Arxiv:2007.00297V1 [Cond-Mat.Soft] 1 Jul 2020 Disclination Unbinding for Hexatic- ﬂuid Melting

Two step melting of the Weeks- Chandler- Anderson system in two dimensions

Shubhendu Shekhar Khali,1, ∗ Dipanjan Chakraborty,1, † and Debasish Chaudhuri2, 3, ‡ 1Department of Physical Science, Indian Institute Of Science Education and Research Mohali, Punjab, India 140306 2Institute of Physics, Sachivalaya Marg, Bhubaneswar 751005, India 3Homi Bhaba National Institute, Anushaktigar, Mumbai 400094, India (Dated: July 2, 2020) We present a detailed numerical simulation study of a two dimensional system of particles interacting via the Weeks- Chandler- Anderson potential, the repulsive part of the Lennard- Jones potential. With reduction of density, the system shows a two- step melting: a continuous melting from solid to hexatic phase, followed by a a first order melting of hexatic to liquid. The solid- hexatic melting is consistent with the Kosterlitz-Thouless-Halperin-Nelson-Young (KTHNY) scenario and shows dislocation unbinding. The first order melting of hexatic to fluid phase, on the other hand, is dominated by formation of string of defects at the hexatic- fluid interfaces.

I. INTRODUCTION a continuous melting transition, and a first order transition [17]. This simulation did not allow the formation The high density phase of mono-dispersed particles in of defects but counted all the Monte-Carlo moves that two dimensions (2d) can not support long ranged posi- potentially formed them, to calculate the defect- core- tional order due to thermal fluctuations [1–3]. However, energy, and the unrenormalized Young’s modulus. These the pair correlation function in it does not decay expo- two quantities along with the KTHNY recursion relation nentially, unlike in the fluid phase. It shows a power- law predicted the precise melting point of the 2d hard- disk π 2 decay, characterizing a quasi- long- ranged positional or- solid at η = 4 ρd = 0.719 [17, 18], where ρ = N/A is der. In addition, the correlation between the orientation the particle density and d denotes the hard- disk diam- of the bonds between neighbours, the so-called bond- an- eter. However, the direct numerical simulations of hard gle correlation, turns out to be long- ranged. It remains disks also showed a clear signature of phase coexistence finite at the longest separations in the system. The quasi- at a lower η [19, 20]. For finite systems, the free energy long- ranged translational order, and the long- ranged is not necessarily convex unlike in the thermodynamic bond- orientation order characterise the 2d- solid. limit, and the equation of state may form a stable loop On the other hand, a system of particles interacting due to interfacial free energies. Such a loop was observed via a short- ranged repulsion, e.g., the sterically sta- in the P − η diagram, with a coexistence interval at a bilized colloids, gets into a homogeneous and isotropic density lower than the solid melting point, characteriz- fluid phase at low densities. This phase is devoid of both ing a first order transition [17, 19, 20]. It is only recently the translational and orientational order discussed above. that for hard disks the presence of a continuous melting Apart from these two phases, an intervening third phase, of solid, and the first order transition have been recon- the hexatic phase, is possible in 2d systems. This phase ciled [20, 21]. Ref.20 showed a two- step melting in hard is characterized by a short ranged positional order and a disks, with a continuous (KTHNY) solid- hexatic melt- quasi- long ranged bond- orientational order. ing at η ≈ 0.72, and a first order hexatic- fluid melting The melting of 2d solid attracted an enormous with phase- coexistence in the range of packing fractions amount of attention in literature [4–12]. The Kosterlitz- 0.700 . η . 0.716. Thouless-Halperin-Nelson-Young (KTHNY) theory[9– In hard disks elastic and entropic effects have the same 12] predicted a two stage continuous melting transition origin. For more general potentials, however, they are in 2d, from a quasi- long ranged ordered (QLRO) solid to not strictly related, and other scenarios of phase tran- fluid via an intervening hexatic phase. Within this sce- sition may emerge [20, 22, 23]. Using r−n interactions, nario both the melting are mediated by defect unbinding Ref.23 showed how the boundary of solid to hexatic con- – dislocation unbinding for solid- hexatic melting, and tinuous melting, and the fluid- hexatic coexistence region arXiv:2007.00297v1 [cond-mat.soft] 1 Jul 2020 disclination unbinding for hexatic- fluid melting. In con- depends on the changing values of the steepness of the trast, early simulations showed signatures of a first or- potential, n. At large n the two- step melting, like in der melting of 2d solid [6–8]. The mean field theories hard- disks, survives. At a smaller n, defects become also predicted a first order transition [13–16]. A later ubiquitous and the nature of hexatic changes. For n . 6, and more careful constrained Monte-Carlo simulation of the fluid- hexatic transition becomes continuous. hard disk particles showed presence of signatures of both The numerical implementation of a potential energy of n the form 1/r , involves a cutoff distance rc beyond which particles do not interact. In general, a discontinuity in ∗Electronic address: [email protected] the interaction force is encountered around the cutoff. †Electronic address: [email protected] This could be made small by increasing the cutoff range. ‡ Electronic address: [email protected] However, a change in rc changes the amount of collisions, modifying the value of the pressure [23]. Such an issue At high densities, the initial configuration is chosen to does not arise for the Weeks- Chandler- Anderson (WCA) be a triangular lattice. The system is equilibrated over potential, the repulsive part of the Lennard-Jones poten- 108 steps, following which statistics is collected over a 7 tial, for which the cutoff rc is chosen at the potential further 10 steps. All simulations were performed us- minimum, where the interaction force vanishes [24, 25]. ing massively parallel home-grown codes implemented on Surprisingly, despite its importance in the modeling of Graphics Processing Units (GPU). soft matter systems including polymers, colloids and fluids, [24–28] studies of phase transitions in WCA- system, even in three dimensions, have found only limited atten- III. RESULTS AND DISCUSSION tion [29–31]. A few early attempts with 2d phase transition of the WCA- particles [5, 32, 33] showed a loop in In this section we study the phase transition using the the pressure- density curve, and also formation of topo- equation of state, density fluctuations, and the hexatic logical defects. However, as far as we know, the detailed and solid order parameters. nature of the melting of the WCA- solid remains to be fully understood. In this paper, we consider phase- transitions in the A. Equation of state 2d system of particles interacting via the WCA- potential. Performing large scale molecular dynamics simu- The first evidence of a first order transition comes lations, and careful analysis, we investigate the melting from the pressure-density diagram of the system shown transitions in this system with decreasing density. We in Fig. 1. The thermodynamic pressure is determined find a continuous solid- hexatic melting, followed by a from the molecular dynamics trajectories using the virial first order melting of the hexatic to the fluid phase. The expression solid melting is associated with dislocation unbinding and shows signatures of KTHNY transition. The first order N N ρ X X hexatic melting is associated with a Maxwell’s loop in P = k T ρ + f(r ) · r (1) B 2N ij ij the equation of state, and a clear hexatic- fluid phase i j>i coexistence.

The rest of the paper is organized as follows. In Sec. II where f(rij) is the interaction force and rij is the inter- we describe the model and the numerical simulations. particle separation. At the fixed temperature kBT = The identification of different phases of the system, the 1, the variation of pressure with density is shown in two melting transitions, and formation of topological de- Fig. 1 (a) for three system sizes: N = 1282, 2562 and fects are described in Sec. III. Finally we summarize our 5122. They clearly show Mayer- Wood loops in the equa- main results and conclude in Sec. IV. tion of states [34]. While interfacial free energy associated with phase coexistence in first order transition ex- plains the loop [34], the converse is not always true [35]. II. MODEL AND SIMULATION The interfacial free energy at phase- coexistence is expected to scale as the interface- length, ∆F ∼ N 1/2, We consider a two-dimensional system of N par- in two dimensions. On the other hand, if present, ∆F ticles interacting via the WCA potential U(rij) = in continuous transition gets independent of N for large −12 −6 4[(rij/σ) − (rij/σ) ] + for separation rij < rc = system size, and the equation of state becomes mono- 1/6 2 σ, and U(rij) = 0 otherwise [24]. Here rij denotes tonic [35, 36]. A Maxwell construction on the equa- the separation between i-th and j-th particle. The choice tion of state suppresses the interfacial effect, and gives 2 2 of cutoff separation rc is made such that the repulsion be- the boundary densities ρ1σ = 0.900 and ρ2σ = 0.918 tween particles vanishes at that separation. The energy, for the coexistence interval in system size N = 2562. length and time scales are set by , σ, and τ = σpm/, The interval is only weakly dependent on the system respectively. The mass of the particles is chosen to be size (Fig. 1 (a) ). Integrating the equation of state over m = 1. We use N = 65536 particles in a simulation box the shaded region in Fig. 1 (a) we obtain the interfacial R ρ2 2 of size A = L × L . At a mean density ρ = N/A, a free energy per particle ∆f = ∆F/N = dρ (P/ρ ). x y √ ρ1 triangular lattice√ configuration has a lattice parameter a It shows a scaling form ∆f ∼ 1/ N as in the first order 2 obeying a = 2/ 3ρ. The separation between√ the con- phase transition (see the inset of Fig. 1 (a) ). secutive lattice planes is given√ by ay = 3a/√2. In our We further follow the phase transition utilizing the simulations, we use Lx = Na and Ly = Nay, and coarse- grained vector field of the hexatic bond orienta- ` the periodic boundary condition. The density is con- tional order ψ6(r), and the scalar field of the local num- trolled by changing the box size. We perform molecular ber density ρ`(r). Unless specified otherwise, here and in dynamics simulations using the standard Leap-Frog al- the rest of the paper we present results for a N = 2562 gorithm [27] with step-size δt = 0.001 τ, in the presence system. The fields are obtained by averaging over sub- 2 of a Langevin heat bath characterized by an isotropic systems of size ` = `x × `y with `x = Lx/10 and `y = ` friction γ = 1/τ fixing the temperature T = 1.0 /kB. Ly/10. In obtaining the coarse- grained ψ6(r), we use the

2 FIG. 1: (a) Equation of state: Plot of the thermodynamic pressure as a function of density for three different system sizes N = 1282, 2562, 5122 indicated in the legend. The loops in the equation of state are clearly visible. The horizontal lines 2 2 2 indicate the Maxwell’s equal area construction. The coexisting densities for N = 128 are ρ1σ = 0.899 and ρ2σ = 0.919, 2 2 2 2 2 2 for N = 256 are ρ1σ = 0.900 and ρ2σ = 0.918 and for N = 512 are ρ1σ = 0.901 and ρ2σ = 0.918. The shaded area in the N = 2562 plot corresponds to the interfacial free energy β∆f between the majority and the minority phases. The inset depicts the scaling of the interface free energy with particle number β∆f ∼ N −1/2.(b)–(i) Plots of local hexatic order and density using the N = 2562 system at ρσ2 values indicated above each plot. They show heat- maps of the magnitude of the ` ` coarse-grained hexatic order ψ6(r) for a single configuration (plots labelled with numeral 1) and the scalar density field ρ (r) averaged over 200 configurations ( plots labelled with numeral 2) for the whole system. The coarse graining is performed by 2 ` averaging over sub-systems of size ` = Lx/10 × Ly/10. Superimposed on these fields are shown the orientations of ψ6 denoted by arrows.

k Pn i 6θkj ` hexatic order of each particle ψ6 = (1/n) j=1 e , heat maps. The orientations of the ψ6(r) vector are de- where the angle θkj denotes the orientation of the bond noted by arrows in these figures. The configurations in vector rkj between the test particle k and its topologi- the coexistence interval, as pointed out by the Maxwell cal neighbor j with respect to the x-axis, and n is the construction, are shown in Fig. 1 (d)–(h). number of topological neighbors. For densities above ρσ2 = 0.920 (Fig. 1 (b) and (c)), ` In Fig. 1 (b)–(i) we show the magnitude of the coarse- the magnitude of ψ6(r) remains uniform throughout the ` grained field ψ6(r) for a single configuration, and the cor- system and its orientations remain aligned along the x- responding time-averaged scalar density field ρ`(r) using axis, suggesting a long- ranged hexatic order. The time-

3 (a) (b) (c)

(d) (e) (f)

FIG. 3: Plots of hψqi around a quasi- Bragg peak at densities 2 ρσ = 0.920 (a) and 0.922 (b). Its actual location qA, the position of the largest value of hψqi (red patch), is shifted from the expected peak- position qp for a perfect triangular FIG. 2: The static structure factor for diﬀerent densities of lattice (cross). The shift is larger at lower density. the system as indicated in the legend. hψqi is calculated in the Fourier plane with resolution interval of 0.01σ−1. With reduction of density, the plots show characteristic features of the solid, hexatic, and isotropic ﬂuid phase. 1. Structure factor

The static structure factor, defined as hψqi = −1 PN iq·rj ∗ averaged density field shows little fluctuation. The mean N hρqρ−qi where ρq = j=1 e with ρq = ρ−q is density ρσ2 = 0.918 starts to show appearance of low- shown in Fig. 2 at different densities. It clearly distin- density and low hexatic order droplets in the otherwise guishes between the solid, the hexatic and the fluid phase. ` ` In the solid phase, the six quasi- Bragg peaks in hψ i cap- ordered background of large ψ6(r), and ρ (r) (Fig. 1 (d)). q As the density is decreased further, the system- wide ori- ture the characteristic six- fold symmetry corresponding ` to the underlying triangular lattice. In contrast, in the entational order of hexatic field ψ6(r) starts to dwindle. In Fig. 1 (e) and (f), still a system spanning band of high fluid phase one obtains the characteristic ring structure hexatic order maintaining a high degree of orientational of hψqi capturing the isotropy of the system (Fig. 2 (a)). correlation is observed, coexisting with surrounding low- At the intermediate densities, the intensity modulation density fluid domains characterized by low hexatic or- on the fluid-like ring in hψqi shows a six- fold symme- der and randomized orientations. The associated density try but with broadened peaks (Fig. 2 (b), (c)). This is a field further highlights the coexistence of high and low characteristic of the hexatic phase [37]. density regions, corresponding to the high and low hex- Here, it is important to note that the actual position of atic order, respectively. The curved interfaces between peaks qA in the structure factor hψqi of solid are shifted the hexatic and fluid regions can be clearly seen from from the expected peak- positions of a perfect triangu- 2 these plots. At even smaller density, ρσ = 0.906 in lar lattice, qp = (0, ±2π/ay(ρ)), (±2π/a(ρ), ±π/ay(ρ)). Fig. 1 (g), the largest hexatic cluster can not span the This is shown in Fig. 3, focussing on a single quasi- Bragg system any more. Regions of significantly low hexatic peak. The shift is both in amplitude and orientation, ` order |ψ6(r)| . 0.05, which already started to appear and the amount of shift depends on the system density. at ρσ2 = 0.910, proliferates further at lower densities. Similar shifts were previously observed in other mod- At ρσ2 = 0.900 in Fig. 1 (e), we observe small hexatic els [20, 38]. As it has been pointed out before [38], it is clusters coexisting with disoriented fluid regions having important to use the actual positions of the quasi- Bragg low hexatic order. The local density plot captures the peaks qA, instead of qp, in all calculations involving po- density fluctuations. Finally, at the smallest density of sitional order and correlation. 2 ρσ = 0.890, the whole system shows a loss of hexatic To quantify the solid melting transition, we use the order, and ρ`(r) displays a uniform profile corresponding solid order parameter hψGi, which is an average of hψqi to the homogeneous fluid. over the six quasi- Bragg peaks at G := {qA}. In Fig. 4 we show the variation of hψGi as a function of the mean density of the system. At very high densities, hψGi remains large, and drops sharply near ρσ2 ≈ 0.92 to van- ishingly small values. The fluctuations of the order pa- B. Solid melting 2 rameter quantified by the mean squared deviation h∆ψGi shows a pronounced maximum at ρσ2 = 0.920 (see the in- Having established a coexistence interval of 0.900 ≤ set of Fig. 4 ), identifying the melting point of the solid. ρσ2 ≤ 0.918, here we proceed to investigate the solid Here we emphasize that the melting point ρσ2 = 0.920 melting using the structure factor, solid- order parame- remains above the interval of phase coexistence identified ter, the positional order and its correlation function. in Fig. 1 (a).

4 à 1.0 à à à ç 0.402

2 0.4 à 0.4 10 ´

\ ç 2

L 0.2 ç G ç ç 0.1 ç -1 3

DΨ ~r

H ç X 0.0ç çççççççç ç L r \

H G 0.89 0.90 0.91 0.92 0.93 G Ψ g X 0.2 ΡΣ2 - 10 2 — 0.924 — 0.920 — 0.919 à — 0.918 à à à à à à à à à à 0.0 10-3 10 100 0.89 0.90 0.91 0.92 r Σ ΡΣ2

FIG. 6: Positional correlation function gG(r) for a system of 2 2 FIG. 4: Plot of the solid order parameter (a) and its fluctu- N = 512 particles at densities ρσ specified in the legend. It −ηG 2 ation (inset) as a function of density. The peak of the order shows power law decay r with ηG < 1/3 at densities ρσ > 2 parameter fluctuation appears at a density ρσ2 = 0.920. 0.920. At the solid melting point, ρσ = 0.920, the correlation −ηG shows a power law decay r with an exponent ηG = 1/3, indicated by the dashed line r−1/3. At further lower densities, ρσ2 < 0.920, the correlation shows exponential decay. 10 ì à 0.922 ì ì á 0.920 ì

\ ç 0.919 −ηG õ ì with system size, hψG(`)i ∼ ` . Within the KTHNY L õ õ ∗ L ì õ 0.916 theory, the exponent ηG approaches η = 1/3 from be- H 5 õ G õ G õ ì ì 0.914 low, as the solid approaches the melting point. After Ψ

X õ melting, hψG(`)i is expected to show an exponential de- ì

\ ç ç ç cay with `, characterizing the short ranged order. The L ç ç õ { decay in hψ (`)i with `, as shown in Fig. 5, is consistent H ç G -1 3 G ~ { L ç ì with this picture. At densities higher than the melting

Ψ õ ç 2 X point ρσ = 0.920, hψG(`)i decreases with ` as a power áá 2 á á á ç law with exponent ηG < 1/3. At ρσ < 0.920 the system- àà à à á á H Là à à á size dependence of hψG(`)i shows a stronger exponential à àá 1 ìõçàá decay. 0.1 0.3 0.5 1.0 { L 3. Positional correlation

FIG. 5: Finite size scaling of solid order parameter hψG(`)i shown in log- log scale. Here ` denotes block-sizes. The The change in order is further characterized by the dashed black line is the plot of (`/L)−1/3. iG·rij positional correlation gG(r) = he δ(r − rij)i, where rij = ri − rj is the inter- particle separation vector, rij = |rij|, and G denotes the the reciprocal lattice 2. Finite size scaling vectors corresponding to the six quasi- Bragg peaks. To explore the power law nature of the correlation in The change in the nature of order across the solid melt- QLRO solid over a longer length scale, we use system 2 ing can be examined using a finite size scaling analysis. size N = 512 in plotting gG(r) in Fig. 6. It is plot- We calculate the solid order parameter hψG(`)i over sub- ted at different densities across the melting transition. 2 systems of varying size ` = `x×`y. The solid order calcu- The change in the decay in correlation from power law −ηG 2 lated over the whole system is denoted here by hψG(L)i. gG(r) ∼ r in the QLRO solid phase at ρσ = 0.924, to 2 In Fig. 5 we show plots of hψG(`)i/hψG(L)i with system exponential after melting ρσ ≤ 0.919 is clearly observed. size `/L, at different mean densities ρσ2 across the solid At the melting point ρσ2 = 0.920, the correlation shows melting. The solid order parameter corresponding to the power law decay consistent with the KTHNY prediction QLRO solid phase is expected to show a power law decay r−1/3.

5 8 æ 0.922 æ á 0.920 æ 6 áá ò 0.918 æ ì L 0.916 á G { 4 æ Ψ á H ò

P áò ò ò òìæì òì ì 2 ìòìá ì ìò òæ ìììììòìò æ ì ì òòò á á ì òò ááæ ì ì ò áá æ ò 0 ìæòáìæòáæáòæáòæáæáæáæáæáæáæææ ììæòáìæòáìæòáìæòáìæòá 0.0 0.2 0.4 0.6 0.8 { ΨG

FIG. 8: Probability distribution of the solid order parame- ` 2 ter P(ψG) calculated over sub-systems of size ` = Lx/10 × Ly/10, for different densities as indicated in the legend. The distribution remains unimodal across the solid melting point 2 at ρmσ = 0.920. At a lower density the peak of the distribu- ` FIG. 7: Plots of the positional order χi for densities indicated tion shifts to a lower ψG. above each plot. In the solid phase, ρσ2 ≥ 0.920, the plots show high positional order ((a) and (b)). Immediately below 2 the solid melting point, at ρσ = 0.918, alternating broad 5. Distribution function of solid order bands of parallel and anti-parallel alignment of χi appear in the system (c). As the density is further reduced, these bands ` get narrower and shows up in the form of alternating thin Here we calculate the solid order parameter ψG over 2 stripes. subsystems of size ` = Lx/10 × Ly/10, and obtain ` their distribution function P(ψG) (Fig. 8). The distribution function remains unimodal across the melting point ρσ2 = 0.920. The peak of the distribution shifts to lower values as the solid melts. This unimodal nature of the distribution function signifies the absence of any 4. Local positional order metastable phase on the other side of the transition, a characteristic feature of continuous transitions. The results obtained in this section shows that the melting of solid is a continuous transition, consistent The local positional order can be visualized using with the KTHNY melting scenario. The structure factor i G·r e i , where ri denotes the position vector of i-th parti- shows that the solid melts to a hexatic. The solid melt- cle. This is a unit vector in two dimensions, and plotted ing point is obtained at a density ρσ2 = 0.920 clearly for typical equilibrium configurations in Fig. 7 in terms separated from, and larger than the coexistence interval i G·r of a heat map of the projection χi = e i · ˆx with re- obtained from the Mayer-Wood loop in the equation of spect to the x-axis, ˆx = (1, 0). In calculating this, we use state. the peak position of structure factor in the first quad- rant. The fluctuations of χi is small in the solid phase. The positional order remains mostly oriented over the system size (Fig. 7 (a) and (b)). The appearance of small C. Hexatic melting patches with anti- parallel orientation of positional order is associated with the QLRO nature of the solid. After In Fig. 9, we plot the mean amplitude of the hextic or- PN i 2 melting, the positional order gets short- ranged. This der parameter hψ6i = h | (1/N) i=1 ψ6 | i as a function can be seen from the formation of striped patterns at of density. The hexatic order reduces continuously with ρσ2 = 0.918 (Fig. 7 (c)). The size of the stripes cor- lowering of mean density to vanish near ρσ2 = 0.900. The 2 responds to the correlation length over which the order fluctuations in hexatic order, h∆ψ6i, plotted in the inset remains oriented. The widths of the stripes get narrower of Fig. 9 shows a pronounced maximum at ρσ2 = 0.906, with the reduction of correlation length at lower density, identifying the hexatic melting point. Note that this as can be seen from the plot at ρσ2 = 0.906 in Fig. 7(d). melting point ρσ2 = 0.906 is right inside the coexistence

6 0.8 ç ì à 0.910 3 0.6 ç ìì 10 ì á 0.908 ´ 0.4 5 \ ç ç 2 ç ç ì L

6 ç 0.906 0.6 \ 0.2 ç à ì ç 0.525 à L DΨ àà H çç à õ 0.904 L X çççç à ì 0.0ç çç à H 6 ì 0.902 0.89 0.906 0.92 à

Ψ 3 \ ì 6 0.4 2 à X ΡΣ Ψ \

X à L

{ - à H õõ ~ L 1 4 6 õ { ì õ

à Ψ ç 0.2 ç ç õ X ç ç õ à ç õ à áá á ç á á H L õç àà á á à 0.016 à à à à á õç à à à á 0.0 à 1 à ìõçàá 0.89 0.90 0.91 0.92 0.93 0.1 0.3 0.5 1.0 ΡΣ2 ; { L

FIG. 9: Plot of the bond orientational order parameter and FIG. 10: System size scaling of orientational order parameter its ﬂuctuation (inset) as a function density. The peak of the ψ6 for the densities indicated in the legend. The dashed line 2 order parameter ﬂuctuation appears at a density ρσ = 0.906. shows the power law (`/L)−1/4.

1.0 interval obtained from the Mayer-Wood loop. Thus the hexatic melting is associated with phase coexistence.

1. Finite size scaling ~r-1 4 0.1 L r

We use a finite size scaling analysis to establish the H 6 — 0.922 change in the nature of hexatic order across the hexatic- g — 0.920 melting transition. We calculate the order parameter 2 — 0.910 hψ6(`)i over subsystems of varying size ` = `x×`y. The hexatic order calculated over the whole system is denoted 0.01 — 0.908 by hψ6(L)i. In Fig. 10 we show plots of hψ6(`)i/hψ6(L)i — 0.906 with system size `/L, at different densities. Above the — 0.904 hexatic melting point, ρσ2 ≥ 0.096, the ratio shows — 0.902 −η6 power law decay hψ6(`)i/hψ6(L)i ∼ ` . The single 1.00 10 100 hexatic phase is characterized by such power laws. The r Σ KTHNY theory predicts that η6 approaches 1/4 from below as one approaches the hexatic melting point. Fig. 10 shows apparent consistency with such expectation. At FIG. 11: Plots of the hexatic correlation function g (r) for a lower densities it transforms to exponential decay. 6 system size N = 5122 at densities ρσ2 indicated in the legend. However, the apparent power- law decay at the hex- The dot-dashed line is a plot of r−1/4. atic melting point is not consistent with the evidence of phase coexistence obtained from the Mayer- Wood loop, and the local heat- map of hexatic order param- 2. Hexatic correlation eters in Fig. 1 (a). We note that, in finite sized simulations (N = 2562 here), if hexatic clusters span the system size even at coexistence, hψ6(`)i may display such In Fig. 11 we plot the hexatic correlation g6(r) = i∗ j i∗ an apparent power-law decay governed by the hexatic do- hψ6 ψ6δ(r − rij))i, where ψ6 denotes the complex con- i mains. In large enough systems, the domain boundaries jugate of ψ6, the local bond orientational order and rij between hexatic and fluid phases modfies the behavior is the separation between particles i and j. In the solid leading to exponential tail. In the following, we show phase g6(r) is expected to remain independent of r, due this in the hexatic correlation function, utilizing a larger to a long- ranged hexatic order. The QLRO single hex- −η6 system size. atic is expected to show a power law decay g6(r) ∼ r .

7 Within the KTHNY theory η is expected to approach 6 2.0 1/4 from below as one approaches the hexatic melting ì ì ì point, and after melting g6(r) is expected to show expo- ì ì æ nential decay. à ì æ æ ààà à ææ ì 1.5 ààà àæàà ì àà æ To investigate the power law nature of g6(r), and de- ææ æ æà à æ ææ æà ìæà à ææææ ìììæàæàà viations from it, we use a larger system size of N = ì ìì à 2 L ìì æì 6 512 (Fig. 11). At densities above the solid melting point { ì ììììì 2 Ψ à ρσ = 0.920, the correlation remains unchanged over the H 1.0 æ whole system size, a characteristic of the solid phase hav- P ì 0.908 àì ing long ranged hexatic order. At lower densities, but æ æ 0.906 above the hexatic melting point, the correlations show à 0.5 ì power- law decay characteristic of hexatic phase. At à 0.904 æ 2 à the hexatic- melting transition ρσ = 0.906, note that ì the correlation shows an apparent power-law like decay æà ìæà −1/4 ìæàæà g6(r) ∼ r up to r . 70σ, which crosses over to 0.0 ììæà exponential decay for longer separations r. Thus sim- 0.0 0.2 0.4 0.6 0.8 ulations with system sizes smaller than the above- men- Ψ{ tioned cross- over length, may show apparent power-law 6 decay of g6(r) which could appear to be consistent with KTHNY theory. This we have checked separately for FIG. 12: Probability distribution of the coarse- grained orien- 2 2 ` smaller systems, N = 128 and 256 (data not shown). tational order parameter P(ψ6) at densities indicated in the legend. However, the observed exponential tail in g6(r) of large systems is due to the presence of coexisting hexatic and fluid domains. At densities lower than the hexatic melt- 2 ing point ρσ < 0.906, the correlation shows exponential to fluid phase. We use Voronoi tessellation to identify the decay characterizing the fluid phase. topological neighbors of a given particle. In a perfect triangular lattice, each particle has ν = 6 neighbors. A particle with ν 6= 6 neighbors is identified as a ν-fold defect, 3. Distribution function of hexatic order: phase coexistence e.g., ν = 5- or 7- fold defects. Fluctuations in solid can accommodate 5−7−5−7 bound quartets, corresponding The nature of the hexatic melting is further character- to bound dislocation- anti-dislocation pairs. In the previ- ` ized in terms of the probability distribution P(ψ6) of lo- ous section, we established a continuous melting of solid ` cal heaxtic order denoted here by ψ6 := ψ6(`), calculated to hexatic. Within the KTHNY theory, solid- hexatic 2 over subsystems of size ` = (Lx/10) × (Ly/10). The melting transition is mediated by unbinding of 5 − 7 and distribution function at and around the hexatic- melt- 7 − 5 pairs signifying unbinding of dislocations, line de- ing point ρσ2 = 0.906 clearly show pronounced multi- fects in 2d solid. Within the same theory, hexatic melting modality due to phase coexistence (Fig. 12). The pres- is expected to be continuous and mediated by unbinding ` ence of metastable maximum in P(ψ6) across the hexatic of dislocations to free 5-fold or 7-fold disclinations. In melting is a characteristic of the first order phase tran- contrast, as we have shown the hexatic- fluid melting for 2 ` sition. At ρσ = 0.906, the peak near ψ6 = 0.56 corre- WCA system is a first order transition characterized by ` sponds to the hexatic regions, while that near ψ6 = 0.25 phase coexistence. is due to the coexisting fluid phase. In Fig. 13 we show defect formation along the melt- Thus we have shown that the hexatic melting in WCA ing transitions. Each figure in Fig. 13(a)-(f) shows a system is clearly a first order transition characterized by heat map of the hexatic order for each particle pro- the coexistence of hexatic and fluid phases. The system i jected along the x-axis, hi := ψ6 · ˆx, where the vec- remains in pure hexatic phase in the range of densities, tor ˆx = (1, 0). By definition, for a perfect triangular 0.918 < ρσ2 < 0.920, lower than the solid- melting point, i i i lattice ψ6 = (Reψ6, Im ψ6) = (1, 0). Thus the blue re- and above the upper limit of the coexistence interval. At gions in the heat maps denote large hexatic order h ≈ 1. 2 i lower densities, 0.900 < ρσ < 0.918, the system gets The system shows a high degree of hexatic alignment for into hexatic- fluid coexistence, with the hexatic melting ρσ2 ≥ 0.918 as is shown in Fig. 13(a)-(c). The patches of 2 2 point identified at ρσ = 0.906. Below ρσ = 0.900 the other colors signifying disordered fluid droplets, grow to system gets into the pure fluid phase. significant fraction of the system size at lower densities, see Fig. 13(d)-(f). The interfaces between coexisting domains can be identified by noticing the change in color. D. Topological defects In addition, we have plotted the ν = 4, 5, 7, 8- fold defects on the heat maps of hi in Fig. 13(a)-(f). These are In this section, we discuss defect formation as the sys- shown more clearly by focussing on small regions of the tem undergoes the two stage melting from solid to hexatic system in Fig. 13(g)-(i).

8 i FIG. 13: (a)–(f): Plots of the hexatic orientation hi = ψ6 · ˆx in a region of size 200σ × 200σ. Superimposed on these plots, we show the particles with topological neighbors ν = 4 (blue), 5 (black), 7 (green) and 8 (red). (g)–(i): Close up views of the regions indicated in the panels (a), (c) and (e) are shown to highlight the locations of ν- fold defects. The particles with ν = 6 topological neighbors are indicated by gray. In (g) and (h), the Voronoi tessellation is shown, and the quartets, dislocations and clusters of defect are highlighted by shading the corresponding Voronoi cells with gray, pink, and green, respectively. In (i), strings of defect are observed at the interfaces between hexatic and disordered ﬂuid.

In Fig. 13(g),(h), we show the Voronoi tessellation cor- the plots, spatial positioning of the defects are related to 2 responding to densities ρσ = 0.922, 0.920, respectively. locally low hexatic order hi. Fig. 13(i) magnifies a small We identify the particles with five neighbors in red, and region of Fig. 13(e) corresponding to the hexatic melting those with seven neighbors in blue. Moreover, we use point ρσ2 = 0.906. Fig. 13(i) shows strings of defects shades of gray, green and pink to identify Voronoi cells located predominantly on the interfaces. corresponding to bound quartets, clusters, and disloca- A quantitative analysis of defect formation in the sys- tions (separated 5-7 defect pairs). As can be seen from tem is obtained from the estimated number of different

9 25 tem. Rather, the monotonic increase and dominance in à a à Total Defects b æ the fraction of defect- clusters across the hexatic melt- 20 æ Clusters 40 æ Quartet æ à ing (see Fig. 14(a)) is due to the appearance of strings ì Quartet à Dislocation æ L à + L 15 æ Dislocation ôDisclination % à % æ of defects at the interfaces of coexisting hexatic and ﬂuid H H L à + H H L \ à DisclinationH \ æ domains (Fig. 13(i)), a characteristic of the ﬁrst order

d à à æ c à à à à æ à à æà 10 d

N à æ à 20 æ à X X ì æ æ hexatic- fluid transition. æ à æ æ à ì ì æ à L æ à ì ì æ à 5 ì æ à æ ì ì æ ì æ à à ì ìæ ìæ ìæ ô 0 0 ô ô ô ô ô ô ô ô ô ô ô ô IV. CONCLUSION 0.89 0.90 0.91 0.92 0.89 0.90 0.91 0.92 2 2 ΡΣ ΡΣ In summary, we have presented a detailed study of the melting transitions of a system of particles interacting via FIG. 14: (a) Variation of percentage defect-fractions the Weeks- Chandler- Anderson (WCA) potential, using 2 hNdi (2), hNci (◦), hNsi (), with system density ρσ .(b) The large scale molecular dynamics simulations. With reduc- percentage defect fraction hdci of quartets, dislocations and tion of density, the system shows two- step melting. The disclinations are shown as a function of the system density. solid melts into a hexatic phase via a continuous transition at the density ρσ2 = 0.920. This melting is associated with unbinding of dislocations, like the solid- melt- types of defects by keeping track of particles with ν 6= 6 ing within KTHNY theory. The system remains in the neighbors. In Fig. 14(a) we show the mean percentage single hexatic phase only up to ρσ2 = 0.918. The hexatic 2 fraction of defects hNdi = (1−hn6i/N)×100 as a function melts into a fluid at ρσ = 0.906 via a first order phase of density, where at each density the averaging is done transition, characterized by hexatic- fluid coexistence, over 200 independent configurations. This gives the es- which is unlike the KTHNY prediction. This melting timate of the percentage fraction of all the defects. In is associated with formation of large strings and clusters addition, we separately consider hNsi, a similar percent- of defects located mainly at the boundaries of coexisting age fraction of a total sum of quartets, dislocations and domains. A wide range of densities 0.900 ≤ ρσ2 ≤ 0.918 disclinations taken together. Further, we separately con- shows this coexistence. Finally at densities ρσ2 < 0.900 sider hNci, the percentage fraction of clusters and strings the system gets into a single fluid phase. of defects. The quartets are not counted in clusters. Vari- The solid melting is characterized using the structure ation of all theses defect fractions, hNdi, hNsi and hNci factor, solid order parameter, positional order and corre- with density are shown in Fig. 14(a). We clearly observe lation function. The probability distribution of the local that above the melting point of the solid, ρσ2 = 0.920, solid order remains unimodal across the transition, show- the percentage fraction of all these defects are negligi- ing consistence with the continuous melting. Finite size bly small. They increase with reduction of density, and scaling of order parameter, and the behavior of the corre- in the fluid phase, hNdi increases by almost an order of lation function gG(r) shows consistency with the KTHNY −1/3 magnitude. The plot shows hNci > hNsi that clusters prediction, particularly a power law decay gG(r) ∼ r and string of defects dominate all through. at the solid- melting point. The hexatic melting is stud- Finally, to get a better insight into the relative role of ied using the bond- orientational hexatic order param- specific kind of defects in the phase transitions, we exam- eter. The probability distribution of local hexatic order ine what is the contribution of quartets, dislocations and shows bimodality across transition, due to the presence of disclinations in the total defect fraction hNdi. We esti- metastable state across the first order melting. The cor- mate this using the quantity hdci = hnd/NNdi × 100, relation function of hexatic order g6(r) shows a power- where nd denotes the number of defects contributing law decay with r within the hexatic phase, capturing its to quartets, dislocations and disclinations, respectively. quasi- long- ranged nature of order. However, at the In Fig. 14(b) we show their variation with density. At hexatic melting point g6(r) shows exponential tail due high density solid phase the fraction of quartets domi- to the coexistence of hexatic and fluid, a behavior un- nate. This fraction decrease monotonically with decreas- like the KTHNY prediction. The Mayer- Wood loop in ing density. In contrast, the dislocation fraction remains the equation of state is due to the phase- coexistence at low in the solid, and increases with decreasing density to first order transition, and identifies the coexistence inter- saturate to ∼ 20%. The increase (decrease) in dislocation val. The solid melting point remains clearly at a density fraction (quartet fraction) across the solid melting point higher than this interval. ρσ2 = 0.920 is consistent with the KTHNY picture of dis- Thus we established a continuous solid- hexatic melt- location mediated melting, in which bound dislocation- ing followed by a first order hexatic- fluid melting in the anti-dislocation pairs (quartets) are expected to unbind WCA system. The detailed analysis of different defect forming free dislocations. On the other hand, unlike the types and visualization of their locations showed that the KTHNY melting of hexatic, disclinations, the fraction of continuous solid- melting is associated with dislocation which remains low all through, play no significant role unbinding, whereas strings of defects that localize near in the first order hexatic melting of the 2d WCA sys- the hexatic- fluid domain boundaries dominate the first

10 order hexatic melting transition. port through grant numbers MTR/2019/000750 and EMR/2016/001454. This research was supported in part by ICTS during a visit for participating in the pro- Acknowledgements gram - 7th Indian Statistical Physics Community Meet- ing (Code: ICTS/ispcm2020/02). D.C. thanks ICTS-TIFR, Bangalore, for an as- sociateship, and SERB, India for ﬁnancial sup-

[1] N. D. Mermin and H. Wagner, Phys. Rev. Lett. 17, 1133 lation: from algorithms to applications (Academic Press, (1966). NY, 2002). [2] N. D. Mermin, Phys. Rev. 176, 250 (1968). [28] M. Kroger, Phys. Rep. 390, 453 (2004). [3] B. I. Halperin, J. Stat. Phys. 175, 521 (2019). [29] A. Ahmed and R. J. Sadus, Phys. Rev. E 80, 061101 [4] H.-H. von Grünberg, P. Keim, and G. Maret, Soft Mat- (2009). ter (Wiley-VCH Verlag, Weinheim, Germany, 2014), [30] A. De Kuijper, J. A. Schouten, and J. P. Michels, J. Vol. 118, pp. 41–86. Chem. Phys. 93, 3515 (1990). [5] M. A. Glaser and N. A. Clark, Adv. Chem. Phys. 83, 543 [31] S. Hess, M. Kröger, and H. Voigt, Physica A 250, 58 (1992). (1998). [6] B. J. Alder and T. E. Wainwright, Phys. Rev. 127, 359 [32] S. Toxværd, Physical Review Letters 51, 1971 (1983). (1962). [33] M. A. Glaser, N. A. Clark, A. J. Armstrong, and P. D. [7] J. Lee and K. J. Strandburg, Phys. Rev. B 46, 11190 Beale, Springer Proceedings in Physics: Dynamics and (1992). Patterns in Complex Fluids (Springer-Verlag, Berlin, [8] J. A. Zollweg and G. V. Chester, Phys. Rev. B 46, 11186 1990), Vol. 52, p. 141. (1992). [34] J. E. Mayer and W. W. Wood, J. Chem. Phys. 42, 4268 [9] J. M. Kosterlitz and D. J. Thouless, J. Phys. C 6, 1181 (1965). (1973). [35] J. J. Alonso and J. F. Fernández,Phys. Rev. E 59, 2659 [10] B. I. Halperin and D. R. Nelson, Phys. Rev. Lett. 41, (1999). 121 (1978). [36] J. Lee and J. M. Kosterlitz, Phys. Rev. B 43, 3265 (1991). [11] D. R. Nelson and B. I. Halperin, Phys. Rev. B 19, 2457 [37] P. M. Chaikin and T. C. Lubensky, Principles of Con- (1979). densed Matter Physics (Cambridge University Press, [12] A. P. Young, Phys. Rev. B 19, 1855 (1979). Cambridge, 2012). [13] T. Ramakrishnan and M. Yussouff, Phys. Rev. B 19, [38] Y.-W. Li and M. P. Ciamarra, Phys. Rev. E 100, 062606 2775 (1979). (2019). [14] A. R. Denton and N. W. Ashcroft, Phys. Rev. A 39, 4701 (1989). [15] X. C. Zeng and D. W. Oxtoby, J. Chem. Phys. 93, 2692 (1990). [16] V. N. Ryzhov and E. E. Tareyeva, Phys. Rev. B 51, 8789 (1995). [17] S. Sengupta, P. Nielaba, and K. Binder, Physical Review. E 61, 6294 (2000). [18] P. Nielaba, K. Binder, D. Chaudhuri, K. Franzrahe, P. Henseler, M. Lohrer, A. Ricci, S. Sengupta, and W. Strepp, J. Phys. Condens. Matter 16, S4115 (2004). [19] A. Jaster, Phys. Rev. E 59, 2594 (1999). [20] E. P. Bernard and W. Krauth, Phys. Rev. Lett. 107, 155704 (2011). [21] M. Engel, J. A. Anderson, S. C. Glotzer, M. Isobe, E. P. Bernard, and W. Krauth, Phys. Rev. E 87, 042134 (2013). [22] A. Hajibabaei and K. S. Kim, Phys. Rev. E 99, 022145 (2019). [23] S. C. Kapfer and W. Krauth, Phys. Rev. Lett. 114, 035702 (2015). [24] J. D. Weeks, D. Chandler, and H. C. Andersen, J. Chem. Phys. 54, 5237 (1971). [25] D. Chandler, J. D. Weeks, and H. C. Andersen, Science 220, 787 (1983). [26] G. S. Grest and K. Kremer, Phys. Rev. A 33, 3628 (1986). [27] D. Frenkel and B. Smit, Understanding molecular simu-