Arxiv:2004.02896V2 [Gr-Qc] 25 Nov 2020 Clcntn Rbe) Nteohrhn,Tevleof Value the Hand, Today Other Cosmolog- Parameter the the Hubble on on the Overview Problem)

arXiv:2004.02896v2 [gr-qc] 25 Nov 2020 clcntn rbe) nteohrhn,tevleof value the hand, today other cosmolog- parameter the the Hubble on On the overview problem). an for constant reasons 7] ical chapter several [4, for e.g., rel- physics, (see, general particle in ordinary explanation and convincing small ativity a and compatible find origin is not its do Λ, acceleration value constant cosmic cosmological picture. plain if the a even in with fit hand, quite one energy not do On dark that of enigmas nature unsolved particu- the are In and [3]. acceleration speculation. today theoretical late-time however, for lar, until theory, room astrophysical Einstein 2] still by is the [1, there reaped of times success and early the description universe Despite from accurate the within, and of objects valid history so-called a the the gives of in incarnated model, relativity, ΛCDM general that fact rcs esrmnso h aetm xaso,frin- for expansion, late-time more to for the wait contribute of we measurements only while precise would models phenomenological again Per- up once pile energy. trick dark this to forming explanation alternative can acceler- motivation, an generate appropriate provide the easily relativity with can general which, field phases twist. from ating scalar new deviations canonical a that a with known with but well al- literature, is models the It dark-energy in exotic proposed of ready amount enormous the underlying the model. to ΛCDM imputed the be de- of might standard assumption that three fact than a (Cepheids- more viations, local by in [5] found observations one based) the with incompatible is [2] i irwv akrudi the in background microwave mic ∗ † [email protected] [email protected] odt,alcsooia bevtosareo the on agree observations cosmological all date, To u i nti oki oadaohrcnedrto contender another add to is work this in aim Our HSCLRVE D REVIEW PHYSICAL ieso .Atragigta hsfidn ih o eanum a be effec not the might as finding regarded this be geometry. const could that acceleration is arguing present-day that spacetime After of 4. to dimension small-scale dimension just Hausdorff from way, the natural Surprisingly, nu a in big-bang theory, and the extend structure-formation a with late-time d a compatible sustain as being to act able is cannot alone geometry geometry derivatives, therefore, and, relativity general egtddrvtvs nteter with theory the In derivatives. weighted ihtesae npriua,i h ocle multifracti so-called the in particular, in scale, the with esuytepsiiiyt bancsooia aetm a late-time cosmological obtain to possibility the study We .INTRODUCTION I. 2 etrfrGaiainlPyis uaaIsiuefrThe for Institute Yukawa Physics, Gravitational for Center 102 H akeeg nmlircinlspacetimes multifractional in energy Dark 0 059(2020) 103529 , optdfo h cos- the from computed Planck inuaCalcagni Gianluca 1 nttt eEtutr el aei,CSIC, Materia, la de Estructura de Instituto yt nvriy 0-52 yt,Japan Kyoto, 606-8502, University, Kyoto ern 2,Mdi,206Spain 28006 Madrid, 121, Serrano eayrelease Legacy [ arXiv:2004.02896 Dtd pi ,2020) 6, April (Dated: q 1, drvtvs h uioiydsac stesm sin as same the is distance luminosity the -derivatives, ∗ n noi eFelice De Antonio and rmnwo,i tl ne osrcin[,1] Unfor- 14]. quantum shown [7, perturbative was construction gravity of under renormalizability still received the is ignore tunately, have will on, we three which now first fourth, from the The while [7], attention action. much dynamical scenar- the multifractional weighted,in of ordinary, types with four ios: only theo- are independent there as model- Taken ries, a via operator. dynamics kinetic Multi- the universal dependent in a and admits 13], which 9–12]. [7, kinemati- measure, parametrization [7, spacetime feature the scales in this cally implement short spacetimes at renormalizability fractional gravity the quantum to phenomenon related of a flow scale, dimensional probed called the quantum with from originated changes spacetimes geometry effective the of sion between geometry. investigating relation spacetime worth and unsuspected acceleration be an late-time to illuminates out it turn cosmo- because ensuing will se- The will scenario possibilities. that logical several here, among more macroscopic considered one the therefore, lect models for the and, happens on rigid also constraints This scales theoreti- falsifiable. all their easily at make geometry predictions four-dimensional usually cal a scales recover post-microscopic to at satisfy to appeal- more have be may than approaches ing top-down acceleration. of late-time class about devia- say This to and as have level they fundamental presented what the check at originally relativity multifrac- general 8], named from we tions [7, theories, here gravitational spacetimes energy, of tional dark class of a model take phenomenological a ing from stance r ihwihe eiaie)cnb eyinteresting very a the- be (the scenarios can these derivatives) of weighted one with that this ory out of results point the micro- will However, fundamental paper Nature. a of provide description not scopic indicated do which theories result these a that 15], [7, derivatives weighted and nalkontere fqatmgaiy h dimen- the gravity, quantum of theories known all In ] clrto hs ihu n uig while tuning, fine without phase cceleration nltere with theories onal r nry nteter ihweighted with theory the In energy. ark fanwrsoainlwfrspacetime for law restoration new a of t lolresaemdfiain fgravity. of modifications large-scale also loytei ons hssget to suggests This bounds. cleosynthesis andt ecoet h topological the to close be to rained clrto rmagoer changing geometry a from cceleration dhoc ad Euclid rclcicdne econclude we coincidence, erical oesbcueterqieetthey requirement the because models not rtclPhysics, oretical 2, † 6.Hwvr nta fpropos- of instead However, [6]. q oipoei h hoiswith theories the in improve to drvtvsadwith and -derivatives q rfatoa derivatives fractional or - YITP-20-44 q - 2 infrared (IR), rather than (or on top of) ultraviolet (UV), energy phase, contrary to the more interesting case with modification of gravity. It is important to stress that this weighted derivatives. We will place constraints on the extension of multifractional theories is admitted by a gen- parameter space of this theory and find its compatibil- eral theorem governing the form of the spacetime mea- ity with observations only when the Hausdorff dimension sure [13] and, therefore, it is not introduced by hand. In of spacetime is kept at the constant value dH = 4, i.e., fact, this holds for any theory of quantum gravity admit- the number of topological dimensions. We interpret this ting dimensional flow in the Hausdorff dimension and a result as the manifestation of something we might call continuum approximation [7, 13]. The main implication a restoration law, not apparent in previous works. It of this result is that, even if the theories we will study is as if late-time acceleration was the expression of the here are not fundamental in the sense specified above, the tendency of the Hausdorff spacetime dimension to stay feature responsible for late-time acceleration is expected close, or at least get back, to 4. The value of 4 is not to be present in other theories of quantum gravity with conserved exactly but it is restored asymptotically. We a scale-dependent Hausdorff dimension.1 Therefore, our give an argument why this might not be just a numer- findings may be interesting not only because they com- ical coincidence, although it is too weak to establish a plete the study of multifractional spacetimes and enrich full-fledged physical principle yet. their phenomenology to unsuspected directions, but also In Secs. II A–II D, we briefly introduce some well- because they could be kept in mind when exploring the known cosmological quantities and review multifractional cosmology of quantum gravities with dimensional flow. theories, quoting their main features on a homogeneous The very idea to link dark energy to quantum gravity background. The reader interested in an in-depth discus- goes against the grain of effective field theory, since it sion can consult [7] and references therein. In Sec. IIE, is usually believed that quantum gravity can only have we extend the spacetime measure of the theories to in- Planck-size corrections, which cannot possibly have an clude beyond-general-relativistic correction terms, which impact on the late-time universe. However, dimensional will play a crucial role in the interpretation of our nu- flow is essentially a nonperturbative phenomenon where merical results. The cosmological equations of the the- other elements apart form the UV/IR divide enter the ory with q-derivatives are reviewed in Sec. IIIA, while in game, such as the discrete scale invariance responsible for Sec. IIIB we show that geometry alone cannot acceler- the cosmic log-oscillations we will describe in due course. ate the universe in this theory. This result is exact. The The starting point for the study of the cosmol- FLRW equations of the theory with weighted derivatives ogy of these theories are the Friedmann–Lemaˆıtre– are presented in Sec. IV A, while in Sec. IVB we en- Robertson–Walker (FLRW) equations obtained from the code all the effects of multiscale geometry into an effec- full background-independent equations of motion [16, 17]. tive dark-energy component obeying standard Friedmann While there are some studies about inflation [18, 19] and equations. The numerical analysis and its discussion are gravitational-wave propagation [20, 21] that constrain or presented in Secs. IVC–IVE, while Sec. V is devoted to propose possible effects in the cosmic microwave back- conclusions. ground and in gravitational-wave signals in multifractional spacetimes, the problem of dark energy has received less attention, except in the case with ordinary II. GENERAL SETUP derivatives. The absence of cosmological vacuum solutions in the theory with ordinary derivatives [16] led some A. Cosmology to explore scenarios with more or less exotic dark-energy components [22–29] (see also [30, 31] for a thermodynam- We work in four topological dimensions (D = 4) and ical approach). However, our point of view is that a sensi- with signature ( , +, +, +). The Hubble parameter is ble theory beyond Einstein should be able to describe cos- − mic acceleration without introducing dark-energy fields a˙ H := , (1) by hand. In the case of multifractional spacetimes, this a acceleration should come only from geometry [17]. The theory with ordinary derivatives is less attractive than where a dot denotes a derivative with respect to cosmic the others due to some technical issues [7]; for this rea- time t. The evolution of the universe can be parametrized son, we will turn to the theories with q-derivatives and by the redshift 1 + z = a0/a, where a0 = a(0) = 1 is the with weighted derivatives, which admit a more rigorous scale factor today. We use a subscript 0 for any quantity structure. evaluated today: the Hubble constant H0, the age of the In particular, we will show that geometry alone in the universe t0, and so on. theory with q-derivatives is unable to sustain a dark- The universe is assumed to be filled by perfect fluids with energy-momentum tensor

Tµν = (ρ + P )uµuν + gµν P (2) 1 An example is the class of theories based on pre-geometric structures with discrete labels, such as group ﬁeld theory and loop and equation of state P = wρ. In all models, the content quantum gravity. of the universe will be nonrelativistic matter (dust, ρ = 3

ρm, w = 0) plus a radiation component (w =1/3). Both change, but the relation between conformal time and the curvature and the cosmological constant are set to zero, dynamics is modified and (7) receives several corrections, k = 0 = Λ, the first from observations [1, 2] and the which we will discuss in due course. second because we want to get acceleration purely from geometry. The latter will give an effective contribution we will dub as ρde, where de stands for dark energy. For each matter component, we will use the dimension- B. Metric and geometric-harmonic structures less energy-density parameter

2 ρm ρde 3H Let D = 4. Multifractional spacetimes are endowed de Ωm := , Ω := , ρcrit := 2 , (3) ρcrit ρcrit κ with two mutually independent structures, a metric one and a geometric-harmonic one. The metric structure is where κ2 = 8πG is Newton’s constant. The late-time given by the metric gµν , in this work the ﬂat FLRW standard Friedmann equation in general relativity in the metric with components absence of curvature is

2 3 3(1+wde) g = 1 , g = a (t) , i =1, 2, 3 . (8) H H0 Ωm,0 (1 + z) +Ωde,0 (1 + z) , (4) 00 − ii ≃ q for a dark-energy component with constant barotropic The geometric-harmonic structure is the integro- index wde. This equation is modified in multifractional differential structure of the theory and is determined by spacetimes, as we will see below. the choice of spacetime measure To compare the theory with supernovæ data, we will use the luminosity distance dL, defined by imposing that the flux of light reaching an observer is the power L d4q(x)= d4x v(x) (9) emitted byF a distant source per unit of area, measured on a sphere of radius dL: in the action and by the type of derivatives in kinetic L terms. In standard general relativity, v = 1 and the ac- =: . (5) tion measure is d4x (times the g volume weight, which F 4πd2 | | L is determined by the metric structure).p The action mea- This is a definition and it is completely model inde- sure weight v(x) is the easiest to understand. The Haus- pendent. The flux is measured from Earth, while dorff dimension of spacetime is defined as the scaling of L is determined by theF astrophysical properties of the the Euclidean-signature volume (ℓ)= ℓ d4x v(x) V ball,cube source. Standard candles such as supernovæ are sources of a 4-ball or 4-cube with linear size ℓ, R for which L is known. From this, one finds dL. In turn, the luminosity distance can be expressed in terms of the spacetime d ln (ℓ) dynamics and, in particular, of the dark-energy equation dH := hV i , (10) of state. It is this last part that depends on the specific d ln ℓ cosmological model. In standard general relativity, the luminosity distance where means average over the harmonic structure (see h·i 4 a photon traveled from some source at redshift z to Earth below). In ordinary spacetime, the volume scales as ℓ , so spacetime (z = 0) is a0r, as measured by the observed at t0. Since that dH = 4. However, when the Hausdorff dimen- 2 2 dτ = dr on the light cone, dL is proportional to confor- sion of spacetime varies with the probed scale and this mal time τ0 τ(z). Taking into account the redshift of variation is slow and smooth at large scales or late times, power L = (energy)− /(time) a/(1/a) = a2 of photons its form is given by a unique parametrization [7, 13], that reaching the observer at different∝ times, one gets [32] we present here only in the time direction because we will be interested in homogeneous solutions on a homo- a0 dL = r = (1+ z)[τ0 τ(z)] (6) geneous metric background. The full homogeneous mea- a − sure weight t0 dt 1 da =(1+ z) = (1+ z) Z a Z Ha2 t(z) a(z) v(t) =q ˙(t) (11) z dz =(1+ z) . (7) H Z0 is, in the most general case, an infinite superposition of αl+iωl Observations of standard candles such as type I super- complex powers of time, q(t) = l γl t/tl , where R C | | novæ allow one to determine the luminosity distance dL, αl,ωl and γl are dimensionlessP constants and the local Hubble parameter H and, from (4), a con- t R∈are fundamental∈ time scales of the geometry. How- 0 l ∈ straint on the dark-energy barotropic index wde. ever, reality of the measure constrains the parameters γl In multifractional theories, the redshift law for time and ωl of this sum to combine into a real-valued expres- intervals and frequencies is the same and (6) does not sion, a generalized polynomial deformed by logarithmic 4 oscillations: in their presence. The positive parameter α∗ is related

+∞ ∗ to the Hausdorff dimension of spacetime in the UV, t α ,l−1 v(t)= Fl(t) , (12a) 3 tl Xl=0 spacetime UV dH α∗ + αi , (15) +∞ ≃ Xi=1 Fl(t)= A0,l + Fñ,l(t) , (12b) nX=1 where αi are the fractional exponents in the spatial direc- D 0 D−1 i t t tions. The full measure is d x v(x ) i=1 vi(x ), where Fñ,l(t)= An,l cos nωl ln + Bn,l sin nωl ln , i tl tl each profile vi(x ) along the spatialQ directions has the i same parametric form as Eq. (14) with t replaced by x (12c) and different parameters α∗ α , t∗ ℓ , and so on. → i → i where l runs over the number of fundamental scales of For a fully isotropic measure, α∗ = αi. In virtually all the geometry and 0 < An,l,Bn,l < 1 are constant amplitudes. literature on the subject, the range of values is chosen as 0 < α∗ < 1 because these theories are thought of as UV modifications of ordinary field theories or of general rela- C. UV binomial approximation tivity where, as in most quantum gravities, the dimension of spacetime decreases in the UV. In this case, (15) is the The most common approximation of (12) used in phys- Hausdorff dimension of spacetime at early times t t∗. ≪ ical applications includes only two terms, (i) one corre- In the presence of log oscillations, the time scale inside sponding to the general-relativistic limit v = 1 (obviously the logarithms in Eq. (14c) was originally denoted as t∞ necessary to get viable phenomenology) and (ii) a correc- and regarded as independent from t∗. Usually, it was set −44 tion term. to the Planck time tPl 5.3912 10 s if 0 < α∗ < 1, but later theoretical arguments≈ × led to the identification (i) Since, as we will see in Sec. II D, the theory is t∞ t∗ [13]. Therefore, in this paper we will not consider defined in such a way that the dimensionality of any≡ such extra scale. The frequency spacetime coordinates is the one of our standard µ clocks and rulers ([x ]= 1 in energy dimensions), 2πα∗ − ω = , N =2, 3, 4,..., (16) then the l = 0 term corresponding to the general- ln N relativistic limit has α∗,0 = 1. In other words, general relativity is the asymptotic IR limit of the the- takes one among a countable set of values, it appears as ory. a consequence of a fundamental discrete scale invariance and parametrizes the typical time length of the discrete (ii) In a scale hierarchy of spacetime made of only one scale symmetry scale t∗ (l takes only two values l = 0, 1, where t0 − 2π is not a physical scale because it is undetermined, t λ t := e ω t . (17) → ω since α∗,0 = 1), the correction term is the UV limit of the measure. If other fundamental scales are also Thus, at scales of order of the log oscillations the time di- present (l takes many values), then t∗ t1 corre- rection takes values on a deterministic fractal constructed sponds to the shortest scale around which≡ general with N similarity maps with similarity ratios all equal to relativity breaks down and UV anomalous scaling λω [33]. effects become visible. The constant A0 in (14b) is the zero mode of the modulation factor and can acquire two values A0 =0, 1 that Taking only these terms and denoting α∗ α∗ , one gets ≡ ,1 can be understood as follows. The mesoscopic-to-large- the binomial (two-term generalized polynomial) expres- scale form of the measure can be obtained with a coarse- sion graining procedure consisting in taking the average over α∗−1 a log oscillation [34]. Restricting to a homogeneous back- t v(t) 1+ (13) ground and denoting as ≃ t∗ 2π 1 ν in the absence of log oscillations, or f(t) := dν f(e ω t) (18) h i 2π Z0 t α∗−1 v(t) 1+ Fω(t) , (14a) the average of a log oscillating function, one has ≃ t ∗ +∞ 2 2 2 2 An + Bn F (t)= A + F˜ (t) , (14b) Fω = A0 , Fω = A0 + . (19) ω 0 n h i h i 2 nX=1 n>X0 t t F˜ (t)= A cos nω ln + B sin nω ln , The case A0 = 0, never explored in the cosmological n n t n t ∗ ∗ phenomenology of these theories, corresponds to a zero (14c) average: starting from mesoscopic temporal scales t ≫ 5 tPl, both the log-oscillatory pattern and the power-law v-dependent contribution. contribution begin to disappear. In the case A0 = 1, the We insist that in these theories the integer and frac- only one considered in the cosmology literature so far tional frame are not physically equivalent and that we [17, 18], the power-law contribution survives longer, up live in the latter, while the integer frame is just a conve- to and beyond scales t & t∗. nient trick to simplify calculations. The multifractional theory with fractional derivatives does not admit a frame mapping, which is the reason why it has been studied D. Physical frame much less than the other two. In general, having a multiscale geometry selects a preferred frame, at least in The form of the kinetic operators in the action is de- multifractional theories. Other multiscale scenarios such termined by the choice of symmetries and the require- as string theory, asymptotically safe quantum gravity, ment that, just like the Hausdorff dimension, in quan- nonlocal quantum gravity and group field theory (which includes spin foams and loop quantum gravity) have a tum gravity also the spectral dimension dS (the scaling of dispersion relations) changes with the probed scale. higher degree of symmetry (Lorentz and diffeomorphism Thus, the system is defined by a nontrivial dimensional invariance are preserved or deformed in a controlled way) and do not require a frame choice, once a low-energy or flow (varying dH and/or dS) and certain action symmetries. In general, the metric is nonminimally coupled semi-classical notion of spacetime is recovered. with matter fields due to the anomalous (i.e., noncon- stant) geometric-harmonic structure, which leads to com- E. Infrared extension of the measure plicated equations of motion where Lorentz invariance is explicitly broken by the time scale t∗ (and the spatial scale ℓ∗, which we do not consider here) appearing in Equations (13) and (14) are the profiles typically con- the measure weight v. However, the problem is greatly sidered in the literature. The constant unit term corre- simplified in two of the admissible cases, the theory with sponds to the measure at the scales on which general rel- q-derivatives and the theory with weighted derivatives. ativity holds, hence its normalization to 1. The physical We briefly recall their main points [7]. picture usually advocated is that multiscale spacetimes In the theory with q-derivatives, the original system is and the dynamics therein are anomalous at short scales recast into a simpler one which is identical to the Stan- t . t∗ and reduce to ordinary spacetimes and general relativity at large scales t t∗. Therefore, the general- dard Model of particle physics where all coordinates are ≫ the geometric profiles qµ(xµ). When gravity is turned on, relativistic regime and its measure weight v = 1 con- the original system becomes equivalent to general relativ- stitute the terminal regime of dimensional flow at large ity in q coordinates, plus matter fields. To perform calcu- scales. lations, one can temporarily forget that the qµ are com- However, this is only one of the possible applications posite coordinates and work in this integer frame until of the dimensional-flow theorem [13] giving rise to the the construction of physical observables, at which point most general profile (12). In principle, ultra-IR terms one must revert to the physical (or fractional) frame dominating the measures at scales larger than those of where the geometry is multiscale. In this theory, the frac- the general-relativistic regime are possible. For example, tional frame is such that clocks and rods do not adapt ignoring log oscillations for the sake of the argument, with the observation scale, while they do in the inte- the measure weight (13) can be replaced by a three-term ger frame. Consequently, in the fractional frame we can profile observe dimensional flow through its imprint on certain t α∗−1 t αc−1 physical observables, while in the artificial integer frame v(t) +1+ , (20) µ µ µ ≃ t t the geometry is constant. The mapping x q (x ) ∗ c connects the two frames and is not a coordinate→ trans- where t is close to Planck time t and 0 < α < 1, while formation in the sense of general relativity. ∗ Pl ∗ t t∗ and In the theory with weighted derivatives, the inte- c ≫ ger frame is defined by field transformations φµν··· → αc > 1 . (21) φ˜µν···(φ, v) involving v. In the absence of gravity, the electroweak-strong model of quantum interactions in the The subscript “c” stands for cosmological. According to fractional frame is mapped into the standard Standard this extended measure, general relativity is only a tran- Model on Minkowski spacetime. In the presence of grav- sient regime between a deep-UV limit, where microscopic ity, the dynamics in the fractional frame is very similar to quantum-gravity effects are important, and an ultra-IR a scalar-tensor theory where v plays the role of the scalar, regime which deviates from standard Einstein theory at but only in a superficial visual analogy at the level of the cosmological scales. In other words, our clocks (and equations of motion: v is not a field, since it is given by rulers, in an inhomogeneous setting) not only changed the nondynamical fixed profile (12) and its spatial gener- in the past, but they will also change in the future. We alization. In the unphysical integer frame, the theory is will see that dark energy can be interpreted as a mani- mapped into general relativity plus a purely geometric, festation of this deviation. 6

Therefore, here we will explore a model where the cor- III. THEORY WITH q-DERIVATIVES rection to general relativity is not a UV term but a “post- IR” one. The UV correction, with 0 < α∗ < 1 and A. Action and FLRW dynamics t∗ t , is totally negligible at late times and can be ig- ∼ Pl nored. The final expression for the measure is therefore In the theory with q-derivatives, the dynamical ac- identical to (14) but with a new set of parameters with tion is the same as in general relativity, except that all prior (21): the coordinates xµ are replaced by the multifractional µ µ µ µ profile q (x ) = dx vµ(x ), where the index µ is t αc−1 v(t) 1+ F (t) , (22a) not contracted. AsR we commented in Sec. II D, this is ≃ t ω c not a coordinate transformation because measurement + ∞ units differ in the fractional and integer frame. Physi- Fω(t)= A0 + Fñ(t) , (22b) cal clocks and rulers, used for actual measurements, are nX=1 defined on the manifold spanned by the coordinates xµ. t t In this fractional frame, the action measure dDq(x) = F˜ (t) A cos nω ln + B sin nω ln , D−1 n ≃ n t n t dqµ(xµ) reflects the basic postulate of the the- c c µ=0 (22c) ory,Q namely, that spacetime geometry changes with the scale. On the other hand, in the integer frame the qµ 2πα ω = c ,N =2, 3, 4,.... (22d) are regarded as noncomposite coordinates and, therefore, ln N the spacetime measure is the standard Lebesgue measure D In this paper, we will consider the following cases in d q, with no dimensional flow. increasing order of difficulty: In the fractional frame, the gravitational action in the absence of a cosmological constant is [17] 1. Binomial measure without log oscillations: 1 D q Sq = d x v g R + Smatter , (25) 2κ2 Z | | A0 =1 , Fñ(t)=0 . (23) p q µν q where the Ricci scalar R := g Rµν , the Ricci tensor q q ρ 2. Binomial measure with log oscillations up to a finite Rµν := R µρν and the Riemann tensor number of harmonics: q ρ 1 q ρ 1 q ρ R µσν := ∂σ Γµν ∂ν Γµσ nmax vσ − vν ˜ q τ q ρ q τ q ρ Fω(t)= A0 + Fn(t) . (24) + Γµν Γστ Γµσ Γντ , nX=1 − q ρ 1 ρσ 1 1 1 Γµν := 2 g ∂µgνσ + ∂ν gµσ ∂σgµν , 2a. One harmonic (nmax = 1) with N = 2 [small- vµ vν − vσ est N in (22d)] and A0, A1 and B1 free pa- are all made with q-derivatives (no summation over µ) rameters. In turn, this and the following can be divided into two subcases, A0 = 0 (no zero ∂ 1 ∂ mode) and A = 1. µ µ = µ µ . (26) 0 ∂q (x ) vµ(x ) ∂x 2b. One harmonic with very large N. We choose The matter action is constructed according to the same the arbitrary value N = e10 to get a suppres- criterion. sion factor 1/10 in ω (small frequency). The Friedmann equations in D = 4 topological dimensions in the presence of a perfect fluid are [17] 3. Binomial measure with many harmonics (nmax > 1). κ2 H2 = v2ρ , (27) 3 3a. Many harmonics with N = 2. Different the- 2 a¨ κ 2 v˙ oretical choices for the n-dependence of An = v (ρ +3P )+ H . (28) a − 6 v and Bn indicate that higher modes are rapidly suppressed and that it is sufficient to con- The continuity equation for a perfect fluid is compatible sider nmax . 10 [19]. We take nmax = 10. with the Friedmann equations: Also, to maximize the effect we take all amplitudes constant and equal to one another: ρ˙ +3H(ρ + P )=0 . (29) A1 = A2 = ... = Anmax and B1 = B2 = ... = Multiscale geometry effects may be implicit in the energy Bnmax . This configuration can be regarded as density and pressure. For instance, for a homogeneous a multiharmonic measure where only the first scalar field O(10) harmonics dominate. 1 1 10 ρ = φ˙2 + W (φ) , P = φ˙2 W (φ) , (30) 3b. Many harmonics with N = e . φ 2v2 φ 2v2 − 7 and its continuity equation is modified both in the fric- exactly as in the cold dark matter (CDM) model of Ein- tion and in the potential term: stein gravity without cosmological constant. The conclu- sion is that this expression cannot possibly fit supernovæ v˙ 2 φ¨ + 3H φ˙ + v W,φ =0 . (31) data and that the theory with q-derivatives cannot make − v the late-time universe accelerate just from pure geometry.

B. No dark energy from geometry IV. THEORY WITH WEIGHTED DERIVATIVES In the multifractional theory with q-derivatives, the geometry of spacetime is such that it eases the slow-roll condition in inflation [17, 18] and, similarly, it could en- For the cosmologist, the theory with weighted deriva- hance the (usually insufficient) acceleration triggered by tives may result more intuitive than the previous one a quintessence field, i.e., it could relax the fine tuning because here the line element is the usual of general rel- in the initial conditions. However, we do not wish to ativity and the only change in the expression for the lu- straighten a general-relativistic model with some extra minosity distance (7) is in the profile H(z). exotic ingredient. A genuine alternative explanation of dark energy should work without relying on a quintessen- tial component. Therefore, we would like to get late-time A. Action and FLRW dynamics acceleration from pure geometry. Unfortunately, this is not possible in the theory with Weighted derivatives are ordinary derivatives with q-derivatives because geometry effects cancel out in the measure-weight factors inserted to the left and to the luminosity distance. The latter receives two corrections, right: one from the definition of conformal time and one from the dynamics. Distances, areas and volumes all must be 1 2 := ∂(vβ ) , β = . (36) calculated with the nontrivial multiscale measure weight D vβ · D 2 v(x) and, in particular, integration along the time direc- − tion includes a factor v(t). In particular, the luminosity The gravitational action in the fractional (physical) 3 distance dL is still proportional to the comoving distance frame is [17] (times the same redshift factor as in general relativity), 0 1 but now the latter is dq(x). The FLRW line ele- S = dDx v g [ γ(v) v µv −q(r) v 2κ2 µ ment R Z p| | R− D D 2 µ µ ν ν 2 2 2 2(D 1) U(v)] + Smatter , (37) ds = gµν dq (x ) dq (x )= dq (t)+ a dq (x) − − 2 2 2 2 −2 = v (t) dt + a v (x) dx , (32) where γ and U are functions of the weight v, Smatter − is the minimally coupled matter action, := gµν , x 1 2 3 µν where v( ) := v1(x )v2(x )v3(x ) is the spatial measure := ρ and R R factorized in the coordinates, vanishes for light rays, so Rµν R µρν that the comoving distance is given by the integration ρ := ∂ βΓρ ∂ βΓρ + βΓτ βΓρ βΓτ βΓρ , of the weighted conformal time dt v(t)/a.2 Thus, the R µσν σ µν − ν µσ µν στ − µσ ντ β ρ 1 ρσ luminosity distance is Γ := g ( µgνσ + ν gµσ σgµν ) . µν 2 D D − D t0 z dt v(t) dz v To avoid confusion with ordinary curvature tensors R..., dL = (1+ z) =(1+ z) . (33) Zt(z) a Z0 H we used the curly symbol ... to denote tensors written in terms of the derivativeR (36). The metric is covari- On the other hand, the Friedmann equation with only antly conserved with respect to the weighted covariant dust and no dark-energy component reads − β τ β τ derivative σ gµν := ∂σgµν Γσµgτν Γσν gµτ = 0, implying that∇ these spacetimes− are Weyl− integrable and 3 H = vH0 Ωm,0 (1 + z) , (34) that the metric is not conserved in the ordinary sense, q g = (β∂ ln v) g . so that ∇σ µν σ µν As we said in Sec. II D, in the integer frame the theory 1+ z z dz resembles a scalar-tensor model in the Einstein frame. dL = , (35) 3 H0 Z0 Ωm,0 (1 + z) Denoting with a bar coordinates and metric quantities p evaluated in this frame, the Friedmann equations in D =

2 Note that this time redeﬁnition does not make the line element conformally ﬂat, unless v(x) = 1. However, we keep the mis- nomer “conformal” for the sake of an easier comparison with 3 Here U has been rescaled by a factor of 2(D − 1) with respect to general relativity. [7, 17] in order to make the Friedmann equations simpler. 8

4 dimensions are [17] hence

2 2 1 1 v˙ κ Ω (∂¯v) U(v) ¯ H¯ 2 = ρ¯ + t + , (38) H = H + , 3 2 v2 v √v 2 v 2 2 ∂2a¯ κ2 U(v) 1 d a¯ 1 a¨ 1 v¨ v˙ v˙ t¯ = (¯ρ +3P¯)+ , (39) = + + H , (44) a¯ − 6 v a¯ dt¯2 v a 2 v v − v2 where derivatives are with respect to time t¯ and Ω = where dots are derivatives with respect to t. To these 3/2+9v2γ(v)/4 is a function of the measure weight. expressions, one must add those for the energy density −One can get the second Friedmann equation (39) either and pressure of the perfect fluid: directly from the equations of motion or by using the ρ P continuity equation ρ¯ = , P¯ = , (45) v2 v2 ∂¯ρ¯ +3H¯ (¯ρ + P¯)=0 . (40) t where the factor 1/v2 stems from the confor- While, in general, U is required to be nonzero for consis- mal rescaling (42), (energy)/(volume) = [(energy) 3/2 tency of cosmological solutions, the function γ was orig- /√v]/[v (volume)]. inally introduced to make an analogy with scalar-tensor Therefore, the Friedmann equations and the master models, with the difference that v is not a dynamical equation in the physical frame read field. However, this term is neither dictated by symme- v˙ κ2 1 1 v˙2 tries nor by consistency of the solutions and it can be 2 H + H = ρ + Ω 2 + U, (46) dropped without loss of generality or creating any theo- v 3v 2 − 2 v retical or conceptual problem. Since our goal is to test a¨ 1 v¨ v˙ v˙2 κ2 + + H = (ρ +3P )+ U, (47) the most minimalistic model with weighted derivatives, a 2 v v − v2 −6v i.e., with log-oscillating measure and no kinetic-like term, κ2 1 3 v˙ 2 v¨ v˙ we can set γ to vanish, so that Ω = 3/2 (this was done ˙ H = (ρ + P )+ Ω 2 + H .(48) only in the numerical code; the equations− written here −2v 2 2 − v − v v are valid in general). It is in these equations, not in (38)–(41), where v(t) is Two major points of departure with respect to scalar- given by (11) and (22). Notice that the contributions tensor theories are, first, that v is not a scalar field but a of the measure weight mimic both a running effective a priori time profile fixed [by (12) in the fractional frame; background-evaluated Newton’s constant see below]. And, second, that the “potential” term U(v) is neither chosen ad hoc nor reconstructed from observa- G G = < G (49) tions, but it is determined from the dynamics itself. In eff v fact, combining (38) and (39) to eliminate the last term in the right-hand side, one gets the master equation and a phantom “scalar” with negative kinetic energy. Combined together, these features can sustain cosmic ac- 2 2 2 ∂ a¯ κ Ω (∂¯v) celeration, without the theoretical inconvenience of phan- t¯ H¯ 2 + (¯ρ + P¯)= t . (41) a¯ − 2 − 2 v2 tom fields (v is not a dynamical degree of freedom and is not associated with classical instabilities or negative- Thus, given the matter content one obtainsa ¯(t¯) and, norm quantum states). from (38) or (39), U(v). Finally, the continuity equation (40) becomes These equations are in the integer (Einstein) frame, 1 v˙ which is unphysical in this theory. The dynamical equa- 0 =ρ ˙ +3H(ρ + P ) (ρ 3P ) . (50) tions in the physical (fractional, Jordan) frame, which − 2 v − we write here for the first time, are obtained by recalling that the Jordan and Einstein metric are related to each Radiation (w =1/3), which is the only conformal invari- other by ant perfect fluid, is conserved also in this frame, while any other fluid experiences an effective dissipation.

g¯µν = vgµν , a¯ = √v a , (42) where v = v(t) is the temporal part of the measure B. Effective dark-energy component weight. Also, having chosen the FLRW metric in the integer frame implies that the same metric holds in the The running of the Hausdorff dimension dH in this the- fractional frame provided dt¯2 = vdt2, i.e., ory can drive one or more phases of acceleration. It is − − convenient to recast the contribution of the multiscale d 1 d geometry and dynamics as a dark-energy component in = , (43) dt¯ √v dt standard general relativity. Thus, the first Friedmann 9 equation (46) is written as Since v is a given function of time, it is also a given func- 2 tion of y. Therefore, one could consider y as the indepen- 2 κ H = (ρm + ρde) , (51a) dent variable of integration. However, we find it useful to 3 have an autonomous system of differential equations. In 3 1 1 v˙2 v˙ 1 this case, we consider to be the independent variable. ρde := Ω + U H 1 ρ , κ2 2 − 2 v2 − v − − v m On doing so, the systemN of equations can be written as

(51b) 2 2 2 2 2 bhvv,y 6 b h vΩm 2 v Ω 2 vv,yy +3 v h′ = − − ,y − ,y , or, equivalently, as − 4b2v2h (58) Ωm =1 Ωde , (52) − Ω Ω′ = m 6 b2h2v2 6 b2h2vΩ + bhvv where the dimensionless energy densities were defined m 2b2h2v2 − m ,y in (3). This equation replaces (4). Since dust matter 2 2 2 v,yΩ 2 vv,yy +3 v,y , (59) does not obey the standard continuity equation, Ωm = − − 3 6 1 Ωm,0(1 + z) . The effective dark-energy pressure Pde is y′ = , (60) obtained by reformulating the master equation (48) as −bh its Einstein-gravity counterpart: where a prime denotes the derivative with respect to . 2 We also remind the reader that here we will considerN ˙ κ H = (ρm + ρde + Pde) , Ω = 3/2. In the most general case (12), several di- − 2 − 1 1 3 v˙ 2 v˙ v¨ mensionless parameters appear, while in the trinomial de αc−1 P := 2 +Ω 2 2H +3U , (53) case (20) v(y)=1+ bcy + , where the ellipsis is −κ 2 2 v − v − v the UV correction, negligible at··· cosmic scales. This is which allows us to find the dark-energy barotropic index the profile v(y) considered here, y being a function of . Among the three dynamical equations (58)–(60), the firstN Pde wde := , (54) corresponds to the second (modified) Einstein equation, ρde whereas the second one is determined by using the con- a complicated function of the Hubble parameter, the tinuity equation. Finally, the third equation is a direct measure weight v and its time derivatives. For Ω = 3/2, consequence of the definition of the e-folds number. It is the first term in the pressure (53) vanishes. Note− that also simple to show that (54) differs from the effective barotropic index coming h′ q= 1+ , (61) from the contribution of all (matter and dark energy) − h components, defined as w := 1+2ǫ/3, where eff − 1 2q wde = − , (62) H˙ 3 (Ωm 1) ǫ := 2 − −H We can integrate the equations of motion from = 0 up 3 v¨ Hv˙ 1 3 v˙ N = (1 Ωde)+ − Ω (55) to = 6, giving the following initial conditions: 2v − 2H2v − 2 2 − H2v2 N y(0) = 1 , h(0) = 1 , Ω (0) = Ω . (63) is the first slow-roll parameter. Another quantity of in- m m0 terest is the deceleration parameter q (not to be confused In any case, we find that the system goes to matter with the composite coordinates q) domination at = 6, i.e., Ω 1 and q 1/2. Further N → → a¨ equations are for z and for the luminosity distance: q := 2 = 1+ ǫ , (56) −aH − (1 + z)2 d˜′ = d˜+ , d˜:= H d , z′ =1+ z . (64) related to w by w = ( 1+2q)/3. h 0 L eff eff − Let us give some time scales for reference. t0 9 17 ≈ C. Solving the dynamics 14 10 yr 10 s is the age of the universe today and × ≈ H0 = H(t0). The onset of big-bang nucleosynthesis is at t 200s, corresponding to y 4.5 10−17 and 21, The dynamics can be solved numerically from a mini- ≈ ≈ × N ≈4 mal set of differential equations. We choose the number while matter-radiation equality happens at t 7 10 yr, y 5 10−6, 8. Our numerical integration≈ × starts of e-foldings as the main time variable, together with ≈ × N ≈ other dimensionless varying or constant parameters: from today and ends at = 6, at the peak of matter domination. This choice isN due to having focused our at- a t H := ln 0 , y := , h := , tention only to late-time data and, in particular, to data N a t0 H0 which do not need a description in terms of perturbation tc tc H0 dynamics. In particular, we will use the constraints from b := t0H0 , yc := = . (57) t0 b type Ia supernovæ (Union 2.1 set) [35], today’s value of 10

H (given in [5]) and, finally, the constraint on the age of 0.1 Ω < 0.5. The marginalized likelihood 2σ 0 ≤ m0 the universe t0 coming from studying the globular cluster contours in the parameter space (2σ preferred val- NGC 6752 [36]. This choice of data sets is given by the ues of the parameters) are shown in Table I. We fact that we only focus on the late-time evolution of the will discuss the physical implications of the results models we are considering. High-redshift information, of this and the other cases in Sec. IVE. For the both from the background (necessary, e.g., for baryon time being, note that the value of h is in tension acoustic oscillations) and perturbation sides (necessary, with the local observations of H0. e.g., for Planck data), require an investigation of the multifractional theory which goes beyond the scope of • Case 2a: one harmonic, N = 2. The above this study. In particular, we want to see if the model in- results do not change in the presence of only one troduced here could be compatible with data which only harmonic mode. As priors, we used the same as require knowledge of the dynamics at low redshift. As we above plus 0 < A1,B1 < 1. The results for A0 =0 will show, although the above data sets are limited, they and A0 = 1 are in Table I. For some data, there is will be enough to constrain considerably the parameter a large degeneracy, especially in the parameters of space of our model. The reason is that the acceleration the v function, namely yc, αc, A1 and B1. On the of the universe is realized by means of the measure v(y), other hand, the value of h is strongly constrained but the latter does not represent itself a dark-energy com- and it is still in tension with the local observations ponent, i.e., a cosmological constant. Since the function of H0. This situation seems not to depend on the v(y) is given a priori by the dimensional-flow theorem value of N. When A0 = 1, the upper bound on [13], it is not surprising that some of its approximated Ωm0 slightly increases. forms may not succeed to make the universe accelerate. Regarding the time scales appearing in v(y), the UV • Case 2b: one harmonic, N = e10. Having cho- scale t∗ received strong bounds from several observations and experiments [7, 8], ranging from Standard Model sen the same priors as before, we find the results of forces time scales to the Planck scale: Table I. Again, the value of h is in tension with today’s value of the Hubble parameter and the upper −61 t∗ −47 bound of Ωm0 increases when A0 = 1. 10 y∗ = 10 . (65) ≤ t0 ≤ • Case 3a: many harmonics, N =2. This is the The lower bound in (65) corresponds to the Planck scale most promising case. The likelihood contours for t = t , while the upper bound was obtained under the ∗ Pl A = 0 are shown in Fig. 1. Notice in Table I the assumption that 0 < α < 1 [8]. These constraints apply 0 ∗ higher value of h and the lower bound at 2σ for only to the UV part of the trinomial measure weight (20), α . In order to explore this better fit more in de- in a regime (particle-physics scales) where the ultra-IR c tail, we tried to enlarge the prior for A and B to correction in (20), the most important correction in the 1 1 0 < A ,B < 3. However, this resulted in a large cosmological model we will consider here, is negligible. 1 1 degeneracy for these two parameters. The χ2 for Conversely, the UV correction at near-Planckian scales this case (χ2 = 555) is significantly smaller than t is completely negligible at late times and we will min ∗ the one for all the previous cases (values approxi- ignore∼ it here. mately equal to 565). The behavior of the Hubble The general procedure we followed to obtain the nu- factor is shown in Fig. 2, while a more detailed dy- merical constraints below was to sample the background- namics for H/H at low redshifts is depicted in Fig. evolution parameter space for the chosen data sets. We 0 3. At the same time, it is interesting to show also performed an MCMC sampling on the parameters of the the deceleration parameter q (Fig. 4) and wde (Fig. theory, letting the sampler (EMCEE [37]) finding the 5). Thanks to the oscillations, we are able to get minima of the χ2 distribution together with the 2σ con- a higher value for H , although acceleration might straints for each parameter. In order to double check the 0 be transient, i.e., existing today but stopping some MCMC results, we also performed a numerical minimiza- time in the future. The results for A = 1 are given tion of the χ2 (via different methods such as the Newton 0 in Fig. 6. The dynamics of this case in terms of the one) and verified that results were compatible with the Hubble factor, H/H at low redshifts, q and wde is MCMC sampling. 0 virtually indistinguishable from that shown in Figs. 2–5. D. Results • Case 3b: many harmonics, N = e10. Here one • Case 1: no oscillations. In the absence of log os- can appreciate a difference between the N = 2 and the N = e10 results, the latter getting considerably cillations, A0 = 1 and the model has only five free 2 −1 −1 worse. Once more, the χ for the minimum shifts parameters: αc, yc, b, h := H0/(100kms Mpc ) to values close to 565. and Ωm0. As priors, we used 3 αc 10, 10−5 y 20, 0.1 b 3,− 0.6≤ h ≤ 0.8, ≤ c ≤ ≤ ≤ ≤ ≤ 11

min min Case nmax NA0 αc (αc ) yc h b Ωm0 +0.01 +0.21 +0.200 1 0 — 1 5.1 (0.6) 1.7 0.70−0.01 1.04−0.17 0.278−0.400 +0.01 +0.20 +0.040 0 7.0 (3.4) 3.2 0.70−0.01 1.05−0.17 0.270−0.040 2a 1 2 +0.01 +0.22 +0.200 1 6.7 (0.3) 2.0 0.70−0.01 1.04−0.16 0.275−0.040 +0.01 +0.20 +0.050 10 0 5.7 (0.4) 1.7 0.70−0.01 1.05−0.17 0.270−0.080 2b 1 e +0.01 +0.21 +0.130 1 5.6 (1.0) 1.9 0.70−0.01 1.04−0.17 0.276−0.050 +0.01 +0.21 +0.036 0 3.8 (3.7) 3.9 0.73−0.03 1.05−0.17 0.276−0.040 3a 10 2 +0.02 +0.22 +0.040 1 3.8 (3.6) 4.1 0.73−0.04 1.05−0.17 0.276−0.044 +0.01 +0.22 +0.04 10 0 7.2 (3.7) 4.0 0.70−0.01 1.05−0.17 0.273−0.04 3b 10 e +0.01 +0.21 +0.046 1 7.1 (3.4) 4.0 0.70−0.01 1.05−0.17 0.273−0.038

TABLE I. Preferred values of the parameters h, b and Ωm0 with 2σ (95 % conﬁdence level) errors. For αc we indicate the min min preferred value and the 2σ lower bound αc , while for yc we only show the lower bound yc . The values of A1 and B1 are not shown due to strong degeneracy. The shaded row corresponds to the only case reproducing the observed late-time Hubble parameter.

E. Consequence: Extending multiscale spacetimes Concerning the value of the oscillation amplitudes A1 and B1, as one can see in Figs. 1 and 6 their distribution All of the above cases share some common features. is uniform inside the prior 0 < A1,B1 < 3. Therefore, The value of the parameter b is close to 1, which means there is a strong degeneracy in the parameter space and that the inverse Hubble parameter today is approxi- it is not possible to establish a preferred value with the mately equal to the age of the universe, just like in the late-time data we used. standard ΛCDM model: t0H0 1. Similarly, Ωm0 is Our results indicate that the theory with weighted close to the ΛCDM value. ≈ derivatives can serve not only as a microscopic (UV) The value of h is the most important discriminator modification of gravity, but also as a cosmological (IR) selecting viable models. Only those with many harmonics one. However, unlike other IR modifications of stan- and N = 2 (case 3a) are able to recover the estimated dard cosmology, this theory does not rely on the typical value from late-time observations.4 Concentrating only UV/IR divide through a characteristic scale. When the on case 3a, the lower bound on yc is multifractional geometry is of this type (no log oscillations), it fails to fit data, since the scale tc is larger than tc not yc = > 3.9 , (66) the age of the universe. Thus, we do end up with the t0 typical model where general relativity is an adequate de- which means that the characteristic time tc is always scription of the cosmos during most of the history of the much larger than the age of the universe t0. On one universe and the recent acceleration is triggered around hand, this is compatible with the finding that the value a certain IR scale. of A0 has little impact on the numerical analysis, i.e., that Rather, multifractional effects embodied in logarithmic the zero mode is strongly suppressed and hence it does oscillations are endemic to the cosmological picture of the not affect cosmology. On the other hand, the bound (66) theory and mimic, in a subtle way, a standard cosmol- is not in contradiction with the fact that multifractional ogy with a late-time dominating dark-energy component. effects can explain observations at times O(t0), since the Here the characteristic scale is very large (actually, be- modulation of superposing log oscillations of the highest yond the Hubble horizon) but, nevertheless, it modifies frequency ω takes place at much shorter scales. Note the the evolution of the universe via the harmonic structure specifications of “superposing” and “highest frequency”: of the geometry. in the absence of log oscillations (case 1), in the pres- The last piece of information gathered about the ge- ence of only one harmonic (nmax = 1, cases 2a and 2b), ometry of the cosmos reserves a surprise and is about or when the harmonics frequency is too low (large N, the Hausdorff dimension of spacetime. Just as in the cases 2b and 3b), the theory is unable to recover the ob- case of yc, the distribution of values of αc is flat past the served late-time value of the Hubble parameter. Note peak (preferred value) and one cannot establish an upper the importance of having a theoretical upper limit on ω. bound. However, while the peak of yc is rather mild, the Without it, it would have been more difficult to find a one for αc is very pronounced and allows us to clearly modulation comparable with the observable patch of the establish a preferred value. In case 3a, the latter is universe and compatible with late-time data. α 3.8 , (67) c ≈

with a very small error bar. The preferred value of αc 4 Since we have not used early-universe data, we are not in a po- is larger in the other cases. This means that the theory H sition to say anything about the 0 tension. with weighted derivatives cannot accommodate at the 12

FIG. 1. Marginalized constraints on the parameters Ωm0, αc, b, yc, h, A1 and B1 for the theory with weighted derivatives with multiharmonic modes, N = 2 and A0 = 0. Although there is degeneracy in some of the parameters, this model is able to reach higher values of h.

time same time particle-physics constraints and explain the and dH = 1 at all times. However, when A0 = 1 the late-time acceleration of the universe if we consider only Hausdorﬀ dimension of time is no longer constant. With the binomial measure (13) [for which t∗ = tc and (66) a FLRW metric and A0 = 1, the eﬀective action is one- would be in gross contradiction with (65)], while the case dimensional, with many harmonics of maximal frequency can if we allow for an ultra-IR term as in (20).

The result (67) has a clear geometric interpretation based on the calculation of the Hausdorff dimension in 3 SFLRW = dt v(t) a (t) (68) a perfectly homogeneous FLRW background. On this Z L spacetime, the measure weight is (12) but, by defini- time tion (10), the time Hausdorff dimension dH is computed after averaging over log oscillations [7, 33], i.e., we must take (20) instead of its version with log oscillations. When A0 = 0, the averaged measure is v = 1 and the volume from (20) (omitting the metric density 13

10000 500 400

1000 300 200

100 100 0

q 0

H/H -100 10 -200 -300 1 -400 -500 0.1 1 1.2 1.4 1.6 1.8 2 1 10 100 1+z 1+z

FIG. 4. Behavior of the deceleration parameter q for the FIG. 2. Behavior of H/H0 for the multiharmonic best ﬁt with multiharmonic best ﬁt with N = 2 and A0 = 0. We can see N = 2 and A0 = 0. Oscillations start at low redshift values. the “heartbeat” of dark energy playing a nontrivial role at The plot for A0 = 1 is the same. late times. The plot for A0 = 1 is the same.

10 40

10 0

1 DE 0 H/H w

-10

-20

-30

0.1 -40 1 1.2 1.4 1.6 1.8 2 1 1.01 1.02 1+z 1+z

de FIG. 3. Behavior of H/H0 for the multiharmonic best fit with FIG. 5. Behavior of w for the multiharmonic best fit with N = 2 and A0 = 0 at low redshifts. The plot for A0 = 1 is N = 2 and A0 = 0. The plot for A0 = 1 is the same. Thanks the same. to the last oscillation, the universe is able to increase H to values close to the measured value H0 today. Here a dark- energy epoch is reached due to oscillations which make the term, i.e., curvature effects5) is universe accelerate and decelerate alternatively. Acceleration becomes then a transient but today the universe does undergo α∗ αc t 1 t 1 t a period of accelerated expansion. (t)= dt′ v(t′)= + t + , (69) V Z α∗ t∗ α t c c so that regime t tc posterior to the present epoch, one has α∗ αc time ≫ t/t∗ + t + t/tc d α . dtime = | | | | . (70) H c H α∗ αc ≃ (1/α∗) t/t∗ + t + (1/αc) t/tc Our finding is that α 4 in order to reproduce the | | | | c observed late-time acceleration.≈ Here, dark energy is not Assuming 0 < α < 1 and α > 1, when t t one ∗ c ∗ an extra matter component in the universe: it is the has dtime α . At intermediate scales t t≪. t , the H ∗ ∗ c manifestation of a dimensional flow where the limiting Hausdorff≃ dimension of time coincides with≪ the topolog- value of the time Hausdorff dimension is the topologi- ical dimension, dtime 1. This is the regime character- H cal dimension 4, as if dimensional flow were recovering izing most of the history≃ of the universe. In a late-time the spatial dimensions “lost” in the symmetry reduction from a generic background to a perfectly homogeneous spacetime. Of course, the FLRW dynamics is not really 5 These are always excluded when computing the Hausdorff or blind to the value of the topological dimension: the value spectral dimension. αc = 4 is implicitly induced by the D = 4 factors hidden 14

FIG. 6. Marginalized constraints on the parameters Ωm0, αc, b, yc, h, A1 and B1 for the theory with weighted derivatives with multiharmonic modes, N = 2 and A0 = 1. in the Friedmann and continuity equations. However, the To study the dimension of spacetime (10) we have to source of this numerical coincidence is not obvious. average out the log oscillations according to the definition (18), which is valid only in the presence of one frequency. We hereby propose the following picture. At times of In the case of the trinomial measure, we have two sets of order of the Planck scale, the universe is dominated by frequencies governed by the fractional exponents in (16) quantum-gravity effects of UV type. In general, quan- and (22d). According to the regime considered, one will tum gravity is associated with dimensional flow, which, use one or the other frequency. in turn, is described by a generalized polynomial mea- 1−α∗ sure weight (12) [7, 13]. In the literature, only UV terms At very early times t . t∗, v(t) t∗/t , short- have been considered in this polynomial expansion, but scale quantum-gravity effects dominate≃ | and| the Haus- quite generally also IR terms can be conceived. There- dorff dimension of spacetime is given by (15). This is fore, the time part v(t) of the spacetime measure weight the usual regime where multifractional theories have been 1 2 3 v(x)= v(t) v1(x ) v2(x ) v3(x ) would be (20) instead of studied [7, 8]. Here we are in an early, pre-inflationary (13). Similar expressions hold in the spatial directions regime where our causal patch of the universe is inho- i with t, t∗, α∗, tc and αc replaced by x , a spatial scale mogeneous and anisotropic. These effects quickly die ℓ∗, a set of fractional exponents αi, and so on. away before or during inflation: the residual IR correc- 15

time spacetime Regime v(t) dH dH 1−α∗ 3 Quantum gravity |t∗/t| α∗ α∗ + Pi=1 αi General relativity 1 1 1 + 3 = 4 2 (general background) time General relativity 1 1 1 dH (FLRW) 1 αc−1 Late times (FLRW) |t/tc| αc = 4 αc = 4

0.5 TABLE II. Time and spacetime Hausdorﬀ dimension in dif- 10-4 0.01 1 100 104 106 ferent regimes, after averaging out log oscillations. t

tion in (20) survives but is negligible at times t∗

[1] Y. Akrami et al. [Planck Collaboration], Planck 2018 re- [arXiv:1306.0629]. sults. I. Overview and the cosmological legacy of Planck, [16] G. Calcagni, Quantum field theory, grav- Astron. Astrophys. 641, A1 (2020) [arXiv:1807.06205]. ity and cosmology in a fractal universe, [2] N. Aghanim et al. [Planck Collaboration], J. High Energy Phys. 03 (2010) 120 [arXiv:1001.0571]. Planck 2018 results. VI. Cosmological parameters, [17] G. Calcagni, Multi-scale gravity and cosmol- Astron. Astrophys. 641, A6 (2020) [arXiv:1807.06209]. ogy, J. Cosmol. Astropart. Phys. 12 (2013) 041 [3] B.P. Abbott et al. [LIGO Scientific and Virgo [arXiv:1307.6382]. Collaborations], Tests of general relativity with [18] G. Calcagni, S. Kuroyanagi, and S. Tsujikawa, Cosmic GW150914, Phys. Rev. Lett. 116, 221101 (2016) microwave background and inflation in multi-fractional [arXiv:1602.03841]. spacetimes, J. Cosmol. Astropart. Phys. 08 (2016) 039 [4] G. Calcagni, Classical and Quantum Cosmology [arXiv:1606.08449]. (Springer, Switzerland, 2017). [19] G. Calcagni, Complex dimensions and their observability, [5] A.G. Riess et al., A 2.4% determination of the local value Phys. Rev. D 96, 046001 (2017) [arXiv:1705.01619]. of the Hubble constant, Astrophys. J. 826, 56 (2016) [20] G. Calcagni, Lorentz violations in multifrac- [arXiv:1604.01424]. tal spacetimes, Eur. Phys. J. C 77, 291 (2017) [6] https://www.euclid-ec.org. [arXiv:1603.03046]. [7] G. Calcagni, Multifractional theories: an uncon- [21] G. Calcagni, S. Kuroyanagi, S. Marsat, M. Sakel- ventional review, J. High Energy Phys. 03 (2017) 138 lariadou, N. Tamanini, and G. Tasinato, Quan- [arXiv:1612.05632]. tum gravity and gravitational-wave astron- [8] A. Addazi, G. Calcagni, and A. Marcianò, New Stan- omy, J. Cosmol. Astropart. Phys. 10 (2019) 012 dard Model constraints on the scales and dimen- [arXiv:1907.02489]. sion of spacetime, J. High Energy Phys. 12 (2018) 130 [22] K. Karami, M. Jamil, S. Ghaffari, K. Fahimi, [arXiv:1810.08141]. and R. Myrzakulov, Holographic, new agegraphic [9] G. ’t Hooft, Dimensional reduction in quantum grav- and ghost dark energy models in fractal cosmology, ity, in Salamfestschrift, edited by A. Ali, J. Ellis, and Can. J. Phys. 91, 770 (2013) [arXiv:1201.6233]. S. Randjbar-Daemi (World Scientific, Singapore, 1993) [23] O.A. Lemets and D.A. Yerokhin, Interacting dark energy [arXiv:gr-qc/9310026]. models in fractal cosmology, arXiv:1202.3457. [10] S. Carlip, Spontaneous dimensional reduc- [24] S. Chattopadhyay, A. Pasqua, and S. Roy, tion in short-distance quantum gravity?, A study on some special forms of holo- AIP Conf. Proc. 1196, 72 (2009) [arXiv:0909.3329]. graphic Ricci dark energy in fractal universe, [11] G. Calcagni, Fractal universe and quantum ISRN High Energy Phys. 2013 (2013) 251498. gravity, Phys. Rev. Lett. 104, 251301 (2010) [25] S. Maity and U. Debnath, Co-existence of modified Chap- [arXiv:0912.3142]. lygin gas and other dark energies in the framework of [12] S. Carlip, Dimension and dimen- fractal universe, Int. J. Theor. Phys. 55, 2668 (2016). sional reduction in quantum gravity, [26] A. Jawad, S. Rani, I.G. Salako, and F. Gul- Classical Quantum Gravity 34, 193001 (2017) shan, Pilgrim dark energy models in fractal universe, [arXiv:1705.05417]. Int. J. Mod. Phys. D 26, 1750049 (2017). [13] G. Calcagni, Multiscale spacetimes from first principles, [27] E. Sadri, M. Khurshudyan, and S. Chat- Phys. Rev. D 95, 064057 (2017) [arXiv:1609.02776]. topadhyay, An interacting new holographic [14] G. Calcagni, Towards multifractional calculus, dark energy in the framework of fractal cos- Front. Phys. 6, 58 (2018) [arXiv:1801.00396]. mology, Astrophys. Space Sci. 363, 230 (2018) [15] G. Calcagni and G. Nardelli, Quan- [arXiv:1810.03465]. tum field theory with varying couplings, [28] D. Das, S. Dutta, A. Al Mamon, and S. Chakraborty, Int. J. Mod. Phys. A 29, 1450012 (2014) Does fractal universe describe a complete cos- 17

mic scenario?, Eur. Phys. J. C 78, 849 (2018) J. High Energy Phys. 01 (2012) 065 [arXiv:1107.5041]. [arXiv:1811.09674]. [34] R.R. Nigmatullin and A. Le Méhauté, Is there geometri- [29] U. Debnath and K. Bamba, Parametrizations of dark cal/physical meaning of the fractional integral with com- energy models in the background of general non- plex exponent?, J. Non-Cryst. Solids 351, 2888 (2005). canonical scalar field in D-dimensional fractal universe, [35] N. Suzuki et al. [Supernova Cosmology Project Eur. Phys. J. C 79, 722 (2019) [arXiv:1902.01397]. Collaboration], The Hubble Space Telescope cluster su- [30] A. Sheykhi, Z. Teimoori, and B. Wang, Thermodynam- pernova survey: V. Improving the dark energy con- ics of fractal universe, Phys. Lett. B 718, 1203 (2013) straints above z > 1 and building an early-type- [arXiv:1212.2137]. hosted supernova sample, Astrophys. J. 746, 85 (2012) [31] S. Haldar, J. Dutta, and S. Chakraborty, A compar- [arXiv:1105.3470]. ative study of different entropies in fractal universe, [36] A. Renzini et al., The white dwarf distance to the globu- arXiv:1601.01055. lar cluster NGC 6752 (and its age) with the Hubble Space [32] P. Coles and F. Lucchin, Cosmology, 2nd ed. (Wiley, Telescope, Astrophys. J. Lett. 465, L23 (1996). Chichester, United Kingdom, 2002). [37] D. Foreman-Mackey, D.W. Hogg, D. Lang, [33] G. Calcagni, Geometry and field the- and J. Goodman, emcee: The MCMC ham- ory in multi-fractional spacetime, mer, Publ. Astron. Soc. Pac. 125, 306 (2013) [arXiv:1202.3665].