1: \documentclass[onecolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
2: \usepackage{graphicx}% Include figure files
3: \usepackage{dcolumn}% Align table columns on decimal point
4: \usepackage{bm}% bold math
5: %\usepackage{subfig}
6: \def\w6{{$\hat{W}_6$}}
7: \renewcommand{\baselinestretch}{3}
8:
9:
10: \begin{document}
11:
12: \title{Geometrical frustration in liquid Fe and Fe-based metallic glass}
13:
14: \author{P. Ganesh}
15: \author{M. Widom}
16: \affiliation{Carnegie Mellon University\\
17: Department of Physics\\
18: Pittsburgh, PA 15213}
19:
20: \date{\today}
21:
22: \begin{abstract}
23: We investigate short range order in liquid and supercooled liquid Fe
24: and Fe-based metallic glass using {\em ab-initio} simulation methods.
25: We analyze the data to quantify the degree of local icosahedral and
26: polytetrahedral order and to understand the role of alloying in
27: controlling the degree of geometric frustration. Comparing elemental
28: Fe to Cu~\cite{Cupaper} we find that the degree of icosahedral order
29: is greater in Fe than in Cu, possibly because icosahedral disclination
30: line defects are more easily incorporated into BCC environments than
31: FCC. In Fe-based metallic glass-forming alloys (FeB and FeZrB) we
32: find that introducing small concentrations of small B atoms and large
33: Zr atoms controls the frustration of local icosahedral order.
34: \end{abstract}
35:
36: \pacs{61.43.Dq,61.20.Ja,61.25.Mv}
37:
38: \maketitle
39:
40: \section{\label{sec:intro}Introduction}
41:
42: As noted by Frank~\cite{Frank}, the local icosahedral clustering of 12
43: atoms about a sphere is energetically preferred because it is made up
44: entirely of four-atom tetrahedra, the densest-packed cluster possible.
45: However, local icosahedral order cannot be propagated throughout space
46: without introducing defects. Frustration of packing icosahedra is
47: relieved in a curved space, where a perfect 12-coordinated icosahedral
48: packing exists~\cite{SM82,Nelson83,Sethna83}. We adopt this structure
49: as an ideal reference against which actual configurations will be
50: compared.
51:
52: Disclination line defects of type $\pm 72^{\circ}$ may be introduced
53: into this icosahedral crystal and thereby control the curvature. In
54: order to ``flatten'' the structure and embed it in ordinary three
55: dimensional space an excess of $-72^{\circ}$ disclinations is needed,
56: and these cause increased coordination numbers of 14, 15 or 16. Large
57: atoms, if present, would naturally assume high coordination numbers
58: and aid in the formation of a disclination line network. Similarly,
59: smaller atoms would naturally assume low coordination numbers of 8, 9
60: or 10, and have positive disclinations attached to them, increasing
61: the frustration. For a particular coordination number, it may be
62: possible to construct a cluster, known as a Kasper-Polyhedron (see
63: Section~\ref{sec:alloyW6}), made entirely of tetrahedrons.
64:
65: Honeycutt and Andersen~\cite{Honeycutt} introduced a method to count
66: the number of tetrahedra surrounding an interatomic bond. This number
67: is 5 for icosahedral order with no disclination, 6 for a $-72^{\circ}$
68: disclination and 4 for a $+72^{\circ}$ disclination. Steinhardt,
69: Nelson and Ronchetti~\cite{Steinhardt} introduced the orientational
70: order parameter \w6 to demonstrate short range icosahedral order. We
71: employ both methods to analyze icosahedral order in supercooled metals
72: and metal alloys, in addition to conventional radial distribution
73: functions, structure factors and Voronoi analysis.
74:
75: Many simulations have been performed on pure elemental metals and
76: metal alloys using model
77: potentials,~\cite{Egami,Chen-Liu,Sadigh,Goddard}, but do not
78: necessarily produce reliable structures owing to their imperfect
79: description of interatomic interactions. First principles ({\em
80: ab-initio}) calculations achieve the most realistic possible
81: structures, unhindered by the intrinsic inaccuracy of phenomenological
82: potentials and with the ability to accurately capture the chemical
83: natures of different elements and alloys. The trade-off for increased
84: accuracy is a decrease in the system sizes one can study, so only
85: local order can be observed, not long range. Also runs are limited to
86: short time scales. Early ab-initio studies on liquid
87: Copper~\cite{Vanderbilt,Hafner,Valladares} and Iron~\cite{Kresse} were
88: not analyzed from the perspective of icosahedral ordering.
89: Recent ab-initio studies on Ni and Zr~\cite{Jakse_Ni,Jakse_Zr} find
90: that the degree of icosahedral ordering increases with supercooling in
91: Ni, while in Zr BCC is more favored. Studies on binary metal alloys
92: by Jakse {\em et al.}~\cite{jakse_alloy1,jakse_alloy2} and by Sheng
93: {\em et al}~\cite{MaNature} quantify local icosahedral order in the
94: alloys. We previously~\cite{Cupaper} investigated icosahedral order
95: in liquid and supercooled Cu.
96:
97: Elemental metals crystallize so easily that they can hardly be made
98: amorphous at any quench rate. Alloying can improve the ease of glass
99: formation. For some special alloys, a bulk amorphous state can be
100: reached by slow cooling. Pure elemental Fe is a poor glass former, but
101: Fe-based compounds like FeB and especially FeZrB, show improved glass
102: formability. We augment our molecular-dynamics simulation with another
103: algorithm called `Tempering' or `replica exchange method'
104: (REM)~\cite{REM,super_REM} for fast equilibration at low temperatures.
105:
106: In comparison to liquid and supercooled liquid Copper~\cite{Cupaper}
107: which show only weak icosahedral order and very little temperature
108: variation, Fe showed a monotonic increase in icosahedral order, which
109: became very pronounced when supercooled. Analysis of quenched Fe
110: revealed a natural way of introducing 5-fold coordinated bonds plus a
111: single -72$^{\circ}$ disclination line segment into an otherwise
112: perfect BCC environment, without disturbing the surrounding structure.
113: Addition of B to Fe decreased the icosahedral order, due to the
114: positive disclinations centered on the smaller B atom which increased
115: frustration. Further inclusion of larger Zr atoms to form FeZrB found
116: an enhanced icosahedral order compared to FeB. This could possibly be
117: explained by formation of negative disclination line
118: defects~\cite{NelsonWidom} anchored on the larger Zr atoms, which
119: eases the frustration of icosahedral order on the Fe atoms.
120:
121: % We are not
122: %aware of a similar construction for FCC. As mentioned previously, if
123: %one wants to go from a space with positive curvature, where
124: %icosahedrons can be arranged to form a periodic lattice, to flat 3D
125: %space, one needs to introduce disclination defects of the form of
126: %-72$^{\circ}$ to relive the frustration of packing icosahedrons. One
127: %way to introduce this disclination in a one component system would be
128: %to form a Frank-Kasper like defect with a coordination number Z=14,
129: %which it seems can be introduced as a local defect in a BCC
130: %environment~\cite{Khmelevskaya}. It would be difficult to obtain
131: %higher coordination in a single component liquid, whose short range
132: %order is very close to dense packing. It is possible to include a
133: %perfect icosahedron in an otherwise FCC structure, leading to fivefold
134: %bonds. Such a transformation of a cubooctahedron to an icosahedron is
135: %known in crystallography ~\cite{Hargitai,Khmelevskaya}. But as
136: %indicated in reference~\cite{Khmelevskaya}, this restructuring can
137: %only occur in the presence of a vacancy whereby a local icosahedron
138: %can form and remove the density deficit. If a liquid forming FCC
139: %ground state, has local coordination of 12, and is relatively more
140: %densely packed compared to a liquid that forms BCC ground state, it
141: %might be more difficult to introduce an icosahedron in it, as compared
142: %to introducing a Z14 negative disclination in the BCC liquid. The
143: %ease of forming this negative disclination in a BCC environment as
144: %opposed to an FCC environment, can then be thought of as the reason
145: %why we observe more icosahedral order in Fe than in Cu. Even if a Z14
146: %negative disclination be formed as a point defect, it can stabilize
147: %many icosahedrons, and at the same time connect to the hexagonal face
148: %of the neighboring BCC crystal. But there is no way the icosahedron
149: %can propagate easily throughout the system in an FCC.
150:
151: At high temperatures all of our measured structural properties of
152: liquid Cu~\cite{Cupaper} and liquid Fe resembled each other, and also
153: resembled a maximally random jammed~\cite{MRJ} hard sphere
154: configuration. This suggests that a nearly universal structure exists
155: for systems whose energetics are dominated by repulsive central
156: forces.
157:
158: Section~\ref{sec:TMD} describes our combined method of monte-carlo and
159: first principles MD, that we refer to as ``Tempering MD'' and
160: discusses other simulation details. Section~\ref{sec:PureFe} presents
161: our results on pure Fe while (section~\ref{sec:alloy}) compares this
162: with FeB and FeZrB alloys.
163:
164: \section{\label{sec:TMD}Tempering Molecular Dynamics (TMD) and other Simulation Details}
165:
166: One reason alloys form glass more easily is that chemical identity
167: introduces a new configurational degree of freedom that evolves
168: slowly~\cite{Greer,Desre}. Unfortunately, this makes simulation more
169: difficult. It is especially difficult to equilibrate the system at
170: very low temperatures, because the probability to cross an energy
171: barrier drops, trapping it in particular configurations. For this
172: reason we use a Monte-Carlo method, known as tempering or replica
173: exchange~\cite{REM,super_REM} to augment our first-principles MD,
174: allowing us to sample the configurational space more efficiently than
175: conventional MD.
176:
177: In the canonical ensemble, energy fluctuates at fixed temperature. A
178: given configuration $C$ with energy $E$ can occur at any temperature
179: $T$ with probability proportional to $e^{-\beta E}$,
180: ($\beta=1/kT$). Now consider a pair of configurations, $C_{1}$ and
181: $C_{2}$ of energy $E_{1}$ and $E_{2}$ occurring in simulations at
182: temperatures $T_{1}$ and $T_{2}$. We can take $C_{1}$ as a member of
183: the ensemble at $T_{2}$, and $C_{2}$ as a member of the ensemble at
184: $T_{1}$, with a probability
185: \begin{equation}
186: \label{eq:prob}
187: P=e^{-(\beta_2-\beta_1)(E_1-E_2)}
188: \end{equation}
189: without disturbing the temperature-dependent probability distributions
190: of energy (or any other equilibrium property). Because each run
191: remains in equilibrium at all times even though its temperature
192: changes, we effectively simulate a vanishingly low quench rate.
193:
194: In practice we perform several MD simulations at temperatures
195: separated by 100K. We use ultrasoft pseudopotentials~\cite{USPP} as
196: provided with VASP~\cite{USPP2} to perform the MD simulation. All
197: calculations are '$\Gamma$' point calculations (a single 'k' point).
198: All runs use an MD time step of 2 fs, and reach total simulated time
199: of order 1.5-1.8 ps (see Table~\ref{tab:alloys}) with a total of N=100
200: atoms. Every 10 MD steps we compare the energies of configurations at
201: adjacent temperatures and swap them with the above probability.
202: Eventually, configurations initially frozen at low temperature reach a
203: higher temperature. The simulations then can carry the structure over
204: energy barriers, after which the temperature can again drop.
205:
206:
207: \begin{table}
208: \caption{\label{tab:alloys}Details of tempering MD runs.}
209: \begin{tabular}{|r|r|r|r|}
210: \hline
211: Chemical Species & Temperatures(K) & density~(\AA$^{-3}$) & time (ps)\\
212: \hline
213: Fe$_{100}$ & 800-1900 & 0.0756 & 1.5\\
214: \hline
215: Fe$_{80}$B$_{20}$ & 800-1500 & 0.0814 & 1.8 \\
216: \hline
217: Fe$_{70}$Zr$_{10}$B$_{20}$ & 800-1800 & 0.0787 & 1.8 \\
218: \hline
219: \end{tabular}
220: \end{table}
221:
222: In an effort to explore the structures of compounds with differing
223: glass-forming ability we compare pure elemental Iron and two
224: Iron-based glass-forming alloys. Tempering MD requires that we perform
225: simulations at a constant density for all the temperatures, but we
226: have no rigorous means of predicting the density at high temperature.
227: For pure liquid Iron, the density is known
228: experimentally~\cite{Waseda}, and we use this value. For FeB and
229: FeZrB, we took a high temperature liquid structure and quenched it,
230: relaxing positions and cell lattice parameters, to predict a low
231: temperature density. We then decreased the density of the relaxed
232: structure by 6 percent to account for volume expansion, to arrive at
233: the densities used in our liquid simulations.
234:
235: Because of the efficient sampling of our tempering MD method, the
236: structure of pure Fe partially crystallizes at low temperatures after
237: about 1 ps. In the following discussion of our T=800K sample we will
238: refer to different structural features before and after
239: crystallization. We also performed several long (2.0ps) conventional
240: first-principles MD at T=800K yielding results similar to the results
241: of tempering MD prior to crystallization.
242:
243: For all runs we employed spin polarization, reasoning that local
244: magnetic moments exist even above the Curie point. These local
245: moments have a significant influence on the short-range order because
246: ferromagnetic Iron prefers a longer bond length than paramagnetic
247: Iron~\cite{Kresse97}. Of course, the ferromagnetic state of the
248: liquid implies improper long-range correlations. Unfortunately, since
249: our forces are calculated for electronic ground states, we cannot
250: rigorously model the true paramagnetic state of liquid Iron and
251: Iron-based alloys with these methods.
252:
253: \section{\label{sec:PureFe}Pure Fe}
254:
255: \subsection{\label{sec:Fegr}Radial Distribution Function $g(r)$}
256:
257: \begin{figure}
258: \includegraphics[width=3in,angle=-90]{Fegr.ps}
259: \caption{\label{fig:Fegr}Radial distribution function of pure
260: elemental liquid Fe. Simulations (magnetic and non-magnetic) are run
261: at T=1800K, compared with X-ray experiments at T=1833K~\cite{Waseda}
262: and Neutron scattering experiments at 1830K~\cite{Holland2}.}
263: \end{figure}
264:
265:
266: The radial distribution function, $g(r)$, is proportional to the
267: density of atoms at a distance $r$ from another atom and is
268: calculated here by forming a histogram of bond lengths. We use the
269: repeated image method to obtain the bond lengths greater than half
270: the box size and anticipate $g(r)$ in this range may be influenced by
271: finite size effects. Further, we smooth out the histogram with a
272: gaussian of standard deviation 0.05\AA.
273:
274: To evaluate the role of magnetism, Fig.~\ref{fig:Fegr} illustrates the
275: radial distribution functions for liquid Fe simulated at T=1800K, just
276: below melting (T$_m$=1833K). Evidently, the simulation with magnetism
277: yields good agreement in the position and height of the first peak in
278: $g(r)$ with experimental X-ray $g(r)$~\cite{Waseda}, while neglect of
279: magnetic moments results in near neighbor bonds that are too short and
280: too weak. However, magnetism overestimates the strength of long-range
281: correlations beyond the nearest-neighbor peak, while neglecting
282: magnetism yields reasonably accurate long-range $g(r)$. Nevertheless,
283: for the present study of local order, it is necessary to make spin
284: polarized calculations, to get the short range correlations and hence
285: the local order correct. Strangely, a recent experimental neutron
286: $g(r)$~\cite{Holland2} has a shorter and broader first peak compared
287: to both our magnetic $g(r)$ and to the $g(r)$ from the prior X-ray
288: diffraction experiment~\cite{Waseda}. The positions of the different
289: maxima and minima in our simulated magnetic $g(r)$ compare well
290: with both the experiments. The position of the first peak in our
291: magnetic $g(r)$ is shifted by 0.05\AA~to the left of the neutron first
292: peak. The X-ray experiment doesn't have enough data points around the
293: first maximum to determine the peak position accurately.
294:
295: %Experimental $g(r)$'s are obtained by fourier transform of the
296: %experimentally measured liquid structure factor $S(q)$
297: %(Section~\ref{sec:sq}). In the following subsection we show that the
298: %recently obtained neutron $S(q)$ does not obey a sum
299: %rule~\cite{Norman,Krogh-Moe} relating to vanishing $g(r)$ at short
300: %distances, indicating a possible difficulty in normalizing or
301: %background subtraction. We also show that our $S(q)$, obtained from
302: %the simulated $g(r)$ obeys the sum rule.
303:
304:
305: We calculate the coordination number by counting the number of atoms
306: within a cutoff distance from a central atom. We choose the cutoff
307: distance ($R_{cut}$) at the first minimum of $g(r)$. For pure Fe the
308: minimum is at $R_{cut}$=3.5~\AA. The precise location of the minimum
309: is difficult to locate, and its variation with temperature is smaller
310: than the error in locating its position, so that we don't change the
311: value of $R_{cut}$ with temperature. With this value of $R_{cut}$ we
312: find an average coordination number ($N_c$) of 13.2 which is nearly
313: independent of temperature ($N_c$ changes from 13.1 at high
314: temperature to 13.3 with supercooling).
315:
316:
317: \subsection{\label{sec:sq} Liquid Structure Factor S(q)}
318:
319:
320:
321: The liquid structure factor $S(q)$ is related to the radial
322: distribution function $g(r)$ of a liquid with density $\rho$ by,
323: \begin{equation}
324: \label{eq:Sqeq}
325: S(q)=1+4 \pi \rho \int\limits_{0}^{\infty}[g(r)-1] {\sin(qr)\over qr}r^2dr
326: \end{equation}
327: One needs the radial distribution function up to large values of $r$
328: to get a good $S(q)$. In our first principles simulation, we are
329: restricted to small values of $r$, due to our small system sizes, so
330: we need a method to get $S(q)$ from our limited $g(r)$ function.
331: Baxter developed a method~\cite{Baxter,Jolly} to extend $g(r)$ beyond
332: the size of the simulation cell. The method exploits the short range
333: nature of the direct correlation function $c(r)$, which has a range
334: similar to the interatomic interactions~\cite{Ashcroft}, as opposed to
335: $g(r)$ which is long ranged.
336: %The exact relation that connects
337: %these two functions is the Ornstein-Zernike relation,
338: %\begin{equation}
339: %\label{eq:O-Z}
340: %h(r)=c(r)+\rho \int h(|{\bf r-r'}|)c(|{\bf r'}|)d{\bf r'}
341: %\end{equation}
342: %where $h(r)=g(r)-1$.
343:
344: Assuming that $c(r)$ vanishes beyond a certain cutoff distance $r_c$,
345: %\begin{eqnarray}
346: %rc(r)=-Q'(r)+2\pi \rho{\int \limits_{r}^{r_c} Q'(r')Q(r'-r)dr'}
347: %\label{cr}
348: %\end{eqnarray}
349: %and
350: %\begin{eqnarray}
351: %rh(r)=-Q'(r)+2\pi \rho {\int\limits_{0}^{r_c}(r-r')h(|r-r'|)Q(r')dr'}.
352: %\label{hr}
353: %\end{eqnarray}
354: %Here $Q(r)$ is zero for $r> r_c$, and continuous at $r_c$, and
355: %$Q'(r)=dQ(r)/dr$. The continuity
356: %of $Q(r)$ means that,
357: %\begin{equation}
358: %Q(r)=-{\int \limits_{r}^{r_c}dr' Q'(r')}.
359: %\label{qr}
360: %\end{equation}
361: %The remarkable property of this method is that if we know $h(r)$ over
362: %a range $0\le r \le r_c$, then we can obtain $c(r)$ over its entire
363: %range (from 0 to $r_c$), which implicitly determines $h(r)$ over its
364: %entire range (from 0 to $\infty$) through Eq.~(\ref{eq:O-Z}).
365: we solve the Baxter's equations iteratively to obtain the full direct
366: correlation function for $0<r<r_c$. From $c(r)$ we calculate the
367: structure factor $S(q)$ by a standard Fourier Transform.
368: %\begin{equation}
369: %S(q)={1\over 1-\rho \hat c(q)}
370: %\label{OZink}
371: %\end{equation}
372: %where,
373: %\begin{equation}
374: %\label{eq:cq}
375: %\hat c(q)=4 \pi \int \limits_{0}^{\infty}r^2 c(r) {\sin(qr)\over qr}dr.
376: %\end{equation}
377: The $S(q)$ showed good convergence with different choices of $r_c$. A
378: choice of $r_c$=5\AA~ seemed appropriate because it was one half of
379: our smallest simulation cell edge length. Even though in metals there
380: are long range oscillatory Friedel oscillations, our ability to
381: truncate $c(r)$ at $r_c$=5\AA~ shows that these are weak compared with
382: short range interactions. An application of this method to obtain
383: $S(q)$ of Cu~\cite{Cupaper} showed excellent agreement with the
384: experimental $S(q)$.
385:
386: The simulated $S(q)$ for pure Fe at T=1800K (see Fig.~\ref{fig:Fesq})
387: is compared to recent neutron scattering experiments at T=1830K. Even
388: though the positions of the different peaks compare very well, there
389: is serious discrepancy in their heights. Especially, the height of the
390: first peak of our simulated $S(q)$ is higher than that of the
391: experiments. This discrepancy is expected because we include
392: magnetism, which gives accurate short-range correlations while
393: overestimating the long-range ones (see Fig.~\ref{fig:Fegr}). But the
394: cause of the discrepency is not entirely clear since a comparison of
395: the simulated structure factor of Ni~\cite{Jakse_Ni} (done without
396: including magnetism) with neutron scattering
397: experiments~\cite{Holland2} shows similar discrepancies between their
398: $S(q)$'s.
399:
400: A sum rule can be obtained for $S(q)$~\cite{Norman,Krogh-Moe}. By
401: inverting the fourier transform of Eq.~\ref{eq:Sqeq} and then taking
402: the $r \rightarrow 0$ limit, one gets
403: \begin{equation}
404: \label{SumRule}
405: I(Q) \equiv \int \limits_{0}^{Q} q^2[S(q)-1]dq\rightarrow-2\pi^2\rho
406: \end{equation}
407: in the limit $Q\rightarrow \infty$. Further, the integral is supposed
408: to oscillate with $Q$ about the limiting value as $Q \rightarrow
409: \infty$. Using our $S(q)$ we observed that the integral is consistent
410: with the sum rule and oscillates nicely about the limiting value for
411: $Q \ge 3 $\AA$^{-1}$, while using the $S(q)$ from the neutron
412: scattering experiments~\cite{Holland2}, we observe a positive drift in
413: the mean value about which the integral oscillates. Such a drift could
414: indicate the presence of spurious background corrections. The $S(q)$
415: from the X-ray experiment~\cite{Waseda} seems to be in good agreement
416: with the ideal sum rule.
417:
418: \begin{figure}
419: \includegraphics[width=3in,angle=-90]{Fesq}
420: \caption{\label{fig:Fesq} Comparison of simulated and experimental S(q) near the melting temperature of Fe.}
421: \end{figure}
422:
423: As we lower the temperature, the peak heights in $S(q)$ grow,
424: indicating an increase in short range order with supercooling. We also
425: observe a slight shoulder in the second peak of the $S(q)$
426: (Fig.~\ref{fig:Fesq}), which grows with supercooling. The split
427: second peak positions are in the ratios of 20:12 and 24:12 with
428: respect to the first peak positions, just what one would ideally
429: observe if there was icosahedral order~\cite{NelsonWidom,Sachdev}.
430:
431:
432: \subsection{\label{sec:alloyW6} Bond Orientation Order Parameter \w6}
433:
434: %\begin{table}
435: %\caption{\label{tab:W6} $\hat{W}_6$ values for a few clusters. The
436: % Z8-Z16 series are Kasper-Polyhedrons including the perfect
437: % icosahedron (Z12).}
438: %\begin{tabular}{|c||c|c|c||c|c|c|c|c|c|c|}
439: %\hline
440:
441: %Cluster & HCP & FCC & BCC & Z8 &Z9&Z10&Z12&Z14&Z15&Z16\\
442: %\hline
443: %No. of atoms & 12 & 12 & 14 & 8 &9 &10 &12&14&15&16\\
444: %\hline
445: %Voronoi type &
446: %(0,12,0)&(0,12,0)&(0,6,0,8)&(0,4,4)&(0,3,6)&(0,2,8)&(0,0,12)&(0,0,12,2)&(0,0,12,3)&(0,0,12,4)\\
447: %\hline
448: %$\hat{W}_6$ & -0.012 & -0.013 & +0.013& +0.010 &-0.038 &-0.093&-0.169&-0.093&-0.037&+0.013\\
449: %\hline
450: %\end{tabular}
451: %\end{table}
452:
453: \begin{table}
454: \caption{\label{tab:W6} $\hat{W}_6$ values for a few clusters. The
455: Z8-Z16 series are Kasper-Polyhedrons including the perfect
456: icosahedron (Z12).}
457: \begin{tabular}{|c||c|c|c|c|}
458: \hline
459: Cluster &No. of atoms &Voronoi type&$\hat{W}_6$\\
460: \hline
461: HCP&12&(0,12,0)&-0.012\\
462: FCC&12&(0,12,0)&-0.013\\
463: BCC&14&(0,6,0,8)&+0.013\\
464: Z8&8&(0,4,4)&+0.010\\
465: Z9&9&(0,3,6)&-0.038\\
466: Z10&10&(0,2,8)&-0.093\\
467: Z12&12&(0,0,12)&-0.169\\
468: Z14&14&(0,0,12,2)&-0.093\\
469: Z15&15&(0,0,12,3)&-0.037\\
470: Z16&16&(0,0,12,4)&+0.013\\
471: \hline
472: \end{tabular}
473: \end{table}
474:
475:
476: \begin{figure}
477: \includegraphics[width=3in,angle=-90]{w6_Fe_Cu}
478: \caption{\label{fig:w6_Fe_Cu}Distribution of \w6 in liquid and supercooled liquid Fe and Cu. Fe shows more pronounced icosahedral order than Cu with supercooling}
479: \end{figure}
480:
481: \begin{figure}
482: \includegraphics[width=3in,angle=-90]{Fe_w6lowT}
483: \caption{\label{fig:Fe_w6lowT}Distribution of \w6 around Fe atoms, in
484: supercooled Fe, FeB and FeZrB at T=800K}
485: \end{figure}
486:
487:
488: Steinhardt, {\em et al.}~\cite{Steinhardt} introduced the $\hat{W}_l$
489: parameters as a measure of the local orientational order in liquids
490: and undercooled liquids. To calculate $\hat{W}_l$, the orientations of
491: bonds from an atom to its neighboring atoms are projected onto a basis
492: of spherical harmonics. Rotationally invariant combinations of
493: coefficients are then averaged over many atoms in an ensemble of
494: configurations. The resulting measures of local orientational order
495: can be used as order parameters to characterize the liquid
496: structures. For an ideal icosahedral cluster, $l=6$ is the minimum
497: value of $l$ for which \w6 $\ne 0$. Table~\ref{tab:W6} enumerates \w6
498: values for different ideal clusters. The ideal icosahedral value of
499: \w6 is far from other clusters, making it a good icosahedral order
500: indicator.
501:
502: Kasper-Polyhedrons~\cite{Kasper_Doye,Nelson83,Frank-Kasper} are
503: polyhedrons which minimize the number of disclinations for a
504: particular coordination number. The series Z8-Z16 in
505: Table.~\ref{tab:W6} are such Kasper-Polyhedrons but with the added
506: constraint that the surface atoms be triangulated with equilateral
507: triangles. The icosahedron with a coordination of 12 is one such
508: Kasper-Polyhedron with no disclinations. Adding disclinations to the
509: icosahedron, one finds that each disclination increases the \w6 value
510: by the same amount irrespective of its sign.
511:
512:
513: As before, we choose the cutoff distance to specify near neighbors as
514: $R_{cut} =3.5$\AA. For pure Fe at high temperatures, \w6 resembles
515: that of high temperature liquid Cu~\cite{Cupaper}, which in turn
516: resembles the maximally random jammed configuration of hard
517: spheres~\cite{MRJ}(Fig.~\ref{fig:w6_Fe_Cu}). On this basis we suggest
518: that the MRJ configuration represents an idealized structure that is
519: universal for strongly repulsive interactions. All pure metallic
520: systems might approach this ideal structure at sufficiently high
521: temperature.
522:
523: However, as temperature drops the \w6 distribution shifts strongly to
524: the left, with a pronounced increase in \w6 $\leq$ -0.1
525: (Fig.~\ref{fig:Fe_w6lowT}). This indicates a rather high concentration
526: of nearly icosahedral clusters in supercooled liquid Fe. Even at very
527: low percentages of supercooling of Cu ($\sim$ 3\%~\cite{Cupaper}) and
528: Fe ($\sim$ 2\%, not shown), Fe shows a clear enhancement in the
529: negative \w6 distribution as compared to Cu.
530:
531: We checked the \w6 distribution of a non-magnetic simulation to see if
532: it is strongly influenced by magnetism, and found a nearly identical
533: result. In particular, we still found a strong enhancement of the
534: nearly icosahedral clusters relative to liquid Cu or pure Fe at high
535: temperature.
536:
537: \subsection{\label{sec:Vor}Voronoi Analysis}
538:
539:
540: \begin{table}
541: \caption{\label{tab:Voronoi} Voronoi statistics (\% of Fe atoms) for typical Fe-centered clusters in Fe, FeB and FeZrB. Edges smaller than 0.33\AA~ and faces smaller than 0.3~\AA$^2$ have been treated as a single vertex.}
542: \begin{tabular}{|c|c|c|c|c|c|c|}
543: \hline
544: Voronoi&\multicolumn{4}{c}{Fe}
545: &\multicolumn{1}{|c}{FeB}
546: &\multicolumn{1}{|c|}{FeZrB}\\
547: \cline{2-7}
548: type&\multicolumn{2}{c|}{supercooled}
549: %&\multicolumn{1}{c|}{}
550: &\multicolumn{2}{c|}{defective}
551: &\multicolumn{1}{c|}{supercooled}
552: &\multicolumn{1}{c|}{supercooled}\\
553: %&\multicolumn{1}{c|}{} && \\
554: &\multicolumn{2}{c|}{liquid}
555: %&\multicolumn{1}{c|}{}
556: &\multicolumn{2}{c|}{crystal}
557: &\multicolumn{1}{c|}{liquid}
558: &\multicolumn{1}{c|}{liquid}\\
559: %&\multicolumn{1}{c|}{} && \\
560: \cline{2-7}
561: &\multicolumn{1}{c|}{800K}
562: &\multicolumn{1}{c|}{relaxed}
563: &\multicolumn{1}{c|}{800K}
564: &\multicolumn{1}{c|}{relaxed}
565: &\multicolumn{1}{c|}{800K}
566: &\multicolumn{1}{c|}{800K}\\
567: \hline
568: (0,0,12) &8.3 &11.0 &5.0 &0.0 &0.6 &1.4\\
569: (0,0,12,2) &2.1 &3.6 &2.0 &2.0 &0.5 &0.3\\
570: (0,2,10,\{0,1\}) &6.2 &6.8 &2.0 &0.0 &4.4 &7.1\\
571: (0,1,10,2) &3.6 &5.4 &3.0 &0.0 &2.1 &3.1\\
572: (0,4,8,\{0,1,2,3\}) &7.9 &4.8 &10.0 &6.0 &8.6 &9.4\\
573: (0,3,8,\{0,1,2,3\}) &6.8 &6.0 &12.0 &0.0 &4.3 &8.3\\
574: (0,2,8,\{1,2,3,4\}) &3.0 &3.0 &3.0 &0.0 &0.0 &0.0\\
575: (0,3,6,4) &2.1 &2.2 &2.0 &6.0 &0.0 &0.0\\
576: (0,5,4,4) &1.5 &1.2 &3.0 &18.0 &0.0 &0.0\\
577: (0,6,0,8) &0.0 &0.4 &0.0 &56.0 &0.0 &0.0\\
578: \hline
579: \end{tabular}
580: \end{table}
581:
582:
583: To explain the origin of this low \w6 peak in pure Fe, we performed a
584: Voronoi analysis~\cite{Finney} of the liquid before and during
585: crystallization. A Voronoi polyhedron is described by indices
586: $(F_{3},F_{4},F_{5},...)$ where $F_{i}$ denotes the number of faces
587: with $i$ edges. For example (0,0,12) denotes an icosahedron, while
588: (0,0,12,2) denotes a 14-coordinated (Z14) atom, with 12 5-fold bonds
589: and 2 6-fold bonds. The (0,0,12,2) is a characteristic TCP
590: (tetragonal close-packed) structure of the Frank-Kasper type, with a
591: $-72^{\circ}$ disclination line running through an otherwise perfect
592: icosahedron. In a body-centered cubic crystal all atoms are of
593: Voronoi type (0,6,0,8), which is an alternate 14-coordinated
594: structure.
595:
596: Supercooled liquid Fe at T=800K (Table~\ref{tab:Voronoi}) contains a
597: high fraction of icosahedral atoms of type (0,0,12) and (0,0,12,2)
598: . Those with Voronoi type (0,2,10,\{0,1\}) and (0,1,10,2) also have
599: very negative \w6, so that they can be thought of as related to the
600: icosahedron. Together, they explain the enhanced negative \w6
601: distribution. Supercooled Cu (which shows weak icosahedral order) and
602: high temperature Fe (T=1900K), in contrast contain about 1.2\% of
603: (0,0,12) and no (0,0,12,2). The MRJ configuration contains 1.2\%
604: (0,0,12) but no (0,0,12,2).
605:
606: Strikingly, the icosahedral clusters tend to join in pairs and strings
607: of 3 atoms in length. An instantaneous quench of a particular liquid
608: structure at 800K which had a high fraction of icosahedral units,
609: using conjugate gradient relaxation of the atomic coordinates and
610: lattice parameters, shows a clear enhancement of the icosahedral and
611: other closely related units. The strings of icosahedral units found in
612: the supercooled liquid connected to form networks. A quench starting
613: from a different instantaneous supercooled liquid structure at 800K
614: containing fewer icosahedral structures resulted in rapid
615: crystallization. This indicates that the presence of icosahedrons may
616: inhibit crystallization. Similar quenches, starting from
617: instantaneous liquid structures at higher temperatures, also resulted
618: in relaxed structures with some of them showing a high negative \w6
619: distribution comparable to the distribution at 800K. The quenches
620: that partially crystallized showed a high fraction of BCC (0,6,0,8)'s
621: ($\sim$ 40\%) and was always accompanied by a high fraction of
622: (0,5,4,4)'s ($\sim$ 25\%), which are otherwise absent or very low in
623: the liquid.
624:
625:
626: Under TMD the supercooled liquid at T=800K eventually
627: crystallizes, with the \w6 distribution peaked strongly around zero
628: consistent with the value for ordinary crystalline clusters, but
629: retaining a subset of atoms with nearly icosahedral \w6$<$-0.14. Not
630: surprisingly, these were precisely the atoms that had Voronoi type
631: (0,0,12) prior to crystallization. The Z14 (0,0,12,2) atoms (with
632: \w6 $\sim$ -0.093) were mutual near-neighbors, linked along their
633: 6-fold ($-72^{\circ}$) bonds, and also were neighbors of the nearly
634: icosahedral atoms.
635:
636: \begin{figure}
637: \includegraphics[width=3in]{defect.ps}
638: \caption{\label{fig:defect}Fragment of BCC crystal illustrating
639: a central bond in the [1,1,1] (vertical) direction, surrounded by two
640: equilateral triangles rotated 60$^{\circ}$ from each other and
641: displaced on either side of the midplane. Balls and sticks
642: follow ZomeTool convention: Balls are icosahedral, yellow sticks are
643: 3-fold and blue sticks are 2-fold bonds.}
644: \end{figure}
645:
646: We quenched this sample by conjugate gradient relaxation of atomic
647: coordinates and lattice parameters. After relaxation we found 56
648: atoms had BCC Voronoi type (0,6,0,8). Six icosahedral atoms became
649: 12-coordinated (0,4,8) structures surrounding the bond connecting the
650: two (0,0,12,2) disclinated icosahedral atoms, which retained their type.
651: The remaining atoms served to link the cluster of icosahedron-related
652: atoms to the surrounding defect-free BCC crystal.
653:
654: The icosahedron-related structure thus forms a point defect in an
655: otherwise perfect BCC crystal. A simple way to create this defect is
656: to take two consecutive triangles surrounding a near-neighbor bond
657: along the BCC [1,1,1] direction (see Fig.~\ref{fig:defect}), displace
658: them into the perpendicular bisecting plane and rotate by
659: 30$^{\circ}$. The atoms along the [1,1,1] bond are now connected to
660: each other by a six-fold bond ($-72^\circ$ disclination) and are
661: connected to the six displaced atoms by five-fold bonds. This scheme
662: to transform a BCC rhombic dodecahedron into a Frank-Kasper Z14
663: polyhedron was described in Ref~\cite{Dubois,Khmelevskaya} in an
664: attempt to explain certain diffraction anamolies in iron and
665: vanadium-based alloys under ion irradiation.
666:
667: Manually removing the defect, by reversing the above procedure then
668: relaxing, yields a perfect BCC crystal with all atoms in a (0,6,0,8)
669: Voronoi environment. We also embedded the point defect in an
670: otherwise perfect BCC crystal with 128 atoms in a cubic box.
671: Relaxation showed that the defect was stable. The energy of the defect
672: was 6.18eV and the fractional volume increase was 0.013.
673:
674: Since tungsten, like Iron, crystallizes in BCC, we performed a
675: separate first-principles simulation of liquid Tungsten. After
676: supercooling by 13\% to T=3200K, we found the \w6 histogram resembled
677: that of Fe at 1600K (which is 15\% supercooled). A Voronoi analysis
678: revealed a similar percentage of (0,0,12)'s and (0,0,12,2)'s in both
679: these elemental BCC-forming metals. Hence we believe the relationship
680: between BCC and icosahedral structures may be linked to our observed
681: high fraction of icosahedra in liquid Fe as compared to Cu. This may
682: also explain the reason why we are able to supercool BCC Fe more
683: deeply than FCC Cu in our simulations.
684:
685: We also included an icosahedron point defect inside an otherwise
686: perfect FCC crystal of Cu, with 256 atoms in a cubic box. Relaxation
687: showed that the defect was stable. The energy cost of the defect was
688: 4.69eV and the fractional volume increase was 0.012. Even though
689: this FCC defect cost less energy than the point defect in BCC, and
690: also needs little rearrangement of atoms as reflected by the slightly
691: lower fractional density change, we do not see a significant
692: icosahedral order in liquid Cu, when compared to liquid Fe. This could
693: be because the number of icosahedrons that a single Z14 disclination
694: can stabilize (up to six) is greater than one. So even an equal number
695: of the two different defects in BCC and FCC would result in more
696: icosahedral order in BCC than in FCC.
697:
698: \subsection{\label{sec:Fe_HA}Honeycutt and Andersen analysis}
699:
700: \begin{figure}
701: \includegraphics[width=3in,angle=-90]{HA_Fe.ps}
702: \caption{\label{fig:HA_Fe}Honeycutt-Andersen analysis for pure Fe
703: shows a clear increase in five-fold bonds with supercooling.}
704: \end{figure}
705:
706: Honeycutt and Andersen~\cite{Honeycutt} introduced a useful assessment
707: of local structure surrounding interatomic bonds. We employ a
708: simplified form of their analysis, counting the number of common
709: neighbors shared by a pair of near-neighbor atoms. This identifies
710: the number of atoms surrounding the near-neighbor bond and usually
711: equals the number of edge-sharing tetrahedra whose common edge is the
712: near-neighbor bond. We assign a set of three indices to each bond. The
713: first index is 1 if the root pair is bonded (separation less than or
714: equal to $R_{cut}$). The second index is the number of near-neighbor
715: atoms common to the root pair, and the third index gives the number
716: of near-neighbor bonds between these common neighbors. We take the
717: same value of $R_{cut}$=3.5\AA~ as mentioned before. Note that the Honeycutt
718: and Andersen fractions depend sensitively on $R_{cut}$, making precise
719: quantitative comparisons with other prior studies difficult.
720:
721: In general, 142's are characteristic of close packed structures (FCC
722: and HCP) and 143's are characteristic of distorted
723: icosahedra~\cite{Ni-Ag}. They can also be considered as
724: $+72^{\circ}$ disclinations~\cite{SM82,Nelson83,Sethna83}. Likewise,
725: 15's are characteristic of icosahedra, with 155's characterizing
726: perfect icosahedra while 154's and 143's characterize distorted
727: ones. 16's indicate $-72^{\circ}$ disclinations. 166's and 144's are
728: also characteristic of BCC.
729:
730:
731: A Honeycutt and Andersen analysis for pure Fe, with an
732: $R_{cut}$=3.5\AA~ (Fig.~\ref{fig:HA_Fe}), showed that with
733: supercooling the fraction of 15 bonds rises from 0.46 at T=1900K to
734: 0.59 at T=800K in the liquid before crystallization. The fraction of
735: 155's (characteristic of perfect icosahedra) was always larger than
736: 154's (characteristic of distorted icosahedra), and seemed to be
737: steeply increasing with supercooling as opposed to 154's, which were
738: relatively flat. The fraction of 14 bonds drop from 0.32 to 0.30, with
739: the icosahedral 143's being always higher than the cubic 142's. The
740: 144's, which are characteristic of BCC remain nearly flat, even though
741: the 166's show a slight increase. The ease of embedding a Z14
742: disclination in BCC Fe (see section~\ref{sec:Vor}) might explain the
743: slight increase in the 166's.
744:
745: \section{\label{sec:alloy} Fe-B and Fe-Zr-B}
746: \subsection{\label{sec:Alloygr}Radial Distribution Function $g(r)$}
747:
748: \begin{figure}
749:
750: \includegraphics[width=3in,angle=-90]{Alloy_Fegr.ps}
751: \includegraphics[width=3in,angle=-90]{Alloy_ZrBgr.ps}
752: \caption{\label{fig:Alloy_gr}Partial radial distribution functions in FeZrB at T=800K.}
753: \end{figure}
754:
755: Fig.~\ref{fig:Alloy_gr} shows the pair correlation functions of
756: supercooled FeZrB at T=800K. From the heights of the first peaks, we
757: see that the strongest bonds form between the metalloid (B) and the
758: metal (Fe or Zr). The relative bond lengths reveal, as expected, that
759: B behaves as a small atom, Fe is medium sized and Zr is
760: large. Comparing simulations of FeZrB to pure Fe, we find that
761: $g_{FeFe}$ is reduced by alloying with B or ZrB, because Fe prefers to
762: associate with B or Zr rather than with Fe.
763:
764: \begin{table}
765: \caption{\label{tab:NcFeB}Average coordination number in FeB.}
766: \begin{tabular}{|r|r|r|r|r|}
767: \hline
768: &FeFe & FeB & BFe & BB\\
769: \hline
770: $R_{cut}$&3.4&3.0&3.0&2.3\\
771: \hline
772: $N_{\alpha\beta}$&12.0 & 2.2 & 8.9 & 0.4 \\
773: \hline
774: \end{tabular}
775: \end{table}
776:
777: \begin{table}
778: \caption{\label{tab:NcFeZrB}Average coordination number in FeZrB.}
779: \begin{tabular}{|r|r|r|r|r|r|r|r|r|r|}
780: \hline
781: &FeFe& FeZr & FeB & ZrFe & ZrZr &
782: ZrB & BFe & BZr & BB\\
783: \hline
784: $R_{cut}$&3.4&3.8&3.0&3.8&4.0&3.3&3.0&3.3&2.4\\
785: \hline
786: $N_{\alpha\beta}$&9.5 & 2.0 & 2.0 & 14.0 & 0.9 & 2.8 & 6.8 & 1.4 & 0.4 \\
787: \hline
788: \end{tabular}
789: \end{table}
790:
791: To calculate the coordination number, define $N_{\alpha \beta}$ as the
792: average number of atoms of type $\beta$ around an atom of type
793: $\alpha$, We set $R_{cut}$ at the first minima of the partial radial
794: distribution functions (Fig.~\ref{fig:Alloy_gr}). We list the partial
795: coordination numbers of FeB in Table.~\ref{tab:NcFeB} and FeZrB in
796: Table.~\ref{tab:NcFeZrB} (averaged over all the temperatures since the
797: temperature dependence is very weak and non-monotonic). The average
798: value of $N_{FeFe}$ decreases with alloying, due to decrease in the
799: concentration of Fe and the preference to bind with B and Zr. Zr,
800: being a large atom, has a larger coordination number. Also note that
801: B and Zr favor each other more than themselves. We do find some B-B
802: pairs in the liquid state.
803: \subsection{\label{sec:Alloysq} Liquid Structure Factor S(q)}
804:
805: \begin{figure}
806: \includegraphics[width=3in,angle=-90]{Sq_FeB.ps}
807: \caption{\label{fig:Sq_FeB}Simulated partial structure factors of
808: FeB at T=800K}
809: \end{figure}
810:
811: Fig.~\ref{fig:Sq_FeB} shows the Faber-Ziman~\cite{Fischer} partial
812: structure factors of FeB at the lowest simulated temperature of
813: T=800K, defined as,
814: \begin{equation}
815: %S(q)=\sum_{\alpha\beta} x_\alpha x_\beta S_{\alpha\beta}(q)
816: S_{\alpha\beta}(q)=1+4\pi\rho\int\limits_{0}^\infty[g_{\alpha\beta}(r)-1] {\sin(qr)\over{qr}}dr
817: \end{equation}
818: The positions of the first and second peaks in partial $S_{FeFe}(q)$
819: is in very good agreement with the experimental results for amorphous
820: FeB~\cite{Lamparter}. Also, at the position of the splitting of the
821: second peak in $S_{FeFe}(q)$, as observed in the experiments, we
822: observe a slight shoulder. Similarly, the positions of the different
823: peaks in $S_{FeB}(q)$ and $S_{BB}(q)$ are in agreement with the
824: experiments.
825:
826: For FeB, the $q \rightarrow 0$ limit of $S_{FeFe}(q)$ is comparable to
827: the experimental value. The $q \rightarrow 0$ limit in the other two
828: partials differ from the experiment, $S_{BB}(q)$ more seriously than
829: $S_{FeB}(q)$. We think that this discrepancy in the long wavelength
830: regime is probably due to the very low density of B in our system
831: leading to poor statistics. Nevertheless, the excellent agreement in
832: the positions of the different peaks in the partial structure factors
833: shows that we have reasonably good representative structures of FeB at
834: T=800K. Partial structure factors or partial pair distribution
835: functions are not available experimentally to compare with FeZrB
836: simulations.
837:
838: \subsection{\label{sec:AlloyW6} Bond Orientation Order Parameter \w6}
839:
840: \begin{figure}
841: \includegraphics[width=3in,angle=-90]{otherw6}
842: \caption{\label{fig:otherw6}Distribution of \w6 for different chemical
843: species in supercooled FeZrB at T=800K}
844: \end{figure}
845:
846: To define the \w6 distribution in an alloy, we concentrate on central
847: atoms of some particular species (e.g. Fe) but consider the
848: neighboring atoms of all species. We chose the near-neighbor cutoff
849: distances as before. Compare the Fe-based \w6 distributions at their
850: supercooled temperatures in Fig.~\ref{fig:Fe_w6lowT}.
851: % The FeB distribution is similar to high temperature \w6 distribution
852: %of Fe, but the FeZrB curve has significantly shifted to the left,
853: %extending more towards the ideal icosahedral value of \w6.
854: The origin of high negative \w6 values for pure Fe was previously
855: explained in section~\ref{sec:Vor}. Replacing a few medium sized Fe
856: atoms with smaller B atoms causes negative disclination lines to
857: concentrate on Fe, leading to a drop in the ideal icosahedral
858: clustering on these Fe atoms and strongly reducing the extreme
859: negative values of \w6 in FeB. On the contrary, inclusion of Zr in
860: FeZrB causes negative disclinations to attach to Zr, easing
861: frustration, leading to more Fe centered clusters with icosahedral
862: ordering and increasing the negative region of \w6, as compared to
863: FeB. Inclusion of the big Zr atom enhances icosahedral order on the Fe
864: atoms.
865:
866: Fig.~\ref{fig:otherw6} shows the \w6 distributions for FeZrB with
867: centers at Fe, Zr and B. The histogram with the center at Zr is almost
868: symmetric about the value of zero. This suggests that the local
869: environment about Zr atoms is nearly spherical, as is expected given
870: its large size. The B centered \w6 histogram is also asymmetric
871: towards negative \w6 values due to Kasper-Polyhedrons and slightly
872: distorted versions of them (see Table.~\ref{tab:W6} and
873: Section~\ref{sec:alloyW6}).
874:
875: \subsection{\label{sec:AlloyVor}Voronoi Analysis}
876:
877: %\begin{table}
878: %\caption{\label{tab:Voronoi_alloy}Voronoi analysis for supercooled FeB
879: %and FeZrB at T=800K}
880: %\begin{tabular}{|c||c|c|}
881: %\hline
882: %Fe centred Voronoi& percentages in FeB &percentages in FeZrB \\
883: %\hline
884: %(0,0,12)&0.5 &1 \\
885: %(0,0,12,2)& 0.4&0.2\\
886: %(0,2,10,\{0,1,2\})& 3.5&5\\
887: %(0,1,10,2)& 1.7 &2.2\\
888: %(0,4,8,\{0,1,2,3,4\})& 6.9&6.6\\
889: %(0,3,8,\{1,2,3,4\})& 4.4&5.8\\
890: %\hline
891: %\hline
892: %B centred Voronoi & percentages in FeB&percentages in FeZrB\\
893: %\hline
894: %(0,3,6)&3.2&3.6\\
895: %(0,3,6,1)&1.4&0.4\\
896: %(0,5,4)&4.0&4.8\\
897: %(0,4,4)&0.6&1.2\\
898: %(0,2,8)&2.0&1.4\\
899: %\hline
900: %\end{tabular}
901: %\end{table}
902:
903: A Voronoi analysis was performed for FeB in the supercooled liquid at
904: T=800K (Table~\ref{tab:Voronoi}). The Fe environments were mostly
905: (0,4,8,x) or (0,3,8,x) types, where x=\{0,1,2,3,4\}, with the higher
906: coordination polyhedron being more favored. This was followed by the
907: very negative \w6 valued (0,0,12), (0,2,10,x)'s and (0,1,10,2)'s
908: Voronoi types occurring at lower frequency than in pure Fe. Boron
909: mainly had environments of type (0,3,6)'s ($\sim$ 15\% of B atoms)
910: (Kasper polyhedron for Z=9 containing a $+72^{\circ}$ disclination)
911: and (0,5,4)'s ($\sim$ 20\%). These types are typical of the
912: tri-capped trigonal prism (TTP) and the mono-capped square archimedean
913: prism (a slightly distorted variant of TTP) respectively. The TTP is
914: found in the crystal structures of Fe$_3$B with Pearson symbols oP16
915: and tI32. The distorted (0,5,4) version is found in the structure
916: Fe$_{23}$B$_6$ of Pearson type cF116. These structures have been
917: identified as the leading competitors for B-Fe glass~\cite{Marek_Fe}.
918: Boron also took environments of the Kasper polyhedron (0,4,4) ($\sim$
919: 3\%) corresponding to Z=8 and (0,2,8) ($\sim$ 10\%) corresponding to
920: Z=10. The association of B with $+72^{\circ}$ disclinations explains
921: how it increases the frustration of icosahedral order on the Fe atoms.
922: Clearly the improved glass-formability of FeB compared with elemental
923: Fe cannot be due to icosahedral order. Rather, it is presumed to be
924: caused by the deep eutectic at Fe$_{83}$B$_{17}$.
925:
926: A Voronoi analysis of supercooled FeZrB at T=800K shows a clear
927: increase in the very negative \w6 polyhedrons, and also a decrease in
928: the number of Z14 (0,0,12,2) types on Fe atoms indicating a decrease
929: in frustration in the ternary as compared to the binary. Environments
930: around B atoms were roughly similar in the binary and the ternary,
931: with a slight increase in the lower coordinated (Z=8) Kasper
932: polyhedrons at the cost of higher coordinated (Z=10) ones. Zirconium
933: took a variety of polyhedra, with an average coordination of 17.6 and
934: a minimal coordination of 15, owing to its large size compared to the
935: other constituents.
936:
937: \subsection{\label{sec:Alloy_HA}Honeycutt and Andersen analysis}
938:
939: \begin{table}
940: \caption{\label{tab:HA_alloy}HA analysis for supercooled FeZrB at T=800K}
941: \begin{tabular}{|r|r|r|r|r|r|r|}
942: \hline
943: \multicolumn{7}{|c|}{Root Pair} \\
944: \hline
945: & Fe-Fe & Fe-Zr & Fe-B & Zr-Zr & Zr-B & B-B\\
946: & & & & & &\\
947: \hline
948: \bf 14 pairs &\bf 0.30 & \bf 0.19 &\bf 0.54 & \bf 0.11 &\bf 0.55 &\bf 0.75\\
949: \hline
950: 142 pairs & 0.08 & 0.06 & 0.03 & 0.06 & 0.03 & 0.02\\
951: \hline
952: 143 pairs & 0.17 & 0.11 & 0.35 & 0.04 & 0.33 & 0.33\\
953: \hline
954: 144 pairs & 0.05 & 0.02 & 0.17 & 0.01 & 0.19 & 0.41\\
955: \hline
956: \hline
957: \bf 15 pairs &\bf 0.56 &\bf 0.53 &\bf 0.39 &\bf 0.48 &\bf 0.41 &\bf 0.18 \\
958: \hline
959: 154 pairs & 0.22 & 0.27 & 0.03 & 0.36 & 0.04 & 0.0 \\
960: \hline
961: 155 pairs & 0.33 & 0.22 & 0.37 & 0.08 & 0.36 & 0.17\\
962: \hline
963: \hline
964: \bf 16 pairs &\bf 0.10 &\bf 0.24 &\bf 0.01 &\bf 0.36 &\bf 0.01 &\bf 0.01\\
965: \hline
966: 166 pairs & 0.09 & 0.20 & 0.0 & 0.26 & 0.00 & 0.00\\
967: \hline
968: \end{tabular}
969: \end{table}
970:
971: We made a HA analysis of the ternary glassy alloy, by looking at root
972: pairs of chemical species $\alpha$ and $\beta$, choosing $R_{cut}$ in
973: the manner of the \w6 analysis. The frequency is normalized to sum to
974: one for each species pair $\alpha\beta$. Table~\ref{tab:HA_alloy} lists
975: the fraction of different $1x$ pairs in supercooled FeZrB.
976:
977: Among the $1x$ pairs with Fe as one of the root pairs, the 15's are
978: most abundant at all temperatures. The 15's are mainly comprised of
979: 155's and the 154's, with the 155's being always higher than the
980: 154's. The percentage of 15's is similar for FeZrB, FeB as well as
981: pure Fe. Note that the 15's are largest for the Fe-Fe pairs. They
982: also show a steady enhancement with supercooling, unlike the 14's
983: which decrease with supercooling. Among the 14's, the
984: icosahedron-related 143's for all root pairs are always higher than
985: the close-packed 142's, and remain fairly constant with
986: supercooling. The 14's are maximal for FeB root pairs and minimal for
987: FeZr.
988:
989: Supercooled pure Fe has a high percentage of ``16'''s (13\%) compared
990: to pure Cu (7\%). The high number of ``16'''s in Fe is related to the
991: occurance of Z14 (0,0,12,2) environments in which the 6-fold bond
992: carries a $-72^{\circ}$ disclination, rather than the BCC 14 atom
993: arrangement which would also show a high degree of four-fold and
994: six-fold bonds (section~\ref{sec:Vor}). Adding B to Fe, shows an
995: increase of ``16'''s for Fe-Fe pairs to 21\% in FeB (not
996: shown). Adding large Zr atoms to FeB decreases the occurance of 16's
997: on the Fe-Fe pairs, putting them on Fe-Zr pairs and easing the
998: frustration of Fe centers. This causes the geometry about Fe centers
999: to be more icosahedral, and hence the shift of the \w6 towards
1000: negative values.
1001:
1002: Among the root pairs not containing Fe, note that the B-B pair has the
1003: maximum 14's ($+72^{\circ}$ disclinations), especially 144's, while
1004: the Zr-Zr pair has the maximum 16's ($-72^{\circ}$ disclinations),
1005: especially the 166's, emphasizing the role of size in controlling the
1006: frustration in alloys.
1007:
1008: \section{\label{sec:conclude} Conclusion}
1009:
1010: This study quantifies icosahedral and polytetrahedral order in
1011: supercooled liquid metals and alloys. This is the first such analysis
1012: of glass-forming Fe compounds using configurations from
1013: first-principles simulations. While the structural properties of Fe
1014: and Cu strongly resemble each other at high temperature, and indeed
1015: are close to a maximally random jammed structure~\cite{MRJ,Cupaper},
1016: their behavior evolves substantially, and in different manners, as the
1017: liquid is supercooled. Proper modeling of atomic interactions is
1018: essential to capture the differing behavior of each element, and use
1019: of a first-principles simulation is the most reliable means of
1020: achieving this.
1021:
1022: For pure elements we find the degree of local icosahedral order in the
1023: supercooled liquid depends on the low temperature crystal structure,
1024: with BCC metals such as Fe and W accommodating icosahedra more readily
1025: than the FCC element Cu. Alloying with large or small atoms can
1026: further influence the degree of icosahedral order, with small atoms
1027: (e.g. B in Fe) aggravating the frustration by introducing positive
1028: disclination line defects, while large atoms (e.g. Zr in Fe) naturally
1029: stabilize negative disclination line defects, relieving frustration on
1030: the medium-sized Fe atoms. While the enhanced glass-forming ability
1031: of FeB compared to Fe cannot be related to icosahedral order, we
1032: suggest that the formation of icosahedral order and disclination line
1033: network, together with the slow dynamics of chemical ordering in a
1034: complex alloy and the destabilization of competing crystal
1035: phases~\cite{Marek_Fe}, enhances the glass-forming ability of FeZrB
1036: compared with FeB.
1037:
1038:
1039: \begin{acknowledgements}
1040: This work was supported in part by DARPA/ONR grant N00014-01-1-0961.
1041:
1042: \end{acknowledgements}
1043:
1044: \bibliography{ico}% Produces the bibliography via BibTeX.
1045:
1046: \end{document}
1047: