Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

THE JOURNAL OF CHEMICAL PHYSICS 133, 174502 2010

Evidence for a simple monatomic ideal glass former: The thermodynamic glass transition from a stable liquid phase
Mns Elenius,1 Tomas Oppelstrup,1,2,a and Mikhail Dzugutov3
1 2

Department of Numerical Analysis, Royal Institute of Technology, 100 44 Stockholm, Sweden Lawrence Livermore National Laboratory, 7000 East Avenue, Livermore, California 94551, USA 3 Department of Materials Science and Engineering, Royal Institute of Technology, 100 44 Stockholm, Sweden

Received 20 May 2010; accepted 5 September 2010; published online 1 November 2010 Under cooling, a liquid can undergo a transition to the glassy state either as a result of a continuous slowing down or by a rst-order polyamorphous phase transition. The second scenario has so far always been observed in a metastable liquid domain below the melting point where crystalline nucleation interfered with the glass formation. We report the rst observation of the liquid-glass transition by a rst-order polyamorphous phase transition from the equilibrium stable liquid phase. The observation was made in a molecular dynamics simulation of a one-component system with a model metallic pair potential. In this way, the model, demonstrating the thermodynamic glass transition from a stable liquid phase, may be regarded as a candidate for a simple monatomic ideal glass former. This observation is of conceptual importance in the context of continuing attempts to resolve the long-standing Kauzmann paradox. The possibility of a thermodynamic glass transition from an equilibrium melt in a metallic system also indicates a new strategy for the development of bulk metallic glass-forming alloys. 2010 American Institute of Physics. doi:10.1063/1.3493456
I. INTRODUCTION

Avoiding crystallization when cooling a liquid toward the glass transition remains a central problem of glass science. Periodic phases are commonly thought to be thermodynamically favored at sufciently low temperatures by all substances. The relaxation time of a liquid within its domain of thermodynamic stability above the melting temperature Tm is usually much smaller than the value of 102 s that denes the glass transition temperature Tg.1 The glass-forming ability of a liquid is measured in terms of the minimum cooling rate that avoids the interfering crystallization. Besides the thermodynamic factors, viscosity is the key parameter that determines the rate of crystal nucleation and growth within the supercooled liquid domain Tg T Tm. The glassforming ability of a liquid is therefore determined by its viscosity at Tm.2 According to a popular empirical rule,3 Tg / Tm 2 / 3. This implies that the highest viscosity at Tm is expected in liquids with Arrhenius temperature dependence of viscosity, labeled strong in the strong-fragile classication.4 Indeed, good glass formers are commonly found to be strong liquids with high viscosity at Tm, and the current strategy in the design of bulk metallic glass formers5 is focused on the alloys closely approximating the Arrhenius behavior. Another possible approach to the problem of avoiding crystallization in the glass transition is to defy the 2/3 rule and attempt to reduce Tm relative to Tg. For that purpose, the free energy of the equilibrium liquid needs to be reduced relative to that of the respective crystal. In nonpolymeric systems, this can be achieved by tuning anisotropic
a

Author to whom correspondence should be addressed. Electronic mail: tomaso@nada.kth.se.

interaction,6 by a judiciously designed pair potential,7 or by composing multicomponent mixtures with strong chemical order and deep eutectic minima.9,10 The ideal glass former11 is dened as a liquid that remains in a stable thermodynamic equilibrium when attaining Tg upon cooling, which entirely excludes the possibility of its crystallization regardless of the rate of cooling. This class of glass formers has so far been found to include atactic polymers12 and some aqueous solutions of electrolytes.8 A single-component nonpolymeric liquid that does not crystallize upon cooling has so far never been found, and the question of its existence is both intellectually challenging and of a signicant technological interest, particularly for the area of bulk metallic glasses. Crystallization of a liquid upon cooling can be precluded by a direct rst-order polyamorphous phase transition to a glassy state. A system that performs such a transition from the stable equilibrium liquid state would be an ideal glass former. A possible thermodynamic singularity, transforming the supercooled liquid into a distinct low-entropy glassy phase, was discussed by Kauzmann13 as a feasible resolution of his paradox. The Gaussian model14 conjectures that a liquid-liquid rst-order phase transition, accompanied by a crossover from fragile to strong behavior, is a generic feature of fragile liquids. This transition is expected to occur both below and above Tg.15 Formation of a glassy phase as a result of a thermodynamic transition has so far always been observed below Tm,16,17 where the presence of critical uctuations due to the liquid-liquid instability signicantly enhances the rate of crystalline nucleation.18,19 Moreover, a rst-order polyamorphous transition from the equilibrium liquid state has never been observed in a monatomic system upon cooling.20 The Jagla soft-core pair potential21 produces two stable liquid phases. However, because of a signicant
2010 American Institute of Physics

0021-9606/2010/133 17 /174502/7/$30.00

133, 174502-1

Downloaded 07 Dec 2010 to 213.89.193.248. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

174502-2

Elenius, Oppelstrup, and Dzugutov

J. Chem. Phys. 133, 174502 2010

1.5 1 0.5 0 0.5 1 0

Z2 LJ

0.5

1.5 Distance

2.5
Ref. 24 , com-

FIG. 1. Pair potential used in the present simulation Z2 pared with the Lennard-Jones LJ potential.

difference in densities, the transition between these phases has only been found under compression and not upon cooling.22,23 Here, we report a molecular dynamics simulation that provides evidence for the existence of a simple monatomic ideal glass former. The simulation explores a one-component model liquid with a prototype metallic pair potential. It is found that when cooled, the liquid performs a rst-order polyamorphous phase transition to a low-entropy state, while remaining stable with respect to crystallization. The lowentropy state has a mesoscopic-range order and the rate of structural relaxation characteristic of the glassy state. The phase transition is found to be accompanied by the fragile to strong crossover. We discuss some conceptual aspects of this nding and its possible implications for the technology of bulk metallic glasses.
II. COMPUTER MODEL

The potential was designed to imitate effective interionic interaction potentials in liquid metals with characteristic Friedel oscillations.25,26 Earlier investigations of the energy minima congurations of this model24 revealed its strong predisposition for tetrahedral ordering. When simulated in the liquid state under cooling, the model demonstrated a pronounced super-Arrhenius behavior of the diffusivity24 as well as other dynamical anomalies characteristic of fragile glass formers.27,28 The main part of the simulation was performed using a system of 128 000 particles. In some simulation runs, we have also used a smaller system of 3456 particles, in order to test possible size-dependence of the observed phase behavior. All the quantities we report here are expressed in terms of reduced simulation units of length and time energy which are dened by the pair potential, with the particle mass assumed to be unity. We note that the main repulsive part of the present pair potential closely approximates that of the Lennard-Jones potential. Considering the latter as a model potential of argon,29 and the mass of the argon atom as a unit of mass, our reduced units can be interpreted in terms of standard macroscopic units. In this interpretation, the reduced units of length and time can be estimated as 0.34 nm and 2.16 ps, respectively.
III. RESULTS

Energy

The results we report here were produced in a molecular dynamics simulation of a simple one-component system. The simulation utilized an earlier reported pair potential24 named Z2 in that reference ; it is shown in Fig. 1. The functional form of the potential energy for two atoms separated by the distance r is V r =a for 0 r e r cos 2k f r + b r3 r = 0.348, n = 14.5, V0 = 0.133 915 . . . , rc = 2.644 877 . . . .
n

+ V0

rc and 0 otherwise. The parameters are as follows:

a = 1.04, = 0.33, k f = 4.139, b = 4.2 106

The values of V0 and rc are chosen so that U rc = U rc = 0. In Ref. 24, b is reported as 4.2 107. This is a typo. The correct value is b = 4.2 106, as displayed in the table above.

To investigate the glass-forming behavior of this model, we cooled it, using the large system of 128 000 particles, from the high-temperature thermodynamically stable liquid state. The cooling was performed isochorically at the number density = 0.85. The liquids temperature T was reduced in a consecutive stepwise manner, with a comprehensive equilibration at each temperature step. The equilibration run-times increased with cooling ranging from 103 to 104, in reduced units. The variation of the systems enthalpy upon cooling, presented in Fig. 2 a , exhibits a discontinuous drop at T = 0.72. The resulting thermodynamic state was reheated in the same consecutive stepwise as cooling with a comprehensive equilibration at each temperature point. This was found to produce a pronounced hysteresis in the enthalpy variation conned between T = 0.72 and T = 0.78. In both points of the thermodynamic discontinuity, the system was fully equilibrated within the run-time of about 106, in reduced units. That equilibration time-scale was used in cooling the system below T = 0.72 to T = 0.68. Figure 2 b compares the structures of the two phases separated by the observed phase transition. These are presented in terms of the respective structure factors S Q = Q Q , where Q is the Q-component of the Fourier-transform of the systems number density. The apparent close proximity of the two structures indicates that the observed rst-order transition is a polyamorphous transition connecting two liquid phases with a large degree of similarity in the local order. These phases will be referred to as the high temperature liquid HTL and the low temperature liquid LTL . We also note that the pronounced split in second peak of S Q featured by both liquid phases indicates that the local order is predominantly tetrahedral.30

Downloaded 07 Dec 2010 to 213.89.193.248. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

174502-3

A simple monatomic ideal glass former


8 7 6 5 S(Q) 4 3 2 1 0.7 0.72 0.74 0.76 0.78 0.8 0 0 0.9 5 4.8 4.6 P 4.4 4.2 4 3.8 T=0.72 0.8 0.82 0.84 0.86 0.88 0.9 0.8 5 10 Q

J. Chem. Phys. 133, 174502 2010

7.2 7 6.8 6.6 6.4 6.2 6 5.8 5.6 H

Heating Cooling

15

20

T=0.78 T=0.75

S(Q)

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0

0.845 0.86 0.87 0.88 0.89

0.5

1.5

FIG. 2. Liquid-liquid phase transition. The data represent a system of N = 128 000 particles, unless indicated otherwise. a Isochoric temperature variation of the enthalpy H at density = 0.85. Circles, cooling; squares, heating. b The structure factors S Q of the HTL at T = 0.78 dashed line and LTL at T = 0.68 solid line , both at the density = 0.85. c Isotherms crossing the region of the liquid-liquid transition. Open squares: HTL. Filled circles represent the = 0.85 isochore shown in a . Open circles and crosses: LTL points obtained by isothermal compression and expansion, respectively. Small dots: LTL states simulated by the system of N = 3456 particles. The lines are included as a guide to the eye. d Density variation of the low-Q behavior of the structure factor along the T = 0.78 isotherm for the indicated densities.

We now explore the transition domain in the P phase diagram at constant T. Figure 2 c shows the isotherms produced by heating, compression, and/or expansion of the original thermodynamic states produced by the phase transition from the HTL phase as a result of its isochoric cooling at = 0.85. All the data shown in the plots represent comprehensively equilibrated states. At each point, the system was observed in equilibrium within run-times ranging from 105 to 106, in reduced units, which signicantly exceeded the systems relaxation time as assessed from the decay of the density-density correlations. That the system attained equilibrium is also conrmed by the reversibility of the isothermal compression-expansion that can be observed in Fig. 2 c . The isotherms exhibit density regions of innite compressibility, indicating that there exists a domain of spinodal instability interposed between the two liquid phases of distinctly different densities. We also tested the size-dependence of the observed phase behavior by exploring the small system of 3456 particles along the T = 0.75 isotherm. No difference in the general pattern of phase behavior was observed. At the same time, the isotherm for the smaller-size system expectedly demonstrates upon compression a strong metastability. Innite compressibility in the spinodal domain manifests in diverging long-wavelength limit of S Q , a measure of the density uctuations on the system-size scale. In a macroscopic system, S 0 is related to the isothermal compressibility T by the compressibility equation: S 0 = kBT T.29 Fig-

ure 2 d shows the density-dependence of the small-Q behavior of S Q along the T = 0.78 isotherm. A trend for divergence of the small-Q limit of S Q is clearly visible at the densities where the spinodal instability was detected from the respective isotherm in Fig. 2 c . Note that there is no divergence in the small-Q limit of S Q for the LTL in Fig. 2 b , indicating that at T = 0.68, = 0.85 it is below the spinodal domain. As we mentioned above, the presence of a liquid-liquid spinodal in the supercooled metastable liquid domain greatly enhances the crystalline nucleation18,19 by reducing the respective free-energy barrier. Therefore, our observation that the two liquid phases coexist in equilibrium may be regarded as an indication that, within the liquid-liquid coexistence domain, the system remains thermodynamically stable with respect to a possible crystallization. Figure 3 presents a real-space picture of the spinodal decomposition of a system of 128 000 particles along the T = 0.78 isotherm. The plots depict cross-sections of the coarse-grained spatial distribution of energy at different densities. At = 0.84 representing the HTL phase, the energy is distributed uniformly. At = 0.87, precipitation of the LTL phase appears as distinct large-scale low-energy domains. Upon further isothermal compression, the LTL domains grow and eventually percolate as the system leaves the spinodal domain. This is indicated by the reduction of the isothermal compressibility that can be detected from the increase in the

Downloaded 07 Dec 2010 to 213.89.193.248. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

174502-4

Elenius, Oppelstrup, and Dzugutov

J. Chem. Phys. 133, 174502 2010

d c

FIG. 3. The coarse-grained energy distribution in a system of 128 000 particles at the thermodynamic states indicated in the plots. The distribution is shown in a plane crossing the simulation box parallel to the XZ axes. Each distribution has been averaged over 103 time steps.

slope of the isotherms shown in Fig. 2 c . Respectively, the HTL domains shrink and become disconnected. The described transformation of the pattern of phase distribution can arguably account for the development of a low-Q prepeak arising in the respective S Q in Fig. 2 d . This observation is consistent with the conjecture31 that a low temperature phase produced by a polyamorphous transition under cooling must be intrinsically heterogeneous. In order to get further insight into the nature and the microscopic mechanism of the observed phase transition, we performed a detailed analysis of the underlying structural transformation. To remove thermally induced structural uctuations, the investigated liquid congurations were subjected to the steepest descent energy minimization. Because of the generally tetrahedral structure of both phases, statistical geometry of the local order has been analyzed in terms of tetrahedral and icosahedral congurations of the nearest neighbors. These congurations were identied using an earlier developed algorithm.7 All the particles comprising the rst peak of the radial distribution function g r were assumed to be the nearest neighbors. Because of the narrow and sharp shape of the peak, this denition of neighbors involved no signicant ambiguity with respect to the cutoff distance. The structure analysis revealed a remarkable peculiarity of this phase transformation. The total number of icosahedra in the LTL phase was found to be by about 30% smaller than in the HTL phase, whereas the statistics of tetrahedral order remained generally unchanged. The transformation of the local order resulting in this counterintuitive effect can be understood in detail by inspecting a characteristic LTL cluster shown in Fig. 4 a . The cluster was identied within a low temperature domain observed in Fig. 3 that we concluded to be a LTL precipitation. It can be described as a vefold tetrahelical conguration composed of axially stacked pentagonal bipyramides. However, the bipyramides pentagonal symmetry appears to be broken, as shown in Fig. 4 b . This results in a linear strain of nontetrahedral defects that involves two adjacent helical lines of atoms, marked as blue and green in Fig. 4 a . Such a defect reduces the energy of the cluster by removing the geometric frustration inherent to the vefold packing of tetrahedra. Moreover, we found that

FIG. 4. Structure of the LTL. a A representative tetrahelical conguration discerned in a low-energy region observable in Fig. 2 for = 0.87. The colors are a guide to the eye distinguishing the six constituent lines of atoms. b A fragment of the conguration shown in a demonstrating a defect in pentagonal packing of tetrahedra. The colors are the same as in a . c Aggregation of the tetrahelical cluster shown in a green with a similar cluster blue . d The linear strain of nontetrahedral defects interfacing the two tetrahelical congurations is shown in c using the same colors. Atoms fullling the neighbor condition as described in the text are connected by bonds.

these linear defects facilitate aggregation of the tetrahelical clusters into an extended tetrahedral network. Figure 4 c shows two clusters which share a line of helically stacked fourfold defects as depicted in Fig. 4 d . In this way, the observed linearly shaped tetrahedral aggregations, apparently incompatible with periodic order, can form under cooling an extended stable long-lived network, which is expected to result in a dramatic increase in the viscosity. The apparent inability of these aggregations to ll the space uniformly renders the LTL phase intrinsically heterogeneous. We next investigate the evolution of the systems dynamics upon isochoric cooling at = 0.85. Figure 5 shows an Arrhenius plot of diffusivity within the entire range of temperatures explored in this simulation. Within the HTL domain, the temperature variation of systems diffusion is of a perfectly Arrhenius kind above the crossover temperature TA = 1. Below it, the diffusion clearly demonstrates a fragile behavior, which can be described by the VogelTamman Fulcher equation. The temperature variation of the LTL diffusivity below the spinodal transition domain appears to be perfectly exponential. This enables us to conclude that the observed polyamorphous phase transition upon cooling is accompanied by a fragile to strong crossover similar to that observed in other glass formers.16,32 We note that the diffusivity does not signicantly change across the phase transition, which is in contrast to the polyamorphous glass transition in supercooled silicon16 where it was found to drop by more than two orders of magnitude. This difference in behavior between the two systems can presumably be attributed to the structural heterogeneity of the LTL phase that was observed in Fig. 3, with the faster diffusing part of the system being conned to the high-energy uid domains. The data presented in Fig. 5 also demonstrate that the LTL diffusion rate is independent of the system-size. This made it possible to use the small system to explore the low-T diffusion within affordable computer time.

Downloaded 07 Dec 2010 to 213.89.193.248. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

174502-5

A simple monatomic ideal glass former

J. Chem. Phys. 133, 174502 2010


T = 0.60 T = 0.65
0.4 0.3 F(Q,t) 0.2 0.1 0 0 0.5
5

10 10 10 10 10 10 10

1 F(Qpp,t) / S(Qpp) 0.8 0.6 0.4 0.2


10
3

t=0 4 t = 10 6 t = 510

1.5

0 4 10

10

10

10

1.3

1.4

1.5

0.5

1.5 1/T

2.5

FIG. 6. Main panel: the intermediate scattering function F Q pp , t , Q pp being the position of the low-Q prepeak of S Q , for the LTL at = 0.85 and the two indicated temperatures. Inset: the time evolution of F Q , t in the small-Q domain.

Simulation data VFT fit

10 D

10

10

0.4

0.6

0.8

1/T

1.2

1.4

1.6

1.8

FIG. 5. top Arrhenius plot of the diffusivity at the density = 0.85. Crosses, HTL Ref. 24 . Open circles: LTL for the system of N = 128 000 particles. Dots and open squares: LTL for the system of N = 3456 particles, under cooling and heating, respectively. The inset shows an enlargement of the transition area. The straight lines indicate Arrhenius ts to the data. bottom HTL diffusion. Circles: simulation results; solid line: the Vogel FultcherTamman equation D = D0 exp BT0 / T T0 , with D0 = 0.12, B = 3.7, and T0 = 0.386.

The structural relaxation is commonly discussed in terms of dissipating density-density correlations as measured by the intermediate scattering function F Q , t = Q , t Q , 0 . The systems approach to the ergodic equilibrium is thus controlled by the slowest dissipating component of F Q , t , within the structurally relevant range of Q. The initial part small t of the F Q , t decay can be approximated29 as 1 F Q , t / S Q TQ2 / S Q t2. We, therefore, expect the small-Q prepeak of S Q that can be observed in Fig. 2 b to be the slowest dissipating structural feature of the LTL phase. This conclusion is conrmed by the results presented in the inset of Fig. 6. The main panel of Fig. 6 shows the time variation of F Q pp , t , Q pp being the position of the prepeak, for T = 0.65 and T = 0.6. For both temperatures, the relaxation times clearly exceed by several orders of magnitude the time interval shown in the plot which, in terms of argon interpretation of the reduced units of time, correspond to 6.5 105 s. This makes it possible to conclude that the LTL is a structurally arrested glassy state.

We have earlier concluded that the density uctuations giving rise to the prepeak featured by the structure factor of the LTL phase correspond to the spatial distribution of the low-energy and high-energy domains like those shown in Fig. 3. The observed nondissipation of S Q pp implies that the domain pattern gets frozen as the system is cooled below the spinodal. To support this conclusion, we present in Fig. 7 the evolution of the coarse-grained distribution of the potential energy in the LTL phase at T = 0.65 and T = 0.60. For each temperature, the time interval separating the two depicted distributions corresponds to that spanned by the respective F Q pp , t shown in Fig. 6. The apparent time invariance of the general patterns of energy distribution indicates that the tetrahedrally ordered low-energy clusters like those shown in Fig. 4 form a percolating network which remains topologically unchanged on the explored time-scale. Figure 8 shows a cluster discerned within a low-energy domain at T = 0.65. The cluster, percolating through the simulation box, represents a conguration composed of face-sharing tetrahedra. Due to this structure, it is expected to possess a great degree of mechanical stability. We note that the low-dimensional manner of the tetrahedra aggregation forming the cluster enables its unlimited growth, avoiding accumulation of the geometric frustration that limits bulk packing of tetrahedra.30,33
0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4
FIG. 7. Top Two cross-sections of the coarse-grained energy distribution at = 0.85 and T = 0.65 separated by the time interval of 0.5 107. Bottom The same as in top panel but for T = 0.6 and the separating time interval of 3 107.

Downloaded 07 Dec 2010 to 213.89.193.248. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

174502-6

Elenius, Oppelstrup, and Dzugutov

J. Chem. Phys. 133, 174502 2010

FIG. 8. A cluster composed of face-sharing tetrahedra detected within a low-energy domain of the conguration shown in the top panel of Fig. 7. The variation of color indicates the variation of Z-coordinate. The cluster includes 10 456 atoms.

IV. DISCUSSION

The general picture of the thermodynamic liquid-glass transition we report here can be summarized as follows. Upon cooling the system at = 0.85 from the equilibrium liquid state, a domain of spinodal instability is observed below T = 0.72 where two structurally and thermodynamically distinct liquid phases, HTL and LTL, coexist in equilibrium; this is shown in Figs. 2 and 3. Upon further cooling, the LTL clusters grow and percolate around T = 0.68, and, below that temperature, the estimated rate of structural relaxation for the percolating LTL network is found to be smaller than the value that denes the glass transition. Thus, we observe a thermodynamic transition transforming an equilibrium liquid through an equilibrium coexistence domain into the glassy state. The system thus remains in equilibrium within the entire temperature range above the glass transition. Moreover, based on earlier results,18,19 the apparent equilibrium coexistence of the two phases may be viewed as an indication that the system remains thermodynamically stable within the coexistence domain. This makes it possible to classify the present model as a candidate for an ideal glass former. The structure analysis demonstrates that the observed polyamorphous transformation is driven by the pronounced predisposition of this model for tetrahedral local ordering. Based on the earlier results,24 this behavior can be accounted for by the narrow and deep rst minimum of the pair potential restricting the variation in the rst-neighbor distance. Breaking the pentagonal symmetry of the tetrahelical clusters by creating a linear strain of octahedral defects as shown in Fig. 4 introduces an elegant novel way of avoiding the geometric frustration inherent to the tetrahedral packing.30,33 It results in the formation of a network of stable low-energy congurations of perfect tetrahedral order, dramatically extending the range of structural coherence and thereby lowering the entropy. We note that the extended range of structural ordering within the tetrahedrally ordered clusters of the LTL phase gives rise to the anomalously high main peak of S Q , shown in Fig. 2 b . This structural peculiarity, unusual for

simple liquids, can be compared with the structure of mesophases, e.g., those formed by proteins.34 It is of interest to mention that, as a possible resolution of his entropy paradox, Kauzmann conjectured13 the existence of some kind of state of high order for the liquid at low temperature which differs from the normal crystalline state. The arrest of the structural relaxation in the LTL phase is apparently caused by the formation of a percolating network of linearly shaped low-energy domains of tetrahedral order which is presented in Figs. 7 and 8. This can be compared with the glass transition in polymeric ideal glass-formers which too was found to occur as a result of a percolation transition.12 There are reasons to consider this mechanism of classication as generic. It was argued35 that the development of the spatial heterogeneity commonly observed in a fragile liquid attaining the glass transition7,36 resembles the process of equilibrium polymerization, with the respective percolation point hidden below Tg. The percolation transition we observe in this simulation is also reminiscent of gelation in colloids forming a percolating network of linear clusters composed of tetrahedra.37 We mention that a gelationlike percolation transition was observed in the present system at low densities.38 The structural heterogeneity of the LTL phase and its relatively high diffusion rate distinguish it from the typical inorganic glasses forming continuous tetrahedral networks. A feasible conjecture is that these distinctions of the present glass may possibly render it ductile and capable of selfrepair, in contrast to the well-known brittleness of the inorganic glasses. This possibility is of signicant technological interest and deserves further investigation. The metallic nature of the interaction potential used in this model, incorporating characteristic Friedel oscillations, suggests a possible new approach to the design of bulk metallic glass formers. The efforts have so far been focused on the search for eutectic mixtures of atomic species with a strong mismatch in size. This is assumed to increase the packing density and create a strong chemical ordering, thereby reducing the free energy of the melt relative to the respective crystal and inducing strong liquid behavior.5,10 The present result demonstrates an alternative way of glass formation in metallic systems where the entropy of a fragile liquid is reduced discontinuously by a polyamorphous phase transition from the equilibrium stable liquid state driven by a purely topological structural ordering. As a relevant observation, we mention a strong-fragile transition with an indication for hysteresis found in a bulk metallic glass-former.39 Although the pair potential used in this model is quite realistic in describing interionic interaction in a hypothetical single-component metallic liquid, no elemental metal reproduces it with a sufcient accuracy. Nevertheless, it is possible to suggest a feasible strategy in the search for a real metallic system with the pattern of phase behavior as we reported here. The Friedel oscillations26 like the one incorporated in the present pair potential are ubiquitous in effective interionic potentials conjectured for liquid metals.25 These oscillations are controlled by the electron density which is a function of the composition of a metallic alloy. Therefore, it is conceivable that by appropriately manipulat-

Downloaded 07 Dec 2010 to 213.89.193.248. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

174502-7

A simple monatomic ideal glass former


7

J. Chem. Phys. 133, 174502 2010 M. Dzugutov, S. Simdyankin, and F. Zetterling, Phys. Rev. Lett. 89, 195701 2002 . 8 E. J. Sare and C. A. Angell, J. Solution Chem. 2, 53 1973 . 9 W.-L. Johnson, Mater. Res. Bull. 24, 42 1999 . 10 S. Schneider, J. Phys.: Condens. Matter 13, 7723 2001 . 11 C. A. Angell, J. Non-Cryst. Solids 354, 4703 2008 . 12 J. K. Kruger, K.-P. Bohn, R. Jimenez, and J. Schreiber, Colloid Polym. Sci. 274, 490 1996 . 13 W. Kauzmann, Chem. Rev. Washington D.C 43, 219 1948 . 14 D. Matyushov and C. A. Angell, J. Chem. Phys. 126, 094501 2007 . 15 C. A. Angell, Science 319, 582 2008 . 16 S. Sastry and C. A. Angell, Nature Mater. 2, 739 2003 . 17 M. I. Mendelev, J. Schmalian, C. Z. Wang, J. R. Morris, and K. M. Ho, Phys. Rev. B 74, 104206 2006 . 18 P. R. ten Wolde and D. Frenkel, Science 277, 1975 1997 . 19 V. Talanquer and D. W. Oxtoby, J. Chem. Phys. 109, 223 1998 . 20 F. Sciortino, J. Phys.: Condens. Matter 17, v7 2005 . 21 E. A. Jagla, J. Chem. Phys. 111, 8980 1999 . 22 L. Xu, S. V. Buldyrev, C. A. Angell, and H. E. Stanley, Phys. Rev. E 74, 031108 2006 . 23 L. Xu, S. V. Buldyrev, N. Giovambattista, C. A. Angell, and H. E. Stanley, J. Chem. Phys. 130, 054505 2009 . 24 J. P. K. Doye, D. J. Wales, F. Zetterling, and M. Dzugutov, J. Chem. Phys. 118, 2792 2003 . 25 J. A. Moriarty and M. Widom, Phys. Rev. B 56, 7905 1997 . 26 N. W. Ashcroft and N. D. Mermin, Solid State Physics Harcourt Brace, San Diego, 1976 . 27 M. Elenius and M. Dzugutov, J. Phys.: Condens. Matter 21, 245101 2009 . 28 T. Oppelstrup and M. Dzugutov, J. Chem. Phys. 131, 044510 2009 . 29 J.-P. Hansen and I. R. McDonald, Theory of Simple Liquids Academic, New York, 2006 . 30 J.-F. Sadoc and R. Mosseri, Geometrical Frustration Cambridge University Press, Cambridge, 1999 . 31 E. G. Ponyatovsky, J. Phys.: Condens. Matter 15, 6123 2003 . 32 I. Saika-Voivod, F. Sciortino, and P. H. Poole, Phys. Rev. E 69, 041503 2004 . 33 D. R. Nelson, Phys. Rev. B 28, 5515 1983 . 34 C. M. Dobson, Nature London 426, 884 2003 . 35 J. F. Douglas, J. Dudowicz, and K. F. Freed, J. Phys. Chem. 125, 144907 2006 . 36 J. Y. Gebremichael, M. Vogel, and S. C. Glotzer, J. Phys. Chem. 120, 4415 2003 . 37 F. Sciortino, P. Tartaglia, and E. Zaccarelli, J. Phys. Chem. 109, 21942 2005 . 38 M. Elenius and M. Dzugutov, J. Phys. Chem. 131, 104502 2009 . 39 C. Way, P. Wadhwa, and R. Busch, Acta Mater. 55, 2977 2007 .

ing the composition, the effective interionic potentials in real metallic systems can be tuned to approximate the present model.
V. CONCLUSIONS

In summary, these results introduce a novel type of the liquid-glass transition, whereby a rst-order polyamorphous phase transition precludes crystallization upon cooling by directly transforming an equilibrium thermodynamically stable liquid phase into a glass. The transition has been observed in a molecular dynamics simulation of a one-component system with a metal-like pair potential. This is the rst observation of a thermodynamic glass transition from an equilibrium stable liquid phase. In this way, the present model may be regarded a reasonable candidate for the rst nonpolymeric single-component ideal glass-former observed so far. Its prototypically metallic nature suggests a possible new strategy for designing bulk metallic glass-forming alloys in search for a noncrystallizing metallic liquid. The nding of a simple one-component low temperature amorphous phase that successfully competes in thermodynamic stability with crystalline phases may possibly be insightful for the continuing efforts to resolve the long-standing Kauzmann paradox.
ACKNOWLEDGMENTS

We are grateful to Professor C. A. Angell for reading the manuscript, valuable comments, and several illuminating discussions. We thank Professor F. Sciortino for a discussion. We also thank the Centre for Parallel Computers PDC at the Royal Institute of Technology, Stockholm and the Ekman Consortium for providing computer resources.
1 2

P. G. Debenedetti and F. Stillinger, Nature London 410, 259 2001 . D. R. Uhlmann, J. Non-Cryst. Solids 7, 337 1972 . 3 R. G. Beaman, J. Polym. Sci., Polym. Phys. Ed. 9, 472 1952 . 4 C. A. Angell, J. Non-Cryst. Solids 73, 1 1985 . 5 R. Busch, JOM 52, 39 2000 . 6 V. Molinero, S. Sastry, and C. A. Angell, Phys. Rev. Lett. 97, 075701 2006 .

Downloaded 07 Dec 2010 to 213.89.193.248. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

You might also like