cond-mat0010204/PRL.tex
1: % A Discotic Disguised as a Smectic
2: % last modified on 13th of Oct 2000
3: 
4: \tolerance = 10000
5: \documentstyle[prl,aps,epsf,multicol]{revtex}
6: %\documentstyle[eqsecnum,aps,prb,multicol,epsf]{revtex}
7: %\documentstyle[preprint,prl,aps,epsf]{revtex}
8: %\documentstyle[prl,aps,epsf,twocolumn]{revtex}
9: 
10: \begin{document}
11: \draft
12: \tighten
13: 
14: \title{A Discotic Disguised as a Smectic: A Hybrid Columnar Bragg Glass}
15: 
16: \author{Karl Saunders$^1$, Leo Radzihovsky$^2$, 
17: John Toner$^1$}
18: \address{$^1$ Dept. of Physics,
19: Materials Science Inst., and Inst. of Theoretical Science, University
20: of Oregon, Eugene, OR 97403}
21: \address{$^2$ Department of Physics, University of Colorado,
22: Boulder, CO 80309}
23: 
24: \date{\today}
25: \maketitle
26: \begin{abstract}
27:   
28:   We show that discotics, lying deep in the columnar phase, can
29:   exhibit an x-ray scattering pattern which mimics that of a somewhat 
30:   unusual smectic liquid crystal. This exotic, new glassy phase of columnar 
31:   liquid crystals, which we call a ``hybrid columnar Bragg glass'', can be
32:   achieved by confining a columnar liquid crystal in an anisotropic
33:   random environment of e.g., strained aerogel. Long-ranged
34:   orientational order in this phase makes {\em single domain} x-ray
35:   scattering possible, from which a wealth of information could be
36:   extracted. We give detailed quantitative predictions for the
37:   scattering pattern in addition to exponents characterizing anomalous
38:   elasticity of the system.
39: 
40: 
41: \end{abstract}
42: \pacs{64.60Fr,05.40,82.65Dp}
43: 
44: %\twocolumn
45: \vspace{-0.5cm}
46: \begin{multicols}{2}
47: \narrowtext
48: 
49: Until now, the x-ray scattering pattern given in Fig.\ref{scattering}
50: would be identified with a system in a somewhat unusual smectic phase
51: with short-ranged translational order and long-ranged orientational
52: order {\em within} the smectic layers; i.e., a smectic composed of
53: {\it nematic}, rather than {\it liquid}, layers.  The set of on-axis
54: quasi-sharp Bragg peaks along $q_h$ is a signature of the
55: quasi-long-ranged translational order (i.e., the periodicity of the
56: layering) that is characteristic of the bulk smectic phase. The
57: presence of the other, broadened, peaks and the azimuthal anisotropy
58: about the $q_h$ axis respectively indicate the incipient {\it
59: short-ranged} translational order and the long-ranged orientational
60: order {\it within} the smectic layers oriented perpendicular to $q_h$.
61: % 
62: \begin{figure}[bth] 
63: \centering
64: \setlength{\unitlength}{1mm} 
65: \begin{picture}(15,42)(0,-12)
66: \put(-35,-45){\begin{picture}(20,20)(0,0) 
67: \special{psfile=NewScattering.ps vscale=35 hscale=35} 
68: \end{picture}} 
69: \end{picture} 
70: \caption{X-ray scattering pattern in the $\perp$ plane for a class of hybrid
71: columnar Bragg glass.} 
72: \label{scattering} 
73: \end{figure} 
74: % 
75: 
76: In this Letter we predict the existence of a remarkable new ``hybrid columnar
77: Bragg\cite{NS,RFXY} glass'' phase (HCBG) which, despite differing
78: fundamentally from the smectic phase described above, shares the same
79: qualitative scattering pattern illustrated in Fig.\ref{scattering}. Such 
80: mimickry of one phase by a completely different
81: other phase is unprecedented.
82: 
83: Columnar phases in pure, bulk (i.e., quenched-disorder free) liquid
84: crystals are phases that have long-ranged translational order in
85: two-directions, and short-ranged translational order (i.e.,
86: liquid-like correlations) in the third. I.e., they are
87: regular two-dimensional lattices of one-dimensional 
88: liquid columns (Fig.\ref{discoticside}(a)). In this Letter,
89: we show that when such a system is confined in an {\em anisotropic}
90: quenched random environment, e.g., strained aerogel\cite{Clarke}, it becomes
91: translationally disordered, but remains topologically ordered (i.e.,
92: free of topological defects like dislocations). This novel state is
93: the HCBG.
94: 
95: Like the smectic phase, the HCBG has translational order that is
96: quasi-long-ranged in one direction and short-ranged in another, as implied by
97: Fig.\ref{scattering}. As illustrated in Fig.\ref{discoticside}(b), the columns
98: remain in roughly equidistant rows perpendicular to the stretching direction,
99: but lose long-ranged translational order within each row. Nevertheless, the
100: hexagonal orientational order is preserved, albeit uniaxially distorted due to
101: the strectch.  However, there are a number of fundamental differences between
102: the two phases.  First, unlike the smectic phase, the exponent $\eta_{\bf G}$
103: characterizing the shape of the quasi-long-ranged translational order peaks at
104: ${\bf G}$, which are given by $I({\bf q})\propto |{\bf q}-{\bf G}|^{ -3+
105: {\eta_{G} } }$, 
106: % 
107: %\begin{eqnarray} 
108: %I({\bf q})\propto |{\bf q}-{\bf G}|^{-3+\eta({\bf G})} 
109: %\label{I} 
110: %\end{eqnarray} 
111: % 
112: is independent of temperature.  This property would allow the two phases'
113: scattering patterns to be distinguished through comparison of the lineshapes
114: as temperature is varied. Secondly, the correlations of the quasi-long-ranged
115: order in the HCBG scale isotropically in space, in contrast to the well-known
116: strongly anisotropic scaling of these correlations in the smectic.  The third
117: and most crucial difference between the two phases is their topolological
118: order, which distinguishes their elasticities but not their scattering. 
119: Specifically, the absence of the extra direction of long-ranged
120: translational order, in the smectic phase, is caused by free topological
121: defects, namely unbound dislocations with Burgers vectors along the smectic
122: planes. Although the HCBG also exhibits translational order that is 
123: quasi-long-ranged in one
124: direction and short-ranged in another, at long length scales it is
125: distinguished from the smectic by being free of these unbound dislocations,
126: which would otherwise destroy the columnar phase topology of HCBG. One
127: important experimental consequence of this absence of free dislocations is
128: that the HCBG retains elastic resistance to distortions in the extra
129: direction, albeit, as we discuss below, of a very strange, anomalous sort.
130: 
131: Of course, for {\em sufficiently} strong disorder 
132: free dislocations {\em will} eventually proliferate. 
133: The anisotropy, imposed by the strained aerogel, leads to
134: the interesting possibility that dislocations with their Burgers
135: vectors in the soft direction may unbind (with increasing disorder) 
136: before those with their Burgers vectors in the hard direction, leading 
137: to the sequence of disorder-driven phase transitions : HCBG $\rightarrow$ m=1 smectic Bragg
138: glass{\cite {JSRT}} $\rightarrow$ nematic elastic glass\cite{RTpr} with
139: increasing aerogel density.
140: 
141: The rest of this Letter gives a more detailed theoretical description of
142: the HCBG phase, including x-ray correlation lengths and universal
143: exponents characterizing the anomalous elasticity. We relegate
144: the technical details to a future publication\cite{SJRT}.  
145: 
146: Our model for a columnar phase consists of
147: disc-shaped molecules with normals aligned along the $\hat {\bf z}$
148: direction. The discs form a hexagonal lattice in the $xy$ ($\perp$)
149: plane and have liquid like correlations along $\hat {\bf z}$ as
150: illustrated in Fig.\ref{discoticside}(a).  We assume, and verify a
151: posteriori, that despite considerable distortion, for sufficiently
152: weak quenched disorder, the columnar phase topology is stable, i.e.
153: our discotic liquid crystal remains free of unbound dislocations.
154: Consequently this system can be described within an elastic theory,
155: with a two-component ($x$ and $y$) lattice site displacement vector
156: $\bf u(\bf r)$ and the discotic director $\hat{\bf n}(\bf r)$ (the normal to
157: the discs) as the only important long length-scale degrees of freedom. The
158: disordering tendency of the aerogel is two-fold: the strands act both
159: to randomly pin the columnar lattice ($\bf u(\bf r)$) and to distort
160: the orientations of the disc normals ($\hat{\bf n}(\bf r)$).  Our starting
161: Hamiltonian is that of a pure hexagonal discotic in {\em isotropic}
162: aerogel
163: %
164: \begin{eqnarray}
165: H&=&\int_{\bf r} \bigg[ {B_\perp\over2} |\partial_z {\bf u} - {\bf
166: {\delta}} {\bf n}|^2 +{1\over 2}\lambda u_{ii}^\perp u_{jj}^\perp +\mu
167: u_{ij}^\perp u_{ij}^\perp\nonumber\\ &+&{\cal R}e \sum_i V_i({\bf
168: r})e^{i {\bf G}_i \cdot {\bf u}({\bf r})} -({\bf g}({\bf
169: r})\cdot{\hat{\bf n}})^2 \bigg] + H_F [\hat{\bf n}]\;.
170: \label{H}
171: \end{eqnarray}
172: %
173: where $u_{ij}^\perp = {1\over2}(\partial_i^\perp u_j +\partial_j^\perp
174: u_i - \partial_k u_i \partial_k u_j)$ is the rotationally invariant
175: symmetric strain tensor, $ \delta {\bf n}({\bf r}) \equiv {\bf \hat
176:   n}({\bf r}) - \hat {\bf z}$, the $B_\perp$ term reflects the
177: tendency of the molecular director (disc normal) $\hat{\bf n}(\bf r)$ to lie
178: along the local tangent ${\bf \hat t} \approx {\bf \hat z} +
179: \partial_z {\bf u} $ to the liquid-like columns, $H_F [{\hat{\bf n}}]$
180: is the Frank free energy of the molecular directors, and $V_i({\bf
181:   r})$ is a complex random pinning potential that couples to lattice
182: site fluctuations along the reciprocal lattice basis vector, ${\bf
183:   G}_i$. At long length-scales its correlations can be accurately
184: represented as zero-mean with {\em short-ranged}, Gaussian statistics:
185: $\overline{V_i({\bf r})V_j^*({\bf r'})} = \tilde{\Delta}_V
186: \delta_{ij}^{\perp} \delta^d({\bf r}-{\bf r'})$,\cite{RTpr} 
187: throughout this paper $\overline{x}$ denotes a quenched average
188: over the disorder of the quantity $x$, while $\langle x\rangle$
189: denotes a thermal average. The last term
190: describes the tendency of the disc normals $\hat{\bf n}({\bf r})$ to
191: align along the random local aerogel strand directed along ${\bf g} ({\bf r})$. 
192: This ``random tilt'' disorder is described by {\em short-ranged,
193:   isotropic} correlations $\overline{g_i({\bf r})g_j({\bf r'})} =
194: 1/2\sqrt{\Delta}\delta_{ij}\delta^d({\bf r}-{\bf r'})$\cite{RTpr}. 
195: $\Delta_V$ and $\Delta$ are phenomenological parameters
196: which, in the simplest microscopic model, are proportional
197: to the aerogel density only, $\rho_A$. As for smectics
198: \cite{RTpr}, these two types of
199: disorder have important long distance effects.
200: %
201: \begin{figure}[bth]
202: \centering
203: \setlength{\unitlength}{1mm}
204: \begin{picture}(30,42)(0,0)
205: \put(-30,-33){\begin{picture}(20,20)(0,0)
206: \special{psfile=discoticside.ps vscale=40 hscale=40}
207: \end{picture}}
208: \end{picture}
209: \caption{(a) Alignment of the lattice for a uniaxial stretch. 
210: (b) Schematic of the distorted lattice in real space.}
211: \label{discoticside}
212: \end{figure}
213: %
214: A detailed analysis\cite{SJRT,RTpr} has shown that this system
215: exhibits a columnar Bragg glass phase with only short-ranged translational order. 
216: However, this changes if the aerogel
217: is anisotropic. Aerogel anisotropy could be realized e.g., by
218: applying a strain to the strands. For heterotropic alignment between
219: the disc normals and strands (assumed throughout) a uniaxial {\em
220: compression} will lead to a phase in the same universality class as
221: the Bragg glass phase of an Abrikosov flux lattice\cite{NS,RFXY}, with
222: quasi-long-ranged translational order in both directions of the
223: $\perp$ plane\cite{SJRT}. The more interesting HCBG
224: phase with quasi-long-ranged order in only one $\perp$ direction can
225: be obtained by applying a uniaxial {\em stretch} to the strands. For
226: homeotropic alignment of strands and disc normals the two phases
227: reverse with respect to stretch and compression leaving all of our
228: other predictions unchanged.
229: %
230: %\begin{figure}[bth]
231: %\centering
232: %\setlength{\unitlength}{1mm}
233: %\begin{picture}(36,85)(0,0)
234: %\put(-15,-70){\begin{picture}(80,80)(0,0)
235: %\special{psfile=discoticside.ps vscale=100 hscale=100}
236: %\end{picture}}
237: %\end{picture}
238: %\caption{Model for the hexagonal discotic.}
239: %\label{fig2}
240: %\end{figure}
241: %
242: 
243: Uniaxial stretch of the aerogel strands(along $\hat {\bf e_h}$ causes
244: the disc normals to align $\perp$ to the axis of stretch
245: (Fig.\ref{discoticside}).  This can be accounted for by the addition of
246: a term $\int_{\bf r}\gamma (\hat {\bf e_h} \cdot \hat {\bf n})^2$ to the
247: Hamiltonian, where $\gamma$ is a phenomenological parameter, which we
248: assume to be a monotonically increasing function of $\rho_A$ and
249: strain. The uniaxial stretch also breaks both the hexagonal symmetry
250: of the lattice, and its rotation invariance.  Thus, we are forced to
251: consider both a more general lattice structure and an elastic
252: Hamiltonian that need not, and indeed, will not be invariant under
253: rotations of the lattice. We can take into account both of these
254: effects by replacing the harmonic pieces of the hexagonal elastic 
255: energy $\int_{\bf r}
256: ({1\over2}\lambda u_{ii}^\perp u_{jj}^\perp + \mu u_{ij}^\perp
257: u_{ij}^\perp)$ with the more general harmonic elastic energy ${1 \over
258:   2}\int_{\bf r} C_{ijkl} \partial^{\perp}_i u_j \partial^{\perp}_k
259: u_l$, where the elastic constant tensor $C_{ijkl}$ is {\it not}
260: symmetric under interchange of its first two or second two indices,
261: due to the lack of in-plane rotation invariance just discussed.
262: The terms cubic and quartic in $\bf u$ in Eq.\ref{H} must also be so generalized of course.
263: Because of the $B_\perp$ term in Eq.\ref{H}, fluctuations of $\hat{\bf n}$ 
264: from the local column tangent $ \hat {\bf t} $ 
265: are small, i.e., $\delta{\bf n} \approx \partial_z \bf u$. 
266: 
267: We have analysed\cite{SJRT} this generalization of the model, Eq.\ref{H}, using
268: renormalization group (RG) methods\cite{RTpr}. One of the most surprising 
269: conclusions of this analysis is that, at long length scales, fluctuations 
270: $u_h \equiv \hat{\bf e}_h \cdot  {\bf u}$ along the direction of stretch decouple 
271: from those $u_s \equiv \hat{\bf e}_s \cdot  {\bf u}$ orthogonal to this direction, where we
272: have denoted the axis of stretch, $\hat {\bf e}_h$, as ``$\em hard$''($\em h$)
273: and called the other $\perp$ axis, orthogonal to $\hat {\bf e}_h$ ``$\em
274: soft$''($\em s$), i.e. ${\bf r}_\perp=( r_h, r_s)$. That is, all couplings
275: between $u_s$ and $u_h$ that are present in the full elastic tensor $C_{ijkl}$
276: flow to zero upon renormalization, leaving $C_{ijkl}$ in the form $C_{ijkl}
277: =B_{ss} \delta_{is} \delta_{js} \delta_{ks} \delta_{ls} + B_{sh} \delta_{is}
278: \delta_{jh} \delta_{ks} \delta_{lh} + B_{hs} \delta_{ih} \delta_{js}
279: \delta_{kh} \delta_{ls} + B_{hh} \delta_{ih} \delta_{jh} \delta_{kh}
280: \delta_{lh}$.
281: 
282: The total Hamiltonian for the system can therefore be expressed as a
283: sum of {\em decoupled} Hamiltonians for $u_h$ and $u_s$: $H_{tot}[u_h ,
284: u_s]=H_{XY}[u_h] + H_{m=1}[u_s]$, with
285: %
286: \begin{mathletters}
287: \begin{eqnarray}
288: H_{XY}&=&{1 \over 2}\int_{\bf r} \bigg[ \gamma|\partial_z u_h|^2 
289: +B_{sh}|\partial_s u_h|^2 + B_{hh}|\partial_h u_h|^2\nonumber\\ 
290: &+&{\cal R}e \sum_i V_i({\bf r})e^{i ({\bf G}_i 
291: \cdot \hat{\bf e}_h) u_h({\bf r})}\bigg]\;,
292: \label{H_XY}\\
293: H_{m=1} &=&  \int_{\bf r}\bigg[  {K\over2}
294: (\partial_z^2 u_s)^2 + B_{ss}\Big(\partial_s u_s - {1\over
295: 2}(\partial_z u_s)^2\Big)^2 \nonumber\\ 
296: &+& B_{hs}(\partial_h u_s)^2 - 
297: g_z ({\bf g} \cdot \hat{\bf e}_s)(\partial_z u_s)\bigg]\;.
298: \label{H_m=1r}
299: \end{eqnarray}
300: \end{mathletters}
301: %
302: These two parts of $H_{tot}$, which describe the fluctuations of the
303: ``hard'' and ``soft'' phonon fields $u_h$ and $u_s$, are not new.
304: Hamiltonians of precisely this form have been previously used to
305: describe the random field XY-model\cite{RFXY} and the ``m=1 smectic
306: Bragg glass''\cite{JSRT}, respectively, and have been studied
307: extensively.  However, a columnar phase confined in
308: anisotropic aerogel, whose Hamiltonian, $H_{tot}$, is a combination, or
309: {\em hybrid}, of the two, is entirely novel.
310: 
311: The $u_h$ fluctuations of our system are the same as those of a
312: slightly anisotropic (although isotropic in {\em scaling}) random
313: field XY-model and are given by $\overline{\langle(u_h({\bf
314:     r})-u_h({\bf 0} ))^2\rangle}= C(d)\ln{r}/G_{0h}^2$, where $G_{0h}$ is the
315: lattice spacing of the projection of the discotic reciprocal lattice
316: onto the hard axis.
317: %, and we've defined 
318: %$r' \equiv \sqrt{(r_s/a_s)^2+(r_h/a_h)^2+(r_z/a_z)^2}$.
319: They diverge logarithmically as a function of distance, implying that
320: the translational order along the hard direction is quasi-long-ranged.
321: While these elastic distortions are reminiscent of the famous
322: Landau-Peierls $\ln{r}$ fluctuations of {\em bulk} smectics they
323: differ crucially in two ways. Firstly, they are $\em disorder$, rather
324: than $\em thermally$, driven with $C(d)$ {\em universal} ($C(3)\approx 1.1$)
325: and the logarithm persisting in all $2<d<4$. Secondly, they are $\em
326: isotropic$ in their scaling. In contrast, in a bulk smectic the layer
327: fluctuations within the layers scale differently than those along the
328: normal to the layers.
329: %Here, the only anisotropy is in the different values of 
330: %$a_{s,h,z}$. 
331: 
332: In $H_{m=1}$ the combination of relevant anharmonic terms and large
333: disorder-induced $u_s$ fluctuations leads to strong anomalous
334: elasticity\cite{RTpr,SJRT}. By anomalous elasticity we mean that
335: the full, anharmonic theory with {\it constant} $K$,
336: $B_{ss}$ and $\Delta$ can, at small wavevector $k \ll \xi_{NL}^{-1}$ 
337: (where $\xi_{NL}$ is a non-universal length determined by material parameters, 
338: e.g., aerogel density), be effectively replaced by a harmonic theory with 
339: wavevector dependent $K$, $B_{ss}$ and $\Delta$. These are given by
340: %
341: \begin{mathletters}
342: \begin{eqnarray}
343: K({\bf k})&=&K k_z^{-\eta_K}
344: f_K\left({k_h/k_z^{\zeta_h}}, {k_s/k_z^{\zeta_s}}\right)\;,
345: \label{K}\\ 
346: B_{ss}({\bf k})&=&B_{ss} k_z^{\eta_B}
347: f_B\left({k_h/k_z^{\zeta_h}}, {k_s/k_z^{\zeta_s}}\right)\;,
348: \label{B_{ss}}\\ 
349: \Delta({\bf k})&=&\Delta k_z^{-\eta_{\Delta}}
350: f_{\Delta}\left({k_h/k_z^{\zeta_h}}, {k_s/k_z^{\zeta_s}}\right)\;.
351: \label{Delta}
352: \end{eqnarray}
353: \label{anom_elasticity}
354: \end{mathletters}
355: %
356: \noindent$B_{hs}$ is
357: not significantly renormalized; that is $B_{hs}({\bf k})=B_{hs} , $
358: independent of wavevector.  Here the anisotropy exponents
359: $\zeta_s\equiv 2-(\eta_B+\eta_K)/2$ and $\zeta_h\equiv 2-\eta_K/2$.
360: The exponents, evaluated using the renormalization group and a high
361: precision $\epsilon$-expansion were found to be: $\eta_K=0.50$,
362: $\eta_B=0.26$, and $\eta_{\Delta}=0.13$.\cite{SJRT,JSRT} We also
363: predict that the anomalous exponents will obey the following {\em exact} scaling relation:
364: %
365: \begin{eqnarray}
366: 1+\eta_\Delta=\eta_B/2 + 2\eta_K .
367: \label{escale}
368: \end{eqnarray}
369: %
370: The translational order of the system along the soft direction is
371: short-ranged and is characterized by the algebraic and anisotropic
372: divergence of $u_s$ correlations,
373: %$\overline{\langle(\delta u_s({\bf r}))^2\rangle}$
374: %
375: \begin{mathletters}
376: \begin{eqnarray}
377: \overline{\langle(\delta u_s({\bf r}))^2\rangle}
378: &=&\cases{({K\over B_{ss}})({ r_z\over\xi_z})^{\chi_z}, &
379: $r_z \gg r_{s,h}$\cr
380: ({K\over B_{ss}})({ r_s\over \xi_z^2}({K\over B_{ss}})^{1/2})^{{\chi_s
381: }}, & 
382: $r_s \gg r_{h,z}$\cr 
383: ({K\over B_{ss}})({ r_h\over \xi_z^2}({K\over B_{sh}})^{1/2})^{{\chi_h}}, &
384: $r_h \gg r_{s,z}$}\nonumber\\
385: \label{xiX}
386: \end{eqnarray}
387: \end{mathletters}
388: %
389: where we have defined $\delta u_s({\bf r}) \equiv u_s({\bf r})-u_s(0)$, $\chi_z \equiv
390: 1-\eta_K+\eta_B/2+\eta_{\Delta} = \eta_B + \eta_K $, $\chi_{s ,
391: h} \equiv {\chi_z/\zeta_{s , h}}$ , and $\xi_z\equiv {K^2
392: B_{sh}^{1/2}/(\Delta B_{ss}^{1/2})}$. In the second equality for
393: $\chi_z$, we have used the {\em exact} scaling relation between the
394: $\eta$'s. The exact scaling relation Eq.\ref{escale}, could be experimentally tested by
395: using the more general expressions for $\chi_z$, $\chi_s$, $\chi_h$ in terms of all three
396: exponents, and verifying that $\eta_B, \eta_K, and \eta_{\Delta}$ obey Eq.\ref{escale}.
397: Our $\epsilon$-expansion results for the $\eta$'s imply $\chi_z = 0.76$ , $\chi_s = 0.47$ 
398: and $\chi_h = 0.43$. The fluctuations given in
399: Eq.\ref{xiX}, like those along the hard direction, are disorder-,
400: rather than thermally-driven.
401: 
402: Despite this lack of translational order, our detailed
403: calculations\cite{SJRT} indicate that dislocation loops remain bound for
404: weak disorder, and therefore the low temperature phase replacing the
405: columnar phase must be distinct from the smectic and hexatic,
406: separated from them by a thermodynamically sharp dislocation unbinding
407: phase transition.
408: 
409: The stability of this exotic glass phase is contingent upon our
410: assumption of long-ranged orientational order.  We validate this
411: assumption by calculating $\overline{\langle| {\bf n}({\bf
412:     r})- {\bf n}({\bf 0})|^2\rangle} = \overline{\langle|\partial_z
413:   {\bf u}({\bf r})-\partial_z {\bf u}({\bf 0})|^2\rangle}$ and showing that
414: it does {\it not} diverge as ${\bf r} \rightarrow \infty$ \cite{SJRT}.  
415: Although equilibration into the ground state might be slow and therefore
416: require field alignment, this
417: orientational order would allow experimentalists to investigate {\em
418:  single domain} samples of HCBG. The anisotropic scaling information
419: which is usually lost in a powder averaged x-ray scattering experiment
420: {\em would be retained in a single domain experiment} 
421: allowing detailed tests of our predictions for 
422: $\eta_K$, $\eta_B$ and $\eta_\Delta$.
423: 
424: The scattering pattern in the $\perp$ plane, obtained from a single
425: domain sample, would consist of a set of spots rather than the set of
426: rings that one would expect from a powder sample. This
427: pattern depends crucially on the relative orientations
428: within the $\perp$ plane of the reciprocal lattice and the axis of
429: stretch, $\hat {\bf e}_h$, which could vary from discotic to discotic since it
430: depends on the microscopic interactions between the discs and strands. 
431: The intensity of a Bragg spot at a
432: reciprocal lattice vector ${\bf G}$
433: %
434: \begin{mathletters}
435: \begin{eqnarray}
436: I({\bf G})&\propto&\int_{\bf r} \exp[-\overline{\langle[{\bf G}\cdot 
437: ({\bf u}({\bf r})-{\bf u}({\bf 0}))]^2\rangle}/2]\,
438: \label{I(G)a}\\
439: &\propto& \int_{\bf r}\exp\big\{-\big[G_h^2\overline{\langle(u_h({\bf r})-u_h({\bf 0}))^2\rangle} \nonumber\\
440: &+& G_s^2\overline{\langle(u_s({\bf r})-u_s({\bf 0}))^2\rangle}\big]/2\big\}\,
441: \label{I(G)b}
442: \end{eqnarray}
443: \end{mathletters}
444: %
445: Unless $G_s=0$, the algebraically diverging $u_s$ fluctuations
446: dominate the logarithmically diverging $u_h$ fluctuations and the
447: integrand is exponentially damped, leading to an anisotropically
448: broadened Bragg peak.  If however, $G_s=0$ then the exponential
449: becomes $r^{-0.55 n^2}$, where $n=G_h/G_{0h}$, with $G_{0h}$ being the
450: magnitude of the {\it smallest} ${\bf G}$ lying {\em on} the hard
451: axis, and for $n<3$ the integral diverges as $r \rightarrow \infty$,
452: leading to quasi-sharp peaks for those $n$'s.  We therefore predict
453: two classes of hybrid columnar Bragg
454: glasses. The first, which we call a {\em commensurate} HCBG, has some
455: reciprocal lattice vectors that lie {\em along} the hard axis, and
456: will exhibit a scattering pattern with peaks lying on the hard axis,
457: with the first two quasi-sharp.  In the second, {\em
458:   incommensurate} HCBG, class, all the peaks lie off the hard axis,
459: and are anisotropically broadened by the contribution from the $u_s$
460: fluctuations.  The smectic-like scattering pattern
461: (Fig.\ref{scattering}), with a quasi-sharp peak on the first ring,
462: will therefore only be observed for commensurate HCBG's.
463: 
464: The dependence of the anisotropically broadened peak widths on the
465: bare elastic constants is the same for both classes of HCBG.  Setting
466: the $u-u$ correlation functions, given in Eq.\ref{xiX}, equal to
467: $G_s^{-2}$ and solving for $(r_z=\xi_z^{X})^{ -1}$, $(r_s=\xi_s^{X})^{
468:   -1}$ and $(r_h=\xi_h^{X})^{ -1}$ gives the width of the peak at $\bf
469: G$ along each of the $z$, $s$ and $h$ directions:
470: \begin{mathletters}
471: \begin{eqnarray}
472: (\xi_z^{X})^{ -1}&=&\xi_z^{-1}({G_s}^2 K/B_{ss})^{\chi_z^{-1}}\;,
473: \label{z}\\ 
474: (\xi_s^{X})^{ -1}&=&\xi_z^{-2} (K/B_{ss})^{1/2}({G_s}^2 K/B_{ss})^{\chi_s^{-1}}
475: \;,
476: \label{s}\\ 
477: (\xi_h^{X})^{ -1}&=&\xi_z^{-2} (K/B_{sh})^{1/2}({G_s}^2 K/B_{ss})^{\chi_h^{-1}}\;.
478: \label{h}
479: \end{eqnarray}
480: \label{peakwidths}
481: \end{mathletters}
482: %
483: The temperature dependence of $\xi^X_{z,s,h}$ could be used to determine
484: the exponents $\eta_K$, $\eta_B$ and $\eta_{\Delta}$ since the {\em bulk}
485: $K(T)$, $B_{ss}(T)$ and $B_{sh}(T)$ in Eq.\ref{peakwidths} have temperature
486: dependences that can be extracted from data on bulk materials.  A more
487: direct way to observe the anomalous elasticity would be a direct measurement
488: of the $u-u$ correlation function
489: $I({\bf q})\propto\overline{\langle | u_s (\delta {\bf q} ) |^2\rangle}$, 
490: which can be obtained\cite{SJRT} for {\it 
491: large} ${\bf q}$ (i.e., {\bf q}'s with at least one component bigger than the
492: corresponding inverse x-ray correlation length quoted above) by looking at an
493: intermediate regime in the ``tails'' of the broad x-ray scattering peaks. In
494: those tails, (i.e., for ${\bf q}={\bf G}+ \delta {\bf q}$ with $\xi_{NL}^{-1}
495: \gg | \delta q_\alpha | \gg (\xi^\chi_\alpha)^{-1} $ for at least one
496: Cartesian direction $\alpha = (h,s,z)$ \cite{N}), it can be shown that
497: %
498: \begin{eqnarray}
499: I({\bf q})&\propto&  {\Delta ({\bf \delta q} ) q_z^2
500: \over (B_{ss} ({\bf \delta q} ) \delta q_s^2 + K({\bf \delta q} ) \delta q_z^4 + B_{hs} \delta q_h^2)^2 } \;.
501: \label{Itail}
502: \end{eqnarray}
503: %
504: Hence, the {\bf q} dependence of $B_{ss}({\bf q} )$, $K({\bf q} )$
505: and $\Delta ({\bf
506:   q} )$ in Eq.\ref{anom_elasticity} could be tested directly by a fit
507: of scattering data to these tails.
508: 
509: A related experimental approach, which has the advantage of {\it not}
510: being restricted to wavevectors larger than the inverse x-ray
511: correlation lengths, but can, rather, explore arbitrarily small {\bf
512:   q}'s, is light scattering, which measures director fluctuations.
513: These can be related to the $u-u$ correlations via our condition
514: $\delta{\bf n} \approx \partial_z \bf u$. This yields
515: %
516: \begin{mathletters}
517: \begin{eqnarray}
518: \overline{\langle| \delta n_s ( {\bf q} ) |^2\rangle} &=&
519: {\Delta ({\bf q} ) q_z^4 \over (B_{ss} ({\bf q} ) 
520: q_s^2 + K({\bf q} ) q_z^4 + B_{hs} q_h^2)^2 } \;,
521: \label{delns}\\
522: \overline{\langle| \delta n_h ( {\bf q} ) |^2\rangle}
523: &=&\cases{C(3)/2 G_{0h}^2 q_z^2/q^3, & commen.\cr k_B T q_z^2 G({\bf
524: q}) +\Delta_h(\bf q) q_z^4 G({\bf q})^2, & incommen.}\nonumber\\
525: \label{delnh}
526: \end{eqnarray}
527: \end{mathletters}
528: %
529: where $\Delta_h(\bf q)$ in Eq.\ref{delnh} is renormalized, ${\bf q}$
530: dependent tilt disorder variance obeying the scaling law
531: $\Delta_h({\bf q})=\Delta_h q_z^{-(\eta_{\Delta}+\eta_B)}
532: f_{\Delta_h}\left({q_h/q_z^{\zeta_h}}, {q_s/q_z^{\zeta_s}}\right)$ and
533: $G({\bf q})=1/(B_{sh} q_s^2 + \gamma q_z^2 + B_{hh} q_h^2)$.  The
534: commensurate and incommensurate cases differ because in the
535: commensurate case there {\it is} a random field acting on $u_h$, while
536: in the incommensurate case there is no random field, leaving the
537: random tilt as the dominant disorder.
538: 
539: L.R. acknowledges support by the NSF DMR-9625111, the MRSEC
540: DMR-9809555, and the Sloan and Packard Foundations. J.T. and K.S. were
541: supported by the NSF DMR-9634596 and DMR-9980123.
542: %
543: \vspace{-.25in}
544: \begin{references}
545: \vspace{-.6in}
546: 
547: \bibitem{NS} For a general discussion of ``Bragg glass'' phases see, 
548: e.g., T. Nattermann and S. Scheidl, to appear in Advances in Physics (2000).
549: 
550: \bibitem{RFXY} T. Giamarchi, P. Le Doussal, Phys. Rev. Lett.
551: {\bf 72}, 1530 (1994); D. S. Fisher, Phys. Rev. Lett. {\bf 78}, 1964 (1997).
552: 
553: \bibitem{Clarke} There has been an immense amount of excellent work on liquid crystals in aerogel. See, e.g.,
554: {\it Liquid Crystals in Complex Geometries}, edited by G.P. Crawford and S. Zumer (Taylor and Francis, London,1996).
555: 
556: \bibitem{JSRT} B. Jacobsen, K. Saunders, L. Radzihovsky and J. Toner
557: Phys. Rev. Lett. {\bf 83}, 1363 (1999).
558: 
559: \bibitem{RTpr} L. Radzihovsky and J. Toner, Phys. Rev. B {\bf 60}, 206 (1999).
560: %;Phys. Rev. Lett. {\bf 78}, 4414 (1997), ibid {\bf 79}, 4214 (1997).
561: 
562: \bibitem{SJRT} K. Saunders, L. Radzihovsky, and J. Toner, (unpublished).
563: 
564: \bibitem{N} We have shown in \cite{SJRT} that the ratio 
565: ${\xi_{NL}}/{\xi^\chi_\alpha}\rightarrow 0$ as the aerogel density
566: goes to zero. Hence, one could make the width of this window in 
567: $\delta q_\alpha$ as large as one likes simply by making the aerogel 
568: density sufficiently small.
569: 
570: \end{references}
571: \end{multicols}
572: \end{document}
573: 
574: 
575: 
576: