1: \documentclass[aps,prl,twocolumn,groupedaddress]{revtex4}
2:
3: \usepackage{graphicx}
4: \newcommand{\be}{\begin{equation}}
5: \newcommand{\ee}{\end{equation}}
6:
7: \begin{document}
8:
9: %Suggested reviewers:
10: %G. V. Chester, A. Griffin, R. B. Hallock, J. K. Johnson, A. J. Leggett, A. D. Migone, O. %E. Vilches, M.D. Miller, R.A. Guyer, A.L. Fetter
11:
12: \title{Bose-Einstein Condensation of Molecular Hydrogen in Nanotube Bundles}
13: \author{Francesco Ancilotto$^{1,2}$, M. Mercedes Calbi$^1$, Silvina M. Gatica$^1$ and Milton W. Cole$^1$}
14: \affiliation{$^1$Physics Department, Pennsylvania State University, University Park, Pennsylvania 16802; \\ $^2$ INFM (UdR Padova and DEMOCRITOS National Simulation Center, Trieste, Italy) \\ and Dipartimento di Fisica ``G. Galilei'',
15: Universit\`a di Padova,\\ via Marzolo 8, I-35131 Padova, Italy.}
16:
17:
18: \date{\today }
19:
20: \begin{abstract}
21:
22: We evaluate the effects of heterogeneity on the density of states of H$_2$ molecules
23: inside interstitial channels within bundles of carbon nanotubes. As temperature
24: (T) falls, the density increases within those tubes having the greatest binding
25: energy. At T $\approx$ 10 mK, the molecules undergo Bose-Einstein condensation,
26: exhibiting a singular heat capacity.
27:
28: \end{abstract}
29: \pacs{}
30: \maketitle
31:
32: The subject of adsorption within bundles of carbon nanotubes has received
33: considerable attention recently, owing to both its fundamental interest and potential
34: applications (e.g. gas storage and isotope separation). One focus of the research is
35: one-dimensional (1D) and quasi-1D phases of matter. These include condensing and
36: crystallizing phases of buckyballs within tubes, He and H$_2$ within the interstitial
37: channels (IC's) between tubes and various gases within grooves on the outside of the
38: nanotube bundles \cite{qsiev,bob,cro,oscCv,rmp,aldo,boro,c60,bienf,carlo,hecond}. Most
39: analyses (including those of our group) have assumed
40: that the tubes are identical and parallel, forming an ordered lattice. Real nanotube
41: bundles, in contrast, consist of a disordered array of tubes with a distribution of
42: radii. A logical question arises: how reliable are predictions that ignore such a
43: variable environment? Shi and Johnson have recently shown that predictions
44: incorporating heterogeneity agree better with adsorption data than do the
45: idealized models \cite{hete}. Stimulated by their work, we have explored the behavior of
46: quantum fluids (He and H$_2$) in such a nonuniform environment at low temperature (T).
47: In this paper, we describe an intriguing result: Bose-Einstein condensation (BEC)
48: of H$_2$ molecules occurs as {\em a consequence of the heterogeneity}. This paper makes
49: predictions about this phenomenon that are testable experimentally. Similar behavior
50: is expected for $^4$He atoms. We note that a superfluid phase of para-H$_2$ in confined
51: geometries has been proposed, but experimental evidence of it is lacking thus far
52: \cite{h2bec}.
53:
54: The occurrence and properties of the BEC transition are determined by the density
55: of states $N(E)$ of the constituent particles. Thus, an initial focus is the determination
56: of that function, derived from the single particle energy spectrum. We assume that
57: particles do not interact (based on the assumption of low density and weak inter-IC
58: interactions). A typical IC presents a highly confining geometry for the molecules
59: if the bundle is close-packed. Because of this confinement (in the x-y plane,
60: perpendicular to the tubes), the molecules have a large zero-point energy of motion.
61: Theoretical values \cite{interc} of this energy are of the order of 500 K. This result is
62: consistent with an enormous isotopic heat difference at low coverage (which equals the
63: binding energy difference) between D$_2$ and H$_2$ in a nanotube bundle observed
64: experimentally \cite{oscar}. That difference was found to be about
65: 250 K, a factor $\approx$ four larger than the difference found on the graphite
66: surface \cite{vid} (due to less localization in that case).
67:
68: Our analysis assumes that molecules can move within the IC's in order to achieve
69: chemical equilibrium with a coexisting vapor phase. At low T, because of the
70: confinement, only the lowest energy state of transverse motion is excited. We
71: call this state's energy $E_t({\bf R})$, where ${\bf R}=(R_1,R_2,R_3)$ is a
72: vector whose components are the radii of the tubes surrounding a particular IC.
73: We have evaluated $E_t({\bf R})$, with interesting results. The calculation
74: assumes that motion parallel to the IC is that of free molecules, so the total
75: energy of a particle with (z component of) momentum p is given by $E(p,{\bf R})=
76: E_t({\bf R}) + p^2/(2m)$. If the potential is corrugated, one must replace the
77: particle's mass $m$ with its band mass \cite{hecond,band}. The density of states
78: for H$_2$ is obtained by summing over the IC's present in the given sample and
79: integrating over p:
80:
81: \begin{eqnarray}
82: N(E)&=&\sum_{{\bf R},p}\; \delta[E - E(p,{\bf R})] \nonumber \\
83: &=& \frac{L}{\hbar \pi} \, \left(\frac{m}{2}\right)^{1/2} \int_0^E \;
84: dE_t \,\frac{g(E_t)}{\sqrt{E - E_t}}
85: \end{eqnarray}
86:
87: Here $g(E_t)$ is the density of states for the transverse oscillation problem
88: and $L$ is the length of the tubes. If the length is not constant, the variation
89: can be included in $g(E_t)$.
90:
91: The energy $E_t({\bf R})$ is evaluated from the potential energy $V({\bf r,R})$
92: of the molecule at position {\bf r}. To compute $V$, we add contributions
93: from the three neighboring tubes, ignoring corrections from more distant tubes
94: and many-body effects associated with the screening of van der Waals interactions
95: by the adjacent tubes \cite{milen}. Although these approximations introduce some errors
96: in the values of $E_t$ they do not affect the key predictions of this work, which
97: are sensitive to the variation of $E_t({\bf R})$. The potential from each neighboring
98: tube was derived with the method of Stan et al \cite{uptake}. Because $V({\bf r,R})$
99: varies
100: rapidly with {\bf r}, values of $E_t({\bf R})$ include large anharmonic and
101: small anisotropic contributions.
102:
103: The form of $N(E)$ for a given collection of nanotube bundles depends on sample
104: preparation. A sample is represented by an ensemble of points (one for each IC)
105: in {\bf R} space. The density of points in {\bf R} space, a function $f({\bf R})$,
106: is defined so that $f({\bf R})\,d{\bf R}$ is the number of IC's within an
107: infinitesimal volume $d{\bf R} = dR_1\,dR_2\,dR_3$, centered at {\bf R}. The function
108: $f$ enters the transverse density of states through this expression:
109:
110: \be
111: g(E) = \int \,d{\bf R}\, f({\bf R})\, \delta[E - E_t({\bf R})]
112: \ee
113:
114: At very low T, we need the value of $f({\bf R})$ only in the immediate vicinity
115: of the global minimum of the energy ($E_m$ at ${\bf R}_m$), but at higher T
116: the specific experimental distribution affects the results quantitatively.
117: Here, we have assumed that the IC's are uniformly distributed in {\bf R}
118: space within a radius spread of width 3 \AA$\,$ near ${\bf R}_m$. More general
119: results will be described in a complete report of this work . We find, as one
120: might expect, that ${\bf R}_m$ occurs along the diagonal, symmetry line
121: ($R_1=R_2=R_3$). Along this line, which we call the (1,1,1) line, there occurs
122: a global minimum $E_m=-1052.97$ K at {$R_i$}=9.95 \AA. To derive $g(E)$,
123: it is important to know
124: the variation of $E_t({\bf R})$ near this minimum. Consider a particular
125: displacement (in the $R_2-R_3$ plane) from ${\bf R}_m$ to a neighboring point
126: for which $R_1$ has the same value, while $R_2$ and $R_3$ are slightly different:
127: $R_2 = R_1+ \delta$ and $R_3=R_1 - \delta$. This change, parallel to the (0,1,-1)
128: direction, yields an extremely small increase in the H$_2$ energy, indicative
129: of a very slow variation of the function $E_t({\bf R})$ near ${\bf R}_m$. This
130: behavior, shown in Fig.1 reveals a long valley of low energy states in this
131: direction; the same behavior occurs along the five equivalent directions, e.g. (1,0,-1).
132:
133: \begin{figure}
134: \includegraphics[height=3in]{contour.eps}
135: \caption{Contour plot showing the variation $E_t({\bf R})-E_m$ near the
136: minimum ${\bf R}_m$ (which is chosen as the origin). From the contour closest to the
137: minimum (the closed one) to the more distant ones, the energy contours correspond to
138: $E_t({\bf R})-E_m = 0.005, 0.01, 0.05, 0.1, 0.5, 1$ and 5 K.}
139: \end{figure}
140:
141: The transverse density of states $g(E)$, from Eq. 2, is shown in Fig. 2. Note
142: that $g(E)$ is proportional to $\sqrt{E - E_m}$ for small $E - E_m$; the
143: prefactor is determined by the principal axes of curvature of the function
144: $E_t({\bf R})$. This square root behavior is identical to that found near a van
145: Hove singularity in the phonon density of states of a 3D system near a minimum
146: in the Brillouin zone, for the same reason- phase space topology \cite{ziman}.
147: At higher energy, instead, the behavior of $g(E)$ switches to $1/\sqrt{E - E_m}$.
148: This (-1/2) power law follows from the fact that the higher energy displacements
149: from ${\bf R}_m$ are quasi-1D. As seen in Fig.1, the iso-energy contours are perpendicular
150: to the diagonal, so that the energy gradient in {\bf R} space is along the diagonal,
151: with essentially constant transverse variation, a 1D situation.
152:
153: \begin{figure}
154: \includegraphics[height=4in]{dos.eps}
155: \caption{Upper panel: Calculated transverse density of states $g(E)$.
156: Lower panel: Total density of states $N(E)$, obtained from Eq.1 in the text.
157: $E_m=-1052.97$ K is the lowest energy level corresponding to $R_1=R_2=R_3=9.95$ \AA.}
158: \end{figure}
159:
160:
161: Fig. 2 also presents $N(E)$, derived with Eq. 1, which convolutes the transverse
162: spectrum with the 1D motion along the axis. The resulting power law behavior
163: can be understood from realizing that if $g(E)$ is proportional to $(E-E_m)^n$,
164: for some $n$, then $N(E)$ is proportional to $(E - E_m)^{n+1/2}$ . Hence, we find
165: that $N(E)$ is proportional to $(E-E_m)$ near threshold. This linear behavior is
166: that characteristic of a 4D gas in free
167: space; the result implies that {\em this system exhibits 4D gas behavior at low T}. For
168: $E>20$ mK above threshold, instead, $N(E)$ becomes approximately constant,
169: corresponding to the density of states of a 2D gas. Thus, there arises a
170: dimensionality crossover originating from the anomalous transverse density
171: of states. We emphasize that the 4D regime is a direct consequence of the
172: existence of a minimum in the function $E_t({\bf R})$, a result that is not sensitive
173: to the details of the calculation \cite{groove}.
174:
175: The thermodynamic behavior of the system is derived with the usual bose gas theory.
176: For a given number of molecules, $N$, the chemical potential $\mu$ is determined
177: from the relation
178:
179: \be
180: N = \int \,dE \,\frac{N(E)}{e^{\beta(E-\mu)} - 1}
181: \ee
182:
183: At a specified $T=1/(k_B \beta)$, this relation yields a maximum value $N_{max}$
184: when $\mu$ equals the lowest energy of the system, $E_m$, at which point BEC
185: begins. That is, a macroscopic fraction of the molecules fall into the lowest
186: energy state when $N > N_{max}$. Equivalently, at fixed $N$, BEC occurs when T
187: falls below the inverse function $T_c = T(N_{max})$. The resulting dependence
188: on $N$ of $T_c$ is shown in Fig. 3. As seen there, $T_c$ is of order 10 mK, which
189: is experimentally accessible. Fig. 4 shows the specific heat $C_N(T)$, calculated
190: from the energy as a function of $\mu$ and T. The novel behavior observed in
191: the figure is a result of the unusual form of $N(E)$. At low T, $C_N$ is
192: proportional to $T^2$, a consequence of the 4D (linear) variation of $N(E)$ at
193: low $E$. Note the presence of singular behavior of $C_N(T)$ as one approaches
194: the transition from above and a cusp at $T_c$ itself. At relatively high T,
195: $C_N(T)/(N k_B)$ is essentially unity because $N(E)$ is 2D-like at high $E$;
196: a nondegenerate 2D gas has $C_N(T)/(N k_B)=1$. Behavior for $T>0.2$ K,
197: not shown, is very sensitive to the distribution of nanotube radii.
198:
199: \begin{figure}
200: \includegraphics[height=2.5in]{n_vs_t.eps}
201: \caption{Density of H$_2$ molecules as a function of BEC transition temperature.}
202: \end{figure}
203:
204: \begin{figure}
205: \includegraphics[height=2.5in]{cv.eps}
206: \caption{Temperature dependence of the specific heat (per molecule)
207: for molecular density $n=1\times 10^{-5}$ \AA$^{-3}$.}
208: \end{figure}
209:
210: Discussions of these results with colleagues have led to several questions
211: concerning the transition. One is this: since each IC contributes a density
212: of states characteristic of a 1D system, why does non-1D behavior arise here?
213: The answer is that particles can exchange between different IC's because
214: of their common access to the vapor region. In practice, this may be a very
215: slow process, leading to nonequilibrium behavior. The presence of breaks or
216: holes in the tubes may alleviate this kinetic problem without invalidating
217: the model, since heterogeneity is an essential aspect of the model. Another
218: question is whether similar behavior occurs for other bose gases. Indeed,
219: BEC of $^4$He is predicted by a similar analysis to occur if the sample's
220: distribution of tubes includes those with some near its energy minima in {\bf R}
221: space, which occur near 8.5 \AA. Finally, one might wonder about the
222: effects of interparticle interactions, which have been ignored up to this point
223: \cite{he4}. Indeed, some previous studies of H$_2$ in IC's have found that
224: a liquid-vapor
225: condensation occurs at a higher temperature ($\approx$ K) (in the absence
226: of heterogeneity) \cite{hecond}. A very recent study, however, found that
227: nanotubes' screening of the intermolecular interaction reduces $T_c$ to about
228: 10 mK \cite{milen}. However, that calculation omitted the role of heterogeneity,
229: which is relevant,
230: according to the Harris criterion, since the 3D specific heat critical exponent
231: is positive \cite{harris}. We expect that this condensation temperature is further
232: reduced by disorder, enabling the BEC transition to occur.
233:
234: We summarize our results as follows. Heterogeneity alters the qualitative
235: behavior of the low energy spectrum of H$_2$ molecules. The lowest-lying states
236: of the system are those of particles in that channel. As T falls, particles
237: aggregate
238: in the (essentially 4D) space of quantum states, ({\bf R},p), with bose
239: statistics having a dramatic effect, i.e. BEC, below a transition temperature
240: of order 20 mK. Anomalous behavior is predicted for the specific heat, a
241: consequence of the unusual density of states, which is 4D-like at very low
242: energy and 2D-like at somewhat higher energy. An experimental probe of the
243: real-space molecular density should reveal the needle-like concentration,
244: below $T_c$, of a macroscopic fraction of the particles within the lowest
245: energy channel.
246:
247: Most intriguing to us is that this transition is a direct consequence of
248: disorder, since the perfectly uniform system of identical nanotubes yields
249: strictly 1D, nonsingular behavior. Such a dramatic effect of heterogeneity
250: occurs elsewhere in low temperature physics. Examples include the spin-glass
251: transition \cite{schiffer}, the effect of tunneling states on thermal behavior of
252: glasses \cite{glasses} and the effects of disorder on monolayer films \cite{hedisorder}.
253:
254: This research has benefited from discussions with Moses Chan, Vin Crespi,
255: Susana Hern\'andez, Jainendra Jain, Peter Schiffer, Paul Sokol, Flavio Toigo
256: and David Weiss and support from NSF and the Hydrogen Storage program at Penn State.
257: F.A. acknowledges funding from MIUR-COFIN 2001.
258:
259: \begin{references}
260:
261: \bibitem{qsiev} Q.Y. Wang, S. R. Challa, D. S. Sholl, and J. K. Johnson, Phys. Rev.
262: Lett. {\bf 82}, 956 (1999).
263:
264: \bibitem{bob} Y.H. Kahng, R.B. Hallock, E. Dujardin, and T.W. Ebbesen, J. Low Temp.
265: Phys.{\bf 126}, 223 (2002).
266:
267: \bibitem{cro} J.C. Lasjaunias, K. Biljakovi\'c, J.L. Sauvajol, and P. Monceau,
268: cond-mat/0301584 (2003); J.C. Lasjaunias, K. Biljakovi\'c, Z. Benes, J.E. Fischer,
269: and P. Monceau, Phys. Rev. B {\bf 65}, 113409 (2002).
270:
271: \bibitem{oscCv} T. Wilson and O.E. Vilches, Proceedings of the CryoCrystals 2002 meeting in Freising, Germany, October 2002, to appear in J. Low Temp. Phys. {\bf 29} (2003).
272:
273: \bibitem{rmp} M.M.Calbi, S.M. Gatica, M.J. Bojan, G.Stan and M.W. Cole, Rev. Mod. Phys.
274: {\bf 73}, 857 (2001).
275:
276: \bibitem{aldo} A.D. Migone and S. Talapatra, "Gas Adsorption on Carbon Nanotubes",
277: to appear in the Encyclopedia of Nanoscience and Nanotechnology, published by
278: American Scientific Publishers, editor H. S. Nalwa.
279:
280: \bibitem{boro} M.C. Gordillo, J. Boronat, J. Casulleras,
281: Phys. Rev. B {\bf 65}, 014503 (2002).
282:
283: \bibitem{c60} D.E. Luzzi and B. W. Smith, Carbon {\bf 38}, 1751 (2000);
284: L.A. Girifalco and M. Hodak, Phys. Rev. B {\bf 65}, 125404 (2002);
285: M.M. Calbi, S.M. Gatica, and M.W. Cole, Phys. Rev. B {\bf 67}, 205417 (2003).
286:
287: \bibitem{bienf} M. Muris, N. Dupont-Pavlovsky, M. Bienfait, and
288: P. Zeppenfeld, Surf. Sci. {\bf 492}, 67 (2001).
289:
290: \bibitem{carlo} C. Carraro, Phys. Rev. Lett. {\bf 89}, 115702 (2002) and Phys.
291: Rev. B {\bf 61}, R16351 (2000).
292:
293: \bibitem{hecond} M.W. Cole, V.H. Crespi, G. Stan, C. Ebner, J.M. Hartman,
294: S. Moroni, and M. Boninsegni, Phys. Rev. Lett. {\bf 84}, 3883 (2000).
295:
296: \bibitem{hete} W. Shi and J. K. Johnson, submitted to Phys. Rev. Lett.
297:
298: \bibitem{h2bec} O.E. Vilches, J. Low Temp. Phys. {\bf 89}, 267 (1992);
299: H.J. Maris, G.M. Seidel, F.I.B. Williams, Phys. Rev. B {\bf 36}, 6799 (1987).
300:
301: \bibitem{interc} M. M. Calbi, F. Toigo and M. W. Cole, J. Low Temp. Phys.
302: {\bf 126}, 179 (2002).
303:
304: \bibitem{oscar} T. Wilson, A. Tyburski, M.R. DePies, O.E. Vilches,
305: D. Becquet, and M. Bienfait, J. Low Temp. Phys. {\bf 126}, 403 (2002).
306:
307: \bibitem{vid} L. Mattera, F. Rosatelli, C. Salvo, F. Tommasini, U. Valbusa
308: and G. Vidali, Surf. Sci. {\bf 93}, 515 (1980).
309:
310: \bibitem{band} See M. Boninsegni, S.Y. Lee, V.H. Crespi, Phys. Rev. Lett.
311: {\bf 86}, 3360 (2001).
312:
313: \bibitem{milen} M.K.Kostov, J.C.Lewis and M.W.Cole, in Condensed Matter Theories,
314: Vol. 16, edited by S. Hernandez and J. Clark, Nova Science Publishers, NY, 2001,
315: pp.161, http://xxx.lanl.gov/abs/cond-mat/0010015.
316:
317: \bibitem{uptake} G. Stan, M.J. Bojan, S. Curtarolo, S.M. Gatica and M.W. Cole,
318: Phys. Rev. B {\bf 62}, 2173 (2000).
319:
320: \bibitem{ziman} J.M.Ziman, {\em Principles of the Theory of Solids}, (Cambridge
321: U.P., 1964).
322:
323: \bibitem{groove} In contrast, the transverse energy for the analogous problem of motion
324: in the groove (on the outside of a bundle) exhibits a local maximum on the $R_1=R_2$
325: symmetry line.
326:
327: \bibitem{he4} Recall that the 3D ideal gas BEC transition of $^4$He is preempted
328: by the interaction-driven condensation to a liquid below 5.2 K, resulting in
329: a ground state condensate fraction about 10\%.
330:
331: \bibitem{harris} D.S.Fisher, G. M. Grinstein and A. Khurana, Physics Today {\bf 41},
332: 56 (1988).
333:
334: \bibitem{schiffer} P. Schiffer et al, Phys. Rev. Lett. {\bf 74}, 2379 (1995).
335:
336: \bibitem{glasses} P.W. Anderson, B.I. Halperin, and C.M. Varma, Philos. Mag.
337: {\bf 25}, 1 (1972); W. A. Phillips, J. Low Temp. Phys. {\bf 7}, 351 (1972).
338:
339: \bibitem{hedisorder} C.E. Campbell, J.G. Dash and M. Schick, Phys. Rev.
340: Lett. {\bf 26}, 966 (1971).
341:
342:
343: \end{references}
344:
345:
346:
347:
348:
349: \end{document}
350: