This action might not be possible to undo. Are you sure you want to continue?

# The ability to infer the mechanical properties of glass from its structural

**composition is important in the glass industry, where accurate prediction of
**

properties such as viscosity, hardness, and chemical durability have obvious

benefits . So far, the main method for deducing the optimal composition to

maximize certain mechanical properties has been empirical [1]. For example, in

1915, scientists used empirical analysis to discover a very durable, heat and shock

resistant glass known commonly today as Pyrex®[18]. While this method does

arrive at the answer, a theoretical understanding of the connection between a

glass’s composition and its mechanical properties is of more interest in modern

glass science.

One direct approach for explaining a glass behavior based on its structure is

Molecular Dynamics. In short, this process examines the Coulomb interactions

between networks of atoms. Since there are no other types of interactions between

atoms, examining the Coulomb forces is certain to work. However, the computation

forces due to atomic bonding requires a tedious numerical integration with a time

step of 10-15 s [2]. These computational limits preclude the analysis of sufficiently

large networks.

In order to get more useful methods of analysis, some simplification needs to

be made. Topological Constraint Theory [4,6,14] does not calculate or examine any

atomic forces, rather, it examines the mere existence or absence of a bond. This

eliminates the need for a tedious strength

calculation. Constraint Theory makes a

further simplification by discarding any

extraneous forces and only calculating the

main influences on atomic motion, called

constraints. This idea of considering

constraints was first suggested by Maxwell

[5] and was later developed into a

breakthrough theory by Phillips in 1979

[4]. These atomic constraints fall into the

Fig 1. Types of Intermolecular Constraints categories: α Constrains and β Constraints

(Fig 1.) α constraints describe forces

associated with bond stretching, while β constraints describe forces associated with

bond bending.

α

β

**Depending on the number of constraints each atom experiences, the nature
**

of its motion can be deduced. If an atom has two few constraints, it is said to be

floppy, or under-constrained and if has atom has too many, it is said to be rigid

or over-constrained. Extra constraints in a rigid system are called redundant

bonds. There is also a critical number of constraints where the atom is neither overconstrained nor under-constrained. Such an atom is said to be isostatically

constrained [6] (Fig 2.). The number of constraint determines the nature of the

topological connection. Since we want information about the network, not just one

atom, we need to figure out the number constraints of some average particle in the

network. We define the mean coordination number with this in mind.

and floppy modes do not. and each addition atom bonded adds two β constraints [4]. The number of atomic constraints in the system can be expressed as N co = m +2 m−3 2 First term: Since each bond connects two atoms.m=x N cn ( A )+ y N cn ( B ) +(1− y−x ) N cn (C) This is a simple weighted average. Second term: This term counts angular constraints. defined here for a 3 component system with N cn denoting the coordination number (number of neighboring atoms) and x and y denoting the molar percentage of A and B. respectively. each bond adds two α constraints to the system. it is useful to express to fraction of floppy modes as a function of the p. the probability that there will exist a bond between two nodes. The initial approximation made by Maxwell [5] is r 1 Nzp−dN n ¿ 2 ¿ dN −¿ f =¿ . not two. here we consider one half constrain per α constraint because we are only focused on the constraints of one average particle. The Nature of Topological Connections Next we must discover how many constraints are affecting this average particle. Floppy Two Bonds are Redundant Isostatically Constrained Rigid Fig 2. Since rigid bonds add the network rigidity. When an atom has a coordination number of two. there in one β constraint.

where d is the number of dimensions. and the 1 Nzp−dN n ¿ indicated how many of them 2 ¿ form constraints. The left-hand simplification shows that f approaches 0 linearly as p is increased. the number of redundant bonds per degree of freedom. further accuracy can be achieved by using the series expansion f =3(1−p)5−2 ( 1− p )6 +O( (1− p )8 ) which also approaches zero as p increases. This approximation closely fits computer simulations of percolation thresholds [8]. This equation defines the useful value nr. N is the total number of atoms in the network. but this time the approach is asymptotic. p *=2d/z. hence the difference between these two values gives you the number of non-constraining bonds. z is the average coordination number. . dN basicalls gives r you to total number of bonds.

the Glass redundant bonds serve only to add internal stresses to the system [8]. The average number of constraints per atom. there exists many local minima. if passes a critical point. All these stresses provide enough energy to overcome the Crystal activation barrier and. not at a single Floppy Critical Rigid Fig 4. Floppy-Rigid Spectrum point. In theory. since thermal energy has not yet entered the discussion. This . we obtain ideal glass behavior at m =2.3] worked to expand topological theory to account for the current temperature of the network.The nature of the floppy and rigid constraints has a profound impact on the behavior of the network. does not only apply to glass-forming ability. N co . they do represent local minima in free energy (Fig 3. Mauro and Gupta [2. each representing a mechanically stable configuration. if a glass is to become a crystal. If the network is mostly floppy. Up until now. Energy Barrier to Crystallization easily formed. atoms arrive at these local minima. The two extreme ends Amount of Atomic Order of Fig. This processes is called rigidity percolation [14]. again. Although glasses are not the most favorable thermodynamic state. Therefore the optimal glass forms when N co =d . such as extremely low minimum cooling rates (easily forms glass) and no internal stresses (no extra stress from redundant bonds). So as N co is varied from one extreme to another. 9]. Phillips reasoned that ideally constrained system has enough constraints to prevent motion but not too many as to stress the network rigid. 4 represent regions where glass is not Fig 3. guided by the gradient of the Nd-dimensional hyperspace. the network becomes a crystal. where local energy considerations are minimized.45. ideal glass-forming ability was discovered to occur in a critical range. They reasoned that. Glasses in this region are said to be in a Boolchand Intermediate Phase. and the network can Energy Barrier easily overcome the low energy barrier to form glass. named after its discoverer [10]. the number of dimensions. For now we will relate this behavior to its glass forming ability. Glasses in the Boolchand Phase have interesting properties. Setting d = 3. That is to say. for a system of N atoms and d dimensions. As the Topological Theory was developed further [1. we have only considered network at absolute zero. This point represents the topological prediction of a network with ideal glass-forming ability.). it must reconfigure its topology into Gibbs Free Energy an ordered network. it takes almost no energy to reconfigure the network. and that takes some amount of activation energy. as introduced by Phillips [4]. which are in accordance with the predictions made by Topological Theory. When a network is over-constrained.

but as the system is cooled down. and the point at which there is no inter-basin travel is dubbed the glass transition temperature. As Applied to Specific Networks Now we shall apply the principles of Topological Constrain Theory to several specific glass networks. Since ) and the vt obs is generally large. knowing N co also tells you how many bonds are present in the system. and t obs is the timescale of the observation. −∆ F vt obs kT ] q ( T )=[ 1−e ∆F is the activation energy needed for breaking a constraint. k is Boltzmann’s constant. This function is reminiscent of the classic Fermi energy distribution function. T >T q where Tq is the temperature of rigidity percolation. and a function of temperature. but some models such as the Yun-Bray model for binary and ternany borosilicate glasses have been developed with some success [17]. or NMR data in more complex systems. A topological analysis begins with the identification of the main configurational units in the network. Borate classes are complex and their behavior is still under debate[12]. T <T q 0.descent guided by the slope of the surface is termed basin. The hyperspace is independent of temperature. the system can flow freely among basins. and there exists only one basin for every stable mechanical configuration. This reasoning was the basis for the temperature considerations of topological theory. which correlates directly to Shear and Elastic Moduli [2]. where q is rigid at T =0 ( network becomes floppy as it gets hotter ( lim q ( T )=0 T→∞ lim q ( T )=1 T→0 ). When the thermal energy in the system is accounted for. This knowledge either comes from direct reasoning in simple cases. knowledge of the network behavior much be applied. such alkali/alkali earth silicates[16]. After the proper fractions of network-forming species can be mathematically deduces as a function of x (the change in the molar . One the units are identified. but the way in which the network navigates this space does depend on temperature. inter-basin travel is reduced. topological theory can be used to predict the viscosity and glass transition temperature of glass networks [1-3]. called network-forming species. Both forms of q(T) are used. v is the vibrational attempt frequency. q(T) also has the discreet form { q ( T )=θ ( T q−T )= 1. Also. At high temperatures. We introduce q(T) the network rigidity. depending on the needs of the situation.

x R2 O∙ ( 100−x ) B 2 O3 . Note: Silica tetrahedral occur in pairs. N (O )=x ( 1 ) +( 100−x)( 2) The number of Si3 and Si4 groups sum to the total amount of Si. Alkali Silicates We begin with an analysis of binary alkali silicate glasses of the composition x R2 O∙ ( 100−x ) Si O2 For x = 0. Each oxygen is then considered a ‘bridging-oxygen’ (BO) linking two Si atoms. constraint counting is applied and can be fit to experimental parameters with good agreement [11-13]. First we assume that only Si4 or Si3 units occur in the network. that is a safe assumption [17]. count the number of oxygen atoms in the network. Rearranging the above equation leads to 3 Si =2 x Si 4 =100−3 x This implies that each addition of an R 2O group creates two Si3 groups and takes away three Si4 groups. Si 4 + Si 3=100−x We can also deduce the number of oxygen atoms from the connectivity of each Si unit 4 3 N (O )=2 Si +2. For relatively low values of x. As network modifiers are introduced into the system. To molar amount of oxygen is used to deduce to ratio of Si4/Si3 in the following way[16]. one for each unit of R2O and two for each unit of SiO2. disorder is introduces into the network by a random bond angle between the tetrahedra. Alkali Borates and Alkali Borosilicates For borates there is a similar compositional formula. the structure consists exclusively of bridging-oxygencontaining SiO4 tetrahedra which lack long range order [11-13].composition). four coordinated Si (Si4) become three coordinate (Si3). Next. pure vitreous silica.5 Si With NBO counting as a full oxygen and a BO counting as half an oxygen atom because it will ultimately be counted twice.

Rmax = K/16 + . The Yun-Bray model [17] and random pair model [15] seem to be conventional choices for a mathematical production of network-forming-species ratios.7). Table 1.5 K≥8. Clockwise from K top. two B4 tetrahedra (this is the only insta Fig 7.17]. The random pair model uses similar reasoning as before to derive to following concentrations of two. The random pair model is used here to demonstrate a theoretical binary borate glass calculation. 6.95R) B4 = R (or . The Configurational Units of a Borate Glass.95R) R ≥ Rmax B4 = Rmax – 0. Ternary Diagram for Na2O-B2O3-SiO2 system. Rmax = 1 R ≤ Rmax B4 = R (or . which are given in Fig.As mentioned earlier. Table 1 shows the YunBray Prediction [17]. 3 N (O )=2 B 4 + B 3=x+ 3(100−x ) 2 B 3+ B 4=2(100−x) Therefore . There are fewer explanations because the process is similar for SiO 2. which can be found experimentally using NMR analysis and deviates minimally with experimentation for extreme values of K [12] (refer to Fig. R ≡ Na/B and K ≡ Si/B. these systems of glasses are still under study and debate [12.25(R – Rmax)/ (1+K) B4 = 1 Na2O R B2O3 SiO2 Fig 6. three and four coordinated Boron units. and relies of the knowledge of the possible network forming species of borate systems. Yun-Bray Prediction of Four Coordinated Boron (N 4) K≤8.

as shown by the Yun-Bray model. the network runs out of available space to host B4 units. The boron anomaly must be part of our theoretical considerations [3. As more network modifiers are added to a borate network. This leads to the follows piecewise equations crossing the x = . The existence on NBOs must be considered for x ≥.5. the coordination initially increases.33).33 critical value (the derivations are left as exercises for the reader). there is one B3. ≤∧x ≤ 31−38 x 3 2 3 2 f ( B )= 0. ≤∧x ≤ 31−38 x 3 2 .∧x ≤ 5−4 x 3 f ( B )= 6−8 x 1 1 . but at a certain point (x=. the further addition of network modifying cations transform B 3 species into B2 which is a three-coordinated boron atom with two BOs and one NBO. Note that it is possible to have concentrations of B4 higher than . This is due to random pair model [15] .N ( B 4 )=2 x N ( B 3) =200−2 x And it follows that the number of bridging oxygens must be N ( OB ) =300−2 x And so the total is given by N ( OB + B4 + B3 ) =5−4 x Calculating the fractions of each network-forming species follows.∧x ≤ 5−4 x 3 f ( B )= 10−20 x 1 1 . ≤∧x ≤ 31−38 x 3 2 4 4−2 x 1 .16]. due to B3 to B4 modifications. After the x=. This is again due to random pair model theory. which states that NBO can only occur on the three coordinated Boron atoms.33 point. which states that the only instance of a B4 units is attached at the corner to only one other B 4 unit. this occurs when there are the same number of B 3 units as there are B4 units. { { { 2x 1 . or for every B4 unit.∧x ≤ 1 3 18 x−6 1 1 .33. This implies that for every double B4 unit there are 6 other bonds made with non B 4 units.

We first employ constrain counting to find the average number of constraints for a given network.K. . B3 and B 4 units. 106.C. ≤∧x ≤ 31−38 x 3 2 B For concentrations above x = 0. To do this we recognize that each oxygen is necessarily bonded to two non-oxygen atoms. P.. and would perhaps be the subject of future reports. Angular constraints are counted using B2. three and five constraints. Y.S. usually around n = 2. I. 094503 (2009). A Topological Model incorporating Temperature Dependent Constraints.” Proc.” J.{ 3−2 x 1 . P. References 1.4.J. A.∧x ≤ 5−4 x 3 f ( O )= 21−28 x 1 1 . We will start by counting the number of α constraints. Fractional network-forming species of other glass systems (ternary or quaternary) follow similar methods. J. Gupta. Hence the average number of constraints is n ( x )=2 (3−2 x )+ f ( B2 )+3 f ( B 3 ) +5 f ( B4 ) Now linearization of some arbitrary property A can be used to predict the compositional dependence property A.C Mauro. The number of α constraints is not affected by the BO/NBO properties of the oxygen.C. Gupta and D. or x and y as in the Yun-Bray model. Acad. 2. 19780-84 (2009). These methods are verified experimentally by Smedskjaer [11-13]. “Viscosity of Glass Forming Liquids.5. Topological Prediction of Mechanical Properties Now that the fractions of each network forming species are known as a function of x. Chem.. Phys. Natl.K. having one. Mauro. Sci. it would be appropriate to use silica network topological calculations. Allan. Linearization of A: A ( x) = Where dA (n ( x )−n crit ) dn dA dn is determined empirically and ncrit is chosen to be in the Boolchand intermediate phase. 130. “Composition Dependence of Glass Transition Temperature and Fragility. Ellison. respectively. U.A. J. so counting the α constraints on the oxygens alone will account for every linear constraint. Yue.

Bray. Non-Cryst. W. W. Phillips. R. C.. “Principles of Pyrex® Glass Chemistry: Structure Property Relationships. Y. N. Smedskjaer.J. Fei. 116. 1. 703-720 (2001). Phys. “Introduction to Glass Science and Technology. Solids. “Nuclear Magnetic Resonance Studies of the Glasses in the System Na2O-B2O2-SiO2.K. Y. 6. Zachariasen. Chem. Shelby. “Composition Dependence of Glass Transition Temperature and Fragility. Mauro. II.” Appl. 12930-12946 (2011). “Topological Principles of Borosilicate Glass Chemisty. 10.. Boolchand. “Constrain Theory. “Pressure Raman Effects and Internal Stress in Network Glasses. Y. Chem.H.F..P. Soc. Bull. 34.M..K. A.J.M. 4. R. J. Yue.” Front.. Mauro. Maxwell. Mater. A. Proceeding of the International Congress on Glass. 2005.L. Mater. D. 1999. D.M Smedskjaer. “Topological Model for Boroaluminosilicate Glass Hardness.C Phillips M. 12. 7.J. J.” J. A Topological Model of Alkali Borate Liquids. 3. 9.. J. 699-702 (1985) 15. 130.” J.C. 31. S. 17.G. Georgiev. Mauro.” Am. J. Boolchand. 363-380 (1978). Yun.. 174201 (2005). 53. “Prediction of Glass Hardness Using Temperature-Dependent Constraint Theory. 115503 (2010). Gupta.C. Smedskjaer...H. Philos Mag.. New Delhi.”Phys. 18. 491-504 (2014) . M. Youngman. J. Phys.. Chubynsky. Hogue. 2 nd Ed. Vector Percolation and Glass Formation. P. 90 (4). B. 3841-51 (1932).E. Jacobs. in Rigidity Theory and Applications. P.” J.E. Potuzak. “Discovery of the Intermediate Phase in Chalcogenide Glasses. 16. 14. P. Ceram.” J. J.3. 8. Generic Rigidity of Network Glasses. Yue. 23 (2014). New York: Kluwer Academic/Plenum.F. Thorpe. Phys. 2011. P.C Mauro. Cambridge: Royal Society of Chemistry.C.V. 71.” Solid State Commum. M. Gupta.” Physical Review.C. Lett. “The Atomic Arrangement in Glass. Smedskjaer. Non-Cryst. Optoelectron Adv. Goodman. 27. J. 153-81 (1979) 5. M. M. 13. 234503 (2009). 27m 294-299 (1864). C. M.” J. Solids.M. Mamedov. Mauro. J. Chem. Thorpe. (1986). Soc. 11.M. 54. 105. Rev.. B. “Topological Constrain Theory.. “Topology of Covalent Non-Crystalline Solids I: Short Range Order in Chalcogenide Alloys. Youngman. J. 115.C. P. J. and Rader.