![Arxiv:2103.14989V2 [Cond-Mat.Soft]](http://data.docslib.org/img/0a3a1a7b0cd4d4a2d1ccd65e6a425580-1.webp)
arXiv:2103.14989v2 [cond-mat.soft] 30 Mar 2021 otoei yia iodrdsses uha liquids as such systems, [ disordered glasses relative structural typical and fluctuations in density those to large-scale of suppression ulda space Euclidean ed ozr [ zero to tends ompitpteni n nwihtesrcuefactor structure the which in one is S pattern point form ainfunction lation orlto ucin[ function correlation where vrcytlieoe,sc suiu rnal optimal, [ nearly defects or against robustness unique and properties as physical such direction-independent ones, advantages crystalline have over can materials matter. hyperuniform of states Disordered amorphous exotic per- and crystals, in quasicrystals, perfect order fect all long-range include to of systems notion many-particle traditional the generalizes tutrlCaatrzto fMn-atceSseso A on Systems Many-Particle of Characterization Structural ( yeuiompitcngrto in configuration point hyperuniform A k ) ≡ h ˜ 1+ ( k steFuirtasomo h oa corre- total the of transform Fourier the is ) ρ h ˜ iy nldn h yeuiomt index hyperuniformity the including mity, opttoal-eeae ape htaencsaiyo necessarily are particula that degree be hyperu samples the should computationally-generated perfect regimes detect scaling from to hyperuniform deviations used identify cause be p can systems are measures hyperuniform in measures diagnostic hyperuniformity and same hyperuniform diagnostic The various disordered, are our interac states that pair ground long-ranged the “stealthy” s where with hard-sphere interacting “sticky” h tems and perfectly rods are hard that d of states o end systems function with equilibrium a as temperature, models or solvable density exactly different three study R tt ihraogtesal rsa rmtsal disorde ratio metastable the sp Using or hard crystal hyperuniform. identical exactly stable of be the packings along equilibrium either of state pressure al We the theoretically. of or experimentally available is mation sraeae”coefficient “surface-area” h coefficient the eie swl stecosvrpitbtente ntrso terms in them between radius point of crossover the as hyperunif well of extent as the regimes ascertain to criteria quantitative ersnaini em ftesrcuefactor structure the of terms in representation fifiieetn ntelimit the in extent infinite of uniaiedsrposta eadwe many-particle a when T herald biology. that and descriptors mathematics quantitative sciences, physical the in topics ( d k 1 safnto fdniy eas xmn o-eprtr ex low-temperature examine also We density. of function a as h td fhprnfr ttso atri neegn mul emerging an is matter of states hyperuniform of study The , ed ozr stewavenumber the as zero to tends ) prahsahprnfr tt safnto fterelevan the of function a as state hyperuniform a approaches h 4 .INTRODUCTION I. 2 ( – R ,i.e., ], r 16 d ) ]. | ≡ k R scaatrzdb nanomalous an by characterized is lim 3 |→ .Tehprnfriyconcept hyperuniformity The ]. h agrteratio the larger The . 1 g , 0 2 2 ( S B .Mr rcsl,ahyperuni- a precisely, More ]. r ( rneo nttt o h cec n ehooyo Mater of Technology and Science the for Institute Princeton ) ntrso h oa orlto function correlation total the of terms in k − 0 = ) rneo nvriy rneo,NwJre 84,USA 08544, Jersey Mathematics, New Computational Princeton, and University, Applied Princeton in Program and 1 and eateto hmsr,Dprmn fPhysics, of Department Chemistry, of Department , B soitdwt h oa ubrvariance number local the with associated g B/A 2 ( r d stepair- the is ) -dimensional B/A ∞ → Dtd pi ,2021) 1, April (Dated: avtr Torquato Salvatore h agrtehprnfr cln eie hc becomes which regime, scaling hyperuniform the larger the , k ocmlmn h nw ietsaerpeetto of representation direct-space known the complement To . B/A | ≡ H (1) States k n h ietcreainfnto eghscale length function direct-correlation the and S swl sohrdansi esrso hyperunifor- of measures diagnostic other as well as , | ( k ,wihi seilyueu hnsatrn infor- scattering when useful especially is which ), o yeuiomsses h exponent the ( systems, hyperuniform For soitdwt h number the with associated atrwt ailpwrlwfr ntevcnt fthe of vicinity the in i.e., form power-law origin, radial a with factor h oa ubrvariance number local the ntelarge- the in iesoa peia bevto idwo radius of window observation spherical dimensional hr nua rcesdnt nesml vrg.A average. ensemble in an pattern point denote brackets angular where h ubrvrac clslk h volume the where like liquids, scales and variance systems, gases number patterns, disordered the point typical Poisson of as those such with contrasted be bevto idw hc sgvnby given is which window, observation > α osdrssesta r hrceie yastructure a by characterized are that systems Consider neuvln ento fhprnfriyi ae on based is hyperuniformity of definition equivalent An r n ohprnfr distance-scaling nonhyperuniform and orm odmntaeta h revlm theory free-volume the that demonstrate so nt size. finite f l sfli nlzn xeietly or experimentally- analyzing in useful rly e rnhdcae htsc n states end such that dictates branch red h efcso hswr steepoainof exploration the in is system work this of focus he )adisvledtrie he ieetlarge- different three determines value its and 0) in stetmeauetnst zero, to tends temperature the as tions ∗ ( ee htapoc tityjammed strictly a approach that heres h eeatcnrlprmtr either parameter, control relevant the f r stvl orltdwt n another. one with correlated ositively prnfr.Seicly eanalyze we Specifically, yperuniform. iomt.Mroe,tecpct to capacity the Moreover, niformity. ,w eieiscrepnigFourier corresponding its derive we ), sesi rirr pc dimension space arbitrary in ystems ntl eeeae edemonstrate We degenerate. finitely oto aaee.W establish We parameter. control t R ie ttso aypril sys- many-particle of states cited h vlm”coefficient “volume” the f owihipretosi nearly in imperfections which to iicpiayfil,ipnigon impinging field, tidisciplinary ii lwrthan slower limit S d R dmninlEcienspace Euclidean -dimensional ( σ k v d 2 1 ) ( ( shprnfr fisvrac grows variance its if hyperuniform is R | ∼ R poc oHyperuniform to pproach o peia window spherical a for ) = ) k ials, | α ( + Γ(1 σ 2 π N for ( d/ R ( R 2 ) R | R d/ k fpit ihna within points of ) h ≡ d d → | 2) hsbhvo sto is behavior This . N . A ξ 0 c ( we , . and R ) 2 α v h − i 1 spositive is ( R N fthe of ) ( R (3) (2) ) R d i 2 - , 2
R scaling behaviors of the number variance [1, 2, 17]: phenomenon, a hyperuniformity length scale ξc and a hy- peruniformity index H. The latter two quantities provide Rd−1, α> 1 (class I) measures of nearness to hyperuniformity. σ2(R) Rd−1 ln R, α = 1 (class II) . (4) In the remainder of the paper, we obtain a variety of ∼ Rd−α, α< 1 (class III) theoretical results to study the problem at hand. First, 2 for a general system, we establish quantitative criteria These scalings of σ (R) define three classes of hype- to ascertain the extent of hyperuniform and nonhyper- runiformity [2], with classes I and III describing the uniform distance-scaling regimes as well as the crossover strongest and weakest forms of hyperuniformity, respec- point between them in terms of the “volume” coeffi- tively. States of matter that belong to class I include all cient A and “surface-area” coefficient B associated with perfect crystals [1, 17], many perfect quasicrystals [17– the variance σ2(R) (Sec. III). Specifically, the ratio 19], and “randomly” perturbed crystal structures [20– B/A determines the crossover length scale Rc. The 23], classical disordered ground states of matter [1, 24, 25] larger the ratio B/A, the larger the hyperuniform scal- as well as systems out of equilibrium [26, 27]. Class II hy- ing regime, which becomes of infinite extent in the limit peruniform systems include some quasicrystals [19], the B/A . This capacity to determine hyperuniform positions of the prime numbers [28], and many disordered scaling→ regimes ∞ is expected to be particularly useful in classical [26, 29–32] and quantum [33–35] states of mat- analyzing experimentally- or computationally-generated ter. Examples of class III hyperuniform systems include samples that are necessarily of finite size. classical disordered ground states [36], random organiza- tion models [37] and perfect glasses [26]. Second, to complement the known direct-space repre- By contrast, for any nonhyperuniform system, it is sentation of the coefficient B in terms of the total corre- shown in Appendix A that the local variance has the lation function h(r)[1], we derive here its corresponding following large-R scaling behaviors: Fourier representation in terms of the structure factor S(k) (Sec. IV). The latter representation is particularly d + useful when the scattering intensity is available experi- 2 R , α =0 (typical nonhyperuniform) σ (R) d−α mentally or if the structure factor is known analytically. ∼ (R , α< 0 (anti-hyperuniform). (5) Third, we show that the free-volume theory of the pres- sure of equilibrium packings of identical hard spheres that For a “typical” nonhyperuniform system, S(0) is approach either a strictly jammed crystal or disordered bounded [2]. In anti-hyperuniform systems, S(0) is un- state dictates that such jammed states be perfectly hype- bounded, i.e., runiform (Sec. V). We describe why this outcome implies that such jammed states must be defect-free. lim S(k)=+ , (6) |k|→0 ∞ Fourth, motivated by a desire to rely on analytical rather than numerical methods, we structurally charac- and hence are diametrically opposite to hyperuniform terize three different disordered-system models as a func- systems. Anti-hyperuniform systems include fractals, tion of the relevant control parameter, either density or systems at thermal critical points (e.g., liquid-vapor and temperature, with end states that are perfectly hyperuni- magnetic critical points) [38–42] as well as certain sub- form. These models are distinguished from most other stitution tilings [43]. models with hyperuniform states in that their pair corre- Our main concern in this paper is the exploration lation functions and structure factors are known exactly of quantitative descriptors that herald when a many- for all values of the control parameter in the thermody- particle system is nearly hyperuniform or approaching namic limit. We purposely avoid the use of simulations a hyperuniform state, whether ordered or not. Eluci- of many-particle systems in finite boxes to draw con- dating such questions not only is expected to lead to a clusions, since hyperuniformity in an infinite-wavelength deeper fundamental understanding of the nature and for- property. The first model that we characterize is an equi- mation of hyperuniform systems but has great practical librium system of hard rods (Sec. VI). Here the control value. For example, since hyperuniformity can endow parameter is the number density ρ (or packing fraction) a system with novel or optimal physical properties, it and its terminal value corresponds to the jammed state is essential to know how close the system must be to that is the integer lattice. The second model studied is perfect hyperuniformity without significantly degrading a certain “sticky” hard-sphere system in arbitrary space its ideal performance. In practice, perfect hyperunifor- dimension d as a function of the number density (or pack- mity is never achieved due to defects that are inevitably ing fraction) (Sec. VII). The third model that we char- present in any real finite-sized system [44], whether crys- acterize are low-temperature excited states of so-called talline, quasicrystalline or disordered [23]. stealthy long-ranged pair interactions in Rd (Sec. VIII). We begin by recalling pertinent previous concepts Here, the control parameter is the temperature T and and results (Sec. II), including the fluctuation- the corresponding ground states at T = 0 are disordered compressibility theorem, hyperuniformity as a critical and degenerate. 3
II. COMPRESSIBILITY, INVERTED CRITICAL where denotes a convolution integral. Fourier trans- POINT, GROWING LENGTH SCALE AND forming⊗ (8) and solving forc ˜(k), the Fourier transform HYPERUNIFORMITY INDEX of c(r), yields
h˜(k) h˜(k) In this section, we briefly review pertinent background c˜(k)= = . (9) material. This outline includes the implications of the S(k) 1+ ρh˜(k) fluctuation-compressibility theorem, hyperuniformity as a critical phenomenon, a length scale that grows on ap- By definition, a hyperuniform system is one in which proach to a hyperuniform state, and the hyperuniformity h˜(k = 0) = 1/ρ, i.e., the volume integral of h(r) ex- − index. ists, implying that h(r) is sufficiently short-ranged in the sense that it decays to zero faster than r −d. Interest- ingly, this means that the denominator on| | the right side A. Fluctuation-Compressibility Theorem of (9) vanishes at k = 0 and thereforec ˜(k = 0) diverges to . This behavior implies that the the volume in- −∞ r Let us now recall the well-known fluctuation- tegral of c( ) does not exist and hence the real-space r compressibility theorem that links the isothermal com- direct correlation function c( ) is long-ranged, i.e., de- cays slower than r −d. We see that this behavior stands pressibility κT of equilibrium single-component many- | | particle ensembles at number density ρ and temperature in diametric contrast to standard thermal critical-point T to infinite-wavelength density fluctuations [3]. In par- systems in which the total correlation function is long- ticular, for “open” systems in equilibrium, one has ranged and the direct correlation function is short-ranged such that its volume integral exists [38–41]. For this rea- 2 2 N ∗ N ∗ son, it has been said that hyperuniform systems are at ρkBTκT = h i − h i = S(k = 0)=1+ρ h(r)dr, N ∗ Rd an “inverted” critical point [1]. As noted earlier, systems h i Z (7) at thermal critical points are anti-hyperuniform. where kB is Boltzmann’s constant and ∗ denotes an There is a class of disordered hyperuniform systems average in the grand canonical ensemble. hi with concomitant critical exponents [1, 2]. For such hype- The fluctuation-compressibility relation (7) enables runiform critical systems, the direct correlation function one to draw useful conclusions about the hyperunifor- has the following asymptotic behavior for large r r mity of equilibrium systems. Any ground state (T = 0) and sufficiently large d: ≡ | | in which the isothermal compressibility κT is bounded 1 and positive must be hyperuniform because the struc- c(r) (r ), (10) ture factor S(k = 0) must be zero according to relation ∼−rd−2+η → ∞ (7)[2]. More generally, we infer from (7) that if the prod- where (2 d) < η 2 is a “critical” exponent associated uct TκT tends to a nonnegative constant c in the limit with c(r)− for hyperuniform≤ systems that depends on the T 0, then the ground-state of this system in this zero- space dimension. The Fourier transform of (10) yields temperature→ limit must be nonhyperuniform if c > 0 or hyperuniform if c = 0. By the same token, this means 1 c˜(k) (k 0), (11) that increasing the temperature by an arbitrarily small ∼−k2−η → positive amount when a system is initially at a hyperuni- form ground state will destroy perfect hyperuniformity, which, when combined with (9), yields the asymptotic since S(k = 0) must deviate from zero by some small form of the structure factor amount determined by the temperature dependence of S(k) k2−η (k 0), (12) the product κT T for small T [2]. This indirectly im- ∼ → plies that phonons or vibrational modes for sufficiently where η =2 α and α is the exponent defined in relation small T generally destroy the hyperuniformity of ground (3). − states [23, 25]. Additionally, in order to have a hyperuni- In what follows, it is assumed for concreteness, that form system that is in equilibrium at any positive T , the the number density is the control parameter. We define isothermal compressibility must be zero, i.e., the system the following dimensionless density: must be thermodynamically incompressible [2]. φ = ρv1(D/2), (13)
B. Inverted Critical Point and Scaling Laws where v1(R) is given by (2) and D is a characteristic “microscopic” length scale. The direct correlation in the The direct correlation function c(r) of a hyperuniform vicinity of a hyperuniform critical state with dimension- less density φ , i.e., for φ φ 1, in sufficiently high di- system behaves in an unconventional manner compared c | c− | ≪ to that of typical liquids. This function is defined via the mensions has the following large-r asymptotic form [1, 2]: Ornstein-Zernike integral equation [45]: exp( r /ξ) c(r) −| | , (14) h(r)= c(r)+ ρc(r) h(r), (8) ∼ r d−2+η ⊗ | | 4 where ξ is the correlation length. If the system ap- where kp is the wavenumber k associated with the largest proaches a hyperuniform state from below the critical peak height of the angularly averaged structure factor. density φc, the correlation length and inverse of the struc- One may empirically deem a system to be nearly or effec- ture factor at k = 0, S−1(0), which is proportional to tively hyperuniform if H is roughly less than about 10−4 c˜(0), are described by the following scaling laws: [32]. The H index has been profitably used to quan- tify the effective hyperuniformity of polymer systems −ν φ − [49, 50], amorphous ices [51], states along the metastable ξ 1 (φ φc ), (15) ∼ − φc → extension of the hard-sphere systems away from jamming [52] and low-temperature states of “quantizer” systems [53, 54]. −γ −1 φ − S (0) 1 , (φ φc ), (16) ∼ − φc → III. HYPERUNIFORM AND where ν and γ are nonnegative critical exponents. Ob- NONHYPERUNIFORM SCALING REGIMES serve that the exponent γ is a measure of how quickly a system approaches a critical point. Combination of the For a large class of ordered and disordered systems, the three previous scaling laws leads to the following interre- number variance σ2(R) has the following large-R asymp- lation between the exponents: totic behavior [1, 2]:
γ = (2 η)ν. (17) R d R d−1 R d−1 − σ2(R)=2dφ A + B + o D D D The specific values of the critical exponents determine " # the universality class of the hyperuniform system. It is (20) where φ is the dimensionless density given by (2) noteworthy that all class III hyperuniform systems are d−1 at critical points with the aforementioned scaling laws. and o (R/D) represents terms of lower order than Specific examples of such a class III critical-point systems (R/D)d−1. Moreover, A and B are “volume” and are nonequilibrium absorbing-state models [37]. “surface-area” coefficients, respectively, which can be ex- pressed as the following volume integrals involving the total correlation function h(r), respectively: C. Growing Length Scale A = lim S(k)=1+ ρ h(r)dr |k|→0 Rd Another important length scale ξc that can be ex- ∞ Z tracted from the direct correlation function and which =1+ d 2d φ xd−1h(x)dx, (21) grows as a hyperuniform state is approached is defined Z0 by [46] d Γ(d/2)ρ 1/d B = r h(r)dr ξc = [ c˜(k = 0)] . (18) −2 π1/2 DΓ[(d + 1)/2] Rd | | − Z d2 2d−1 Γ(d/2)φ ∞ Thus, we see that ξc is the dth root of the volume inte- = xd h(x)dx, (22) r −π1/2Γ[(d + 1)/2] gral of the direct correlation function c( ). At a hyper- Z0 uniform critical point, ξ diverges to + . It was shown c ∞ and x = r/D is a dimensionless distance. Here h(r) is that a precursor to the hyperuniform maximally random the radial function that depends on the distance r jammed (MRJ) state of sphere packings [47] under com- r , which results from averaging the vector-dependent≡ pression was evident for densities far below the jamming quantity| | h(r), i.e., density was reached, as reflected by this static grow- ing length scale [46]. The quantity ξc was also used to 1 r identify length scales in supercooled atomic liquid mod- h(r)= h( ) dΩ, (23) Ω Ω els that substantially grow as the temperature decreased Z [48]. dΩ is the differential solid angle and dπd/2 Ω= (24) D. Hyperuniformity Index Γ(1 + d/2) is the total solid angle contained in a d-dimensional The hyperuniformity index H provides a measure of sphere. In a perfectly hyperuniform system [1], the non- the nearness of a system to a hyperuniform state, which negative volume coefficient vanishes, i.e., A = 0, implying is defined as the sum rule S(0) H , (19) ρ h(r)dr = 1, (25) ≡ S(kp) Rd − Z 5 such that the surface-area coefficient B is nonnegative. intensity is available experimentally or if the structure Thus, such hyperuniform systems fall within class I, since factor is analytically available, as it is when solving the the variance grows like the window surface area (Rd−1). Ornstein-Zernike integral equation (8). On the other hand, when A > 0 and B = 0, the system We define the Fourier transform of a function f(r) that is hyposurficial, implying the sum rule depends on the vector r in d-dimensional Euclidean space Rd ∞ as follows: xdh(x)dx =0. (26) 0 f˜(k)= f(r) exp [ i(k r)] dr, (28) Z Rd − · Z Examples of hyposurficial systems include ideal gases, d Rd where k is a wave vector and (k r) = k r is the certain hard-core systems in [1], non-equilibrium · i=1 i i phase transitions in amorphous ices [51], and certain sys- conventional Euclidean inner product of two real-valued P tems with bounded pair interactions [55]. Appendix A vectors. The function f(r) can generally represent a ten- provides a more general asymptotic expansion of the lo- sor of arbitrary rank. When it is well-defined, the corre- cal number variance, which is used to derive nonhyper- sponding inverse Fourier transform is given by uniform scaling laws. 1 d For a large class of nonhyperuniform disordered sys- f(r)= f˜(k) exp [i(k r)] dk. (29) 2π Rd · tems that are sub-Poissonian [56], such as fluids and col- Z loids without particle clustering, the volume coefficient If f is a radial function, i.e., f depends only on the mod- A is often larger than the magnitude of the surface-area ulus r = r of the vector r, its Fourier transform is given coefficient B. For super-Poissonian configurations [56], by | | such as the one described in Appendix B, the magnitude ∞ d J (kr) of B can be larger than A. When A < 1 and A the structure factor S(k) at the origin of a nearly hy- peruniform system with vacancies is proportional to the concentration of vacancies [23] and hence cannot be hy- peruniform, as predicted by (48). Thus, the free-volume form for the pressure (47) cannot apply to this nonequi- librium vacancy-riddled but strictly jammed state. More generally, Torquato and Stillinger [1] suggested that certain defect-free strictly jammed packings of iden- tical spheres are hyperuniform. Specifically, they conjec- tured that any strictly jammed saturated infinite packing of identical spheres is hyperuniform. A saturated pack- ing of hard spheres is one in which there is no space available to add another sphere. What is the rationale for such a conjecture? First, it recognizes that mechan- ical rigidity is a necessary but not sufficient condition for hyperuniformity. Indeed, requiring the saturation property in the conjecture eliminates the class of strictly jammed crystal states that are riddled with vacancies, FIG. 1. The isothermal phase behavior of the 3D hard-sphere as per the aforementioned example. Moreover, the con- model in the pressure-packing fraction plane, as adapted from Ref. 61. An infinitesimal compression rate of the liquid traces out the jecture excludes packings that may have a rigid back- thermodynamic equilibrium path (shown in red), including a dis- bone but possess “rattlers” (particles that are not locally continuity resulting from the first-order freezing transition to a jammed but are free to move about a confining cage) be- crystal branch (shown in green) that ends at a maximally dense, cause a strictly jammed packing, by definition, cannot infinite-pressure jammed state. Rapid compressions of the liquid (blue curves) generate a range of amorphous metastable extensions contain rattlers [61, 68]. Typical packing protocols that of the liquid branch that jam only at the their density maxima, have generated disordered jammed packings tend to con- which we show here must be perfectly hyperuniform states. tain a small concentration of rattlers [61, 69], and hence the entire (saturated) packing cannot be deemed to be jammed. Therefore, the conjecture cannot apply to cur- behavior of the pressure, even in the infinite-system limit rent numerically-generated disordered packings, even if [61]. Although there is no rigorous proof yet for this the structure factor at the origin is very small, e.g., an H claim, all numerical evidence strongly suggests that it is index of the order of 10−4 for 3D MRJ-like sphere packing correct as the jamming state is approached either along [29]. Thus, rattler-free disordered jammed sphere pack- the crystal [66, 67] or metastable branch [68]. Assuming ings are expected to be perfectly hyperuniform. Indeed, that the system dynamics remain ergodic in the vicin- if the free-volume theory applies along the metastable ity of a jammed state, we can apply the fluctuation- extension ending at the MRJ state, rattlers cannot be compressibility relation (7) along with the free-volume present (due to a type of constrained ergodicity on time form (47) to yield the corresponding asymptotic relation scales much less than relaxation times) and hence should −1 for S (0): be perfectly hyperuniform. It has been suggested that the ideal MRJ state is rattler-free, implying that the packing −1 d S (0) (φ φJ ). (48) is more disordered without the presence of rattlers [32]. ∼ (1 φ/φ )2 → − J Thus, we see that in the limit φ φ , the system be- J The consequences of the free-volume theory requires comes perfectly hyperuniform, which→ appears to validate a modification of the Torquato-Stillinger conjecture be- a more general conjecture of Torquato and Stillinger [1], cause the former eliminates defects of any type in the as elaborated below. jamming limit and the saturation condition may not It is important to note that ergodicity condition re- prohibit all defect types. For example, the saturation quired to obtain (48) implies that the system must be property may not exclude dislocations in strictly jammed defect-free. To stress this point further, consider an fcc crystal states. A more refined variant of the conjecture packing in R3 at the jammed state with packing fraction is the following statement: Any strictly jammed infinite φ = π/√18=0.74048 .... Next, shrink each sphere by an packing of identical spheres that is defect-free is hyper- infinitesimal uniform amount (such that the impenetra- uniform. bility constraints are linerizable) and randomly remove a small but statistically significant fraction of spheres so that the resulting vacancies are well-separated from In Appendix C, we obtain exact sum rules and the one another. This vacancy-riddled packing remains in exact large-k asymptotic behaviors of the structure fac- its jamming basin. Now, slowly compress it until it jams, tors of certain general packings of identical spheres in Rd, whereby the system pressure diverges. It is known that whether in equilibrium or not. 8 VI. EQUILIBRIUM HARD RODS The structure factor S(k) is plotted in Fig. 2 for the hard-rod system near the jammed state with φ = 0.98. We first structurally characterize one-dimensional sys- At such high densities, its peak values occur at wavenum- tems of identical hard rods of length D in equilibrium as bers that are approximately integer multiples of 2π/D, a function of the packing fraction φ up to the jammed, which is exactly the case for the integer lattice (jammed hyperuniform state (φ = 1). This hyperuniform state is state) with lattice constant D. The envelope of the the integer lattice Z and hence cannot be regarded to be largest peak values decays with wavenumber approxi- 3/2 a critical hyperuniform state, as described in Sec. IIB. mately as the power law 1/k . It immediately follows Percus [70] obtained the following exact expression for from (52) that the volume coefficient A, defined by (21), the direct correlation function for an equilibrium hard- is exactly given by rod system at packing fraction φ: A = S(0) = (1 φ)2 (53) 1 φr/D − c(r)= Θ(D r) − (49) The surface-area coefficient B is easily obtained for all φ − − (1 φ)2 − from its Fourier representation (44) and use of the exact where Θ(x) is the Heaviside step function. Combination structure factor formula (52) by high-precision numerical of this relation with (46) yields the well-known expression integration. These results for B are in excellent agree- for the pressure of the hard-rod system [71]: ment with previous results using the direct-space repre- p 1 sentation of B [1]. A simple and highly accurate formula = , (50) for B is given by ρkB T 1 φ − φ which we see takes the free-volume form (47) for the en- B = (6 8φ +3φ2). (54) tire range of packing fractions, not just near the jammed 12 − state φJ = 1, which must be hyperuniform. The Fourier This formula is obtained by using exact results at low and transform of c(r) is given by high packing fractions and assuming that B is a cubic 2 [φ (cos(kD) 1)+ kD sin(kD)(φ 1)] polynomial in φ (without a constant term). Specifically, c˜(k)= − − . (51) (1 φ)2(kD)2 the linear and quadratic terms are determined by the ex- − act expansion of S(k) S0 in powers of φ through second Using (9), the corresponding structure factor is given by order in φ, as determined− from the the exact formula 1 (52). The remaining cubic term is found by using the S(k)= . 2 φ [φ (cos(kD) 1)+ kD sin(kD)(φ 1)] fact that B is exactly equal to 1/12 for the integer lattice 1 − − − (1 φ)2(kD)2 at φ = 1 [1]. Thus, combination of the aforementioned − (52) formulas for A and B, yields the following excellent ap- It is seen that the asymptotic large-k behavior of S(k) proximation for the ratio is consistent with the general expression (C6) for d = 1, B φ(6 8φ +3φ2) where the contact value g (D+)=(1 φ)−1 is bounded = − . (55) 2 − A 12(1 φ)2 for all φ, except at its maximal value φ = 1. − The ratio B/A is plotted in Fig. 3 as a function of φ. It 100 3/2 is seen that B is about ten times larger than A at φ =0.9 Equilibrium Hard Rods ~1/k and the ratio dramatically diverges to infinity as φ tends 10 φ = 0.98 to unity. For example, at φ =0.99, B is about 842 times larger than A. Referring to Sec. A, the magnitude of the 1 ratio B/A determines the hyperuniformity scaling regime for the local number variance σ2(R) to be those window −1/d 0.1 radii in the range (ρ ) 1e+08 must increase as φ increases and diverges in the limit that the jamming density is approached from below, i.e., φ 1e+06 1−, which should be distinguished from the case when→ Equilibrium Hard Rods φ = 1 in which S(k) consists of Dirac delta functions 10000 located at wavenumbers that are multiples of 2π/D. The first peak value of the structure factor, which is also the 100 largest value, at a fixed value of the packing fraction is B/A plotted in Fig. 5 as a function of φ, which of course must 1 diverge at the hyperuniform jammed state φ = 1. The 0.01 following rational function provides an excellent fit to the data for all φ up to the jamming density: 0.0001 1e-06 0 0.2 0.4 0.6 0.8 1 1+ a φ + a φ2 + a φ3 φ S(k )= 1 2 3 , (58) p (1 φ)2 − FIG. 3. The ratio of the surface-area coefficient to volume coefficient, B/A, as a function of the packing fraction φ for equilibrium hard-rod systems. where a1 = 1.565536, a2 = 0.470399 and a3 = 0.196196. This− fit uses the exact small-φ expansion (56) 0.1 to ascertain the coefficient a1. This accurate approxi- 0.01 mation when combined with exact formula (53) for S(0) Equilibrium Hard Rods yields the following expression for the hyperuniformity 0.001 index: 2 0.0001 1e-05 -2 ~ R φ = 0.99 (R)/R 4 2 (1 φ) 1e-06 nonhyperuniform H = − , (59) σ 2 3 1+ a1φ + a2φ + a3φ 1e-07 φ = 1 -1 ~ R 1e-08 hyperuniform 1e-09 which is plotted in Fig. 6 as function of φ. 1 10 100 1000 10000 1e+05 R/D FIG. 4. The scaled variance σ2(R)/R2 versus R/D, as com- puted from relation (20), for the jammed hyperuniform state 100 with φ = 1 (as found in Ref. [1]) and for a nonhyperuni- form state near jamming with φ = 0.99, where Rc = B/A = 841.7474996. Because the variance is bounded for the hyper- 80 Equilibrium Hard Rods uniform state [1], σ2(R)/R2 decreases with R for large R like 1/R2. 60 ) It easily follows from formula (52) that the largest peak p value of the structure factor S(kp) through first-order in S(k 40 the packing fraction is exactly given by sin(kpD) 2 S(kp)=1 φ + (φ ) 20 − kpD O =1+(0.434464192 ...)φ + (φ2), (56) O 0 0 0.2 0.4 0.6 0.8 1 where φ 3 1 k D = π + 9π2 16=4.489654702 .... (57) p 4 4 − FIG. 5. The largest peak value of the structure factor S(kp) It is highly nontrivialp to derive an analytical expression versus the packing fraction φ for equilibrium hard-rod sys- for S(kp) for arbitrary φ. Of course, we know that S(kp) tems. 10 1 order unity, and hence B/A and ξc have the same scal- 0.1 ing behavior as the hyperuniform state of the hard-rod system is approached, namely, 0.01 B ξ 1 c . (61) 0.001 A ∼ D ∼ (1 φ)2 Equilibrium Hard Rods − H 0.0001 This is to be contrasted with the inverse of the hyper- uniformity index H−1, which grows exponentially faster, 1e-05 i.e., we see from (6) that it grows like the square of the 1e-06 other two quantities: 2 2 1e-07 B ξc 1 H−1 . (62) ∼ A ∼ D ∼ (1 φ)4 1e-08 0 0.2 0.4 0.6 0.8 1 − φ VII. STICKY HARD SPHERES FIG. 6. The hyperuniform index H versus the packing frac- tion φ for equilibrium hard-rod systems. The rapid growth of We utilize a particular g2-invariant process to analyt- ξc with increasing packing fraction is readily apparent. ically study the approach of another disordered system to a hyperuniform but unjammed state as a function of Using (51), it immediately follows that the length scale the relevant control parameter, the number density. A ξc, defined by (18), for equilibrium hard rods is given by g2-invariant process is one in which a chosen nonneg- ative form for the pair correlation function g2 remains (2 φ)D invariant over a nonvanishing density range while keep- ξ = c˜(k =0)= − . (60) c − (1 φ)2 ing all other relevant macroscopic variables fixed [1, 73]. − The upper limiting “terminal” density is the point above Figure 7 clearly shows that this length scale grows ap- which the nonnegativity condition on the structure fac- preciably with increasing packing fraction. It is already tor, i.e., S(k) 0 for all k, would be violated. Thus, at about an order of magnitude greater than the length of a the terminal or≥ critical density, the system is hyperuni- rod D at φ =0.6 and rapidly takes on even larger values form if the minimum in S(k) occurs at the origin and it as φ increases further, eventually diverging to infinity in is realizable. This optimization problem has deep con- the limit φ 1. → nections to the sphere-packing problem. Specifically, it is the linear program (LP) lower bound on the maximal 100 packing density that is dual to the Elkies-Cohn LP upper bound formulation [74]. A certain g2-invariant test func- 80 Equilibrium Hard Rods tion was employed to conjecture that the densest sphere packings in very high space dimensions are disordered, rather than ordered as in low dimensions [75]. 60 Here we specifically consider the g2-invariant process /D defined by a pair correlation function that imposes a c ξ hard-core constraint via a unit step function plus a delta 40 function contribution that acts at contact r = D: Z g2(r) =Θ(r D)+ δ(r D), (63) 20 − ρs1(D) − where Z is the nonnegative average contact coordination 0 0.2 0.4 0.6 0.8 1 number and φ dπd/2rd−1 s (r)= (64) 1 Γ(1 + d/2) FIG. 7. The dimensionless length scale ξ/D versus the pack- ing fraction φ for equilibrium hard-rod systems. is the surface area of a sphere of radius r. It was pre- viously shown that this “sticky-sphere” pair correlation It is useful to compare and contrast the growth rate function is numerically realizable, to an excellent approx- −1 of the three descriptors B/A, ξc and H as the packing imation, up to the hyperuniform terminal packing frac- fraction φ increases and approaches the jammed hyper- tion for d = 2 (see Fig. 8), implying that such pair uniform state. We see that the numerators of the formu- correlation functions are also realizable in all higher di- las (55) and (60) for moderate to high values of φ are of mensions, as dictated by the decorrelation principle [75]. 11 asymptotic large-k behavior of S(k) is consistent with the general expression (C10) for any d, where Z is given by (65). Relation (67) leads to the power law φ −1 S−1(0) = 1 , φ φ−, (68) − φ → c c which upon comparison to (16) yields the critical expo- nent γ = 1. Figure 9 shows the structure factor S(k) versus kD for sticky hard spheres for d = 2 and d =5 at their respective hyperuniform states. We see that struc- ture factor reflects strong decorrelation in going from two to five dimensions, which is consistent with the decorre- lation principle [75]. 2 Sticky Hard Spheres FIG. 8. A numerically generated two-dimensional configu- 1.5 ration of 500 “sticky” hard disks that realizes the step plus delta function pair correlation function at the terminal pack- ing fraction φc = 1/2, as adapted from Ref. [76]. This critical 1 hyperuniform state consists only of dimers, i.e., Z = 1, and d=2 S(k) hence is unjammed, in contrast to hyperuniform states in the phase diagram of hard spheres, as shown in Fig. 1. d=5 0.5 Here we collect previously established key results for this g2-invariant process [1, 73] in order to characterize how such systems for any d approach a hyperuniform 0 0 5 10 15 20 state as the packing fraction increases up to the critical kD terminal value φc. It was shown that 2dd FIG. 9. The structure factor S(k) versus the dimensionless Z = φ (65) wavenumber kD for sticky hard spheres in two and five di- d +2 mensions at their respective hyperuniform states, i.e., φ = φc, is obeyed in order to constrain the location of the min- where φc is given by (66). imum of the structure factor to be at k = 0, where the packing φ lies in the range 0 φ φ , and ≤ ≤ c In any space dimension d, the volume and surface-area coefficients for 0 φ φ are respectively given by d +2 ≤ ≤ c φc = (66) 2d+1 φ A = S(k =0)=1 (69) is the terminal or critical packing fraction, which has the − φc same asymptotic form as a lower bound on the maximal packing density for lattice packings derived by Ball [77]. and At the hyperuniform critical point, the contact number d2Γ(d/2) φ Z = d/2, indicating that such sticky-sphere systems are B = . (70) c √ never jammed in any dimension. 8 π Γ((d + 3)/2) φc The corresponding structure factor for any d and φ in Thus, ratio B/A is given by the range 0 φ φ is given by ≤ ≤ c 2 d/2 B d Γ(d/2) φ/φc 2 Γ(2 + d/2) φ J(d/2)−1(kD) Jd/2(kD) = . (71) S(k)=1+ . A 8 √π Γ((d + 3)/2) (1 φ/φ ) (kD)(d/2)−1 φ d +2 − kD − c c " # (67) We see that apart from d-dimensional constants, the be- An important mathematical property of the structure havior of the ratio B/A is exactly the same across di- factor for such sticky hard spheres is that its extrema mensions when the packing fraction φ is scaled by the for a fixed d are independent of the packing fraction. terminal packing fraction φc. The ratio B/A versus φ is For example, for d = 2 and d = 5, kpD = 6.38015 ... plotted in Fig. 10 for d = 2 and d = 5 for all φ up to and kpD = 8.18255 ..., respectively. It is seen that the their respective critical points. 12 1e+06 relation for sticky spheres: d/2 d 10000 (2π) D J(d/2)−1(kD) Jd/2(kD) (d/2)−1 (kD) " d +2 − kD # 100 d=5 c˜(k)= 2d/2Γ(2 + d/2) φ J (kD) J (kD) 1+ (d/2)−1 d/2 d=2 (kD)(d/2)−1 φ d +2 − kD 1 c " # B/A (73) 0.01 It immediately follows that at zero wavenumber, we have Sticky Hard Spheres φ −1 0.0001 c˜(0) = 2v (D) 1 , (74) − 1 − φ c 1e-06 0 0.1 0.2 0.3 0.4 0.5 and, hence, the length scale defined by (18) yields φ φ −1/d ξ = [2v (D)]1/d 1 , (75) c 1 − φ FIG. 10. The ratio of the surface-area coefficient to volume c coefficient, B/A, as a function of the packing fraction φ for sticky hard spheres in two and five dimensions. indicating that it grows more slowly as the space dimen- sion increases. Figure 7 shows how the length scale ξc grows with φ for d = 2 and d = 3. We see from Fig. 9 that the peak value of the structure The length scale ξc should be contrasted with the cor- factor is not strongly dependent on the space dimension relation length ξ, defined by (15), which, when combined and is of order unity. Thus, the hyperuniformity index with relation (74), yields H is primarily determined by the inverse of A [cf. 69)] D φ −1/4 and so ξ = 1 , φ (76)φ−. [8(d + 2)(d + 4)φ ]1/4 − φ → c c c 1 H . (72) Comparison of (76) to the power law (15) yields the ex- ∼ 1 φ/φ − c ponent ν = 1/4. It should not go unnoticed that since the exponents γ, η and ν are either integers or a simple The actual values of H are plotted in Fig. 11 as a function fraction, independent of dimension, these systems always of φ for d = 2 and d = 5 up to their respective critical behave in a mean-field manner at their critical points. points. 10 1 9 Sticky Hard Spheres Sticky Hard Spheres 8 0.8 7 6 d=5 0.6 /D 5 c d=2 ξ H 4 d=2 0.4 3 d=5 2 0.2 1 0 0 0.1 0.2 0.3 0.4 0.5 0 φ 0 0.1 0.2 0.3 0.4 0.5 φ FIG. 12. The dimensionless length scale ξc/D versus the pack- ing fraction φ for sticky hard spheres in two and five dimen- FIG. 11. The hyperuniformity index H as a function of the sions. packing fraction φ for sticky hard spheres in two and five dimensions. −1 All three descriptors B/A, H and ξc are increasing functions of the packing fraction φ, but how do they cor- Combination of (9) and (67) yields the following ex- relate with one another? From formulas (71) and (72), pression for the the Fourier transform of the direct cor- we see B/A and H−1 have the same scaling behavior as 13 the critical hyperuniform state of the sticky hard-system stealthy ground states in the canonical ensemble behave is approached, namely, like “pseudo” equilibrium hard-sphere systems in Fourier space with an “effective packing fraction” φ that is pro- B 1 H−1 . (77) portional to χ, one can employ well-established integral- A ∼ ∼ (1 φ/φc) equation formulations for the pair correlation function − of hard spheres in direct space [3, 57] to obtain accu- According to (75), the length scale ξc for d 2 grows r k ≥ rate theoretical predictions for g2( ; φ) and S( ; χ) for a more slowly than the other two quantities as the critical moderate range of χ about χ = 0 [25], i.e., state is approached, i.e., S(k; χ)= gHS (r = k; φ). (81) B 2 (ξ )d H−1. (78) c ∼ A ∼ For example, the total correlation function h(r) in the limit χ 0 for any d, which is exactly given by → VIII. STEALTHY DISORDERED GROUND K d/2 STATES ρh(r)= J (Kr) (χ 0), (82) − 2πr d/2 → We are also interested in theoretically understanding corresponds to the following unit-step function for the how vibrational modes in excited low-temperatures states structure factor: of matter impact the approach to ground states, which S(k)=Θ(k K) (χ 0). (83) are necessarily hyperuniform (see Sec. II A), as the tem- − → perature T tends to zero. For this purpose, we consider Our interest in this paper are the excited states asso- stealthy hyperuniform many-particle systems [24, 78], ciated with the long-ranged stealthy potentials [25] close which are a subclass of class I hyperuniform systems to the ground states, i.e., when the absolute temperature [25] in which the structure factor is zero for a range of T is small. It was shown that the structure factor at wavevectors around the origin, i.e., the origin S(0) varies linearly with T for excited states. S(k)=0 0 k K, (79) Importantly, this positive value of S(0) is the uniform ≤ | |≤ value of S(k) for 0 k K for the special case of the ≤ | |≤ where K is some positive constant. Such systems are step-function power-law potential for small χ, which was called “stealthy” because they completely suppress sin- verified by simulation results in various dimensions [25]. gle scattering of incident radiation for the wave vectors Because we are interested in qualitative trends as the dis- within an exclusion sphere of radius K centered at the ordered ground state is approached as the temperature origin in Fourier space. It has been shown that certain is decreased, it suffices for our purposes to consider the bounded (soft) long-ranged pair interactions have classi- following simple form for the structure factor at positive cal ground states that are stealthy and hyperuniform [25]. but small temperatures and sufficiently small χ: The nature of the ground-state configuration manifold S(k)= T ∗Θ(K k)+Θ(k K), (84) (e.g., the degree of order) associated with such stealthy − − pair potentials depends on the number of constrained where T ∗ is a dimensionless temperature that is much wave vectors. The dimensionless parameter χ measures smaller than unity. The form (84) is obtained by com- the relative fraction of constrained degrees of freedom bining the aforementioned uniform value for 0 k K compared to the total number of degrees of freedom and, with the ground state structure factor as χ tends≤ | to|≤ zero, counterintuitively, is inversely proportional to the num- i.e., Eq. (83). ber density [25]; specifically, it is explicitly given by the It follows from (21) and (84) that the volume coefficient formula is given by v (K) ρχ = 1 , (80) A = T ∗ (85) 2d (2π)d Moreover, substitution of (84) into the Fourier represen- where v1(K) the volume of a d-dimensional sphere of ra- tation of the surface-area coefficient (44) yields dius K [cf. Eq. (2)]. For d = 2 and d = 3, the ground states are disordered for 0 χ< 1/2. At χ =1/2, there d ≤ B = (1 T ∗) (86) is a phase transition to a crystal phase. The “collective- πK − coordinate” optimization procedure represents a power- ful approach to generate numerically disordered stealthy Thus, ratio B/A is given by hyperuniform many-particle systems [24, 25, 78]. B d (1 T ∗) A statistical-mechanical theory for disordered stealthy = − , (87) A πK T ∗ ground states has been formulated in the canonical en- semble in the zero-temperature limit. By exploiting which is plotted in Fig. 13 as a function of the inverse an ansatz that the entropically favored (most probable) of T ∗ for selected space dimensions. We see that at fixed 14 temperature, the ratio B/A increases with the space di- 1e+05 mension. From relation (84), we see that the peak value d=1 of the structure factor is trivially unity (S(kp) = 1), and 10000 d=3 hence the hyperuniformity index is simply linear in T ∗, d=6 i.e., 1000 K c ξ H = T ∗. (88) 100 10 Toward Stealthy Ground States 1 1e+06 100 1000 10000 1e+05 * d=1 1/T 1e+05 d=3 d=6 FIG. 14. The dimensionless length scale ξcK versus the in- ∗ 10000 verse of the dimensionless temperature 1/T for excited states associated with the long-ranged stealthy pair potential for di- mensions d = 1, 3 and 6. 1000 B/A −1 The three descriptors B/A, H and ξc are increasing 100 functions of the inverse temperature. From formulas (87) and (72), we observe that B/A and H−1 have the same 10 Toward Stealthy Ground States scaling behavior as the inverse temperature tends to in- finity, i.e., the disordered hyperuniform ground state is 1 approached, namely, 10 100 1000 10000 1e+05 1e+06 * B 1 1/T H−1 . (91) A ∼ ∼ T ∗ FIG. 13. The ratio of the surface-area coefficient to volume coefficient, B/A, as a function of the inverse of the dimen- This is to be contrasted with the generally slower growth ∗ ∗ sionless temperature 1/T for excited states associated with rate of ξc with T for d 2, i.e., the long-ranged stealthy pair potential for dimensions d = 1, 3 ≥ B and 6. (ξ )d H−1, (92) c ∼ A ∼ where we have used relation (90). Combination of (9) and (84) yields the following ex- pression for the the Fourier transform of the direct cor- IX. CONCLUSIONS AND DISCUSSION relation function for the excited states: We have derived a Fourier representation of the surface-area coefficient B in terms of the structure factor k ρc˜(k)=(1 T ∗)Θ(K k). (89) S( ), which is especially useful when scattering informa- − − tion is available experimentally or theoretically. In a dis- ordered system that is nearly hyperuniform, we showed that the ratio of surface-area to volume coefficients, B/A, Thus, evaluating this relation at zero wavenumber and enables one to ascertain hyperuniform and nonhyperuni- use of (9) yields the length scale form distance-scaling regimes of the variance σ2(R) as a function of R as well as the corresponding crossover distance between these regimes. While the analysis of 1/d 1/d the ratio B/A was applied to systems that ultimately 2dχ 1 ξ = (2π) 1 . (90) approached a class I hyperuniform state, the correspond- c v (K) T ∗ − 1 ing extensions to other hyperuniformity classes is very straightforward. Using the ratio B/A, as well as other diagnostic mea- This length scale is plotted in Fig. 14 as a function of sures of hyperuniformity, including the hyperuniformity the inverse temperature for selected space dimensions. index H and the direct-correlation function length scale 15 ξc, we characterized the structure of three different ex- and actly solvable models as a function of the relevant con- φ d Γ(d/2) trol parameter, either density or temperature, with end BN (R)= h(r) r dr. −2 π1/2, D v (D/2)Γ[(d + 1)/2] r | | states that are perfectly hyperuniform. In the case of the 1 Z| |≤2R sticky hard-sphere and stealthy models, we studied the (A3) effect of dimensionality. For all three models, we showed Observe that when the volume coefficient AN (R) and that these diagnostic hyperuniformity measures are pos- surface-area coefficient BN (R) converge in the limit R → itively correlated with one another. Thus, the quantities , they are equal to the constants A and B, defined by ∞ H and ξc can also be employed to infer the hyperuniform (21) and (22), respectively. For this reason, A and B and nonhyperuniform distance-scaling regimes. This ca- are called the global volume and surface-area coefficients, pacity to determine hyperuniform scaling regimes is ex- respectively [2]. pected to be of great utility in analyzing experimentally- An anti-hyperuniform system is one in which the expo- or computationally-generated samples that are necessar- nent α [cf. (3)] is negative α< 0, resulting in a structure ily of finite size. factor that diverges in the zero-wavenumber limit. The r In the same way that there is no perfect crystal, due to corresponding total correlation function h( ) decays like 1/rd+α for large r r . This power-law tail controls the the inevitable presence of imperfections, such as vacan- ≡ | | cies and dislocations, there is no “perfect” hyperuniform growth rate of AN (R) with R; specifically, carrying out system in laboratory practice, whether it is ordered or the integration in (A2) yields not [23]. It is clear that the same diagnostic measures A (R) R−α. (A4) explored in the present work can be used to detect the N ∼ degree to which such “imperfect” nearly hyperuniform Substitution of this scaling behavior in the leading-order systems deviate from perfect hyperuniformity. term of (A1) gives that the local number variance for We showed that the free-volume theory of the pressure an anti-hyperuniform system scales like σ2(R) Rd−α, of equilibrium packings of identical hard spheres that ap- which proves the anti-hyperuniform scaling given∼ in (5). proach a strictly jammed state, either along the stable Since α is negative, the number variance grows faster crystal or metastable disordered branch, dictates that than the window volume, i.e., faster than Rd. Up to a such end states be exactly hyperuniform. This implies trivial constant in the leading-order term in the asymp- that the packing on approach to the jammed state must totic expansion of the number variance given by (A1), one be ergodic in a generalized sense and hence free of any can view a “typical” nonhyperuniform system in which defects, e.g., vacancies or dislocations along the crystal S(0) is bounded as one in which α =0+ (approaches zero branch or rattlers along the metastable fluid branch. It from above), even if S(k) is not described by a power-law is an outstanding problem for future research to place form in the zero-wavenumber limit. This enables one to these implications of free-volume theory for the hyperuni- conclude that AN (R) converges to a constant in the limit formity of strictly jammed packings on a firmer rigorous R , thus yielding the Rd scaling behavior for σ2(R) theoretical foundation. in→ (5). ∞ Appendix B: Evaluation of the coefficients A and B Appendix A: Number Variance Scalings for for a Super-Poissonian Point Pattern Nonhyperuniform Systems Any nonhyperuniform point process for which S(0) > 1 2 2 The local number variance σN (R) is generally a func- has a large-R asymptotic number variance σ (R) that is tion that can be decomposed into a global part that grows larger than that for a Poisson point process [S(0) = 1] with the window radius R and a local part that oscillates with the same mean N(R) is called super-Poissonian on small length scales (e.g., mean nearest-neighbor dis- [56]. We evaluate theh volumei and surface-area coeffi- tance) about the global contribution. The more general cients A and B for a specific model of a super-Poissonian large-R asymptotic formula for the variance is given by point process, namely, the Poisson cluster process, which [2] is characterized by strong clustering of the points with a large but finite value of S(0), and hence is far from d d−1 d−1 being hyperuniform. The higher-order moments of the 2 d R R R σ (R)=2 φ AN (R) +BN (R) +o number, fluctuations as well as the corresponding prob- D D D " #ability distribution for this model were recently studied (A1) [56]. The construction of the cluster process starts from where the R-dependent volume and surface-area coeffi- a homogeneous Poisson point process of intensity ρp [79]. cients are respectively given by Each point of the Poisson point process is the center of a cluster of points. The number of points in each cluster φ is independent and follows a Poisson distribution with AN (R)=1+ h(r)dr (A2) v (D/2) r mean value c. Following Ref. [56], we consider here the 1 Z| |≤2R 16 special case in which the positions of the points relative For large k, this Fourier transform is given by to the center of the cluster follow an isotropic Gaussian cos[ka (d + 1)/4] 1 distribution with standard deviation r0, which can be re- m˜ (k; a) Rd(2π)d/2 2/π − + (ka ). ∼ (ka)(d+1)/2 O (ka)(d+3)/2 → ∞ garded to be the characteristic length scale of a single cluster. In the infinite-volume limit, the structure factor p (C4) for any d is exactly given by [56] For any packing of spheres of diameter D, h(r) = m(r; D) for r sin(kD) 1 S(k) 1+ Z + (d = 3) . their careful reading of the manuscript. This work was ∼ kD O (kD)2 supported by the National Science Foundation under (C13) Award No. DGE-2039656. Again, we see that the rate of decay of S(k) increases as d increases, but less rapidly than in unjammed packings at the same dimension in which the jump discontinuity in g2(r) at contact is finite. ACKNOWLEDGMENTS The author is grateful to Jaeuk Kim, Michael Klatt, Charles Maher, Murray Skolnick and Haina Wang for ∗ Email: [email protected] [13] S. Gorsky, W. A. Britton, Y. Chen, J. Montaner, [1] S. Torquato and F. H. Stillinger, “Local density fluctua- A. Lenef, M. Raukas, and L. Dal Negro, “Engineered tions, hyperuniform systems, and order metrics,” Phys. hyperuniformity for directional light extraction,” APL Rev. E 68, 041113 (2003). Photonics 4, 110801 (2019). [2] S. Torquato, “Hyperuniform states of matter,” Physics [14] A. Sheremet, R. Pierrat, and R. Carminati, “Absorption Reports 745, 1–95 (2018). of scalar waves in correlated disordered media and its [3] J. P. Hansen and I. R. McDonald, Theory of Simple Liq- maximization using stealth hyperuniformity,” Phys. Rev. uids, 4th ed. (Academic Press, New York, 2013). A 101, 053829 (2020). [4] M. Florescu, S. Torquato, and P. J. Steinhardt, “De- [15] J. Kim and S. Torquato, “Multifunctional composites signer disordered materials with large complete photonic for elastic and electromagnetic wave propagation,” Proc. band gaps,” Proc. Nat. Acad. Sci. 106, 20658–20663 Nat. Acad. Sci. 117, 8764–8774 (2020). (2009). [16] S. Yu, C.-W. Qiu, Y. Chong, S. Torquato, and N. Park, [5] C. De Rosa, F. Auriemma, C. Diletto, R. Di Giro- “Engineered disorder in photonics,” Nature Rev. Mater. lamo, A. Malafronte, P. Morvillo, G. Zito, G. Rusciano, 6, 226–243 (2021). G. Pesce, and A. Sasso, “Toward hyperuniform disor- [17] C. E. Zachary and S. Torquato, “Hyperuniformity in dered plasmonic nanostructures for reproducible surface- point patterns and two-phase heterogeneous media,” J. enhanced Raman spectroscopy,” Phys. Chem. Chem. Stat. Mech.: Theory & Exp. 2009, P12015 (2009). Phys. 17, 8061–8069 (2015). [18] C. Lin, P. J. Steinhardt, and S. Torquato, “Hyper- [6] O. Leseur, R. Pierrat, and R. Carminati, “High-density uniformity variation with quasicrystal local isomorphism hyperuniform materials can be transparent,” Optica 3, class,” J. Phys.: Cond. Matter 29, 204003 (2017). 763–767 (2016). [19] E. C. O˘guz, J. E. S. Socolar, P. J. Steinhardt, and [7] T. Ma, H. Guerboukha, M. Girard, A. D. Squires, R. A. S. Torquato, “Hyperuniformity of quasicrystals,” Phys. Lewis, and M. Skorobogatiy, “3D printed hollow-core Rev. B 95, 054119 (2017). terahertz optical waveguides with hyperuniform disor- [20] A. Gabrielli, “Point processes and stochastic displace- dered dielectric reflectors,” Adv. Optical Mater. 4, 2085– ment fields,” Phys. Rev. E 70, 066131 (2004). 2094 (2016). [21] A. Gabrielli and S. Torquato, “Voronoi and void statistics [8] G. Zhang, F. H. Stillinger, and S. Torquato, “Transport, for superhomogeneous point processes,” Phys. Rev. E 70, geometrical and topological properties of stealthy disor- 041105 (2004). dered hyperuniform two-phase systems,” J. Chem. Phys [22] A. Gabrielli, M. Joyce, and S. Torquato, “Tilings of 145, 244109 (2016). space and superhomogeneous point processes,” Phys. [9] G. Gkantzounis, T. Amoah, and M. Florescu, “Hyper- Rev. E 77, 031125 (2008). uniform disordered phononic structures,” Phys. Rev. B [23] J. Kim and S. Torquato, “Effect of imperfections on the 95, 094120 (2017). hyperuniformity of many-body systems,” Phys. Rev. B [10] L. S. Froufe-P´erez, M. Engel, J. Jos´eS´aenz, and F. Schef- 97, 054105 (2018). fold, “Transport Phase Diagram and Anderson Localiza- [24] O. U. Uche, F. H. Stillinger, and S. Torquato, “Con- tion in Hyperuniform Disordered Photonic Materials,” straints on collective density variables: Two dimensions,” Proc. Nat. Acad. Sci. 114, 9570—-9574 (2017). Phys. Rev. E 70, 046122 (2004). [11] D. Chen and S. Torquato, “Designing disordered hyper- [25] S. Torquato, G. Zhang, and F. H. Stillinger, “Ensem- uniform two-phase materials with novel physical proper- ble theory for stealthy hyperuniform disordered ground ties,” Acta Materialia 142, 152–161 (2018). states,” Phys. Rev. X 5, 021020 (2015). [12] H. Zhang, W. Wu, and Y. Hao, “Luneburg lens from [26] G. Zhang, F. H. Stillinger, and S. Torquato, “The perfect hyperuniform disordered composite materials,” in 2018 glass paradigm: Disordered hyperuniform glasses down IEEE International Symposium on Antennas and Prop- to absolute zero,” Sci. Rep. 6, 36963 (2016). agation & USNC/URSI National Radio Science Meeting [27] Q-L. Lei and R. Ni, “Hydrodynamics of random- (2018) pp. 2281–2282. organizing hyperuniform fluids,” Proc. Nat. Acad. Sci. 18 116, 22983–22989 (2019). [48] E.´ Marcotte, F. H. Stillinger, and S. Torquato, [28] S. Torquato, G. Zhang, and M. de Courcy-Ireland, “Hid- “Nonequilibrium static growing length scales in super- den multiscale order in the primes,” J. Phys. A: Math. & cooled liquids on approaching the glass transition,” J. Theoretical 52, 135002 (2019). Chem. Phys. 138, 12A508 (2013). [29] A. Donev, F. H. Stillinger, and S. Torquato, “Unex- [49] A. Chremos and J. F. Douglas, “Particle localization and pected density fluctuations in disordered jammed hard- hyperuniformity of polymer-grafted nanoparticle materi- sphere packings,” Phys. Rev. Lett. 95, 090604 (2005). als,” Annalen der Physik 529 (2017). [30] C. E. Zachary, Y. Jiao, and S. Torquato, “Hyperuni- [50] A. Chremos and J. F Douglas, “Hidden hyperunifor- form long-range correlations are a signature of disordered mity in soft polymeric materials,” Phys. Rev. Lett. 121, jammed hard-particle packings,” Phys. Rev. Lett. 106, 258002 (2018). 178001 (2011). [51] F. Martelli, S. Torquato, N. Giovambattista, and R. Car, [31] Y. Jiao and S. Torquato, “Maximally random jammed “Large-scale structure and hyperuniformity of amor- packings of Platonic solids: Hyperuniform long-range phous ices,” Phys. Rev. Lett. 119, 136002 (2017). correlations and isostaticity,” Phys. Rev. E 84, 041309 [52] Y. Zhou, B. Mei, and K. S. Schweizer, “Integral equation (2011). theory of thermodynamics, pair structure, and growing [32] S. Atkinson, G. Zhang, A. B. Hopkins, and S. Torquato, static length scale in metastable hard sphere and weeks- “Critical slowing down and hyperuniformity on approach chandler-andersen fluids,” Phys. Rev. E 101, 042121 to jamming,” Phys. Rev. E 94, 012902 (2016). (2020). [33] R. P. Feynman and M. Cohen, “Energy spectrum of the [53] M. A. Klatt, J. Lovri´c, D. Chen, S. C. Kapfer, F. M. excitations in liquid helium,” Phys. Rev. 102, 1189–1204 Schaller, P. W. A. Sch¨onh¨ofer, B. S. Gardiner, A-S. (1956). Smith, G. E. Schr¨oder-Turk, and S. Torquato, “Uni- [34] L. Reatto and G. V. Chester, “Phonons and the proper- versal hidden order in amorphous cellular geometries,” ties of a Bose system,” Phys. Rev. 155, 88–100 (1967). Nature Comm. 10, 811 (2019). [35] S. Torquato, A. Scardicchio, and C. E. Zachary, “Point [54] T. M. Hain, M. A. Klatt, and G. E. Schr¨oder-Turk, processes in arbitrary dimension from Fermionic gases, “Low-temperature statistical mechanics of the Quantizer random matrix theory, and number theory,” J. Stat. problem: Fast quenching and equilibrium cooling of the Mech.: Theory Exp. 2008, P11019 (2008). three-dimensional Voronoi liquid,” J. Chem. Phys. 153, [36] C. E. Zachary and S. Torquato, “Anomalous local coordi- 234505 (2020). nation, density fluctuations, and void statistics in disor- [55] G. Zhang and S. Torquato, “Realizable hyperuniform and dered hyperuniform many-particle ground states,” Phys. nonhyperuniform particle configurations with targeted Rev. E 83, 051133 (2011). spectral functions via effective pair interactions,” Phys. [37] D. Hexner and D. Levine, “Hyperuniformity of critical Rev. E 101, 032124 (2020). absorbing states,” Phys. Rev. Lett. 114, 110602 (2015). [56] S Torquato, J. Kim, and M. A. Klatt, “Local number [38] B. Widom, “Equation of state in the neighborhood of the fluctuations in hyperuniform and nonhyperuniform sys- critical point,” J. Chem. Phys. 43, 3898–3905 (1965). tems: Higher-order moments and distribution functions,” [39] L. P. Kadanoff, “Scaling laws for Ising models near Tc,” Phys. Rev. X (2021). Physics 2, 263–272 (1966). [57] S. Torquato, Random Heterogeneous Materials: Mi- [40] M. E. Fisher, “The theory of equilibrium critical phe- crostructure and Macroscopic Properties (Springer- nomena,” Rep. Prog. Phys. 30, 615 (1967). Verlag, New York, 2002). [41] K. G. Wilson and J. Kogut, “The renormalization group [58] B. J. Alder and T. E. Wainwright, “Phase transition for and the ǫ expansion,” Phys. Rep. 12, 75–199 (1974). a hard sphere system,” J. Chem. Phys. 27, 1208–1209 [42] J. J. Binney, N. J. Dowrick, A. J. Fisher, and M. E. J. (1957). Newman, The Theory of Critical Phenomena: An Intro- [59] D. Frenkel, “Entropy-driven phase transitions,” Physica duction to the Renormalization Group (Oxford Univer- A 263, 26–38 (1999). sity Press, Oxford, England, 1992). [60] S. C. Mau and D. A. Huse, “Stacking entropy of hard- [43] E. C. O˘guz, J. E. S. Socolar, P. J. Steinhardt, and sphere crystals,” Phys. Rev. E 59, 4396–4401 (1999). S. Torquato, “Hyperuniformity and anti-hyperuniformity [61] S. Torquato and F. H. Stillinger, “Jammed hard-particle in one-dimensional substitution tilings,” Acta Cryst. Sec- packings: From Kepler to Bernal and beyond,” Rev. tion A: Foundations & Advances 75 (2019). Mod. Phys. 82, 2633 (2010). [44] R. Dreyfus, Y. Xu, T. Still, L. A. Hough, A. G. [62] S. Torquato and F. H. Stillinger, “Multiplicity of genera- Yodh, and S. Torquato, “Diagnosing hyperuniformity tion, selection, and classification procedures for jammed in two-dimensional, disordered, jammed packings of soft hard-particle packings,” J. Phys. Chem. B 105, 11849– spheres,” Phys. Rev. E 91, 012302 (2015). 11853 (2001). [45] L. S. Ornstein and F. Zernike, “Accidental deviations of [63] A. Donev, S. Torquato, F. H. Stillinger, and R. Connelly, density and opalescence at the critical point of a single “A linear programming algorithm to test for jamming in substance,” Proc. Akad. Sci. (Amsterdam) 17, 793–806 hard-sphere packings,” J. Comput. Phys. 197, 139–166 (1914). (2004). [46] A. B. Hopkins, F. H. Stillinger, and S. Torquato, [64] S. Torquato and F. H. Stillinger, “Toward the jamming “Nonequilibrium static diverging length scales on ap- threshold of sphere packings: Tunneled crystals,” J. proaching a prototypical model glassy state,” Phys. Rev. Appl. Phys. 102, 093511 (2007), Erratum, 103, 129902 E 86, 021505 (2012). (2008). [47] S. Torquato, T. M. Truskett, and P. G. Debenedetti, [65] S. Torquato, A. Donev, and F. H. Stillinger, “Break- “Is random close packing of spheres well defined?” Phys. down of elasticity theory for jammed hard-particle pack- Rev. Lett. 84, 2064–2067 (2000). ings: Conical nonlinear constitutive theory,” Int. J. Solids 19 Structures 40, 7143–7153 (2003). (2017). [66] Z. W. Salsburg and W. W. Wood, “Equation of state of [73] S. Torquato and F. H. Stillinger, “Controlling the short- classical hard spheres at high density,” J. Chem. Phys. range order and packing densities of many-particle sys- 37, 798–1025 (1962). tems,” J. Phys. Chem. B 106, 8354–8359 (2002), Erra- [67] F. H. Stillinger and Z. W. Salsburg, “Limiting polytope tum 106, 11406 (2002). geometry for rigid rods, disks, and spheres,” J. Stat. [74] H. Cohn and N. Elkies, “New upper bounds on sphere Phys. 1, 179–225 (1969). packings I,” Annals Math. 157, 689–714 (2003). [68] A. Donev, S. Torquato, and F. H. Stillinger, “Pair cor- [75] S. Torquato and F. H. Stillinger, “New conjectural lower relation function characteristics of nearly jammed disor- bounds on the optimal density of sphere packings,” Ex- dered and ordered hard-sphere packings,” Phys. Rev. E perimental Math. 15, 307–331 (2006). 71, 011105: 1–14 (2005). [76] O. U. Uche, F. H. Stillinger, and S. Torquato, “On the re- [69] G. Parisi and F. Zamponi, “Mean field theory of hard alizability of pair correlation functions,” Physica A 360, sphere glasses and jamming,” Rev. Mod. Phys. 82, 789— 21–36 (2006). 845 (2010). [77] K. Ball, “A lower bound for the optimal density of lattice [70] J. Percus, “Pair distribution function in classical statis- packings,” Int. Math. Res. Notices 68, 217–221 (1992). tical mechanics,” in The Equilibrium Theory of Classical [78] R. D. Batten, F. H. Stillinger, and S. Torquato, “Clas- Fluids, edited by H. L. Frisch and J. L. Lebowitz (Ben- sical disordered ground states: Super-ideal gases, and jamin, 1964). stealth and equi-luminous materials,” J. Appl. Phys. [71] L. Tonks, “The complete equation of state of one, two and 104, 033504 (2008). three-dimensional gases of hard elastic spheres,” Phys. [79] G. Last and M. Penrose, Lectures on the Poisson Process, Rev. 50, 955–963 (1936). Institute of Mathematical Statistics Textbooks (Cam- [72] J. Kim and S. Torquato, “Effect of window shape on bridge University Press, Cambridge, United Kingdom, the detection of hyperuniformity via the local number 2017). variance,” J. Stat. Mech.: Th. and Exper. 2017, 013402