Renormalization footprints in the phase diagram of the Grosse-Wulkenhaar model

Dragan Prekrat∗ University of Belgrade, Faculty of Physics, P.O. Box 44, SR-11001 Belgrade, Serbia (Dated: May 18, 2021) We construct and analyze a phase diagram of a self-interacting matrix field coupled to curvature of the non-commutative truncated Heisenberg space. In the infinite size limit, the model reduces to the renormalizable Grosse-Wulkenhaar’s. The curvature term is crucial to renormalization. When turned off, the triple point collapses into the origin as matrices grow larger. When turned on, the triple point shifts away proportionally to the coupling strength and matrix size. Coupling attenu- ation that renormalizes the Grosse-Wulkenhaar model cannot subdue the shifting, and the trans- lational symmetry-breaking stripe phase escapes to infinity, taking away the problematic UV/IR mixing.

I. INTRODUCTION managed to evade the UV/IR mixing. It features a self- interacting real scalar field on the Moyal space equipped We are often challenged to explain what we do in just with a ?-product a few words. What springs to mind in connection to non- i/2 ∂θ~ ∂~ µ ν µν commutative QFT is: CERN in Wonderland. A realm f ? g = f e g ⇒ [x , x ]? = iθ . (2) where space and time behave oddly and the rules are a bit off. As if that were not enough, even the chit-chat1 Its potential is enhanced by the external harmonic oscil- at the Hatter’s tea party seems to be inspired by the lator term of a possible gravitational origin. Namely, the non-commutative quaternions [2]. model can be reinterpreted [15] as that of a scalar field Some seven decades later, non-commutativity (NC) ar- in a curved NC space rived from Wonderland to Quantumland, settling first in Z  2  √ 1 2 ξ 2 M 2 Λ 4 the canonical relation between position and momentum SR = g (∂φ) − Rφ + φ + φ . (3) and shortly after into NC space-time. It was first invoked 2 2 2 4! in the hope of resolving the confusion about the infinities in the nascent quantum field theory [3], but the trick of The oscillator term, which holds a key to renormalizabil- renormalization beat it to the punch. Since then, it oc- ity, is now seen as a coupling to the curvature of the htr casionally reemerged, both in fundamental and effective underlining truncated Heisenberg algebra form, from condensed matter physics to quantum gravity [µx, µy] = i(1 − µz), (4a) [4,5]. Finally, when NC was discovered in the low energy sector of the string theory at the turn of the millennium [x, z] = +i{y, z}, [y, z] = −i{x, z}, (4b) [6], various new NC models followed. The parallel with Wonderland continues in an uncanny projected onto z = 0 section. Another possible source manner. Just as the bites from opposing sides of the of the oscillator term was presented in [16], where it ele- magic mushroom threaten to shrink or stretch Alice in a gantly appears in the expansion of the kinetic term of the runaway fashion, divergencies of non-planar Feynman di- free scalar field that is situated in the Snyder–de Sitter agrams spill from high energy to low energy sector, spoil- space. This model also predicts the running of the cur- ing the renormalizability of NC QFTs [7–10]. And just vature coupling, which is an essential ingredient of the like the problem with the mushroom, UV/IR mixing can GW-renormalizability. It would be interesting to see if be resolved only by proper balancing of large and small similar conclusions could be reached in fuzzy de Sitter arXiv:2104.00657v2 [hep-th] 16 May 2021 scales, in the guise of the Langman-Szabo duality [11]. space as well [17, 18]. Grosse-Wulkenhaar (GW) model [12–14] UV/IR mixing still poses a problem for gauge fields Z 2 2 on NC spaces [19]. Hoping to build on the GW model’s 1 2 Ω −1 2 m 2 λ 4 SGW = (∂φ) + ((θ x)φ) + φ + φ (1) success, [20, 21] tried to adapt it to a gauge field on 2 2 2 4! htr. Still, after extensive treatment, we found non- renormalizability lurking in the form of divergent non- local derivative counterterms [22]. Apart from the trivial vacuum, this model contains another, which breaks the ∗ [email protected] 1 “Then you should say what you mean,” the March Hare went translational invariance: on. µ2y µ2x µ “I do,” Alice hastily replied; “at least—at least I mean what I A1 = − ,A2 = + , φ = . (5) say—that’s the same thing, you know.” g g g “Not the same thing a bit!” said the Hatter. “You might just as well say that ‘I see what I eat’ is the same thing as ‘I eat what This echoes the translational symmetry-breaking stripe I see’!” [1] phase that seems to be at the root of UV/IR mixing. 2

“Stripes” refer to patterns of spatially non-uniform mag- hence their matrix counterparts are netization, which appear when the field oscillates around 1 different values at different points in space [23–25]. They P1 = −Y,P2 = X,P3 = . (8) also seem to shatter the symmetry between large and 2 small scales that keeps the UV/IR mixing at check: lo- These identifications associate (3) with a matrix model cally, vacuum appears ordered, but globally, watched through the lenses of spatial-averaging, it looks smudged 2 2 4
SN = Tr ΦKΦ − crRΦ − c2Φ + c4Φ , (9) into a disordered zero. It would be interesting to find out what happens with the stripe phase in the GW model. in which the field Φ is a N × N hermitian matrix, K We would like to see how its renormalizability plays out the kinetic operator and R the diagonal curvature of htr from the phase transition point of view. space Phase diagrams on NC spaces have been extensively ( studied in various matrix models, since they regular- 31 16i, i < N, ize corresponding continuum theories in a numerical KΦ = [Pα, [Pα, Φ]],Rii = − (10) 2 8N, i = N. simulation-friendly fashion [26–46]. They generically fea- ture three phases that meet at a triple point. Two are readily present in commutative theories: in the disor- All originally dimensionful quantities are expressed in µ dered phase, field eigenvalues clump around zero, and in units of NC mass scale . We chose the minus sign of c the ordered phase around one of the mirror image-minima the mass term to enable positive 2 to parameterize the c of the potential. The third is a matrix counterpart of the relevant portion of the phase diagram, while positive 4 S NC stripe phase: eigenvalues there gather both around ensures that is bounded from below. They will be ac- positive and negative minimum at the same time. companied by the rescaled model parameters In [44], we started a numerical comparison of the two- c2 c4 c2 = , c4 = . (11) dimensional GW-model matrix regularization with (R- e N e N on) and without (R-off) the curvature term, focusing mainly on the latter. Here, we wish to present more We performed parallel hybrid Monte Carlo simulations details on the former and to see how it bares under the to measure various thermodynamic observables, most im- oscillator term switching off procedure that ensures the portant being φ4-model’s renormalizability [12]. Since the triple point ? C = Var SN 2 controls the extension of the problematic stripe phase, • heat capacity , we will try to pinpoint its location.  • magnetic susceptibility χ = Var |Tr Φ| N, The paper is organized as follows. We first reintro- duce the model and present its detailed N = 24 phase • distributions of eigenvalues and traces of Φ, diagrams. Then we track the R-off triple point as we in- crease the matrix size. Finally, we present the effects of where expectation value hOi and variance Var O of the the curvature coupling variation on the phase diagram, observable O are given by look at the coupled model as we turn the coupling off, R −S dΦ O e 2 2 and compare its limit with the uncoupled one. hOi = , Var O = O − hOi . (12) R dΦ e−S

II. MATRIX MODEL Phase transitions in finite systems form smeared finite peaks and edges in profiles of free energy derivatives. We here continue inspection of the matrix regulariza- Different quantities yield slightly different estimates of tion SN of (3) started in [44]. Let us reintroduce the transition points, but they ultimately converge for large model and walk through its main features. enough matrices. To locate them, we scanned through If the NC strength in htr is set to  = 1, coordinates parameter space by varying c2 at fixed c4, which played µx and µy can be represented by finitely-truncated ma- a role of temperature, and searched for peaks in C and trices of the Heisenberg algebra in the energy basis of the χ. harmonic oscillator Classical equation of motion for SN √ √ 1 ! - 1 ! 2
1 √ √ i √ √ 2KΦ − cr{R, Φ} + Φ −2c2 + 4c4Φ = 0, (13) X = √ 1 √ 2 ,Y = √ 1 √ - 2 . (6) 2 . 2 . 2 . . 2 . . N×N N×N gives us an idea what kind of phases to expect. Its kinetic, Derivatives in model (3), analysed in the frame formal- curvature and pure potential parts are respectively solved by ism, are realised as commutators ∂µ = [pµ, · ] with mo- menta pµ  0, c2 ≤ 0, 2  p1 p2 p3 1 Φ ∝ 1, Φ = 0, Φ = c2 1 (14) = +µy, = −µx, = µz − , (7)  , c2 > 0, iµ iµ iµ 2 2c4 3

The ↑↓-phase is a matrix equivalent of the stripe phase. Large mass parameter lives in ↑↑-phase, and large quartic coupling in l-phase, with ↑↓-phase nested in between. When the kinetic term is negligible (e.g. field near ∝ 1) and c2 ≥ maxj{cr|Rjj|}, a diagonal solution exists that combines the effects of the curvature and the potential and which deforms the vacuum of the ordered phases:

c2 1 +crR Φ2 = . (15) 2c4 Both transitions out of the stripe phase in the R-off case (FIG.1) follow the square root behaviour for larger quartic coupling. For ec4 > 1, they are well approximated by: p l→↑↓: ec2 = 2.67(6) ec4 − 0.55(9), (16a) p ↑↓→↑↑: ec2 = 3.99(5) ec4 − 0.90(7). (16b)

For comparison, a pure potential√ model would show a l→↑↓ transition line ec2 = 2 ec4 in the infinite N limit. In [44] we presented a simple argument that, due to di- agonality, curvature acts as a quasi-mass term and should shift transition lines proportionally to the cr by δec2 rel- ative to the R-off case:

1 δc2 63 ≤ e ≤ 16 − . (17) 2N cr 2N We previously demonstrated this by numerical simulation at a token value of quartic coupling and with the absent kinetic term; here, we expose this effect in full in FIG.1. Similar shifting is in the meantime also reported on the fuzzy sphere after adding a modification to the kinetic term [46]. The detailed analysis of the curvature’s effects on the phase diagram is ongoing and will be presented elsewhere, while here we concentrate only on the aspects relevant to the position of the triple point. We wish to simultaneously inspect two finite limits of FIG. 1. Contour plots of N = 24 phase diagram for c = 0 our matrix model, which zoom-in on different portions of r the phase diagram: (top) and for cr = 0.2 (bottom). Darker colors depict lower and lighter colors higher values of specific heat, bright stripes hS (c , c , c )i being the transition lines. The dotted line on the bottom plot N 2 4 r S = lim 2 , (18a) indicates a diagram shift relative to the R-off case. Diagrams N→∞ N are constructed based on more than 4000 points. hSN (c2, c4, cr)i Se = lim e e , (18b) N→∞ N 2

2 corresponding to three phases depicted in FIG.1: S0 and Se0 referring to cr = 0. S is closely related to SGW up to a light adjustment of coefficients (AppendixA) and • disordered l-phase: hΦil = 0, is focused on the neighbourhood of the origin, while Se0 is the limit we addressed in [44] which is focused on the • uniformly ordered ↑↑-phase: hΦi↑↑ ∝ 1, 3rd order l→↑↓ transition line. † • non-uniformly ordered ↑↓-phase: hΦi↑↓ ∝ U 1± U , UU † = U †U = 1 12 = 1 |Tr 1 | < N where , ± and ± . III. R-OFF TRIPLE POINT

In [44], we found that the triple point of Se0 lies at 2 Throughout this text, we will use Wolfram Mathematica bluish ec4(T ) . 0.005 (c4 . 0.14 at N = 28) and established the StarryNightColors scheme for R-off plots and reddish SunsetCol- descending trend of ec4(T ) with increase in matrix size. ors scheme for R-on plots. In the meantime, we collected more data for matrix sizes 4

c2 It is important to notice that even if the triple point of S0 does not lie precisely at the origin, the triple point 2.0 of Se0 will, due to eci = ci/N. This is in contrast with the φ4-model on the fuzzy sphere [41, 43]. 1.5

1.0 IV. R-ON SHIFT AND RENORMALIZATION

0.5 Coupling with curvature pushes the cusp of the stripe 1 phase away from the origin proportionally to its strength 0.01 0.02 0.03 0.04 0.05 N (FIG.3). The measured shift of the triple point in the N = 24, R-on case c4 0.8 ec2(T |r) = 0.18(8) + 15.5(7)cr (21) R 0.6 relative to the -off value

c2(T |/r) = 0.14(5), (22) 0.4 e agrees well with the maximal prediction allowed by (17): 0.2  63  1 max δec2 = 16 − cr ≈ 14.7cr. (23) 0.02 0.04 0.06 0.08 0.10 N 2N The slight overshoot is discussed in AppendixB. FIG. FIG. 2. Shrinking of the R-off l→↑↑ transition line with 2 shows that intercept of ec2(T |r) goes to zero with the an increase in matrix size. (top) Linear fit of the end- increase in matrix size, so it is safe to assume that pro- point’s c2 as a function of the inverse matrix size 1/N: portionality to cr becomes exact in the infinite size limit. c2(T |/r) = +0.07(5) + 37(2)/N. (bottom) Quadratic fit of the endpoint’s c4 as a function of the inverse matrix size 1/N: 2 c4(T |/r) = −0.001(13)+4.9(8)/N +20(11)/N . Data gathered ˜ c2 from susceptibility χ for N ≤ 60.

3 up to N = 60, allowing us to track the shrinking rate of the l→↑↑ transition line. Unexpectedly, this transition 2 disappears entirely and the triple point collapses into the origin (FIG.2). AppendixB provides details on locating the triple 1 point from raw data and different attempted data fits

(TableI). We modeled small aberrations from the linear cr trend set by larger matrices by quadratic and power-law 0.05 0.10 0.15 0.20 functions of 1/N. All the estimates agree with triple ˜ point lying at the origin, and the best one bounds its c4 coordinates to 0.25 (c2, c4)T ≤ (0.15, 0.005) (19) 0.20 with 95% probability each, which is an order of magni- 0.15 tude improvement in precision. In addition, linear extrapolation of the slopes of tran- 0.10 sition lines for N = 24, 32, 40, 50 shows that they radiate 0.05 from the triple point as c 0.05 0.10 0.15 0.20 r l→↑↓: c2 = 7.1(8)c4, (20a) ↑↓→↑↑: c2 = 17(1)c4. (20b) FIG. 3. Shift of the cusp of the stripe phase, with rising This is also how the phase diagram of Se0 looks like curvature coupling: ec2(T |r) = 0.18(8) + 15.5(7)cr, ec4(T |r) = zoomed-in around the origin,√ while away from it, its tran- 0.014(6) + 1.19(8)cr. Data gathered from specific heat C for N = 24. sition lines bend into ∝ ec4. 5

In the GW approach [12], renormalizability of the lim- spreading towards larger values of the mass and quartic 4 iting φ?-model is assured by taking parameters. We compared the model with and without s the curvature term since its inclusion is crucial to the 1 − Ω2 1 model’s renormalizability. 2 = 1 − 2 , (24) 1 + Ω (1 + log(Λ0/ΛR)) We first refined the estimate [44] of the triple point position in the case of no coupling with the curvature. √ We conclude that it collapses to the origin in the infinite which for large cutoff Λ0 ∝ N switches Ω off as matrix size limit, completely erasing the l→↑↑ transition 1 line. Ω ∼ . (25) log N When the curvature coupling activates, it shifts the triple point and the transition lines towards larger val- Since we in AppendixA identify Ω and cr as ues of the mass parameter, proportionally to its strength Ω2/8 and the matrix size. GW coupling attenuation proce- cr = 2 , (26) dure does not reverse this effect. Instead, the triple point 1 − Ω /2 seems to be expelled to infinity, taking the stripe phase away with it. we consider the limit where cr decreases as This leads to different phase diagrams for renormalized 4 1 and non-renormalized φ?-model. The former seems to be cr ∼ . (27) log2 N stripe phase-free, while in the latter, the stripe phase is tethered to the origin of the parameter space. Combining this with δec2 ∝ cr, or rather δc2 ∝ crN, We are currently completing the exploration of the R- would effectively swipe the stripe phase off to infinity as on model, and we plan to further inspect the R-off triple N > 60 N point for to make sure it lies at the origin. Al- 2 , (28) though we observed the convincing curvature-mediated log N linear shift for N = 24, it would be prudent to confirm leaving the renormalized model with a completely dif- the effect for larger matrices as well. We would also like ferent phase diagram from the one obtained by simply to simultaneously decrease the coupling and increase the matrix size since their combined effect was indirectly de- setting cr = 0. Looking back at the equation of motion (13) and its so- duced. lutions (14), we see that curvature term prefers the trivial Additionally, we mean to revisit the GW-inspired over striped vacuum. The action (9) also shows that the gauge model [22] in the hope of uncovering the reverse ef- curvature itself compensates the attenuation of the cou- fect: non-renormalizability due to retention of the stripe pling. For nearly ordered field configurations Φ2 ∝ 1, the phase. After fixing the NC strength and scale, the model curvature term dominates the potential by factor is left with only one adjustable parameter — the field strength coupling g. Its additional stripe-like vacuum Tr RΦ2 Tr R  31  (5) transforms into a trivial one for the infinite coupling, ≈ = −8N 1 − ∼ N, (29) Tr Φ2n Tr 1 16N implying the phase diagram’s possible structure: stripe phase for small g, disordered phase for large g. which multiplied by the coupling leads once more to Another possible line of investigation could be the phase diagram of renormalizable spinor model on htr [47] N 2 . (30) in context of fermionic matrix models [48, 49]. log N If this correspondence between renormalizability and phase structure proves to hold across models, it might This somewhat unusual situation is perhaps best un- be helpful when constructing new ones. It could be derstood with the help of The Little Prince. If we, by possible to search numerically for early signatures of chance, stumble in the dark upon a snake from the nar- (non)renormalizability in their phase diagrams, assessing rator’s drawing, judging by its thin tail, we would con- the new model’s renormalizability potential, even before clude that it is pretty hungry. But once the light is on, the involved and time-consuming analytical exploration. we would see that it has, in fact, swallowed an elephant.

V. CONCLUDING REMARKS ACKNOWLEDGMENTS

This paper aimed to see if the GW-model’s renormal- This work was supported by the Serbian Ministry izability is reflected in its corresponding phase diagram of Education, Science and Technological Development and if it affects the extent of the stripe phase connected Grant ON171031 and by COST Action MP1405. The to the UV/IR mixing. With that in mind, we tracked author would like to thank Prof Maja Burić, Prof Denjoe the positions of the triple point of the matrix regulariza- O’Connor, and Dr. Samuel Kováčik for valuable discus- tion of the model: this is where the stripe phase starts, sions and DIAS for hospitality and financial support. 6

˜ Appendix A: Model correspondence c2 ↑↑ According to [39], mapping 0.05 last ↑↓ Z √ 0.04 φ ←→ Φ, ←→ det 2πθ Tr (A1) 0.03 0.02 connects field theory on Moyal space and matrix field theory with the same parameters. Also, [15] provides a 0.01 first S S ↕ ˜ correspondence between GW and R: c 0.001 0.002 0.003 0.004 0.005 4  Ω2  SGW = 1 − SR, (A2a) ˜ 2 c2  2    2 Ω 2 15 2 0.4 ↑↑ ↑↓ m = 1 − M − ξµ , (A2b) 2 2 0.3  Ω2  λ = 1 − Λ, (A2c) 2 0.2  Ω2  Ω2 = 82 1 − ξ. (A2d) 2 0.1 ↕ ˜ c From these, by comparing (3) and (9), it is easy to con- 0.01 0.02 0.03 0.04 0.05 4 clude that SGW and SN are connected by

 Ω2  FIG. 4. (top) Change in slope of the edge of ↑↑-phase indicates SGW = π 1 − SN , (A3a) 2 a triple point. Arrows mark the last point on l→↑↑ and the first point on l→↑↓ lines. Constructed from χ-data for  2    2 Ω 15 N = 40. (bottom) Extrapolated transition lines with 83% m = − 1 − c2 + cr , (A3b) 2 2 confidence intervals. Constructed from C-data for N = 40.  Ω2  λ = 12 1 − c4, (A3c) 2 C  2  1.5 stationary point 2 Ω Ω = 8 1 − cr, (A3d) 2 1.0 in the large N limit (θ12 = 1/µ2, units: µ = 1). Further- more, action multiplier can be absorbed into the field during expectation value integration and will affect only 0.5 c4: ˜ 0.0 c2 hκS(c2, c4, cr)i = hS(c2, c4/κ, cr)i , (A4a) 0 1 2 3 4 5 √ κS S κ hΦi (c , c , c ) = hΦi (c , c /κ, c ), κS 2 4 r S 2 4 r (A4b) C ˜ yielding 1.0 c 4 = 0.25

CκS(c2, c4, cr) = CS(c2, c4/κ, cr), (A5a) 0.8 κχκS(c2, c4, cr) = χS(c2, c4/κ, cr). (A5b) 0.6 Since we are interested in the position of peaks of C and χ, this means that phase transition diagrams for κS and ˜ ˜ 0.4 c 4 = 0.30 c 4 = 0.20 S will be the same up to a reparametrization ˜ c 1.5 2.0 2.5 3.0 3.5 2 (c2, c4, cr) ←→ (c2, c4/κ, cr). (A6)

For phase diagrams of SGW and S in the Ω → 0 limit FIG. 5. (top) Triple point region for N = 24, cr = 0.2 at (cr → 0), this means: ec4 = 0.25 resolved into two peaks. The stationary point at the plateau is chosen as the triple point proxy. (bottom) 2 (m , λ) ←→ (−c2, 12c4/π). (A7) Plateau at ec4 = 0.25 mounts above plateaus at ec4 = 0.20 and ec4 = 0.30, building a wall between phases. 7

TABLE I. Different models of R-off triple point position fitting. All intercepts are consistent with the triple point located at the origin. A linear fit is performed for data subsets with higher N, where nonlinearities are imperceptible.

data model c4-fit c2-fit

linear c4 = −0.012(11) + 5.9(4)/N c2 = +0.07(5) + 37(2)/N

2 2 χ quadratic c4 = −0.001(13) + 4.9(8)/N + 20(11)/N c2 = +0.04(8) + 39(5)/N − 42(60)/N

1.25(13) 0.93(14) power law c4 = +0.016(17) + 12(4)/N c2 = −0.0(2) + 31(10)/N

C linear c4 = −0.12(10) + 16(3)/N c2 = +0.9(9) + 93(21)/N

Appendix B: Triple point proxies and fits In the R-on case, we used contour diagrams (e.g. FIG. 1) to detect the cusp of the ↑↓-phase from C-data. We looked at the bright triple point peak position and then We here discuss triple point proxies used for FIG.2 checked the neighboring raw data to pinpoint its exact and FIG.3. location. As it turns out, the peak resolves into two very In FIG.4, the split of the profile of χ into two separate closely spaced convoluted peaks – which presumably co- peaks changes the slope of the ↑↑-phase border ∂ ↑↑. The incide when matrix size increases – joined by a wall that midpoint between the last point in one-peak and the first separates phases (FIG.5). For FIG.3, we used the posi- point in the two-peak regime served as the triple point tion of the protruding peak, which gave a slightly higher proxy for FIG.2. We used the standard deviation of the estimate for the slope than (23). The stationary point triangular distribution ending at these two points as the on this wall seems a more realistic estimator of the triple triple point position uncertainty. For N ≤ 24, the two point position, but it is also more difficult to measure. A peaks are not completely separated in the triple point rough estimate using the stationary point region, so we took the intersection of extrapolated tran- sition lines instead. ec2(T |r) = 13.2(11)cr + 0.24(9), (B2) For consistency, we also checked the C-data, which has less predictive power due to larger uncertainties and fits within the interval (17) and is close to its upper distance from the triple point region. We extrapolated bound. For comparison, the slope of the line connect- l→↑↓ transition line to its intersection with ∂ ↑↑. To get ing the smaller peaks is around 12. 68% confidence intervals of the intersection point coor- Looking at FIG.1, we see a small elliptical local dinates, we used 83% confidence intervals of transition minimum-region with a bright triple point peak at its line fits, since the probability of triple point belonging to lower left edge. Eigenvalue distribution there has the their intersection is given by characteristics of the ↑↓-phase. We do not believe this to constitute a separate phase but a finite-size effect that collapses into a triple point as the matrix size increases. P (l→↑↓ ∩ ∂ ) = P (l→↑↓)P (∂ ) ↑↑ ↑↑ (B1) This should be, of course, checked at larger N. The different extrapolations of the triple point position and 0.68 ≈ 0.832. in the infinite matrix limit are collected in TABLEI.

[1] L. Carroll, Alice’s Adventures in Wonderland (The [9] B. P. Dolan, D. O’Connor, and P. Presnajder, JHEP 03, Project Gutenberg EBook, 2008). 013 (2002), hep-th/0109084. [2] M. Bayley, New Scientist 204, 38 (2009). [10] C. P. Martin, J. Trampetić, and J. You, UV/IR mix- [3] H. S. Snyder, Phys. Rev. 71, 38 (1947). ing in Noncommutative SU(N) Yang-Mills theory, 2020, [4] J. Bellissard, A. van Elst, and H. Schulz- Baldes, 2012.09119. Journal of Mathematical Physics 35, 5373 (1994), [11] E. Langmann and R. J. Szabo, Physics Letters B 533, https://doi.org/10.1063/1.530758. 168–177 (2002). [5] K. Fujii, (2005), quant-ph/0502174. [12] H. Grosse and R. Wulkenhaar, JHEP 12, 019 (2003), [6] N. Seiberg and E. Witten, JHEP 09, 032 (1999), hep- hep-th/0307017. th/9908142. [13] M. Disertori, R. Gurau, J. Magnen, and V. Rivasseau, [7] S. Minwalla, M. Van Raamsdonk, and N. Seiberg, JHEP Phys. Lett. B 649, 95 (2007), hep-th/0612251. 02, 020 (2000), hep-th/9912072. [14] Z. Wang, Annales Henri Poincare 19, 2435 (2018), [8] C.-S. Chu, J. Madore, and H. Steinacker, JHEP 08, 038 1805.06365. (2001), hep-th/0106205. [15] M. Buric and M. Wohlgenannt, JHEP 03, 053 (2010), 8

0902.3408. [32] A. P. Polychronakos, Phys. Rev. D 88, 065010 (2013), [16] S. A. Franchino-Viñas and S. Mignemi, Eur. Phys. J. C 1306.6645. 80, 382 (2020), 1911.08921. [33] J. Tekel, JHEP 10, 144 (2014), 1407.4061. [17] M. Buric, D. Latas, and L. Nenadovic, Eur. Phys. J. C [34] B. Ydri, JHEP 03, 065 (2014), 1401.1529. 78, 953 (2018), 1709.05158. [35] S. Rea and C. Sämann, JHEP 11, 115 (2015), [18] M. Buric and D. Latas, Phys. Rev. D 100, 024053 (2019), 1507.05978. 1903.08378. [36] J. Tekel, Acta Phys. Slov. 65, 369 (2015), 1512.00689. [19] D. N. Blaschke, Fortsch. Phys. 62, 820 (2014), 1402.5980. [37] J. Tekel, JHEP 12, 176 (2015), 1510.07496. [20] M. Buric, H. Grosse, and J. Madore, JHEP 07, 010 [38] B. Ydri, K. Ramda, and A. Rouag, Phys. Rev. D 93, (2010), 1003.2284. 065056 (2016), 1509.03726. [21] M. Buric, M. Dimitrijevic, V. Radovanovic, and [39] B. YdriLectures on Matrix Field Theory Vol. 929 M. Wohlgenannt, Phys. Rev. D 86, 105024 (2012), (Springer, 2017), 1603.00924. 1203.3016. [40] P. Sabella-Garnier, JHEP 08, 121 (2017), 1705.01969. [22] M. Burić, L. Nenadović, and D. Prekrat, Eur. Phys. J. [41] J. Tekel, Phys. Rev. D 97, 125018 (2018), 1711.02008. C 76, 672 (2016), 1610.01429. [42] K. Hatakeyama, A. Tsuchiya, and K. Yamashiro, PTEP [23] S. S. Gubser and S. L. Sondhi, Nucl. Phys. B 605, 395 2018, 063B05 (2018), 1805.03975. (2001), hep-th/0006119. [43] S. Kováčik and D. O’Connor, JHEP 10, 010 (2018), [24] P. Castorina and D. Zappala, Phys. Rev. D 77, 027703 1805.08111. (2008), 0711.2659. [44] D. Prekrat, K. N. Todorović-Vasović, and D. Ranković, [25] H. Mejía-Díaz, W. Bietenholz, and M. Panero, JHEP 10, JHEP 03, 197 (2021), 2002.05704. 056 (2014), 1403.3318. [45] M. Šubjaková and J. Tekel, PoS CORFU2019, 189 [26] D. J. Gross and E. Witten, Phys. Rev. D 21, 446 (1980). (2020), 2006.12605. [27] X. Martin, JHEP 04, 077 (2004), hep-th/0402230. [46] M. Šubjaková and J. Tekel, JHEP 06, 088 (2020), [28] M. Panero, JHEP 05, 082 (2007), hep-th/0608202. 2002.02317. [29] D. O’Connor and C. Saemann, JHEP 08, 066 (2007), [47] M. Burić, J. Madore, and L. Nenadovic, Class. Quant. 0706.2493. Grav. 32, 185018 (2015), 1502.00761. [30] F. Garcia Flores, X. Martin, and D. O’Connor, Int. J. [48] G. W. Semenoff and R. J. Szabo, Int. J. Mod. Phys. A Mod. Phys. A 24, 3917 (2009), 0903.1986. 12, 2135 (1997), hep-th/9605140. [31] F. Lizzi and B. Spisso, Int. J. Mod. Phys. A 27, 1250137 [49] I. Aref’eva and I. Volovich, JHEP 10, 114 (2019), (2012), 1207.4998. 1902.09970.