Landau poles in condensed matter systems

Shao-Kai Jian,1 Edwin Barnes,2 and Sankar Das Sarma1 1Condensed Matter Theory Center and Joint Quantum Institute, Department of Physics, University of Maryland, College Park, Maryland 20742, USA 2Department of Physics, Virginia Tech, Blacksburg, Virginia 24061, USA The existence or not of Landau poles is one of the oldest open questions in non-asymptotic quantum field theories. We investigate the Landau pole issue in two condensed matter systems whose long-wavelength physics is described by appropriate quantum field theories: the critical quantum magnet and Dirac fermions in graphene with long-range Coulomb interactions. The critical quantum magnet provides a classic example of a quantum phase transition, and it is well described by the φ4 theory. We find that the irrelevant but symmetry-allowed couplings, such as the φ6 potential, can significantly change the fate of the Landau pole in the emergent φ4 theory. We obtain the coupled beta functions of a φ4 + φ6 potential at both small and large orders. Already from the one-loop calculation, the Landau pole is replaced by an ultraviolet fixed point. A Lipatov analysis at large orders reveals that the inclusion of a φ6 term also has important repercussions for the high-order expansion of the beta functions. We also investigate the role of the Landau pole in a very different system: Dirac fermions in 2+1 dimensions with long-range Coulomb interactions, e.g., graphene. Both the weak-coupling perturbation theory up to two loops and a low-order large-N calculation show the absence of a Landau pole. Furthermore, we calculate the asymptotic expansion coefficients of the beta function. We find that the asymptotic coefficient is bounded by that of a pure bosonic φ4 theory, and consequently graphene is free from Landau poles if the pure φ4 theory does not manifest a Landau pole. We briefly discuss possible experiments that could potentially probe the existence of a Landau pole in these systems. Studying Landau poles in suitable condensed matter systems is of considerable fundamental importance since the relevant Landau pole energy scales in particle physics, whether it is quantum electrodynamics or Higgs physics, are completely unattainable.

I. BACKGROUND turbation theory, there is a second fundamental ‘problem’ with QED, first emphasized by Landau [3], and often re- ferred to as the Landau pole (it also is sometimes called Quantum electrodynamics (QED) is the most success- ‘Landau ghost’ or ‘Moscow zero’) [4]. The term Landau ful theory in physics. The most recent QED calcula- pole refers to the divergence of a running (or renormal- tion up to 10th order in the fine structure coupling con- ized) coupling constant at a finite energy in a field theory. stant (i.e. the fifth order in the QED perturbation the- Landau poles happen only in field theories that are not ory) involves the accurate determination of 389 differ- asymptotically free, where the running coupling increases ent high-dimensional integrals contributed by 6354 Feyn- with increasing energy, in contrast to asymptotically free man vertex diagrams, with a resultant electron anoma- field theories where the coupling constant decreases with lous magnetic moment agreeing quantitatively with ex- increasing energy. If the Landau pole is a true feature periments up to 10 significant digits [1]. Obviously, we of QED, then the only way to obtain a reasonable the- have come a long way (although it took 70 years for ory would be to set the bare charge to zero, resulting in this progress) from Schwinger’s ground-breaking analyt- a theory that is completely trivial (or noninteracting). ical work of 1948 calculating just the first order QED Landau poles have been discussed extensively in the con- correction, which obtained the electron anomalous mag- text of both QED and scalar field theories, such as the netic moment correct to two significant digits [2]. This φ4 theory, which is relevant for Higgs bosons in the stan- amazing agreement between theory and experiment is the dard model. The problem is, however, that the existence most impressive success of quantum mechanics, and the of a Landau pole is theoretically established only within common belief is that this astounding success will con- the leading-order perturbative approximation (extended tinue up to many orders in the QED perturbation theory, to very high energies) and therefore, whether the Landau arXiv:2003.05098v2 [cond-mat.str-el] 9 Jun 2020 with theory and experiment agreeing certainly up to well pole is real or an artifact of the perturbation theory is over 100 significant digits, although neither experiment unknown. In the context of QED, direct numerical simu- nor theory is likely to get to such a high precision in the lations have been used to explore the existence of Landau foreseeable future. Because the QED perturbation the- poles, but whether the poles have any physical relevance ory is asymptotic, it will eventually break down, with in the sense of imposing triviality remains unclear [5–7]. the perturbation series eventually diverging at some very Unfortunately, approaches based on lattice simulations high order ( 137), but this is not a serious concern at are hindered by chiral symmetry breaking, making the this stage. The infinite-order perturbative QED result parameter regime where a Landau pole occurs inaccessi- is thus bound to be incorrect since it gives a divergent ble [5]. Numerical methods have also been applied to φ4 answer. theory, with the conclusion that the continuum theory is In addition to the asymptotic nature of the QED per- most likely trivial. This finding has in turn been used 2 to impose bounds on the Higgs mass [8,9]. However, learn from condensed matter systems about the possible the triviality question has not been fully resolved here presence/absence of Landau poles would be valuable for either, and a stronger bound on the mass comes from future progress in the subject. unitarity [10], obscuring the role of the Landau pole. It is also unclear to what extent these findings carry over to effective scalar theories, where higher-order couplings II. INTRODUCTION allowed by symmetry can affect renormalization group (RG) flows at high energies. It is possible that a fully The concept of Landau poles is a long-standing issue nonperturbative, strong-coupling theory would be neces- in quantum field theory that raises questions about fun- sary to eventually settle the question since weak-coupling damental aspects of the renormalization group (RG), in perturbation theories simply may not be applicable in particular the asymptotic behavior of renormalized cou- predicting the physics of a divergent renormalized cou- plings [3, 12, 13]. In a quantum field theory that is not pling. The Landau pole problem is intrinsically tied to asymptotically free, consider the beta function that gov- the asymptotic behavior of the QED perturbation ex- erns the running of a dimensionless coupling constant λ, pansion and to the nonperturbative effects of instantons. dλ Although there has been recent progress in developing a = β(λ), (1) more rigorous treatment of nonperturbative effects like d log µ instantons using resurgent trans-series [11], the question of the existence or not of the Landau pole in QED re- where µ is the energy scale of interest. One can integrate mains as open today as it was in the early 1950s when over both sides to get Landau first proposed it. Z ∞ dλ Z µ∞ = d log µ. (2) One thing is, however, clear. Even if the Landau pole λ(µ1) β(λ) µ1 exists in QED and/or quartic scalar field theories, there is no hope for its experimental manifestation in particle µ∞ is the energy scale at which the running coupling be- physics because the energy scale for the Landau pole is comes infinite. Suppose the beta function has the asymp- unphysically huge (e.g. way above the Planck scale in totic behavior β(λ) ∝ λa, λ  1, where a is a constant QED and the Higgs mass in the φ4 theory). Quite apart independent of λ, then from the fact that such large energy scales are experi- ∞ ∞ ( mentally unattainable (not only now, perhaps ever), new Z dλ Z ∞, a ≤ 1 ∝ λ−adλ = physics, outside the scope of QED, comes in at high en- β(λ) log µ∞ < ∞, a > 1. λ(µ1) λ(µ1) µ1 ergy scales, and the predictions of QED for the Landau (3) pole become academic since QED itself (quite apart from a perturbation theoretic analysis of QED leading to the In the first case, a ≤ 1, there is no Landau pole. The Landau pole) is no longer a correct description of nature running coupling reaches infinity only at infinite energy at such high energies. scales. In the second case, a > 1, however, the coupling This is the context of the current theoretical work, diverges at a finite energy scale µ∞. This is a Landau where we investigate the condensed matter analogs of the pole. Clearly, the existence of a Landau pole depends on Landau pole in the theories of critical quantum magnets the asymptotic form of the beta function at λ  1. and the physics of graphene. It is well known that the φ4 Because Landau poles were first discovered theoreti- scalar field theory describes the long-wavelength critical cally in quantum field theories relevant for high energy behavior of quantum magnets, and the two-dimensional physics, such as QED, research into this issue has re- massless Dirac theory describes the long-wavelength be- mained largely within the high-energy physics commu- havior of graphene. Thus, graphene is an example (with nity [5–7, 14–17]. The simplest solution to the Landau suitable modifications) of QED, whereas quantum mag- pole problem is the idea of quantum triviality [6, 14–16], nets are examples (with suitable modifications) of the in which the pole is avoided by setting the coupling to φ4 scalar field theories. Our goal is to study the pres- zero at all scales, yielding a trivial, noninteracting the- ence/absence of Landau poles in these two concrete con- ory. This is of course unsatisfactory if one is interested densed matter physics examples to motivate further the- in the effect of interactions (which are clearly present in oretical work that could shed light on the fundamental QED experiments), so a more phenomenological resolu- issue of triviality in the quantum field theories which are tion that is often adopted is to argue that the theory is not asymptotically free. Another equally important ob- incomplete and gets replaced by another theory at high jective of our work is to motivate experimental work in energies before the Landau pole is reached. For instance, condensed matter systems to directly probe the existence QED is usually believed to be just one part of a more or not of Landau poles in these two systems, which are fundamental electroweak theory. Moreover, the Landau described by continuum field theories containing Landau pole in QED can be estimated to occur at an energy poles in the leading-order perturbative analysis. The scale on the order of 10286 eV, which is far beyond the question of the existence or not of Landau poles is of Planck scale 1028 eV, suggesting that it is a purely aca- sufficient fundamental significance that anything we can demic issue. However, the reliability of such estimates is 3 questionable given that the Landau pole is intrinsically set dimensionally by the lattice spacing. It is possible tied to the large-coupling regime, whereas most analyses that the cutoff is in fact somewhat shorter (longer) than rely on weak-coupling perturbation theory carried out to the lattice spacing, in which case the effective field theory first or second order. Many attempts have been made would apply to energies above (below) the inverse lattice to go beyond small-order perturbation theory starting spacing. In the context of Landau poles, the hope is that from Lipatov’s method [18, 19] for calculating large or- the pole energy here would not be much higher than the ders in an asymptotic series. In fact, a resummation inverse lattice spacing energy (within an order of mag- procedure based on such large-order expansions suggests nitude) so that there is some reason to believe that the that the Landau pole does not exist at all in the φ4 the- effective field theory does not completely break down at ory [20, 21]. But the Landau pole question is by no means the Landau pole energy. In such a situation, it is sensible settled since the resummation technique is essentially a to consider condensed matter experiments to investigate Borel interpolation, and the exact strong coupling theory the presence/absence of Landau poles. If the pole hap- remains elusive, so one cannot be sure that the Landau pens to be at an energy much larger than the natural pole does not exist. In particular, numerical simulations ultraviolet cutoff for the lattice in the system, there is no seem to indicate its existence [6]. hope for experimentally studying the Landau pole in the corresponding condensed matter system since the effec- In addition to having wide application in particle tive field theory is unlikely to apply at that high-energy physics, continuum field theories and the renormalization scale. group are also important for condensed matter systems, especially in the context of phase transitions, where the Here, we consider two condensed matter systems: universal physics is controlled by long-wavelength fluctu- the critical quantum magnet and Dirac fermions in ations at a critical point [22, 23]. (In fact, Wilson de- 2+1 dimensions with long-range Coulomb interactions veloped his RG theory motivated by condensed matter (graphene). Both systems are believed to be well de- considerations in critical phenomena, and the first prob- scribed by effective continuum field theories. The critical 4 lem he solved using his momentum-shell RG theory is quantum magnet is described by a bosonic φ theory as 4 the Kondo problem, a celebrated condensed matter prob- dictated by symmetry. In four dimensions, the φ theory lem involving singular spin-flip scattering of electrons in has a coupling that grows logarithmically with energy metals from quenched magnetic impurities [24].) Thus, scale, eventually leading to a Landau pole as determined it is natural to consider the Landau pole problem in con- from perturbation theory [30]. However, this theory is an densed matter systems. Unlike most cases in high-energy effective field theory, meaning that all terms allowed by n physics, here the ultraviolet completion is well under- symmetry, e.g., all even-order terms φ , where n ∈ 2Z, 4 stood as it comes from the lattice, which brings a natu- are in principle present in the action. Although the φ ral energy scale cutoff set by the inverse lattice constant. term is the most relevant one deep in the infrared, the The question then becomes whether a Landau pole arises n > 4 terms can become important at the higher energy above or below this scale. This question is particularly scales where the Landau pole potentially arises. This important in systems where the coupling constant can be motivates us to investigate to what extent Landau poles 4 large, as is the case in graphene, where the fine structure are affected by infrared-irrelevant terms in the φ effec- constant away from the Dirac point is α ∼ 1 [25], and tive field theory. To answer this question, we study the 4 6 even more so in synthetic twisted bilayer graphene, where beta function of a φ + φ potential and find that the α ∼ 10 [26–29]. The interactions in these systems are Landau pole can be strongly affected by the presence of 6 substantially stronger than in QED, where α ≈ 1/137, the φ potential at both small and large orders of the suggesting that if there is a Landau pole, then its energy asymptotic expansion. scale could occur below the cutoff, potentially leading Dirac fermions in graphene constitute a second impor- to consequences that are experimentally accessible. We tant case study because the Fermi velocity approaches emphasize that in condensed matter physics, all contin- zero in the ultraviolet, enhancing the role of Coulomb uum field theories are, by definition, effective field theo- interactions at high energies and potentially inducing ries since the lattice explicitly breaks the continuum by a Landau pole [25]. The theory is infrared-stable be- providing a short-distance cutoff length scale (or equiv- cause the Fermi velocity increases as the energy scale ap- alently, an effective ultraviolet energy/momentum cutoff proaches zero, leading to a decreasing running coupling for the system which does not have to be put in by hand (which is cut off at some exponentially low energy scale in the theory). The field theory is valid up to this ultra- when retardation effects due to the finite speed of light violet energy/momentum scale (∼the inverse lattice con- come into consideration). The marginal Fermi liquid be- stant), which is a physical constraint defining the domain havior that results from the logarithmic growth of the of validity of the effective field theory. Our current un- Fermi velocity near the Dirac point is well studied [31– derstanding of all field theories as effective field theories 33]. We review the result for the small-order beta func- up to some scale is explicitly obeyed in condensed matter tion in graphene [34, 35] and show that the Landau pole physics and does not have to be artificially introduced as, problem is not resolved by either weak-coupling or large- for example, in lattice gauge theories. Of course, one does N perturbation theories at one loop. On the other hand, not know the precise cutoff scale except that it should be a two-loop analysis using either ordinary perturbation 4 theory or a large-N expansion indicates the absence of a 0.25

Landau pole. We also compute the large-order expansion 0.20 coefficients nonperturbatively, showing that the asymp- 4 totic coefficients are smaller than those of the pure φ 0.15 theory. This indicates that if the pure φ4 theory does not

6 0.10 manifest a Landau pole, then neither does the Coulomb- λ interacting graphene field theory. Our main, but ten- 0.05 tative, conclusion in this work is that in all likelihood, Landau poles do not exist in condensed matter systems 0.00 ● ● at any energy scale, although the leading-order pertur- bative RG theory may imply their existence, often at -0.05 -0.2 -0.1 0.0 0.1 0.2 energy scales beyond which the corresponding effective λ4 field theory is applicable.

The paper is organized as follows. We discuss Landau FIG. 1. The flow diagram of λ4-λ6. The blue arrow indicates poles in the critical quantum magnet and in graphene in the RG flow to higher energies. The black and red dots are Secs.III andIV, respectively. In each section, we com- the Gaussian and the new fixed point in Eq. (7), respectively. pute both the small- and large-order terms of the beta The dashed line which separates different flow regimes is given function for the running coupling. While the small-order by Eq. (8). We take N = 1 for simplicity. beta function is obtained through the standard pertur- bation theory, the large orders are evaluated through the saddle-point approximation following the Lipatov at the critical surfacer ¯ = 0, wherer ¯ is the dimensionless method [18, 19]. In Sec.V, we briefly discuss possible ex- mass, are periments and conclude the paper. The appendixes con- dλ 4 = −(3N + 12)λ + (4N + 32)λ2, (5) tain technical details and an instructive zero-dimensional d log µ 6 4 toy model to illustrate the steps of the Lipatov method dλ 32N + 832 for the φ4 + φ6 potential. 6 = 2λ + (12N + 168)λ λ − λ3,(6) d log µ 6 4 6 3 4  2 A3 g4 A3 g6 2 III. CRITICAL QUANTUM MAGNETS where λ4 = (2π)4 4 and λ6 = (2π)4 6 Λ are dimen- sionless couplings, and we have set d = 4. Ad is the area In this section, we compute the small- and large-order of a d-sphere with unit radius. terms of the asymptotic expansion of the beta function From Eqs. (5) and (6), besides the Gaussian fixed for the effective theory consisting of both φ4 and φ6 point, there is also an ultraviolet fixed point at terms. We use the Wilsonian renormalization group to (λ4, λ6) obtain the one-loop beta function. We show that while 3 4  N + 8 (N + 8)  the pure φ theory exhibits a Landau pole, the inclusion = − , of the φ6 term removes the pole and replaces it with an ul- 2(N 2 + 6N + 128) 3(N + 4)(N 2 + 6N + 128)2 traviolet fixed point. In addition, we compute the large-  1 1  ≈ − + O(1/N 2), + O(1/N 3) . (7) order expansion of the beta function by using a saddle- 2N 3N 2 point approximation in the regime of negative couplings. We find that the expansion coefficients of the φ4 coupling The fixed point has two relevant directions at the criti- 2 λ4 grow as k! while those of λ6 grow as (k!) . From both cal surface. One should distinguish it from the tricritical small-order and large-order expansion coefficients, we see point. For d ≥ 3 spacetime dimensions, the tricritical that including the φ6 term strongly affects the fate of the point is the same as the Gaussian fixed point, because Landau pole compared to a pure φ4 theory. the upper critical dimension is three for the φ6 theory. Moreover, a tricritical point has only one relevant direc- tion at the critical surface. A. Small orders The flow diagram of λ4-λ6 is shown in Fig.1, where the black and red dots are the Gaussian fixed point and We consider the effective theory, the ultraviolet fixed point in Eq. (7), respectively. The dashed curve is given by Z   d 1 2 g4 4 g6 6 3 S[g4, g6; φ] = d x (∂φ) + φ + φ , (4) 16(N + 26)λ4 2 4 6 f(λ4) = , λ4 > 0. (8) 3(6Nλ4 + 84λ4 + 1) where φ , i = 1, ..., N is an O(N) field, and the sum- i If one fixes λ = 0 to reduce to the pure φ4 theory, as in- 2 PN 2 2 6 mation over i is implicit: φ = (φi) , and (∂φ) = i dicated by the black line in Fig.1, λ4 will become infinite PN Pd 2 i µ(∂µφi) . Using the standard Wilsonian RG tech- at the Landau pole. This is consistent with the one-loop nique (see AppendixA), the perturbative beta equations beta function for λ4. However, in the global flow diagram 5 of both λ4 and λ6, the points at λ6 = 0 are not stable Since the unstable saddle-point solution at g4 < 0 will at higher energies. Depending on the initial values, the contribute to the imaginary part, we use the saddle-point RG flow regimes are separated by the dashed line given approximation. When g6 = 0, the saddle-point equation by Eq. (8). Denote the initial values by λ40 and λ60. On is the one hand, at λ40 < 0 ∪ [λ40 > 0 ∩ λ60 > f(λ40)], it 2 3 will flow to the new fixed point (if one can fine tune −∂ ϕc + g4ϕc = 0. (12) λi = 0 for i = 8, 10, ...). On the other hand, at ξ(x) 2 3 λ40 > 0 ∩ λ60 < f(λ40), λ6 will flow to negative val- Let ϕc(x) = √ , where ξ is the solution of ∂ ξ +ξ = 0 ues, indicating that the theory is not stable because the −g4 and is independent of g4. In d = 4 dimensions, there are energy is not bounded from below. Therefore, if one in- √ two solutions: ξ(x) = ± 2 2 (see AppendixC). Note cludes the symmetry allowed φ6 term, the Landau pole x2+1 that the saddle-point equation is scale-invariant in the is replaced by a fixed point of the one-loop beta function. sense that if ξ(x) is a solution, then b−1ξ(x/b) is also a Note that even if the φ6 term is not included initially, it 4 solution for any rescaling parameter b. We will remove will be generated by the φ potential. R d 2 this scaling redundancy later on. Because d x[(∂ϕc) + 4 g4ϕc ] = 0, the on-shell action is B. Large orders g Z S S[g , 0; ϕ ] = − 4 ddxϕ4 = − 0 , (13) 4 c 4 c 4g We have shown that the inclusion of the φ6 term in 4 the effective theory will replace the Landau pole by an R d 4 where S0 ≡ d xξ is a positive constant independent ultraviolet fixed point even in the one-loop calculation. of g . In d = 4 dimensions, S = 32π2/3. We now extend the analysis to large orders. Namely, we 4 0 The fluctuations around the saddle-point solution can calculate the asymptotic expansion coefficients βk, k  1 P k be expanded in terms of an orthonormal basis {ϕj(x)}, of beta functions, β(λ) = k βkλ . We first calculate the asymptotic expansion of the correlation function, and ∞ then relate it to the beta function. In AppendixB, we X φ(x) = ϕc(x − x0) + cjϕj(x). (14) present a toy model that provides a simple illustration of j=1 the basic steps we employ.

Here, x0 is an arbitrary point in spacetime. Using this basis, one can change the functional integration into a 1. Asymptotic expansions of the correlation function c-number integration. Note that to make the integration variable linearly independent, one requires Consider the n-point correlation function of the mass- 4 6 Z less φ + φ theory defined by Eq. (4), d d xϕj(x)∂µϕc(x) = 0, µ = 1, ..., d. (15) n Z Y G (g , g ; X ) = Dφ φ(x )e−S, (9) n 4 6 n j Moreover, the Jacobian that results from the change of j=1 integration variables from the functional measure to x0 where Xn = (x1, ..., xn) denotes all n spacetime argu- and cn is given by ments, and in this section we consider N = 1 for sim- 1/2 plicity. We are interested in computing the asymptotic " d Z #  d/2 Y d 2 S0 expansion of the n-point correlation function Gn(g4, g6) J = d x(∂µϕc) = − , (16) dg4 (in the following, we suppress the argument Xn for no- µ=1 tational simplicity). In general, the expansion has the following form: where we have used the spherical symmetry of the saddle- point solution. The quadratic kernel around the saddle X (n) k X (n) k0 Gn(g4, g6) = Ak g4 + Bk0 (g4)g6 . (10) point is k≥0 k0≥1 δ2S[g , 0; ϕ ] In four dimensions, the asymptotic expansion of 0 4 c  2 2  0 M(x, x ) = 0 = −∂x − 3ξ (x) δ(x − x ), 4 δϕc(x)δϕc(x ) Gn(g4, 0) as a function of g4 is known for the pure φ theory with g6 = 0 [18, 19]. We first review this calcu- (17) lation and then generalize it to g6 > 0. When g6 = 0, P (n) k which is independent of g4. Note that ∂µϕ(x) is the eigen- Gn(g4, 0) = k Ak g4 has a branch cut at g4 ∈ (−∞, 0). As shown for the toy model in Eq. (B8), the expansion function of the kernel with zero eigenvalue. The above coefficient is related to the branch cut by orthonormal basis {ϕj} can be chosen to be the eigen- functions of M(x, x0). Z 0 (n) dz ImGn(z, 0) Putting everything together, the imaginary part of the Ak = k+1 . (11) −∞ π z correlation function under the saddle-point approxima- 6 tion is given by which is independent of g6. We evaluate the imaginary part of the correlation function using the saddle-point 1 X Z Z ImG (g , 0) = Jddx e−S[g4,0;ϕc] Dϕ approximation, which yields n 4 2i 0 ⊥ ϕc 1 Z Z X d −S[φc] n ImGn(0, g6) = Jd x0e Dφ⊥ R d d 0 0 0 2i Y − d xd x ϕ⊥(x)M(x,x )ϕ⊥(x ) φ × ϕ(xj)e c n j=1 Y R d d 0 0 × φ(x )e− d xd x φ⊥(x)M(x,x )φ⊥(x) d/2 S0 j S  e 4g4 F (X ; ξ) = 0 n n , (18) j=1 d/2+n/2 1/2 S6 d (−g4) i(det M) d/2 − √ S  e 3 −g6 F (X ) = 6 n n .(25) R d Q d/4+n/4 1/2 where Fn(Xn; ξ) = d x0 j ξ(xj − x0), n ∈ 2Z, and d (−g6) i(det M) P ϕ⊥ denotes the functions orthogonal to ∂µϕ. de- ϕc where F (X ; χ) = R ddx Q χ(x − x ) and n ∈ 2Z. notes the summation over the two saddle points. The n n 0 j j 0 expansion coefficient is The expansion coefficients are B(n)(0)  d/2 k (n) k S0 iFn(Xn) S A = (−1) d/2 0 − √6 k 1/2 S  iF (X ) Z dz e 3 −z d (det M) = (−1)k 6 n n 1/2 k+d/4+n/4+1 0 S0 d (det M) −∞ π (−z) Z dz e 4z × d 2k+ n k+d/2+n/2+1 k 2  3  2 iFn(Xn; χ)  3  2  d n −∞ π (−z) = (−1) Γ 2k + + . 1/2 d n π d (det M) S6 2 2 1  4  2 iF (X ; ξ)  4 k+ 2 d n = (−1)k n n Γ(k + + ). 1/2 (26) π d (det M) S0 2 2 (19) The fact that the kernel M has a negative eigenvalue can be seen by noting that Because M(x, x0) has a unique negative eigenvalue [19], Z Z i ddxddx0χ(x)M(x, x0)χ(x0) = −4 ddxχ6(x) < 0. (det M)1/2 is a positive number. 4 6 We now generalize this procedure to the case of φ +φ (27) theory. We first treat the pure φ6 theory nonperturba- tively and then include perturbative corrections from the It remains to be shown that there is an odd number of φ4 term. The saddle-point equation of a pure φ6 theory negative eigenvalues. We assume this is true, in which i is case (det M)1/2 is a real number. Now we can consider the correction from a finite −∂2φ + g φ5 = 0. (20) R d 4 c 6 c d xg4φ term. The first-order perturbation is given by the correction to the on-shell action, χ(x) 2 5 The solution is φc(x) = 1/4 , where −∂ χ − χ = 0. (−g6) Z R d 2 6 S6 d 4 S6 g4 Because d x[(∂φc) + g6φc ] = 0, the on-shell action is S[g4, g6; φc] ≈ √ + g4 d xφc = √ − S4, 3 −g6 3 −g6 g6 Z g6 d 6 S6 (28) S[0, g6; φc] = − d xφc = √ , (21) 3 3 −g6 R d 4 where S4 = d xχ is a positive constant independent R d 6 of g4 and g6. Then the asymptotic coefficient is where S6 ≡ d xχ (x) is a positive constant indepen- (n) dent of g6. Bk (g4) As before, fluctuations around the saddle-point solu- d S6 g4 0 − √ + S4 S  2 iF (X ; χ) Z dz e 3 −z z tion can be expanded in an orthonormal basis {φj(x)}, = (−1)k 6 n n d (det M)1/2 π (−z)k+d/4+n/4+1 ∞ −∞ X d 2k+ n φ(x) = φ (x − x ) + c φ (x), (22) 2  3  2 iFn(Xn; χ)  3  2 d n c 0 j j = (−1)k Γ(2k + + ) 1/2 j=1 π d (det M) S6 2 2 2 R d  d n 1 S6 1  and d xφj(x)∂µφc(x) = 0, for µ = 1, ..., d. In this case, ×U k + + , , , (29) the Jacobian is 4 4 2 36S4 g4 where we have defined " Z #1/2  d/2 Y d 2 S6 a J = d x(∂µφc) = √ . (23) U(a, b, c) = c U(a, b, c), (30) d −g6 µ and U is Tricomi’s (confluent hypergeometric) function. The quadratic kernel around the saddle point is One can further expand Tricomi’s function to get the j k expansion coefficients of g4g6 , as we will do in the next 0  2 4  0 M(x, x ) = −∂x − 5χ (x) δ(x − x ), (24) section. 7

√ 2. Beta functions at large orders 2 2 solution, namely, ξ(x) = ± x2+1 . These modifications are taken into account by working with the function In the previous section, we computed the asymptotic Z Z 4 2 d −3 3y − x0  expansion of the n-point correlation function of the φ + F˜ξ(y) = S0 d log b d x0b ξ . (35) φ6 theory. In this section, we use this result to compute b the high-order expansion of the beta function. Putting everything together, we get We define the renormalized couplings as the full n- 2 2 p  p  X (n) (n) j point vertex functions, λn = −Γn(¯g4, g¯6; 2 ), where k k µ Γn g¯4, g¯6; = Ak g¯4 + Bj,k g¯4g¯6 , (36) 2 µ2 n = 4, 6, andg ¯4 = g4 andg ¯6 = g6µ are the dimensionless k couplings. Here, p is the momentum at the renormaliza- tion point, and µ denotes the energy scale. From the and the asymptotic forms of the coefficients are given by equations defining the saddle points, Eqs. (12) and (20), k d (−1)  4  2 iF˜ (p) we have A(n) = ξ k π d (det M)1/2 Z d 3 k+ n ξ(x) = d y∆ (x − y)ξ (y), (31)  4  2 d n 0 × Γ(k + + ), (37) Z S0 2 2 d 5 k d 2 χ(x) = d y∆ (x − y)χ (y), (32) 2(−1)  3  2 iµ F (p) 0 B(n) = χ m,k π d (det M)1/2 1 n m where ∆0(p) = 2 is the free propagator. We then have  3 2k+ 2 d n ∂ U p × Γ(2k + + ) , (38) S 2 2 m! Z n ˜ 6 d Y where Fs(p), Fs(p)(s = ξ, χ) are the Fourier transforms Fn(Xn; s) = d yFs(y) ∆0(xj − y), (33) of Fs(x), F˜s(x). j=1 The function Eq. (30) has the expansion R d 3 where s = ξ, χ, e.g., Fξ(y) = d x0ξ (y − x0) and ∞ R d 5 X (a)n(a + 1 − b)n Fχ(y) = d x0χ (y − x0). Equation (33) converts the U(a, b, c) = (−c)−n, (39) n! correlation function to the vertex function [18]. There is n=0 a subtlety for d = 4 dimensions because the φ4 potential where (a)0 ≡ 1, (a)n ≡ a(a + 1)...(a + n − 1). The two is scale invariant in this case. If ξ(x) is a solution to the m 2 3 ∂ U saddle-point equation, ∂ ξ + ξ = 0, then it follows that limits of m! are given by b−1ξ(x/b) is also a solution for any rescaling factor b. We  2m  m m k 36S4 can remove this trivial scaling redundancy by using the ∂ U  − 2 , k  m  1, m! S6 following identity: → 2k  m (40) m! m!m 36S4  (k!)2 − S2 , m  k  1. Z  Z x2  6 1 = S d log b2δ d4xξ4(x) log , (34) 0 b2 (n) from which we see that the asymptotic behaviors, Ak (n) (n) in conjunction with the rescaling transformation ξ(x) → and Bk,m for fixed m increase factorially as k!, while Bm,k b−1ξ(x/b). The δ function above fixes the scaling of the for fixed m increases faster as (k!)2, i.e.,

k (n)  4  d n k 2 + 2 −1 Ak ∝ (−1) k k!, k  1 (41) S0  d+n−3  2k m +2m  k+m 3 36S4 k 2 k 2 (n) (−1) S S2 m! 4 (k!) , k  m  1 B ∝ 6 6 (42) m,k  2k m d+n−3 k+m 3 36S4 2 k 2k (−1) 2 k 4 m m!, m  k  1 S6 S6

4 6 We are interested in the φ +φ theory in four dimensions, from which we can reexpandg ¯n as a function of λn. The the coupling constants λ4 and λ6 are leading orders are given by ∞ X (4) k X (4) j k λ4 =g ¯4 − Ak g¯4 − Bj,k g¯4g¯6 , (43) k=2 j,k ∞ (4) 2 2 X (6) k X (6) j k g¯4 = λ4 + B0,1λ6 + O(λ4, λ6, λ4λ6), (45) λ6 = − A g¯4 +g ¯6 − B g¯4g¯6 . (44) k j,k 2 2 k=2 j,k g¯6 = λ6 + O(λ4, λ6, λ4λ6). (46) 8

The beta functions have the asymptotic expansion, the function B(g) with a branch cut at −α,

∞ −β−1  g  X k dλ4 X (4) k X (4) j k B(g) = 1 + = Bkg , (51) = − ∂Ak g¯4 − ∂Bj,k g¯4g¯6 , α d log µ k k=2 j,k (k + β)! X (4) k (4) k k k β ≈ − ∂A [λ + (B λ ) ] Bk = (−α) → (−α) k , (52) k 4 0,1 6 k!β! k P k X (4) (j + k)! (4) to a series W (g) = Wkg whose asymptotic coeffi- − ∂A λj (B λ )k k j+k j!k! 4 0,1 6 cient is given by j,k k β X (4) j k Wk → (−α) k k!. (53) − ∂Bj,k λ4λ6 , (47) k Comparing Eq. (53) to Eqs. (41) and (42), a naive ∞ 6 dλ6 X (6) X (6) Borel sum without considering the φ potential is given = − ∂A g¯k − ∂B g¯jg¯k, d log µ k 4 j,k 4 6 by k=2 j,k −4 X (6) (4) (0)  S0  ≈ − ∂A [λk + (B λ )k] B (λ ) ∼ 1 + λ . (54) k 4 0,1 6 4 4 4 k X (6) (j + k)! (4) The presence of λ6 leads to a different (naive) Borel sum − ∂A λj (B λ )k k j!k! 4 0,1 6 that at k-th order has the form j,k  S2 −2k−1 X (6) j k (k) k 6 − ∂B λ λ . (48) B (λ4) ∼ λ6 1 + λ4 . (55) j,k 4 6 36S4 k It is apparent that these Borel sums lead to different ∂ where ∂A = ∂ log µ A. asymptotic behaviors. Nevertheless, we need to em- Equations (41), (42), (47) and (48) are the main results phasize that instead of having multiple Borel sums, like of this section. We see from these results that the inclu- Eqs. (54) and (55), for each power in λ6, it is more ap- sion of the φ6 term, which is allowed by symmetry in con- propriate to treat the asymptotic expansions of λ4 and densed matter systems (and, in any case, is automatically λ6 on equal footing, because λ6 has a vanishing radius of generated in the theory from the φ4 term), dramatically convergence. changes the asymptotic behavior of the beta function for λ4. On the one hand, if λ6 ∼ λ4, the asymptotic coef- IV. DIRAC FERMIONS WITH COULOMB ficient of the beta function for λ4 changes qualitatively from k! to (k!)2: INTERACTIONS

2k (4)  6  d+1 In this section, we consider d-dimensional Dirac 2 2 B0,k ∝ k (k!) , k  1. (49) S6 fermions with Coulomb interactions. Ultimately, we are interested in Dirac fermions in graphene, where d = 2+1, Thus, we see that the inclusion of λ6 completely changes with Coulomb interactions. This is the condensed mat- the asymptotic behavior of the beta function for λ4. In ter analog of QED in the context of Landau poles. To light of Eq. (3), this will in turn change the fate of the have a local quantum field theory, one can implement Landau pole. On the other hand, even if λ6  λ4, as a Hubbard-Stratonovich transformation to decouple the 6 long as the φ term is nonzero λ6 > 0, the asymptotic Coulomb interaction by introducing a scalar field. The form of the second term in Eq. (47) is comparable to the theory under consideration is then given by [34] first term of the pure λ4, i.e., Z d−1 k d ¯ 0 ¯ X i ¯ 0 (4)  4  d +1 S = d x(ψγ ∂0ψ + vψ γ ∂iψ + igφψγ ψ) ∂A ∝ k 2 k!, k  1, (50) 1+k i=1 S0 Z 3 1 4 X 2 which can also qualitatively change the behavior of the + d x (∂iφ) , (56) 4 2 beta function compared to the pure φ theory. In particu- i=1 lar, including the φ term will result in a new asymptotic 6 where ψ (φ) refers to the four-component Dirac fermion series, so even if the pure λ4 series does not end up with 6 (the Hubbard-Stratonovich scalar field), and v (e) refers a Landau pole, it is possible that including the φ term µ will lead to a Landau pole. to the Fermi velocity (the coupling strength). γ , µ = 0, ..., d − 1, denotes the 4 × 4 γ matrices. The effective We are not aware of a technique to resum an asymp- Ng2 totic series with multiple variables, but it is still useful interaction strength is given by α = 16v (also known as the fine structure constant), where N = 2 is the spin to illustrate how the Borel sum of λ4 is affected by the 2 6 degeneracy, and g2 = 2e , where ε is the background φ potential. Notice that a Borel transform [36] will take (1+ε)0 9

to lie in a region of the parameter space made inaccessi- ble by a strong-coupling chiral symmetry-breaking transi- tion.) However, evidence for the two-loop critical point in Eq. (57) has not been observed in graphene experiments, (a) suggesting that the second-order perturbation theory in α may not be reliable, as was argued in Ref. [35]. In par- ticular, the existence of such a possible strong-coupling fixed point preempting the Landau pole may necessitate a stability analysis going to higher orders. For example, if the two terms in Eq. (57) were the leading terms of (b) (c) (d) a geometric series, then the result to all orders would be [35] FIG. 2. (a) The polarization of the scalar field. The solid 2 line (single-dashed) represents the Dirac fermion propagator dα 2 α = f1 , (58) (the bare scalar propagator). (b-d) The 1/N order Feynman d log µ f1 − f2α diagrams. The double-dashed line represents the screened scalar propagator. which gives rise to β(α  1) ∝ α2 at small α, and β(α  1) ∝ α at large α. This time there is no ultraviolet fixed point in the theory defined by Eq. (58), and the dielectric constant (which in general can be > 1 in solid coupling grows indefinitely with energy scale. There is state systems), and 0 is the vacuum permittivity. (Note still no Landau pole, but the mechanism by which it is that the definition of α used here contains an additional avoided is quite different. A one-loop theory with just the factor of πN/4 compared to that used in Ref. [35].) first term in Eq. (57) clearly has a Landau pole whereas As in the previous analysis of critical quantum mag- neither the full Eq. (57) with both terms or the resummed nets, we calculate the behavior of the beta function of theory of Eq. (58) has a Landau pole (albeit for very the fine structure constant at both small and large or- different reasons). This illustrates how leading-order one- ders of perturbation theory. Small-order terms can be loop perturbative calculations can be misleading when calculated using either the standard perturbation theory it comes to determining the existence of a Landau pole in α or a large-N analysis in which 1/N is the expansion in a given theory. Similar trends can occur in large-N parameter. Here, we review the results obtained previ- calculations, as we now discuss in the case of graphene. ously for both methods [34, 35, 37], and we discuss their Besides expanding in the coupling strength, it is often implications for the Landau pole problem. To our knowl- useful to consider a large-N expansion (in the number of edge, the large-order terms of the asymptotic series in α fermion flavors rather than a perturbative expansion in had not been computed prior to the present work. Here, the coupling constant) instead, especially when the bare we obtain these by first integrating out the Dirac fermion coupling parameter is not small. This is the situation for and then using a saddle-point approximation similar to free-standing graphene, where αbare ≈ 3.4, suggesting the one employed in the previous section on quantum that an expansion in powers of 1/N with N = 2 would magnets. be more reliable than an expansion in powers of αbare (simply by virtue of the fact that α > 1 and 1/N < 1). This was confirmed in Ref. [37]. In the following, we A. Small orders review the large-N calculation for graphene and show that a Ward identity guarantees that only one parameter, The two-loop beta equation was calculated in Ref. [35] the Fermi velocity, renormalizes. Moreover, we show that for the case of graphene (where d = 2 + 1 and we have the beta function obtained from the leading-order large- N = 2 flavors of four-component Dirac fermions), N calculation does not exhibit a Landau pole. We believe that the same remains true for the next-to-leading-order dα theory in 1/N also, which is likely to produce only small = f α2 + f α3, (57) d log µ 1 2 quantitative corrections to the leading order 1/N theory. The leading large-N diagram shown in Fig. 2(a) gives g2 1 2 log 2−8/3 rise to the dressed boson propagator, where α = 8v , and f1 = 2π , and f2 = π2 . Im- portantly, we see that f2 < 0, so when α becomes large, 3  g2N |~q|2 −1 the negative α term dominates and prohibits α from be- D(q) = 2|~q| + , (59) coming larger. Therefore, the Landau pole is replaced by 8 pq2 a fixed point at α∗ = − f1 . This is similar to the φ4 + φ6 f2 2 2 2 theory in the previous section. (A similar strong-coupling where q = q0 + |~q| . The Feynman diagrams at the next fixed point is found in the direct nonperturbative numer- order are shown in Figs. 2(b)-2(d). Before calculating ical simulation of the usual (3+1)-dimensional QED on these diagrams, it is worth noting that there is a residual the lattice in Ref. [5], where the Landau pole was found gauge invariance in Eq. (56) corresponding to the trans- 10 formation It is apparent that there is no Landau pole because 1 β(α  1) ∝ α. Instead, the coupling diverges as the ψ → eiθ(τ)ψ, φ → φ − ∂ θ(τ), (60) g τ energy scale goes to infinity. However, if we expanded perturbatively in α (and also in 1/N), and retained terms where θ is an arbitrary function of τ. This residual gauge up to order α3, we would erroneously conclude that there symmetry leads to the Ward identity, is a fixed point at α = 3π/8. This is similar to the fixed −1 −1 p0Γ3(p0, ~p; q0, ~q) = ig[G(q0 + p0, ~q) − G(q0, ~q) (61)], point [c.f. Eq. (57)] we found from our original perturba- −1 tion theory in α above. Of course, if we instead stopped where Γ is the vertex function, and G is the inverse 2 3 at order α2, we would obtain β(α  1) = α , and we fermion propagator. This Ward identity is similar to πN the Ward identity in standard QED where the photon would conclude that there is a Landau pole as in the is dynamical, except that here it only involves the tem- leading-order perturbative 1-loop theory [i.e., Eq. (57) poral component. The Ward identity relates the charge with just the first term on the right-hand side]. Thus, renormalization factor to the anomalous dimension of the we see that truncating the series at a low order in α scalar field. can produce misleading results about the Landau pole While the Dirac fermion is confined to a two- issue. This is again an indication that the Landau pole dimensional plane, the scalar field in Eq. (56) can propa- in graphene, which is apparent in the 1-loop perturbative gate in the full three-dimensional space. Thus, the back- RG theory as in the original QED context [3], may be an reaction from the Dirac fermion is restricted within the artifact of perturbation theory. two-dimensional plane, and so it does not lead to an anomalous dimension for the scalar field. Technically, B. Large orders this is because the scalar propagator is nonlocal in the −1 plane D0 (q) ∼ |~q|, and thus it does not receive cor- rections. As a result, according to the Ward identity, Now we turn to the large-order expansion of the beta the charge does not renormalize. The only parameter in function. Because of the fermionic nature of Dirac fields, Eq. (56) that does renormalize is the Fermi velocity v, or the saddle-point analysis is not directly applicable. We can circumvent this problem by integrating over the equivalently, the dynamical exponent z = 1+γv, where γv is the anomalous dimension of Fermi velocity. Note that Grassmanian field, which results in a functional deter- this is not true for the theory in 3+1 dimensions, because minant. In the strong-coupling limit, the determinant the scalar field can have a finite anomalous dimension in can be evaluated using the quasilocal approximation (see that case, leading to two independent parameters in 3+1 AppendixD), which then yields an effective action in- dimensions. volving just the scalar field [19, 38]. In the following, we In light of the above discussion, the only Feynman dia- start from this effective action and apply a saddle-point gram we need to evaluate is Fig. 2(b), because the vertex analysis to obtain the large-order expansion of the beta correction can be inferred from the self-energy correction function. by the Ward identity. Using the results from Ref. [34], The effective action is given by [19, 39, 40] Z Z we get the fermion self-energy, 4 1 2 d d S = d x (∇φ) + d xαd|φ| , (67) 4 0 2 Σ(p) = [f0(α)p0γ + f1(α)~p · ~γ] log µ. (62) Nπ2 2Γ(−d/2) N 2 d/2 where αd = d/2 vd−1 (g ) . Notice that it involves where (4π) √ a different combination of g and v because of the dif- ( 1−α2 π ferent approximation we implement in the large-order − α arccos α − 1 + 2α , α < 1 f1(α) = √ √ α2−1 log(α + α2 − 1) − 1 + π , α > 1 series. However, in 2 + 1 dimensions, since from the α 2α previous subsection we know that the RG correction to (63) the charge g vanishes and all RG effects come from the ( 2−α2 π 1 N 3 − √ arccos α − 2 + , α < 1 renormalization of the Fermi velocity v, α3 = 3π v2 |g| is α 1−α2 α f0(α) = 2 √ directly connected to α in the weak-coupling expansion √α −2 log(α + α2 − 1) − 2 + π , α > 1 α α2−1 α approach. The large-order perturbation series for Dirac (64) fermions with Coulomb interactions is equivalent to that of the effective scalar field theory given by Eq. (67). The from which we can extract the dynamical exponent, 1 − 4 potential term in Eq. (67) leads to a branch cut in the z = 2 [f1(α) − f0(α)], and the beta equation, π N n-point correlation function for αd ∈ (−∞, 0). This al- ∂α lows us to use similar techniques as before to evaluate = (1 − z)α. (65) ∂ log µ the asymptotic expansion of the correlation function. The equation of motion that follows from Eq. (67) is We can expand the beta function at small and large α: −∇2φ + dα sign(φ )|φ |d−1δ(4−d)(z ) = 0, (68)  1  1 2 8 3 4  c d c c ⊥  N π α − 3π2 α + O(α ) , α  1 β(α) = (66) where z⊥ denotes the coordinates perpendicular to the 1  4 4 −1   N π2 α − π + O(α ) . α  1 spacetime where the Dirac fermions live. For instance, 11 z⊥ = z is the direction perpendicular to the plane where what is shown in Eq. (35), we arrive at F˜n(Xn; η) = the Dirac fermions are confined in the case of d = 2 + 1 R 2 R 3 Qn −1 ~xj −~x0 Sd d log b d x0 j=1 b η( b ). The expansion co- dimensions (graphene). The equation of motion is solved efficient is then 1 − d−2 by φc = |dαd| η, where η is a solution to the differen- 3 2 d−1 (4−d) 1 S  2 iF˜(X ; η) tial equation −∇ η − sign(η)|η| δ (z⊥) = 0. Note A(n)(z) = (−1)k d n k n 1/2 that here αd < 0 since we want to apply the saddle-point d d−2 3 (det M) technique. The on-shell action is h (2−d)S − 2 i 0 d d−2 Z dz exp 2d (−z) 2 × d − 2 − k+1+ n+3 Son-shell = |αd| d−2 Sd, (69) −∞ π (−z) d−2 2d 3 2 k d − 2 Sd  2 iF (Xn; η) R d − d−2 d = (−1) where Sd = d xd |η| is a positive constant inde- n 1/2 2πd d−2 3 (det M) pendent of αd. d−2 3+n  2d  2 k+ 2 d − 2 n + 3 At each time τ, the field configuration can be expressed × Γ k + . by an orthogonal basis ϕj(~x) (d − 2)Sd 2 2 X (75) φ(τ, ~x) = φc(~x − ~x0) + fj(τ)ϕj(~x). (70) j Comparing Eq. (75) to Eq. (19), we obtain the large-order expansion of the beta equation: We can change the functional integration measure to ~x0 and fj. The Jacobian associated with the change of vari- dα X = C αk, (76) ables is given by d log µ k d k 1/2 3/2 " Z # " # d−2 k Sd  2d 1  d n+1 d − 2  Y 4 2 k 2 J = d x(∂iφc) = 2 , (71) Ck ∝ (−1) k k !. (77) d−2 d − 2 Sd 2 i 3|αd| and the quadratic kernel around the saddle-point solu- The k-th order expansion coefficient of the beta function d−2 tions is is proportional to ( 2 k)!. It is less than the asymptotic 4 2 expansion coefficient ∼ k! in the pure φ theory in d < 4 0 δ S[φc] dimensions. In the (2 + 1)-dimensional Dirac fermion M(x, x ) = 0 δφc(x)δφc(x ) theory with Coulomb interactions, the asymptotic coef- k h 2 d−2 (4−d) i (4) 0 ficient is proportional to ( )!. Since every term in the = −∇ − (d − 1)|η| δ (z⊥) δ (x − x ).(72) 2 asymptotic expansion is bounded by that of a pure φ4 It is straightforward to show that this kernel has a nega- model, we can conclude that if there is no Landau pole tive eigenvalue by projecting it onto η in a manner anal- in pure φ4 theory, then there is no Landau pole in the ogous to Eq. (27). We again assume that there is an odd Dirac fermion theory with Coulomb interactions. This is number of negative eigenvalues. a direct consequence of the fact that the existence of a We are interested in the n-point correlation function, Landau pole is fully determined by the asymptotic be- havior of the beta function, as shown in Eq. (3). Z n Y −S Gn(αd; Xn) = Dφ φ(xj)e . (73) j=1 V. DISCUSSION AND CONCLUSION In the saddle-point approximation, the imaginary part of this correlation function is Since the seminal works of Wilson [23, 24], it has 1 X Z Z been understood that quantum field theories are essen- ImG (α ) = Jd3x e−S[φc] Dφ n d 2i 0 ⊥ tially effective descriptions of the low-energy and long- φc wavelength behavior of an underlying physical system, n Y R d d 0 0 and the quest for an axiomatic foundation of quantum × φ(x )e− d xd x φ⊥(x)M(x,x )φ⊥(x) j field theories, fashionable during the 1960s, is futile and j=1 unnecessary. Such an effective field theory description is h (2−d)Sd i 3 exp 2 of course particularly germane in condensed matter sys- 1 S  2 F (X ; η) d−2 = d n n 2d(−αd) , tems where the existence of a physical lattice imposes n 1/2 n+3 d−2 3 i(det M) d−2 d (−αd) a real high-energy short-distance cutoff on any contin- (74) uum description. It is remarkable that such theories can give quantitatively accurate predictions over a wide va- R 3 Qn where Fn(Xn; η) = d x0 j=1 η(~xj − ~x0). Because riety of systems and phenomena, ranging from Koster- the effective action is scale-invariant, i.e., invariant un- litzThouless transitions in Josephson-junction arrays to der η(x) → b−1η(x/b), the scaling freedom needs to be the tricritical point where water and gas can no longer fixed in the path integral. Using a similar method to be distinguished. In the context of particle physics, the 12

Landau pole issue is often viewed as purely academic, be- suggest that this may indeed be the case for systems cause when a pole appears, it is typically at an incredibly currently under investigation. The well-studied three- high energy scale, and one normally assumes that even dimensional antiferromagnet TlCuCl3 features an O(3) if the pole is a real feature of the field theory, the theory quantum phase transition realized by tuning the pres- itself will likely become invalid by the time this energy is sure. From Ref. [30], the quartic interaction strength reached. It is difficult to make these statements precise at the quantum critical point of TlCuCl3 is estimated because, in the particle physics context, effective field to be around λ4 ≈ 0.23/4π at 1 meV, and accordingly theories capture the low-energy limit of a system whose the one-loop calculation predicts that the Landau pole high-energy, microscopic degrees of freedom are often un- occurs at 3.5 meV, which is far below the lattice scale. known. In addition, the continuum is presumably real in For comparison, the measured magnon dispersion at zero particle physics since there is no underlying physical lat- pressure (in the disordered phase) reaches as high as 7 tice providing a short-distance cutoff. Appealing to the meV [41], well above the predicted Landau pole energy. possibility of a theory with the Landau pole becoming In the ordered phase, the reported gap of the longitudi- inapplicable at high energies (where the pole presumably nal excitation reaches about 1.2 meV [42]. These facts lies) because some other theories may control the physics imply that the predicted Landau pole is within the reach at the higher energy scale is not aesthetically pleasing of experiments. A possible signature of a Landau pole because the disturbing question still remains about the could come from the fact that a diverging quartic inter- existence of the original theory with the Landau pole action should cause a strong decay of the longitudinal (e.g., QED): Is it a well-defined interacting theory or is mode to transversal modes in the ordered phase. This in it trivial? turn suggests that the decay width should exhibit a sig- The Landau pole problem becomes a bit more con- nificant enhancement if there is a Landau pole. However, crete in condensed matter systems, because the limita- existing data shows no evidence of such an increase in the tions of such theories are typically known and determined line-width [cf. Ref. [43]]. On the other hand, there has by lattice-scale cutoffs. This raises interesting questions: by no means been an exhaustive search, and a systematic Is it possible for a Landau pole to occur below the lattice- experimental survey at energies approaching the lattice scale cutoff? What would be the experimental implica- scale is required to settle this issue. We believe that ex- tions of this? The effective field theory point of view periments in quantum magnets looking for signatures of makes the Landau pole existence question rather subtle Landau poles are needed given that the existence of an though. The fact that all symmetry-allowed terms can infinite number of symmetry-allowed field operators (i.e. 2n become important in the high-energy regime of an ef- all the φ terms) in the theory make a decisive theoret- fective field theory makes it particularly challenging to ical conclusion impossible. determine if a Landau pole arises. Because the Landau We also examined a second type of condensed matter pole is by definition a high-energy phenomenon, it does system (namely, graphene) searching for Landau poles: not matter whether these terms influence the low-energy Dirac fermions in 2+1 dimensions with Coulomb interac- properties of the theory or not. In the critical quantum tions. This is the QED analog of the Landau pole (albeit magnet described by the φ4 theory, φ2n (n > 2) poten- in two spatial dimensions). We argued with the help of tials are present without question. Although irrelevant a Ward identity that the renormalization group flow is to the long-wavelength critical phenomena, the appear- controlled by a single parameter [35], the fine structure ance of a φ6 potential on top of the φ4 theory can signifi- g2 constant α ∼ v (or equivalently the Fermi velocity). cantly change the fate of the Landau pole at high energies Both perturbation theory in α and the large-N expan- (which is a short-distance rather than a long-wavelength sion to leading order suggest that the Landau pole is phenomenon), as we have shown. On the one hand, at absent in graphene. However, keeping only the first few small orders of the beta function, we find that the Lan- orders in perturbation theory is insufficient to address dau pole is removed and replaced by a new ultraviolet the issue. Therefore, we also evaluated the coefficients of fixed point because of the coupled RG flow of λ4 and high-order terms in the asymptotic series using a nonper- λ6. On the other hand, at large orders, the asymptotic turbative approach adapted from the Lipatov’s method. expansion of the beta function gets a large contribution Similarly to the case of relativistic fermions with Yukawa from λ6, which can also alter the fate of the Landau pole. type interactions [19, 38], the asymptotic coefficient is, 2n What happens if other φ terms are also included in the for d < 4 dimensions, bounded by that of the four- analysis? Ultimately, it may have to be decided by exper- dimensional pure φ4 theory, and consequently graphene iments that probe strong-coupling, high-energy regimes is free of Landau poles if the pure φ4 theory does not to finally settle the Landau pole issue in condensed mat- manifest a Landau pole. Moreover, the knowledge of the ter systems. asymptotic series combined with a few small-order co- We note that there is nothing in principle ruling out efficients can be used as the input into a resummation the possible existence of a Landau pole in condensed mat- technique that could potentially lead to a resolution of ter systems at energy scales well below the ultraviolet the Landau pole problem that is independent of pure φ4 lattice cutoff scale, making the issue relevant both the- theory in this particular context [20, 21]. We leave this oretically and experimentally. In fact, rough estimates technically demanding calculation to future work. 13

Finally, we point to another direction for possible fu- as inferred from the running of the coupling implied by ture investigations. As is apparent from the form of the the one-loop perturbative RG theory does not exist (and fine structure constant, small velocities can yield large is purely an artifact of the one-loop theory) since the coupling strengths. It is worth noting that the Fermi one-loop theory seems unable to quantitatively describe velocity at the charge neutrality point in twisted bilayer the running of the coupling at large coupling. These graphene near the magic angle is extremely small [26–29]. preliminary measurements should be repeated in future This suggests that this system could be a particularly experiments for a definitive resolution of the question of interesting place to explore the Landau pole problem, Landau pole in graphene since its implications extend far assuming the continuum Dirac description is still valid beyond graphene all the way to QED, where α ∼ 1/137, here. We can estimate the Landau pole energy scale us- and the RG running of the coupling all the way to the ing the result for the one-loop beta function, which yields Landau pole in laboratory experiments is manifestly √ 2π/α EL ≈ ~vF ne , where n is the carrier density and α impossible. Careful experimental investigation of twisted is the effective coupling (fine structure constant). Typ- bilayer graphene near the Dirac point may finally shed 8 ical values for these quantities are vF ≈ 10 cm/s and light on the 80-year-old question of the existence or not n ≈ 1012 cm−2. For graphene grown on a BN substrate of Landau poles in quantum electrodynamics. 3 where α ≈ 0.4π/2, EL ≈ 10 eV, which is several or- ders of magnitude larger than the lattice cutoff. On the other hand, for suspended graphene where α ≈ 2.2π/2, the Landau pole energy is around EL ≈ 0.4 eV, which is comparable to the lattice scale. Interestingly, for twisted ACKNOWLEDGEMENTS bilayer graphene on hBN where α ≈ 10π/2, the Lan- dau pole energy is further suppressed to EL ≈ 98 meV. S.-K.J. is supported by the Simons Foundation through This suggests that the Landau pole could be very likely the It from Qubit Collaboration. S.D.S. is supported by within the reach of experiments. A possible signature the Laboratory for Physical Sciences and the Microsoft could come from the fact that the Fermi velocity should Corporation. E.B. acknowledges support from the Na- be strongly suppressed as the Landau pole energy scale is tional Science Foundation (DMR-1847078). approached due to the fact that the velocity gets renor- malized and scales as 1/α. This in turn could lead to interaction-enhanced dispersion flattening and strong ef- fective couplings that could persist as the system is tuned Appendix A: Perturbative renormalization group 4 6 slightly away from a magic angle. In fact, it was reported analysis of φ + φ theory in Ref. [29] that the running coupling constant extracted from experiments is not consistent with a one-loop calcu- Since there are two vertices, the structure of Feynman lation, indicating that intriguing strong-coupling effects diagrams is may be taking place in twisted bilayer graphene. Such strong interaction effects, in light of our discussion, may L = I − V4 − V6 + 1, (A1) provide important insights on the Landau pole problem E = 4V + 6V − 2I, (A2) as it pertains to Dirac fermions. On the other hand, if a 4 6 small gap is opening up at the Dirac point in the twisted bilayer graphene, then the Dirac description fails at low where L, I, and E denote the number of loops, internal energies, and the Landau pole issue becomes moot. propagators, and eternal propagators, and V4 and V6 de- note the number of four-vertices and six-vertices, respec- It may be useful to emphasize that the detailed anal- tively. For the RG equation at the one-loop level, the ysis of Ref. [29] focusing on the twisted bilayer graphene Feynman diagrams contributing to the RG of polariza- experiments of Refs. [27, 28] and Ref. [44] came to the tion (V4,V6) = (1, 0), four-point vertex (2, 0), (0, 1), and conclusion that the measured effective mass and Fermi six-point vertex (1, 1), (3, 0) are shown in Figs.3,4,and5. velocity near the Dirac cone of low-angle twisted bilayer The renormalized action coming from the Feynman di- graphene agrees with strong-coupling nonperturbative theories such as the resummed Borel-Pad´eperturbation series and the 1/N expansion, while disagreeing very strongly with the one-loop perturbative RG theory. In particular, the one-loop theory predicts a very large renormalization of the effective mass and the Fermi velocity for the large (α ∼ 10π/2) effective fine structure (a) (b) constant in the system, which is simply not observed experimentally. If these preliminary experimental results FIG. 3. Feynman diagrams contribute to the beta equation of hold in future measurements in flat band twisted bilayer the polarization. The solid line represents the boson propaga- tor of the same index, and the wavy line represent short-range graphene, where the effective interaction strength is very 4 6 large, one inevitable conclusion is that the Landau pole interactions (φ and φ vertices). 14

Using the momentum shell, the loop integrals are Z Z Λ 2 1 Ad−1 Λ l G(k) = 2 = d , (A4) k k k + r (2π) 1 +r ¯ Z Z Λ (a) (b) (c) 2 1 Ad−1 l G(k) = 2 2 = d 2 , (A5) k k (k + r) (2π) (1 +r ¯) Z Z Λ 3 1 Ad−1 l G(k) = 2 3 = d 2 3 , (A6) k k (k + r) (2π) Λ (1 +r ¯)

d R Λ R d k Ad−1 R Λ d−1 (d) (e) where k ≡ Λ (2π)d = (2π)d Λe−l k dk, l > 0 is the −2 running parameter, andr ¯ = rΛ . Ad is the area of FIG. 4. Feynman diagrams contribute to the beta equation the d-dimensional unit sphere. Using these integrals, we of the four-point vertex. arrive at the effective action Z h1 1 A Λ2l S = ddx ∂φ2 + r + g (2N + 4) d−1
φ2 2 2 4 (2π)d 1 +r ¯  A Λ2l + g + g (3N + 12) d−1 4 6 (2π)d 1 +r ¯ 1 A l  − g2(8N + 64) d−1 φ4 2 4 (2π)d (1 +r ¯)2 A l +g − g g (12N + 168) d−1 (a) (b) (c) 6 4 6 (2π)d (1 +r ¯)2 1 A l i + g3(64N + 1664) d−1
φ6 , (A7) 6 4 (2π)d Λ2(1 +r ¯)3 which leads to the beta equations in Eqs. (5) and (6) by settingr ¯ = 0 to restrict to the critical surface.

(d) (e) (f) Appendix B: Zero-dimensional φ4 + φ6 theory: a toy model

Here, we study a toy model [19] to better illustrate the field theory calculation. The following integral is a (g) (h) zero-dimension ‘toy’ version of the φ4 theory: Z 1 −( 1 x2+ g4 x4) FIG. 5. Feynman diagrams contribute to the beta equation Z(g4) = √ dxe 2 4 . (B1) of the six-point vertex. 2π The integral can be done analytically,

1 8g4 1 e K 1 ( 8g ) Z(g ) = √ 4 4 , (B2) 4 2 πg agrams is 4 where K is the modified Bessel function of the second kind. It is apparent that Z(g4) has a branch cut at g4 ∈ (−∞, 0) if one analytically continues g4 to the complex Z Z plane. As a result, the integral is controlled by the branch d h 2 δS = d x g4(2N + 4) G(k)φ cut, namely, k Z Z 0 Z 0 4 dz Z(z + i) − Z(z − i) dz ImZ(z) +g6(3N + 12) G(k)φ Z(g4) = = . k −∞ 2πi z − g4 −∞ π z − g4 Z 1 2 2 4 (B3) − g4(8N + 64) G(k) φ 2 k Z We are interested in getting the asymptotic expansion 2 6 at the strong-coupling limit, −g4g6(12N + 168) G(k) φ k ∞ 1 Z i X k 3 3 6 Z(g4) = Akg4 , k → ∞. (B4) + g4(64N + 1664) G(k) φ . (A3) 6 k k=0 15

To get the answer, it is straightforward to expand the Thus we get the coefficient integral Z 0 k dz 1 1 (−1) k A = − √ e 4z = √ 4 Γ(k) Z k k k+1 1 1 2 X (−1) 2π z 2π − 2 x −k 4k k −∞ Z(g4) = √ dxe 4 x g4 , (B5) 2π k! (−1)k k = √ 4k(k − 1)!. (B14) 2π from which It is precisely the asymptotic expansion shown in k 1 k (−1) Γ(2k + 2 ) (−1) k Eq. (B6). From this lesson, we know that it is the unsta- Ak = √ → √ 4 (k − 1)!, (B6) π k! 2π ble saddle point giving rise to the imaginary contribution, and consequently to the asymptotic expansion. where we have used the Stirling’s formula and the iden- Now let us consider another integral that is the zero- tity dimensional version of the φ4 + φ6 theory, Z 1 (2n)!√ 1 −S(x) Γ(n + ) = π. (B7) Z(g4, g6) = √ dxe , (B15) 2 4nn! 2π 1 g g We see that the asymptotic expansion is not convergent. S(x) = x2 + 4 x4 + 6 x6. (B16) There is another way to get the asymptotic expansion, 2 4 6 4 which relates the expansion coefficient to the branch cut Unlike the zero-dimensional φ integral, Z(g4, g6) is an- of g4. To get the expansion coefficient, we use alytic for all g4 as long as g6 > 0. The branch cut comes from g6 ∈ (−∞, 0). So we can use the same strategy h k Z 0 i Z 0 1 ∂ dz ImZ(z) dz ImZ(z) to get the asymptotic expansion coefficient of g6. The Ak = k = k+1 , k! ∂g4 −∞ π z − g4 g4=0 −∞ π z asymptotic expansion of Z(g4, g6) is given by

(B8) X k X k0 Z(g4, g6) → Akg4 + Bk0 (g4)g6 , (B17) from which we know the expansion coefficient through k k0>0 the imaginary part of Z(g4) at the branch cut. where Ak is given by Eq. (B14). Similar to Eq. (B8), the When g is negative, one needs to continue the variable 4 asymptotic expansion of g6 is given by x to make the integral converged. For g4± ≡ g4 ± , making x ≡ ρeiθ, the φ4 potential in Eq. (B1) is given by Z 0 dz ImZ(g , z) B (g ) = 4 . (B18) k 4 π zk+1 g |g | −∞ 4± x4 = 4 ρ4e±iπ+i4θ 4 4 We can use the saddle-point approximation to get the |g | imaginary part of the integral. In the following, we con- = 4 ρ4[cos(±π + 4θ) + i sin(±π + 4θ)],(B9) 4 sider g4 as a small perturbation. The saddle-point equa- tion is (g4 = 0, g6 < 0) Thus, we can use a different contour at |x|  1 for g4±: 5 x + g6x = 0, (B19) 3π π g4 = g4+, − < θ < − , (B10) and the saddle-points are (for g6 < 0) x = 0, xc = 8 8 1 ± 1/4 . The quadratic fluctuation near the saddle π 3π (−g6) g = g , < θ < . (B11) point solution is given by 4 4− 8 8 g4 1 Instead of evaluating the integral directly, we can use S(δx) ≈ − + √ − 2δx2, (B20) 4g 3 −g the saddle point approximation, which can be extended 6 6 to the field theory. The saddle point equation is g4 4 where we include 4 x as a perturbation. As expected 3 from the unstable solution, the fluctuation has a negative x + g4x = 0, (B12) mass. This contributes to the imaginary part of integral, namely, the imaginary part of the integral is given by where the saddle points are x = 0 and x = ± √ 1 . Ap- −g4 parently, x = 0 does not contribute to the imaginary 1 X 1 Z −i∞ √ −S(δx) part. And the other two saddle points lead to the contri- ImZ(g4, g6) = dδxe 2i 2π i∞ bution x=xc 1 g4 − √1 4g6 3 −g Z −i∞ = − e 6 , (B21) 1 X 1 −( 1 x2+ z x4) x2 2 ImZ(z) = √ e 2 s 4 s dxe 2i 2π i∞ x =± √1 from which we can get the asymptotic expansion s −z k 1 1 (−1)  1 1  4z −k = −√ e . (B13) Bk(g4) = g4 Γ(2k)U k, , , (B22) 2 π 2 9g4 16 where U is the confluent hypergeometric function. The fact that we can set c = ±2 in the last step and still Thus the theory has at least two fates. On the one obtain a valid solution is a consequence of the scale in- hand, if g4  g6, the asymptotic expansion coefficient of variance of Eq. (C1). Does Eq. (C3) contain other valid g6 vanishes, solutions beyond Eq. (C5)? To answer this question, we return to Eq. (C2) and analyze the large y behavior. As lim Bk(g4) = 0. (B23) g4→∞ y → ∞, we require f(y) → 0 because we need the action R d 4 at the saddle point, S0 = d xξ (x), to be finite so that The expansion is controlled by Ak in Eq. (B14) as ex- it makes a non-negligible contribution to the correlation pected. On the other hand, if g4  g6, we recover the function. In this limit, the f 3(y) term either decays more asymptotic expansion of the pure φ6 integral, quickly than the other two terms in Eq. (C2) or it is com- parable to them. If it is comparable, then the solution (−1)k (−1)k62k lim B (g ) = 9kΓ(2k) → (k!)2(B24). behaves as k 4 3/2 g4→0 π 2(πk) √ d − 3 f(y) → ± √ as y → ∞. (C6) To check this, it is straightforward to calculate the coef- y ficient directly, This implies that ξ(x) ∼ 1/|x| as |x| → ∞, which in 1 k Z ∞ (− 6 ) 1 6k − 1 x2 turn leads to a divergent action, S0 → ∞, for dimension Bk(0) = √ dxx e 2 k! 2π −∞ d ≥ 4. Therefore, we discard such solutions and instead consider the case where the f 3(y) decays faster than the (− 4 )k Γ(3k + 1 ) (−1)k62k = √3 2 → (k!)2.(B25) other terms in Eq. (C2). In this case, the asymptotic π Γ(k + 1) 2(πk)3/2 behavior of f(y) is We can see that the results from the saddle-point ap- f(y) → cy1−d/2 as y → ∞, (C7) proximation exactly reproduce the asymptotic expan- 2−d sion. More importantly, different from the φ4 integral, where c is a constant. This yields ξ(x) ∼ |x| as the asymptotic coefficient of the φ6 integral increases as |x| → ∞, and the action is finite for d ≥ 3. When d = 4, (k!)2, which is qualitatively different from that of the φ4 this asymptotic behavior is of course the same as that of theory. Eq. (C5). The constant c in Eq. (C7) is again a reflec- tion of the scale-invariance of the solutions, and so we conclude that the only two solutions with finite action Appendix C: Saddle-point solutions of φ4 and φ6 are those given in Eq. (C5). theories The saddle-point equation for pure, massless φ6 theory is 4 The saddle-point equation for pure, massless φ theory ∂2χ + χ5 = 0. (C8) is We again assume that χ(x) is spherically symmetric, in ∂2ξ + ξ3 = 0. (C1) which case Eq. (C8) reduces to 4yg00(y) + 2dg0(y) + g5(y) = 0, (C9) We assume that ξ(x) is spherically symmetric, i.e., it only 2 Pd 2 depends on x = xi , where d is the dimension of where g(y) = χ(x). We require that the solutions vanish i=1 R d 6 5 the (Euclidean) spacetime. Writing f(y) = ξ(x) where as y → ∞ so that S6 = d xχ (x) < ∞. If g (y) is y ≡ x2, Eq. (C1) becomes comparable to the other two terms in Eq. (C9) as y → ∞, then the asymptotic behavior is 4yf 00(y) + 2df 0(y) + f 3(y) = 0. (C2) (2d − 5)1/4 g(y) → ± √ . (C10) In d = 4 dimensions, the solution to this equation is 2y1/4 r  √  This implies that χ(x) ∼ 1/p|x| as |x| → ∞, which c + 2 c c + 2 f(y) = sn √ log(y/y0) , (C3) in turn leads to a divergent S for d ≥ 3. We therefore y c 6 2 2 conclude that the g5(y) term decays faster than the other two terms in Eq. (C9), which then yields the following where sn(w, m) is the Jacobi elliptic function, and y0 and c are integration constants. For c ≈ 0, this reduces to asymptotic behavior: √ g(y) → cy1−d/2 as y → ∞. (C11) 16 cy0 f(y) = . (C4) Because Eq. (C8) has a scale invariance such that if χ(x) cy + 32y0 is a solution then so is b−1χ(x/b2), we understand that If we then set c = ±2 and y0 = ±1/16, this becomes c is a redundant scale factor. Therefore, there are only √ two distinct solutions of Eq. (C8) that contribute to cor- 2 2 relation functions, and these two solutions are related by f(y) = ± . (C5) 1 + y an overall minus sign. 17

Appendix D: Quasi-local approximation of a Now we include spin, and consider the determinant of determinant the Dirac operator,

0 −1 First consider a determinant [19] D(e) = det(∂/ + ieφγ + m)(∂/ + m) . (D5)

2 2 −1 µ µ det(− 4 +m + gV (x))(− 4 +m ) . (D1) where ∂/ = γ ∂µ, and γ is the 4 × 4 Dirac matrix. Here to regularize the determinant, we introduce a mass m. Pd 2 where 4 = µ ∂µ. A useful formula is We set it to zero in the final step to retain the gapless Dirac dispersion. To evaluate the determinant, note that Z ∞ dt Tr log AB−1 = −Tr (e−At − e−Bt). (D2) under the parity symmetry, 0 t D(e) → det(−∂/ − ieφγ0 + m)(−∂/ + m)−1. (D6) Since we are interested in the large g behavior, the inte- gral is dominated by the small t regime. We can use the As a result, we can evaluate the square of determinant, quasilocal [19, 38] approximation, i.e.,

2 2 e−(−4+m +gV (x))t ≈ e−t(−4+m )e−tgV (x). (D3) log D(e)2 = Tr log(− 4 +e2φ2 + m2)(− 4 +m2)−1, Now we have (D7) Tr log(− 4 +m2 + gV (x))(− 4 +m2)−1 where we have assumed the field φ is large and smooth ∞ to neglect the correction from derivatives and used the Z dt 2 = −Tr e−t(−4+m )(e−tgV (x) − 1) fact the γ matrix is traceless. From Eq. (D4), we note 0 t ∞ d Z Z dt Z d p 2 2 Z h 2Γ(−d/2) i −t(p +m ) d −tgV (x) D(e) = exp − ddx[(m2 + e2φ2)d/2 − md] . = − d e d x(e − 1) d/2 0 t (2π) (4π) ∞ 1 Z Z dt 1 2 2 (D8) = − ddx [e−t(m +gV (x)) − e−tm ] d/2 d/2 Now we can safely send m to zero to get the effective field (4π) 0 t t theory in Eq. (67). Including the Fermi velocity and the Γ(−d/2) Z = − ddx[(m2 + gV (x))d/2 − md]. (D4) fermion species N, the final answer is modified as (4π)d/2 h 2Γ(−d/2) N Z i One can also evaluate the correction to the above approx- D(e) = exp − ddx[(m2 + e2φ2)d/2 − md] . d/2 d−1 imation order by order using the BakerCampbellHaus- (4π) v dorff formula. But we do not consider these corrections (D9) here.

[1] T. Aoyama, T. Kinoshita, and M. Nio, Revised and im- model in the symmetric phase, Nucl. Physics. B290 proved value of the QED tenth-order electron anomalous (1987) 25; Scaling laws and triviality bounds in the lattice magnetic moment, Phys. Rev. D 97, 036001 (2018). φ4 theory: (II). One-component model in the phase with [2] J. Schwinger, On Quantum-Electrodynamics and the spontaneous symmetry breaking, Nucl. Physics. B295 Magnetic Moment of the Electron, Phys. Rev. 73, 416 (1988) 65. (1948). [10] B. W. Lee, C. Quigg, and H. B. Thacker, Strength of [3] L. D. Landau, A. A. Abrikosov, and I. M. Khalatnikov, Weak Interactions at Very High Energies and the Higgs On the removal of infinities in quantum electrodynamics, Boson Mass, Phys. Rev. Lett. 38, 883 (1977). Dokl. Akad. Nauk SSSR 95, 497 (1954). [11] G. V. Dunne and M. Unsal¨ What is QFT? Resurgent [4] See https://en.wikipedia.org/wiki/Landau_pole. trans-series, Lefschetz thimbles, and new exact saddles, [5] M. G¨ockeler, R. Horsley, V. Linke, P. Rakow, G. Schier- PoS LATTICE 2015 (2016) 010. holz, and H. St¨uben, Is There a Landau Pole Problem in [12] L. D. Landau, in W. Pauli, ed. Niels Bohr and the De- QED?, Phys. Rev. Lett. 80, 4119 (1998). velopment of Physics, (London: Pergamon Press 1955). [6] H. Gies and J. Jaeckel, Renormalization Flow of QED, [13] N. N. Bogoliubov and D. V. Shirkov, Introduction to the Phys. Rev. Lett. 93, 110405 (2004). Theory of Quantized Fields, 3rd ed. (Nauka, Moscow, [7] D. Djukanovic, J. Gegelia, and Ulf-G. Meißner, Triviality 1976; Wiley, New York, 1980). of quantum electrodynamics revisited, Commun. Theor. [14] D. J. E. Callaway, Triviality Pursuit: Can Elementary Phys. 69 (2018) 263. Scalar Particles Exist?, Physics Reports 167, 241 (1988). [8] R. Dashen and H. Neuberger, How to Get an Upper [15] D. J. E. Callaway and R. Petronzio, Can elementary Bound on the Higgs Mass, Phys. Rev. Lett. 50, 1897 scalar particles exist?: (II). Scalar electrodynamics, Nu- (1983). clear Physics B277, 50 (1986). [9] M. L¨uscher and P. Weisz, Scaling laws and triviality [16] S. Kim, J. B. Kogut, and M.-P. Lombardo, Gauged bounds in the lattice φ4 theory: (I). One-component 18

NambuJona-Lasinio studies of the triviality of quantum in quantum magnets, Phys. Rev. B 92, 220401(R) (2015). electrodynamics, Phys. Rev. D 65, 054015 (2002). [31] J. Gonz´alez,F. Guineai, and M. A. H. Vozmediano, Non- [17] A. A. Andrianov, D. Espriu, M. A. Kurkov, and F. Lizzi, Fermi liquid behavior of electrons in the half-filled honey- Universal Landau Pole, Phys. Rev. Lett. 111, 011601 comb lattice (A renormalization group approach), Nucl. (2013). Phys. B424, 595 (1994). [18] L. N. Lipatov, Divergence of the perturbation-theory se- [32] M. A. H. Vozmediano, Graphene: The running of the ries and the quasiclassical theory, Zh. Eksp. Teor. Fi. 72, constants, Nat. Phys. 7, 671 (2011). 411 (1977). [33] D. C. Elias, R. V. Gorbachev, A. S. Mayorov, S. V. Mo- [19] J. Zinn-Justin, Perturbation series at large orders in rozov, A. A. Zhukov, P. Blake, L. A. Ponomarenko, I. V. quantum mechanics and field theories: Application to the Grigorieva, K. S. Novoselov, F. Guinea, and A. K. Geim, problem of resummation, Physics Reports 70, 109 (1981). Dirac cones reshaped by interaction effects in suspended [20] I. M. Suslov, Summing divergent perturbative series in graphene, Nat. Phys. 7, 701 (2011). a strong coupling limit. The Gell-Mann-Low function of [34] D. T. Son, Quantum critical point in graphene approached the φ4 theory, Journal of Experimental and Theoretical in the limit of infinitely strong Coulomb interaction, Physics, 93, 1 (2001). Phys. Rev. B 75, 235423 (2007). [21] I. M. Suslov, Renormalization group functions of the φ4 [35] E. Barnes, E. H. Hwang, R. E. Throckmorton, and S. Das theory in the strong coupling limit: Analytical results, Sarma, Effective field theory, three-loop perturbative ex- Journal of Experimental and Theoretical Physics 107, pansion, and their experimental implications in graphene 413 (2008). many-body effects, Phys. Rev. B 89, 235431 (2014). [22] L. P. Kadanoff, Scaling laws for Ising models near Tc, [36] E. Borel, Leccecdon sur les series divergentes, (Gauthier Physics 2, 263 (Long Island City, N.Y. 1966). Villars, Paris 1928). [23] K. G. Wilson, The renormalization group: critical phe- [37] J. Hofmann, E. Barnes, and S. Das Sarma, Why Does nomena and the Kondo problem, Rev. Mod. Phys. 47, Graphene Behave as a Weakly Interacting System?, Phys. 773 (1975). Rev. Lett. 113, 105502 (2014). [24] K. G. Wilson, Renormalization group and critical phe- [38] G. Parisi, Asymptotic Estimates in Perturbation Theory nomena. I. Renormalization group and the Kadanoff scal- with Fermions, Phys. Lett. 66B, 382 (1977). ing picture, Phys. Rev. B 4, 3174 (1971). [39] C. Itzykson, G. Parisi, and J. B. Zuber, Asymptotic esti- [25] A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. mates in quantum electrodynamics, Phys. Rev. D 16, 996 Novoselov, and A. K. Geim, The electronic properties of (1977). graphene, Rev. Mod. Phys. 81, 109 (2009). [40] R. Balian, C. Itzykson, G. Parisi, and J. B. Zuber, [26] R. Bistritzer, and A. H. MacDonald, Moir´ebands in Asymptotic estimates in quantum electrodynamics. II, twisted double-layer graphene, Proceedings of the Na- Phys. Rev. D 17, 1041 (1978). tional Academy of Sciences 108, 12233 (2011). [41] M. Matsumoto, B. Normand, T. M. Rice, and M. Sigrist, [27] Y. Cao, V. Fatemi, S. Fang, K. Watanabe, T. Taniguchi, Magnon Dispersion in the Field-Induced Magnetically E. Kaxiras, and P. Jarillo-Herrero, Unconventional super- Ordered Phase of TlCuCl3, Phys. Rev. Lett. 89, 077203 conductivity in magic-angle graphene superlattices, Na- (2002). ture 556, 43 (2018). [42] Ch. R¨uegg, B. Normand, M. Matsumoto, A. Furrer, [28] Y. Cao, V. Fatemi, A. Demir, S. Fang, S. L. Tomarken, D. F. McMorrow, K. W. Kr¨amer,H. -U. G¨udel,S. N. J. Y. Luo, J. D. Sanchez-Yamagishi, K. Watanabe, T. Gvasaliya, H. Mutka, and M. Boehm, Quantum Magnets Taniguchi, E. Kaxiras, R. C. Ashoori, and P. Jarillo- under Pressure: Controlling Elementary Excitations in Herrero, Correlated insulator behaviour at half-filling TlCuCl3, Phys. Rev. Lett. 100, 205701 (2008). in magic-angle graphene superlattices, Nature 556, 80 [43] Y. Kulik and O. P. Sushkov, Width of the longitudinal (2018). magnon in the vicinity of the O(3) quantum critical point, [29] S. Das Sarma, and F. Wu, Electron-phonon and electron- Phys. Rev. B 84, 134418 (2011). electron interaction effects in twisted bilayer graphene, [44] H. Polshyn, M. Yankowitz, S. Chen, Y. Zhang, K. arXiv:1910.08556. Watanabe, T. Taniguchi, C. R. Dean, and A. F. Young, [30] H. D. Scammell and O. P. Sushkov, Asymptotic freedom Large linear-in-temperature resistivity in twisted bilayer graphene, Nature Physics 15, 1011 (2019).