cond-mat0310381/jutr.tex
1: %\documentstyle[12pt,psfig]{article}
2: \documentclass[12pt]{article}
3: \usepackage{graphicx}
4: \begin{document}
5: 
6: \hsize 14.5 cm
7: \vsize 21 cm
8: 
9: \leftline {\LARGE Quantum Spin Glass Phase Boundary in $\pm J$}
10: 
11: \medskip
12: 
13: \leftline {\LARGE   
14: Transverse Field Ising Systems\footnote{
15: Presented at CMDays-03, Jadavpur Univ., Kolkata, Aug. 2003. Proc. in Ind. J. 
16: Phys. (to be published).}}
17: 
18: 
19: \vspace{0.6cm}
20: 
21: \noindent Arnab Das$^\dagger$, Amit Dutta$^\ddagger$ and 
22: Bikas K. Chakrabarti$^\dagger$
23: 
24: \vspace{0.6cm}
25: 
26: \noindent $^\dagger$ Saha Institute of Nuclear Physics,
27: 1/AF Bidhannagar, Kolkata - 700064, India \\
28: 
29: 
30: \noindent $^ \ddagger$ Physics Department, Indian Institute of
31: Technology, Kanpur - 208016, India
32: 
33: \vspace{1.3cm}
34:  
35: \noindent{\bf Abstract:}\quad Here we study zero temperature 
36: quantum phase transition driven by the transverse field for random $\pm J$
37: Ising model on chain and square lattice.  
38: We present some analytical results for one dimension and some numerical results
39: for very small square lattice under periodic boundary condition.
40: The numerical results are obtained employing exact diagonalization technique 
41: following Lanczos method. \\
42: 
43: \vspace{0.6cm}
44: 
45: \noindent{\bf Keywords:}\quad Quantum phase transition, Spin glasses, 
46: Transverse Ising model, Diagonalization techniques.
47: 
48: \vspace{0.6cm}
49: 
50: \noindent{\bf PACS Nos.:}\quad 05.70.Fh, 42.50.L, 75.10.N      
51: 
52: \vspace{0.6cm} 
53: 
54: \noindent {\bf 1.\qquad Introduction} 
55: 
56: \medskip
57: \noindent  
58: The interest in the
59: study of  transverse Ising spin glass models was revived in early
60:  1990 by the discovery of zero-temperature transition in dipolar Ising 
61: transverse field magnet LiHo$_x$Y$_{1-x}$F$_4$ [1]. 
62: Proton glasses such as mixture of ferroelectric and
63: anti-ferroelectric materials like  Rb$_{(1-x)}$(NH$_4$)$_x$(H$_2$P)$_4 $
64:  [2] also provided earlier useful realizations of such
65:  quantum spin 
66: glasses. \\
67: 
68: \noindent These developments initiated extensive theoretical 
69: studies in quantum spin glass models.
70: Ising model in transverse field has already been studied 
71: extensively in this context
72: through analytical approaches using approximate renormalization 
73: techniques and real space
74: renormalization group method, as well as using numerical methods 
75: like quantum Monte Carlo
76:  and exact diagonalization techniques [3]. Fairly extensive studies 
77: on the quantum spin glass
78: phases have been made so far for transverse field  Edward-Anderson
79:  model and 
80: Sherrington-Kirkpatrick model with
81: random exchange distributions [3, 4] and to some extent on quantum 
82: Heisenberg spin glass models
83: [4]. In view of the rigorous developments in the study of two 
84: dimensional nearest neighbour Ising model with random $\pm J$ 
85: exchange interactions  and the precise knowledge of location 
86: of the Nishimori line [5] in
87: such classical spin glass model (driven by temperature), we 
88: consider here the quantum phase
89: transition (at zero temperature) in the same $\pm J$ Ising model.\\
90: 
91:     
92: 
93: \noindent We have shown analytically that introduction of random $-J$ 
94: impurities cannot affect the zero temperature phase transition in  
95: one dimensional system as they can be transformed away.
96: We have also compared and verified the result numerically
97: for the small system size considered.
98: For two dimensional systems,
99:   we present some priliminary results obtained for a square lattice using
100: exact diagonalization results for very small system sizes following Lanczos 
101: technique [6]. Only the behaviours of configurationally averaged
102:  energy gap $\Delta=(E_1 - E_0)$  between the first excited state   
103: and  the ground state  
104:  and the second order response function 
105: $\chi = \left( \partial^2 E_0  / \partial\Gamma^2 \right)$, 
106: equivalent to specific
107: heat, have been studied here. The variations of $\Delta$ and $\chi$ 
108:   with 
109: respect to transverse field $\Gamma$
110: have been obtained, and the phase 
111: boudary has been estimated from these results. \\
112: 
113: \noindent We work with a transverse Ising
114: system, using only nearest neighbour interactions, whose Hamiltonian  
115: is given by
116: \begin{equation}
117: {\mathcal H} = -\sum_{\langle i,j \rangle } J_{ij} S_i^z S_j^z - 
118: \Gamma\sum_{i=1}^N S_i^x ,
119: \end{equation} 
120: \noindent where the transverse field $\Gamma$ is uniform through out the system
121: and the nearest neighbour exchange constant $J_{ij}$'s are chosen randomly from the binary distribution
122: \begin{equation}
123: P(J_{ij}) = p\delta(J_{ij} + J) + (1 - p)\delta(J_{ij} - J).
124: \end{equation}
125: \noindent Here $J$ is taken positive and $p$ is thus the concentration of 
126: anti-ferromagnetic $-J$ bonds in the system.\\
127: 
128:  
129: \noindent {\bf 2. \qquad Results in One Dimensional System }
130: 
131: \medskip 
132: \noindent Here first we show analytically
133:   that in a one dimensional transverse Ising 
134: Hamiltonian with uniform $J$ and $\Gamma$,
135: if we replace some $J$ bonds by $-J$ bonds randomly, then the resulting 
136: Hamiltonian can be gauge transformed back to one with uniform $J$, 
137: and hence the critical field remains unchanged with randomness concentration. 
138: Simillar result for one dimensional system with distributed $J$ 
139: had been obtained earlier [7]. \\
140: 
141: \noindent Let us take the one dimensional random bond tranverse Ising 
142: Hamiltonian 
143: \begin{equation} 
144: {\mathcal H} =   
145: -\sum_i J_i S_i^z S_{i+1}^z - \sum \Gamma S_i^x, 
146: \end{equation} 
147: where the transverse field $\Gamma$ is uniform throughout the system, and $J_{i}$'s are randomly
148: chosen from the same distribution as given in $(2)$. 
149:  Since the $J_i$'s have same magnitude $J$ all through, 
150: and their randomness is
151: only in their sign, we may write $ J_i = J {\mathrm {sgn}}(J_i)$ , 
152: and thus Hamiltonian (3)
153: takes the form 
154: \begin{equation}
155:  {\mathcal H} = 
156: -J\sum {\mathrm {sgn}}(J_i) S_i^z S_{i+1}^z - \Gamma\sum  S_i^x . 
157: \end{equation}
158: 
159: \noindent Now let us  define a new set of spin variables as below
160: 
161: $$ \tilde{S}_i^z = S_i^z \prod_{k=1}^{i-1} {\mathrm {sgn}}(J_k) $$    
162: 
163: $$ \tilde{S}_i^x = S_i^x $$
164: 
165: $$ \tilde{S}_i^y = S_i^y \prod_{k=1}^{i-1} {\mathrm {sgn}}(J_k). $$
166:  It is easy to see that $\tilde{S}$ 's satisfy the same commutation and 
167: anti-commutation relations as those of $S$'s and hence will exhibit exactly
168: the same dynamical behaviour. Now,
169: 
170:   
171: 
172: $$ \tilde{S}_i^z\tilde{S}_{i+1}^z 
173: = S_i^z S_{i+1}^z \left[\prod_{k=1}^{i-1}[{\mathrm {sgn}}
174: (J_k)]^2 \right]{\mathrm {sgn}}(J_i), $$
175:       
176: \noindent or,
177: 
178: $$ \tilde{S}_i^z\tilde{S}_{i+1}^z = S_i^z S_{i+1}^z {\mathrm  {sgn}}(J_i), $$ 
179: since $ [{\mathrm {sgn}}(J_k)]^2 = 1 $. 
180: Thus in terms of new spin variables, Hamiltonian 
181: (4) becomes
182: \begin{equation}
183: {\mathcal H} = -J\sum_i \tilde{S}_i^z \tilde{S}_{i+1}^z - 
184: \Gamma \sum_i \tilde{S}_i^x. 
185: \end{equation}
186:               The above Hamiltonian describes the same random 
187: system in terms of
188: new variables, and yet, as one can see, it has in itself no randomness at all. 
189: One can use Jordon-Wigner transformation in terms of $\tilde{S}$'s and see that
190: here also
191: quantum phase transition occurs only at $\Gamma \ge \Gamma_c (=J)$ as it occurs
192: in a non-random Hamiltonian in $S$'s. In Fig. 1, we present 
193: some data computed for a
194: chain of size $N=9$, which shows that the gap $\Delta$ vanishes 
195:  at $\Gamma_c  \approx 1$ (the field being scaled by $J$). These data for
196: $\Delta = E_1 - E_0$ is obtained from the computed average value of $E_0$ and
197:  $E_1$, each one averaged over about 10 configurations for $p \ne 0$.
198: For infinite system, $\Delta$ is a linear function of $\Gamma$ for 
199: $\Gamma \ge \Gamma_c$. In our case, linearity is observed at high values of
200: $\Gamma$, and $\Gamma_c$ is determined by backward linear extrapolation from
201: the linear region.
202: \begin{figure}[htb]
203: \resizebox{13.0cm}{!}{\rotatebox{270}{\includegraphics{figap1d.ps}}}  
204: \caption{A numerical estimate of the energy gap $\Delta$ 
205: for a chain with $ N $= 9. Phase boundary is obtained from the 
206: location of $\Delta(\Gamma) = 0$, and is shown in the inset.}
207: \end{figure}
208:  
209: %%\newpage
210: 
211: \noindent From the numerical data in Fig. 1, we see that there is a slight
212: variation of $\Gamma_c$ with $p$ ($\Gamma_c$ varies between 0.9 and 1.0). This
213: variation can be attributed to the very small size of the system. However, 
214: it may be noted that with even number of $-J$ bonds in the 
215: chain, with periodic boundary condition, there is no problem of 
216: incommensuration and $E_0 (\Gamma)$ or $E_1 (\Gamma)$ become strictly identical
217: for such values of $p$. Similarly, in every case of odd number of 
218: $-J$ bonds in the 
219: chain, incommensuration problem always occurs for one spin only, rendering
220: identical values (but different from the even $-J$ case) 
221: for $E_0 (\Gamma)$ and $E_1 (\Gamma)$
222: in all such cases. \\ 
223: 
224: 
225: \noindent {\bf{3. \qquad  Results for Square Lattice}}
226: 
227: \medskip
228: 
229: \noindent We consider now the same system (represented by Hamiltonian (1)) on
230: a square lattice of size 3$\times$3 with periodic boundary condition. We again 
231: calculate the ground state and the first excited state energy $E_0$ and $E_1$ 
232: respectievely as functions of the transverse field $\Gamma$, for different
233: values of $p$. Each value of $E_0$ and $E_1$ is averaged over at least 10 
234: configurations for each $p \ne 0$. Apart from $\Delta$, we also calculate
235: $\chi = \partial^2E_0/\partial\Gamma^2$ and their variations with $\Gamma$
236: as shown in Figs. 2 and 3 respectively.
237: \begin{figure}[htb]
238: \resizebox{13.0cm}{!}{\rotatebox{270}{\includegraphics{figap2d.ps}}}
239: \caption{A numerical estimate of the configurational avrage of the energy
240: gap $\Delta$
241: for a square lattice of size 3$\times$3. Phase boundary
242: obtained from $\Delta (\Gamma) = 0$ is outlined
243:  in the inset.}
244: \end{figure}
245: 
246: %\begin{figure}
247: %\resizebox{12.0cm}{!}{\rotatebox{270}{\includegraphics{figspht2d.ps}}}
248: %\caption{Here variation of $\chi = \partial^2 E_0 / \partial \Gamma^2 $ with
249: %$\Gamma$ is shown. The transition point occurs at $\Gamma = \Gamma_c$ where 
250: %$\chi$ diverges; for finite system one gets only a  peak in $\chi$ at 
251: %$\Gamma = \Gamma_c(p)$.
252: %We have outlined the corrosponding phase boundary in the inset. }
253: %\end{figure} 
254: \noindent Our results here are severely 
255: constrained by the system size. The value of pure ferromagnetic critical field 
256: $\Gamma_c(p=0)$ is found here to be around 2.2, while the series 
257: results [3] or cluster algorithms [8] give the value to be around 3.0. This 
258: discrepancy is attributed to the smallness of our system size 
259: $(N = 3^2 )$. However the qualitative behaviour of the order-disorder phase
260: boundary (between ferro/spin glass and para) seems to be reasonable: 
261: $\Gamma_c(p)$ decreases with $p$ initially, and then increases again as $p$
262: approaches unity (pure anti-ferromagnet). The use of periodic boundary 
263: condition here (to avoid some numerical errors) also restricts the domain 
264: features and thereby affects our results. The absence of the knowledge of
265: the ground-state wave function (and the correlation functions) in this method
266: also forbids us to analyse the structure of the ordered phases. \\
267:  
268: 
269: 
270: \begin{figure}
271: \resizebox{12.0cm}{!}{\rotatebox{270}{\includegraphics{figspht2d.ps}}}
272: \caption{Here variation of $\chi = \partial^2 E_0 / \partial \Gamma^2 $ with
273: $\Gamma$ is shown. The transition point occurs at $\Gamma = \Gamma_c$ where 
274: $\chi$ diverges; for finite system one gets only a  peak in $\chi$ at 
275: $\Gamma = \Gamma_c(p)$.
276: We have outlined the corrosponding phase boundary in the inset. }
277: \end{figure}
278: \vskip 0.6 cm
279: 
280: \noindent {\bf Acknowledgement:} BKC is grateful to Hidetoshi Nishimori 
281: for useful discussions. \\\
282: \vskip 1 cm
283: 
284: \noindent {\LARGE\textbf{References}} \\
285: \vspace{0.7cm}
286: 
287: \noindent [1] W. Wu, B. Ellmann, T.F. Rosenbaum, G. Appeli and D.H. 
288: Reich, Phys. 
289: Rev. Lett. {\bf 67} 2076 (1991);  
290:  W. Wu, D. Bitko, T.F. Rosenbaum and G. Appeli, Phys. Rev. Lett. 
291: {\bf 71} 1919 (1993) \\
292: 
293: \noindent [2] R. Pric, B. Tadic and R. Blinc, Z. Phys. B {\bf 61} 69 (1985) \\
294: 
295: \noindent [3] B. K. Chakrabarti, A. Dutta and P. Sen,         
296: {\it Quantum Ising Phases and Transitions in Transverse Ising Models}, 
297: Lecture Notes in Physics, {\bf m41}, Springer, Heidelberg (1996)\\
298: 
299: \noindent [4] R. N. Bhatt, in {\it Spin Glasses and Random Fields}, Ed.
300: A. P. Young, p. 225, World Sc., Singapore (1998) \\
301: 
302: \noindent [5] H. Nishimori, Prog. Theor. Phys. {\bf 66} 1169 (1980);
303:  {\it Statistical Physics of Spin Glasses \& Information Processing : 
304: An Introduction}, Oxford University Press, Oxford (2001); 
305: H. Nishimori and K. Nemoto, J. Phys. Soc. Jpn. {\bf 71} 1198 (2002) \\
306: 
307: \noindent [6] J. Stoer, R. Bulirsch, {\it Introduction to Numerical Analysis},
308: Text in Appl. Maths. {\bf12}, Springer-Verlag, New York, (1993 ) \\
309: 
310: 
311: \noindent [7] B. McCoy, in {\it Phase Transitions and Critcal Phenomena}, vol.
312: II, Eds. C. Domb and M. S. Green, Academic Press, London (1983);   
313: D. S. Fisher, Phys. Rev. B {\bf 50} 3799 (1994) \\
314: 
315: \noindent [8] H. Rieger and N. Kawashima, Euro. Phys. J. B {\bf 9} 233 (1999)
316: 
317:  
318:   
319: 
320: 
321: \end{document}
322: