoms. Science 326, 1683–1685 (2009).
 43.
Gross, N., Shotan, Z., Kokkelmans, S. & Khaykovich, L. Observation of universality in ultracold ^{7}Li threebody recombination. Phys. Rev. Lett. 103, 163202 (2009).
 44.
Lompe, T. et al. Radiofrequency association of Efimov trimers. Science 330, 940–944 (2010).
 45.
Nakajima, S., Horikoshi, M., Mukaiyama, T., Naidon, P. & Ueda, M. Measurement of an Efimov trimer binding energy in a threecomponent mixture of ^{6}Li. Phys. Rev. Lett. 106, 143201 (2011).
 46.
Kunitski, M. et al. Observation of the Efimov state of the helium trimer. Science 348, 551–555 (2015).
 47.
Huang, B., Sidorenkov, L. A., Grimm, R. & Hutson, J. M. Observation of the second triatomic resonance in Efimov’s scenario. Phys. Rev. Lett. 112, 190401 (2014).
 48.
Appelquist, T., Nash, D. & Wijewardhana, L. C. R. Critical behavior in (2 + 1)dimensional QED. Phys. Rev. Lett. 60, 2575–2578 (1988).
Acknowledgements
This work was supported by the Israel Science Foundation Grant No. 924/09. Funding for the experimental work provided by DOEFG0299ER45742 (STM/STS), NSF DMR 1207108 (fabrication and characterization).
Author information
Author notes
O. Ovdat and Jinhai Mao contributed equally to this work.
Affiliations
Department of Physics, Technion, Israel Institute of Technology, Haifa, 3200003, Israel
 O. Ovdat
 & E. Akkermans
Department of Physics and Astronomy, Rutgers University, Piscataway, NJ, 08854, USA
 Jinhai Mao
 , Yuhang Jiang
 & E. Y. Andrei
Authors
Search for O. Ovdat in:
Search for Jinhai Mao in:
Search for Yuhang Jiang in:
Search for E. Y. Andrei in:
Search for E. Akkermans in:
Contributions
O.O. and E.A.: Proposed observing the aforementioned quantum phase transition in graphene with a charged vacancy. They have contributed to interpreting and solving the theoretical model as well as analysing the experimental data and making contact with the theory. J.M., Y.J. and E.Y.A.: Conceived of and designed the experiment, as well as performed the measurements and analyzed the data. All authors contributed to discussions and preparation of the manuscript.
Competing interests
The authors declare no competing financial interests.
Corresponding author
Correspondence to E. Akkermans.
Electronic supplementary material
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Further reading

Efimov effect in Dirac semimetals
Frontiers of Physics (2018)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.