1

diffraction

Martin Wilding1

Chris Benmore2

Rick Weber3

Oliver Alderman2,3

Anthony Tamalonis3

Mark Wilson4

John Parise5

1. Department of Chemistry, University College London, 20 Gordon Street, London

WC1H 0AJ, UK

2. X-ray science division, Advanced Photon Source, Argonne National Laboratory,

Argonne IL 60439, USA

3. Materials Development Inc. Arlington Heights, Il 60004, USA

4. Department of Chemistry, Physical and Theoretical Chemistry Laboratory, University

of Oxford, South Parks Road, Oxford, OX1 3QZ, UK

5. Department of Geosciences Stony Brook University, Stony Brook, NY 11794-2100,

USA

KEYWORDS: X-ray diffraction, containerless techniques, liquid structure

ABSTRACT

The average structure of glass-forming liquids and glasses can be studied by high energy X-

ray diffraction (HEXRD) and can be directly compared to predictions by simulations. X-rays

with high incident energy (>100 keV) are highly penetrating and provide scattering data to high

values of scattering vector (up to 30Å-1). The beam spot sizes are small (typically 50-500

microns) and data acquisition time rapid (typically 0.1 to 0.5s), so relatively small samples can

be studied and transient processes studied. By combining HEXRD with containerless

techniques a series of refractory oxide liquid systems have been studied with the opportunity

available to study deeply supercooled regions, the vitrification process and the influence of

changing the partial pressure of oxygen. The scope of these studies has recently been extended

to include molten salts and lower melting point liquids, these latter studies have been combined

with state-of-the-art Molecular Dynamics simulations which allow full descriptions of the

structure and dynamics of liquids to be developed.

943

Contributed Papers from Materials Science and Technology 2017 (MS&T17)October 8 – 12, 2017, David L. Lawrence Convention Center, Pittsburgh, Pennsylvania USA

Determining the structure of stable and supercooled liquids by high energy X-ray

Copyright © 2017 MS&T17®DOI 10.7449/2017/MST_2017_943_956

2

1. INTRODUCTION

The structure related properties of glass-forming liquids such as viscosity strongly influence

their crystallisation and dynamical behaviour. However, making the connection between

structure and viscosity remains challenging. Diffraction using radiation with wavelengths close

to the interatomic spacing is regarded and as a powerful method of determining the atomistic

structure and is the method of choice for single crystals or powders [1]. The diffraction patterns

of liquids and amorphous materials are significantly different from those of a crystalline

material, comprising broad features that present several overlapping contributions from

different atom-atom pairs. In addition, the intensity of the scattered radiation from an

amorphous or liquid sample is significantly lower than that from crystalline samples. Despite

these apparent drawbacks liquid state diffraction techniques have been developed and applied

to a range of materials over a range of temperatures [2]. In this contribution we will highlight

some recent studies that involve high energy X-rays, combining these with containerless

techniques, and will outline possible future direction for these types of studies.

The structure of liquids and glasses can be described by a range of structural models from the

conventional view of an ionic liquid in which the liquid will comprise a series of anions (or

molecular anions) and cations to three dimensional networks such as those first proposed by

Zachariasen (1932) for silica. These models all define a series of distances characteristic of the

liquid structure. Taking the continuous random network (CRN) model of Zachariasen as an

example, this comprises a network of corner-linked SiO44- coordination tetrahedra which form

different-membered rings. The characteristic distances in SiO2 (Si-O, Si-Si and O-O) will

represent both the short-range order (the coordination polyhedron) and correlations at greater

distances. The distribution of these distances is seen in the atom-atom pair-distribution function

which is a probability of densities at different radial distances (c.f. the radial distribution

function), this is a series of broad peaks, with the most prominent at shorter distance. In the

case of SiO2 the shortest peak is attributed to the Si-O distance in the SiO4 tetrahedron. This

distance does not change since in the CRN model it is the way in which the tetrahedra are

linked that defines the amorphous structure, so correlations at greater radial distance have more

significance. For a system which is not monatomic the number of partial contributions

increases as the number of components increases. This might be seen as limiting the possibility

of liquid state studies to relatively simple systems however relatively complicated systems can

be studied by combining diffraction with other analytical techniques as well as computer

simulation.

The interpretation of an amorphous structure outlined above is in real space, diffraction data

are however obtained in reciprocal space in the form of the total structure factor, S(Q). The

S(Q) is the weighted sum of partial atom-atom contributions to total scattering, where Q is the

scattering vector (Q=4/ sin ). The real the real space pair-distribution function is obtained

by sine Fourier Transform of the S(Q). It will be appreciated that in order to obtain good real

space resolution the value of Q should be maximized. Both neutrons and high energy (>

100keV) X-rays have short wavelengths and are also highly penetrating and both techniques

can be used effectively (an ideally together) in studies of amorphous and glassy materials. For

liquids, especially refractory liquids, there can be advantages to using high energy X-rays since

the higher flux at synchrotron sources means that smaller beam sizes can be used and the data

acquisition is faster, this is especially apparent if the high energy X-rays can be used with

containerless techniques[3].

944

3

Often there are significant differences in the structure of equivalent composition, liquid,

amorphous and vitreous materials. Although glasses are related to the liquid structure they are

not snap shots of the structure and the vitrfication process involves partial relaxation of

structure up until the point of dynamic arrest. Direct measurement of the liquid in situ is

therefore desirable but can involve bulk sample environments such as furnaces and sample

containers which can dominate the scattered signal. Containerless techniques can be effectively

used for in situ measurement, eliminating the crystalline furnace contribution and fully utilise

the small beam size available at high energy X-ray beam lines. Furthermore, since containerless

techniques such as aerodynamic levitation eliminate heterogeneous nucleation, liquids can be

vitrified and diffraction data obtained during this vitrification process[4].

Multi-component liquid structures reflect many complicated interactions which may include

the temperature-dependence of the equilibria of different liquid species, preferred bonding

arrangements and density fluctuations. These are exemplified in arguably the most important

diagram in understanding liquids, the fragility plot, popularised by Angell[5]. All liquids show

a temperature-dependence of viscosity and the simplest description of this temperature-

dependence is the Arrhenius relation. The fragility plot shows the temperature-dependent

viscosity of a liquid with temperature normalised to the calorimetric glass transition

temperature (Tg) and shows successive departure from Arrhenius-Law (strong) liquids by

progressive degrees of curvature (increased fragility). Many different liquid types can be

compared on the fragility plot including complex organic liquids and network-forming liquids

such as SiO2. More fragile (non-Arrhenius) liquids have a temperature-dependent liquid

structure, described formally by the Adam-Gibbs model of structural relaxation with the

temperature-dependence indicated by high configurational entropy (Sconf). The fragility plot

indicates that at high temperature viscosities are low, but still approximately Arrhenius and

viscosity increases when the liquids are supercooled. This indicates that there will be associated

structural changes in the supercooled liquid regime and direct measurement of theses

metastable liquids is key to understanding liquid viscosity. Diffraction data can be obtained

from supercooled liquids by combining high energy X-ray diffraction with containerless

techniques since this can probe structures that change most in fragile liquids within this

metastable regime. However, interpretation of these structures requires sophisticated

modelling.

In this contribution the combinations of high energy X-rays with aerodynamic levitation will

be described for three example systems. The quality of data, combination with other techniques

and possible interpretations will be discussed, as well as the results that combine the diffraction

patterns with state of the art computer simulation of these liquids. Improvements to the

diffraction methodology and the expanded scope and future direction of these measurements

will also be discussed.

2. EXPERIMENTAL METHODS.

2. 1. HIGH ENERGY X-RAY DIFFRACTION

High energy X-ray diffraction (HEXRD) uses X-rays with incident energies of greater than

60keV. These are routinely available at third generation synchrotron X-ray sources and can be

used to study liquid and amorphous materials. Experiments performed for example at the

beamlines 11-ID-C and 6-ID-D at the APS have demonstrated that high energy X-ray

diffraction measurements are an extremely effective in studying the liquid and amorphous

945

4

states. A range of materials have been studied including silicate[6] and aluminate liquids [7],

TiO2 [8], borates[9, 10] and, recently molten uranium oxide [11, 12].

The high incident energies mean that X-rays are highly penetrating and can be considered

“neutron like”. The short wavelengths mean that high values of scattering vector (Q) can be

accessed, which accordingly leads to good real space resolution. Although relatively high

values of scattering vector can be achieved, X-ray form factors are however Q-dependent

which means that neutron diffraction can routinely access higher Q values. X-ray scattering is

from the electron shell and so will increase with atomic number (Z) but does not vary with

isotope and the powerful technique of neutron diffraction with isotopic substitution clearly has

an advantage in extracting partial structural information. However at neutron sources the flux

is substantially lower than at synchrotron X-ray sources, and with high energy X-ray diffraction

smaller beam sizes can be used and data acquisition is faster. The first high energy x-ray super-

conducting undulator (SCU0) was installed at sector 6 (APS) in 2013 [13, 14] and has resulted

in a substantial increase in photon flux at specific energies. The combination of high brightness

and small incident beam size means that data can be collected very rapidly (full liquid

diffraction patterns have been obtained in a little as 0.2s) and offers the opportunity to study

complex fragile liquids where structures change rapidly with temperature.

In a typical HEXRD experiment a transmission geometry is adopted [2]. The incident X-ray

beam is collimated to a size of, routinely 0.5 x 0.5 mm and scattered X-rays detected on a

vertically mounted, 2048 x 2048 flat plate scintillation detector. Parameters such as sample-

detector distance and detector tilt are obtained by calibrating using a NIST crystalline CeO2

standard. Two-dimensional liquid diffraction patterns are collected on this detector and reduced

to one-dimensional patterns by integrating over all the pixels, using programs such as

Fit2D[15] which also corrects for polarization effects and geometry. Programs such as

PDFgetX2[16]are used with appropriate background measurements to obtain the total structure

factor.

The sample can be placed in a simple thin-walled capillary at the sample position if near

ambient temperature measurements of amorphous of vitreous material is all that is required.

For high temperature liquid state studies the HEXRD, the transmission configuration can be

used with containerless techniques and these latter sample environments can be fully integrated

into high energy beam lines.

2.2 CONTAINERLESSTECHNIQUES.

Containerless techniques avoid the use of bulky and crystalline sample containers and, if

temperatures are greater than ambient, furnaces. The diffuse scattering from liquid and

amorphous phase materials is very weak when compared to their crystalline components and

the contribution for the sample environments to the total scattered signal can be large and

difficult to subtract if there are significant absorption effects. The sample environment is

eliminated in the containerless methodology and these techniques enable studies of the liquid

state directly. In this contribution high temperature containerless techniques are discussed.

946

5

High melting point liquids (>2000oC) require laser heating to achieve the temperatures required

to reach melting point and the aerodynamic levitation furnace is designed to access the liquid

by levitating liquid drops using a gas jet and heating by a laser. The aerodynamic levitation

furnace consists of a custom designed, water-cooled conical nozzle in which a bead of ceramic

precursor is levitated by a gas jet. The bead is heated from above by 400W CO2 laser. The

temperature is monitored by pyrometer. The entire levitation furnace is enclosed in a stainless

steel shroud with access ports for video monitoring and temperature measurement and can be

fully integrated into high energy X-ray beam lines (such as at sector 6 at the APS), as shown

in figure 1.

Figure 1: The aerodynamic levitator

installed at the beamline 6-ID-D,

Advanced Photon Source. The conical

nozzle levitator is enclosed in a stainless

steel shroud with access to collimated

incident and scattered X-rays via

KaptonTM windows. The scattered X-rays

are detector by the amorphous silicon

image plate. The laser heating of the

sample (by 400W CO2 laser) is from the

top of the sample. The levitated bead (an

iron-bearing silicate liquid) heated by a

laser is shown in figure 1b.

2.3 FAST DATA ACQUISITION.

In aerodynamic levitation, the sample is suspended in a stream of flossing gas. This means that

when cooled heterogeneous nucleation can be avoided and it is often possible to cool samples

to the point of vitrification. Diffraction data can therefore be obtained, in principle from stable

liquids, supercooled liquids and glasses during this vitrification process. Recently, there has

been significant technical progress that enables rapid acquisition of diffraction data[6, 17]. By

using a fast large area detector, measurements can be made at a rate of 5 Hz and the structure

of liquids in transient, supercooling conditions studied.

The detector can be configured the detector to collect data in triggered mode such that the laser

heating the sample is blocked or regulated to a given cooling rate, to quench the sample and a

947

6

series of frames can be collected as the liquids are vitrified. The imaging control program is set

up to allow continuous data scanning. The controlled program can be triggered to allow a fast

burst of data to be buffered in memory and then written to disk. This arrangement allows blocks

of up to ~50 images to be acquired in a period of a few seconds. The data collection is

synchronised with the laser operation and data can be collected during the quenching of the

liquid. The high flux at sector 6 allows full diffraction patterns to will be collected in as little

as 200ms. The data are then reduced and the process of vitrification therefore studied directly.

2.4 CONTROLLED ATMOSPHERE EXPERIMENTS.

The aerodynamic levitation furnace is enclosed in a chamber (figure 1). The liquid drop itself

is levitated by a gas jet. Earlier studies have used Argon or oxygen as the levitation gas but

more recently, gas mixing has been introduced which means that the composition of the gas

can be changed and the partial pressure of oxygen in the levitating jet and sample chamber can

be controlled. Liquids that contain mixed valence elements can be studies, for example the

introduction of gas-mixing (CO/CO2) has allowed the partial pressure of oxygen to be

controlled and the study of liquids iron- and cobalt-bearing liquids under controlled fO2. This

illustrates the addition of a further dimension (temperature, composition and fO2) to these

combined structural studies.

3. RESULTS

The applications of these various methods are described below for three example systems.

3.1 TELLURITE SYSTEMS

Although Pure TeO2 is known to be a poor glass former, glasses can be formed readily when

small amounts of modifying components are added[18]. In a recent study the structure of

glasses in the system Bi2O3-Nb2O5-TeO2, motivated in part by the similarity in the textures

observed in some of these glasses with the apparent “polyamorphic” texture observed in

yttrium aluminate systems [19]. The segregations formed in the tellurite systems are identified

with the so-called “anti-glass” or partly ordered crystalline materials. The HEXRD studies of

the glasses do not suggest any link between these two phenomena, ternary glasses in the Bi2O3-

Nb2O5-TeO2 system reveal a glassy network based on interconnected TeO4 and TeO3 units that

appears to remain relatively unchanged from the parent TeO2 system. The diffraction data are

dominated by Te-Te correlations at distances greater than those found in pure TeO2 phases but

correspond to those for crystalline tellurites containing modifier cations. To explore this

system further, liquids in the ternary system have been studied in both the stable liquid regime

and, by using rapid data acquisition, as the liquids quench. The goals are to evaluate the

differences in structure on vitrification and also to establish whether there are structural

changes in the deeply supercooled regime that correlate with suggested polyamorphic

transitions.

The S(Q) data and the transform to real space (differential distribution function,

D(r)=4r[G(r)-1]) for a tellurite liquid (80%TeO2) are shown in figure 2. This illustrates the

quality of data achievable at sector 6 using combined HEXRD and containerless techniques,

the maximum Q value of 22Å-1- and oscillations in this S(Q) can be distinguished in these

rapidly acquired patterns to at least 15 Å-1. These data were obtained using the rapid data

acquisition as the liquid was quenched to a glass. Each individual S(Q) represents the integral

948

7

of a 2D pattern acquired in 200ms. The S(Q) for this composition will be dominated by Te-Te

and Te-O correlations and the changes with intensity of the peaks in the S(Q) as the liquid is

cooled will reflect changes in the underlying TeO2 network during vitrification. The first peak

in the diffraction pattern occurs at 2 Å-1. This principal peak [20, 21] can be associated with

spatial correlations over the medium range length-scale. This first peak can be identified as

Te...Te correlations occurring with real space periodicity 2/Q=~3-4 Å.

Figure 2. The results of HEXRD for tellurite liquids (10%

Bi2O3, 10% Nb2O5, 80% TeO2) as the liquids are quenched

to form a glass. The peaks in the S(Q) become sharper as

temperature decreases (a). The transforms to real space (b)

show sharpening of the correlations. The bold black lines

are from the liquid, different colours indicate different

temperatures. The integrals between isosbestic points in the

3-5 Å range (c) indicate the onset of non-ergodic behaviour

and mimic the viscosity temperature plot.

The transform to real space reveals the characteristic distances in the glass and liquid structure

(fig 2b). The peak at 1.95 + 0.05 Å is the Te-O peak which overlaps with broad Nb-O and Bi-

O correlations that form shoulders to this correlation at greater radial distances. The maximum

of this correlation shifts to greater radial distance as the liquid becomes supercooled. At greater

values of radial distances there is a prominent correlation at 3.75 Å, this will be dominated by

Te-Te and Te-O correlations and will reflect the changes in the underlying tellurite network on

cooling. From earlier work, it has been shown that this network in interpreted as a mixture of

3- and 4- coordinate Te-O polyhedra that may be both corner and edge-shared [ref]. The

changes in this peak with temperature will reflect decreases in Te-Te and Te-O distances with

temperature (Debeye-Waller effects) as well as formation of glass-like areas within the liquid.

This can be illustrated by considering integrals between so-called isosbestic points in the D(r),

as has been discussed for silicate liquids[22]. For this composition the integrals show a non-

linear increase as temperatures are decreases and an inflection at ~1.1 Tg (Figure 2c) that could

949

8

reflect the onset of non-ergodic liquid behaviour, this mimics the viscosity temperature

behaviour in these fragile liquids. It is possible that there is a change in structure at ~ 1.2Tg, as

predicted from mode coupling theory, but detailed interpretations cannot be made without a

structural model.

3.2 IRON-BEARING LIQUIDS

Iron is an important component of a variety of liquids. The role of iron in these liquids is

however complicated by the different structural role that can be adopted by iron depending on

its oxidation state. In a study of the geologically important mineral liquid, fayalite,

Alderman[23] has combined high energy X-ray diffraction with aerodynamic levitation under

different fO2 with XANES measurements, also using levitation methods and controlled

atmospheres.

Faylite (Fe2SiO4) is not glass forming and so in situ liquid measurement to access the Fe3+

content and local structural information is required. Subtle but systematic changes in the real

space transform of the diffraction data are shown and qquantitative information on the nearest

neighbor Fe-O pair correlations was extracted by means of peak fitting. This shows a single

Fe-O distribution up to 2.0 Å and provides the Fe-O coordination number of 4.4 to 4.7. X-ray

Absorption Near Edge Structure (XANES) spectra were collected at the Fe K-edge. The spectra

have different centroid energies, corresponding to different Fe3+ contents and all have similar

total areas which lie intermediate between values expected for 4- or 6-fold Fe, indicating

average Fe-O coordination numbers close to 5 in agreement with the HEXRD and shows a

significant fraction of 5 and 6-fold iron but only limited (barely discernible) change in nFeO

with Fe3+ in contrast with the conventional view that Fe3+ favours tetrahedral sites.

The XANES data show a shift in the pre-edge peaks to higher E upon melt-quenching. Such

changes imply oxidation during melt-quenching, which appears to be supported by the

Mössbauer measurements on the recovered solids[24].

The studies of CaSiO3+”FeO” silicate liquid shave also been carried out, motivated by studies

of simple slag systems and the need to evaluate the partial structural role of Fe-O. CaSiO3

compositions with 10, 30 and 50 FeO have been suited by both HEXRD and XAS[23]. Only

the 10% composition is glass-forming, however, a more recent set of experiments has been

carried out on CaSiO3 liquids with 20% FeO, this composition forms a glass but also has

sufficiently high iron to enable the structure contributions to be easily resolved.

The study is informed by the earlier results [24]and aims to address several questions that are

raised from the earlier HEXRD and XANES measurements, the apparent oxidation of iron on

cooling and the possible dependence of oxidation state and structure depend on quench path.

A wider range of fO2 was explored, this includes fO2 buffered CO:CO2 mixtures but also the

use of more oxygen-rich gases so that the structural role of Fe3+ could be more easily assessed.

Since the liquid composition was glass forming we were able to make more direct comparison

between the liquid and glass structures and were also able to monitor the changes in structure

during the process of vitrification. To evaluate the quench rate (fictive temperature)

dependence of the structure we measured the liquid structures at two different temperatures,

one supercooled to approximately 300K below the melting temperature. Separate glasses were

produced from these two different liquids to determine whether the bulk structures showed

950

9

discernible changes that were dependent on thermal history. The recovered glasses are in the

process of being analysed for Fe3+/Fe by Raman spectroscopy, _XANES and also tested for

homogeneity by electron microprobe.

In figure 3 the total structure factors for the CaSiO3 with 20 % FeO composition liquids are

shown for both the “stable” ( 1909-2379K) and “supercooled” (1876-1991K) measurements.

The ferric iron content can be estimated using the calculation of Jayasuriya[25] for the different

gas mixtures and temperatures. In some cases these yield identical ferric iron contents for the

supercooled and stable liquids, yet the structures are different. In figure 3b the real space

transform of the stable and supercooled liquids are shown, again as a function of temperature.

Figure 3: The total structure factor, shown to 6Å-1 (S(Q)) (a) and real space transform (b) of 20% “FeO” CaSiO3

compositions under different levitation gases. These two sets of liquids represent different temperatures, one

supercooled by ~300K below the stable liquid temperature. Even though the differences in Fe3+ content are

minimal in some cases there are small differences in structure.

There are more obvious changes in

structure when these liquids are

quenched. The real space data for

glasses quenched from the stable

and supercooled (metastable)

liquids are shown in figure 4.

Figure 4: The real space transforms of

CaSiO3 + 20FeO glasses quenched from

stable and supercooled liquids under

different redox conditions determined by

different levitation gases.

951

10

3.3 MOLTEN SALTS

In the results presented thus far it will be appreciated that a detailed interpretation of the liquid

and glassy diffraction patterns cannot be made without some form of structural model. Recent

work on molten salts has combined HEXRD with state of the art molecular dynamics

simulations [ref] and illustrates the possibility of significant insight into the structure of these

apparently simple liquids.

The structure factor for molten Na2CO3 was determined recently at temperatures of 1255-1328

K for a range of levitation gases[26]. Molecular dynamics simulation of these liquids has been

used to establish the contribution of each atom pair to the total structure factor. Earlier

simulations of carbonates have assumed the liquids behave as molten salts with rigid carbonate

(CO32-) anions and metal cations. Vibrational spectroscopy however suggests that the anions

are flexible[26]. So in these simulations the carbonate anion is assumed to have a basic trigonal

planar geometry and flexibility is allowed by allowing harmonic springs controlling C-O and

O-O interactions. The development of the simulation models also allows the temperature-

dependent structural properties to be probed.

The temperature dependence of structure is shown in the changes in the partial contributions

to the first three peaks in the diffraction pattern[26] and analysis in real space shows changes

in C-C and C-Na contributions and indicate the formation of a chain-like network of carbonate

anions. The length of these chains decreases as temperature increases and correlates with

diffusion and therefore viscosity.

More recent studies of molten salts include data collected on alkali carboantes and sulphates

and also on alkali nitrates (including ammonium nitrate) and nitrite [ref].

The diffraction data for one of these liquids, lithium sulphate, is shown in figure 5. The main

structural unit is the SO4 tetrahedron. Lithium is weakly scattering in X-ray and so the main

features observed are the sulphate (S-S, S-O and O-O) correlations. A prominent feature in the

diffraction pattern is the peak at 1.5 Å-1. This first sharp diffraction peak (FSDP) is noted in

network forming materials cf. SiO2. Not that expected in a molten salt. Simulation was carried

out using the same approach, modelling the anions using flexible bonds in the sulphate

tetrahedron, the preliminary results show good agreement with the experiment (figure 5b), and

the FSDP has contributions from both the S-O and S-S atom pairs. These indicate intermediate

range order, were these to be SiO2 liquids the Si-O and O-O distances would give correlation.

In the lithium sulphate the contribution from lithium pairs is reduce because lithium scatters so

weakly. This suggests a periodicity for the S-O pairs and S-S pairs, qualitatively similar to the

FSDP in network-forming glasses like SiO2 and other liquids of MX2 stoichiometry. Associated

with intermediate range order on the length scale of 4-5 Å where (2/QFSDP). This two indicates

formation of chains or low-dimensional networks in the sense that there are S-S distance

correlations in real space and an inferred sharing of oxygens in the tetrahedral units in the same

way that SiO4 tetrahedra are shared. The temperature dependence has also been modelled and

suggests a similar diffusional relationships (paddle wheel effects) in the sulphates.

952

11

Figure 5: Diffraction data for liquid Li2SO4 compared with the results of simulation (left). The breakdown of the

total structure factor into partial contributions from the simulation is shown at right.

Molten nitrates have potential applications for thermal storage in solar energy plants and

accordingly there have been extensive studies into the viscosity of alkali and alkaline-earth

bearing nitrates. From a fundamental perspective these liquids (especially Ca-K nitrates)[27,

28] are some of the most fragile liquids known, yet glasses can be formed in these systems

offering the opportunity to explore the stable and metastable structures of these fragile liquids

to the point of vitrification. As with other ionic liquids these are generally viewed as molten

salts with a triangular NO3- anion and will be amenable to modelling using the same approach

(flexible anions tested against diffraction data) to establish whether temperature-dependent

transient networks also dominate these liquids.

The same basic model is used for the simulation using flexible anions (similar to the carbonate).

Na and Li nitrate MD models are in good agreement with the diffraction data. The simulation

allows exploration of the temperature (cf carbonates) and the N-N partial shows a strong

temperature dependence.

Nitrate and nitrite liquids have also been compared, and the liquid structures for the two

different anions (NO3- and NO2

-) are shown for the sodium and potassium nitrate and nitrite

systems (Figure 6)

953

12

Simulations have again been carried out using a flexible anions. These results suggest different

charge distributions and different roles in nitrates and nitrites and both nitrate and nitrite ions

can be distinguished, and form low dimensional networks with different length scales.

Figure 6: Liquid diffraction patterns for potassium (left) and sodium (right) nitrate and nitrite liquids.

The carbonates, sulphates, nitrates and nitrites all suggest that formation of complicated, low

dimensional networks. These liquids have hitherto been modelled as molten salts and simple

ionic models do not seem to be appropriate. However collection of diffraction data and

adjusting the model for every conceivable molten salt is not too productive unless there is some

deeper insight. The overall goal of these studies is to look at these sorts of liquids in a broader

context, with emphasis on the temperature dependence of structures, the formation of low

dimensional networks and also to establish how these structures are related. Simulations have

been carried out on nitrates using flexible charges but rigid ions, however these rigid models

do not have out-of-plane polarization which flexible bonds allow. However it is desirable to

allow the charge to fluctuate since this has been used to effectively model nitrates. Simulations

indicate quite clearly that there are extended length scales and transient networks, such that

other techniques such as NMR would help establish the nature of these longer distance scales.

FUTURE DIRECTION

The combination of HEXRD and containerless techniques allows detailed structural

information to be obtained as a function of composition and y as a function of fO2 in both the

stable and deeply supercooled regimes. This offers potential to develop models of relaxation

processes and the onset of glass formation. The techniques described are not restricted to these

systems and metallic and other exotic liquids can also be studied, a notable recent study is of

UO2[12].

The results of these diffraction studies and the molecular dynamics simulations suggest the

emergence of two different length scales, but cannot discern the influence of longer length

scales on the liquid properties. Combinations of small and wide angle scattering measurements

are desirable so that the longer distance scales can be probed. The ongoing development of

extended range PDF techniques and sector 6 offers the opportunity to probe nanoscale

structures. The combination of overlapping small and wide angle detectors at sector 6 will

enable continuous measurements over a range of Q space from 0.01 Å-1 to a maximum value

of ~15-20 Å-1 . Similar in coverage to novel neutron instrumentation.

954

13

Although at this stage the opportunities for using aerodynamic levitation with SAXS/WAXS

are not available, lower temperature containerless techniques are available and studies of

concentrated aqueous solutions of nitrates, carbonates and perchlorates have been performed

that connect directly with the current molten salt studies.

In conclusion, high energy X-ray diffraction can be combined with containerless processing to

study a range of liquids in situ enabling an exploration of the influence of composition,

temperature and fO2 for a variety of liquid types. This includes liquids that are not necessarily

glass-forming, such that systematic explorations of phase space can be performed. In the case

of glass forming liquids the process of vitrification can be directly observed. These techniques

are especially powerful when combined with recent development in molecular dynamics

simulation and there is potential to use these techniques to make substantive contributions to

the fundamental understanding of the liquid state as well as more direct applications to

materials-motivated studies, such as of liquids of interest to the steel industry, the advanced

ceramics community and to geophysics. Instrumentation development at synchrotron sources

provides the opportunity to expand the length scales studied and observed the formation of

intermediate range structures in a variety of technically important liquids.

ACKNOWLEDGEMENTS

This work was supported by the U.S. DOE Argonne National Laboratory under contract

number DE-AC02-06CH11357. The authors thank Guy Jennings for technical help with the

fast X-ray data acquisition and both Rick Spence and Doug Robinson for their invaluable

support in ensure the success of these measurements.

REFERENCES

1. Fischer, H.E., et al., The bound coherent neutron scattering lengths of the oxygen

isotopes. Journal of Physics-Condensed Matter, 2012. 24(50).

2. Weber, J.K.R., et al., Aerodynamic levitator for in situ x-ray structure measurements

on high temperature and molten nuclear fuel materials. Review of Scientific

Instruments, 2016. 87(7).

3. Weber, J.K.R., et al., Measurements of liquid and glass structures using aerodynamic

levitation and in-situ high energy x-ray and neutron scattering. Journal of Non-

Crystalline Solids, 2014. 383: p. 49-51.

4. Benmore, C.J., et al., Structural and topological changes in silica glass at pressure.

Physical Review B, 2010. 81(5).

5. Angell, C.A., Formation of glasses from liquids and biopolymers. Science, 1995.

267(5206): p. 1924-1935.

6. Benmore, C.J., et al., Temperature dependent structural heterogeneity in calcium

silicate liquids. submitted, 2010.

7. Wilding, M.C., et al., Structural changes in supercooled Al2O3-Y2O3 liquids. Physical

Chemistry Chemical Physics, 2013. 15(22): p. 8589-8605.

8. Alderman, O.L.G., et al., Structure of molten titanium dioxide. Physical Review B,

2014. 90(9).

9. Alderman, O.L.G., et al., Temperature-Driven Structural Transitions in Molten Sodium

Borates Na2O-B2O3: X-ray Diffraction, Thermodynamic Modeling, and Implications

for Topological Constraint Theory. Journal of Physical Chemistry C, 2016. 120(1): p.

553-560.

10. Alderman, O.L.G., et al., Liquid B2O3 up to 1700 K: x-ray diffraction and boroxol ring

dissolution. Journal of Physics-Condensed Matter, 2015. 27(45).

955

14

11. Benmore, C.J., et al., Topological ordering in liquid UO2. Journal of Physics-

Condensed Matter, 2016. 28(1).

12. Skinner, L.B., et al., Molten uranium dioxide structure and dynamics. Science, 2014.

346(6212): p. 984-987.

13. Ivanyushenkov, Y., et al., Development of a superconducting undulator for the APS, in

11th International Conference on Synchrotron Radiation Instrumentation, J. Susini and

P. Dumas, Editors. 2013.

14. Ivanyushenkov, Y., et al., A Design Concept for a Planar Superconducting Undulator

for the APS. Ieee Transactions on Applied Superconductivity, 2011. 21(3): p. 1717-

1720.

15. Hammersley, A.P., et al., Two-dimensional detector software: From real detector to

idealised image or two-theta scan. High Pressure Research, 1996. 14(4-6): p. 235-248.

16. Qiu, X., J.W. Thompson, and S.J.L. Billinge, PDFgetX2: a GUI-driven program to

obtain the pair distribution function from x-ray powder diffraction data. Journal of

Applied Crystallography, 2004. 37: p. 678.

17. Weber, J.K.R., et al., Instrumentation for fast in-situ X-ray structure measurements on

non-equilibrium liquids. Nuclear Instruments & Methods In Physics Research Section

A-Accelerators Spectrometers Detectors And Associated Equipment, 2010. 624(3): p.

728-730.

18. Wilding, M.C., et al., Structural studies of Bi2O3-Nb2O5-TeO2 glasses. Journal of

Non-Crystalline Solids, 2016. 451: p. 68-76.

19. McMillan, P.F., et al., Polyamorphism and liquid-liquid phase transititions in

amorphous silicon and super-cooled Al2O3-Y2O3 liquids, in Liquid Polymorphism, H.E.

Stanley, Editor. 2013. p. 309-353.

20. Susman, S., et al., Intermediate-range order in permanently densified vitreous SiO2- a

neutron diffraction and molecular dynamics study. Physical Review B, 1991. 43(1): p.

1194-1197.

21. Susman, S., et al., Temperature-dependence of the 1st sharp diffraction peak in vitreous

silica. Physical Review B, 1991. 43(13): p. 11076-11081.

22. Benmore, C.J., et al., Temperature-dependent structural heterogeneity in calcium

silicate liquids. Physical Review B, 2010. 82(22).

23. Alderman, O.L.G., et al., Iron K-edge X-ray absorption near-edge structure

spectroscopy of aerodynamically levitated silicate melts and glasses. Chemical

Geology, 2017. 453: p. 169-185.

24. Alderman, O.L.G., et al., Local structural variation with oxygen fugacity in Fe2SiO4+x

fayalitic iron silicate melts. Geochimica Et Cosmochimica Acta, 2017. 203: p. 15-36.

25. Jayasuriya, K.D., et al., A Mossbauer study of the oxidation state of Fe in silicate melts.

American Mineralogist, 2004. 89(11-12): p. 1597-1609.

26. Wilding, M.C., et al., Low-Dimensional Network Formation in Molten Sodium

Carbonate. Scientific Reports, 2016. 6.

27. Ribeiro, M.C.C., Ionic dynamics in the glass-forming liquid Ca0.4K0.6(NO3)(1.4): A

molecular dynamics study with a polarizable model. Physical Review B, 2001. 63(9).

28. Ribeiro, M.C.C., High frequency sound velocity in the glass former 2Ca(NO3)(2)center

dot 3KNO(3): Molecular dynamics simulations. Physical Review B, 2007. 75(14).

956