1: \documentstyle{article}
2: \begin{document}
3: \title{\hspace{5mm}Low Energy Behaviour of XXZ Antiferromagnetic Spin
4: Chain}
5: \author{B. Roostaei\footnote{Email:Bahmanr@mehr.sharif.ac.ir}}
6: \date{\small \it Department of Physics,Sharif University of Technology,P.O. Box 19395-5534,Tehran-Iran \\
7: and Institute for Studies in Theoretical Physics and
8: Mathematics(IPM),P.O. Box 11365-9161,Tehran-Iran} %
9: \maketitle %
10: \centerline{{\bf Abstract}}%
11: {\small The zero temperature phase diagram of XXZ spin chain in
12: external magnetic field is investigated at low energies using
13: path integral approach.It has been shown by spin wave analysis
14: and then by nonlinear sigma model transformation that below some
15: critical field the system undergoes quantum Kosterlitz-Thouless
16: phase transition.Above that critical field for low anisotropies
17: the system is saturated while at strong anisotropy the system has
18: Ising antiferromagnetic order.}
19: \section{Introduction}
20: Considerable effort has been made to understand the long
21: range (low energy) behaviour of quantum spin systems specially in
22: low dimensions.The importance of these studies is in their
23: applicability in various physical phenomena in for example quantum
24: Hall ferromagnets[1], high $T_c$ superconductors[2] and many
25: other magnetic systems.On the other hand we know very few models
26: to be exactly solvable, nevertheless development of field
27: theoretical approaches has greatly changed the language in this
28: part of strongly correlated systems and has provided us with
29: effective methods to be applied to our problems of spin
30: systems.One of the great observations by means of field
31: theoretical approach has been the problem of ground state of 1D
32: Heisenberg antiferromagnet: after separating out the rapidly
33: varying fluctuations we obtain a nonlinear sigma
34: model(NL$\sigma$M) with a topological term which is responsible
35: for destroying the long range order of ground state for integer
36: spin chains according to Haldane's conjecture[3] while we know
37: that this problem is exactly solvable by Bethe's ansatz only for
38: spin half problem[4].The Haldane's conjecture has been now
39: confirmed by strong experimental and numerical evidence[4,5,6].
40: The method of separating out short wavelength fluctuations which
41: resulted in NL$\sigma$M+topological term for antiferromegnetic
42: spin chains is now widely used for various spin systems including
43: Ferrimagnets[7] and spin ladders[8].In this paper I am going to
44: apply this method of obtaining low energy behaviour to anisotropic
45: spin chains.
46: \section{Spin Wave Analysis}
47: For the anisotropic Heisenberg model in magnetic field:
48: \begin{equation}
49: \hat{\mathcal {H}}=J\sum_{<ij>} [{1\over 2}
50: (S_i^{+}S_j^{-}+S_i^{-}S_j^{+})+\lambda
51: S_i^{z}S_j^{z}]+\sum_{j=1}^{N}{\bf B}\cdot{\bf S}
52: \end{equation}
53: where $0\leq\lambda\leq 1$ there are numerical evidence[9] that
54: under some critical value for the magnetic field $B_c$ the $S_z$
55: fluctuations become negligible and the system behaves as XY model
56: in long range.On the other hand we know that the isotropic
57: Heisenberg antiferromagnet becomes massive by turning on magnetic
58: field[8].These two considerations suggest that $B_c\propto
59: (1-\lambda)^{-\nu}$ for XXZ model.This may be examined by some
60: crude spin wave analysis: I Apply the Holstein-Primakov
61: transformation which is some kind of perturbation around
62: saturation point: $S_i^{z}=s-a_i^{\dag}a_i$ and
63: $S_i^{-}=(2s)^{1\over 2}a_i^{\dag}(1-{a_i^{\dag}a_i\over
64: 2s})^{1\over 2}$. Here $a^{\dag}$ creates a boson to lower the
65: $S^{z}$ by one.We are in a situation where the number of bosons
66: are low enough to approximate the square root and obtain:
67: $S_i^{-}\approx (2s)^{1\over 2}a_i^{\dag}$.Applying this
68: transformation to the Hamiltonian (1) for ${\bf B}=B\hat z$ and
69: Fourier expanding we find:
70: \begin{equation}
71: \hat{\mathcal{H}}\approx
72: \sum_{k}[2(\gamma_k-z\lambda)Js-B]a_k^{\dag}a_k
73: \end{equation}
74: in which $\gamma_k\equiv\sum_{\bf a}e^{ik\cdot {\bf
75: a}}=\gamma_{-k}$. ${\bf a}$ represents the translation vectors to
76: the neighboring sites and $z$ is the number of nearest neighbors.
77: \newline
78: In one dimension the spin wave spectrum becomes gappful:
79: $E_k=4Js(\cos ak-\lambda)-B$ with the gap $\Delta
80: =4Js(1-\lambda)-B$.This estimation confirms my suggestion and sets
81: $\nu =1$.Under $B_c$ the saturation point becomes unstable.This
82: elementary analysis will be invalid as $B$ approaches zero and is
83: valid just near $B_c$.This result shows that near $B_c$ the
84: correlation length of $S_z$ fluctuations $\xi \propto \mid
85: B-B_c\mid^{-1}$. In section 3 after separating out short range
86: $Ne\grave{e}l$ field we can see that the correlation length of
87: $S_z$ fluctuations is actually $\xi\propto (1-\lambda)^{-{1\over
88: 2}}$ for $\lambda\leq 1$.
89: \newline
90: For the case of $\lambda\geq 1$ the analysis is a bit
91: complicated.In this regime to find the fluctuations around Ising
92: antiferromagnetic order(IAF) we can again apply the
93: Holstein-Primakov transformation for two sublattices:
94: $S_i^z=s-a_i^{\dag}a_i$ and
95: $S_i^{-}=\sqrt{2s}a_i^{\dag}(1-{a_i^{\dag}a_i\over 2s})^{1\over
96: 2}$ for $i$ in sublattice A and for sublattice B:
97: $S_i^z=-s+b_i^{\dag}b_i$ and
98: $S_i^{-}=\sqrt{2s}(1-{b_i^{\dag}b_i\over 2s})^{1\over 2}b_i$.
99: Here we have two kind of Bosons one $a^{\dag}(b^{\dag})$
100: lowers(adds) the $S_z$ by one in sublattice A(B).By applying the
101: first order approximation: $S_i^{-}\approx \sqrt{2s}a_i^{\dag}$
102: for A and $S_i^{-}\approx \sqrt{2s}b_i$ for B, and inserting in
103: the Hamiltonian (1) we find:
104: \begin{eqnarray}
105: \hat{\mathcal{H}}\approx Js\sum_{i\in A}\sum_{i\in
106: B}[a_ib_j+a_i^{\dag}b_j^{\dag}+\lambda
107: (a_i^{\dag}a_i+b_j^{\dag}b_j)]+B\sum_{i\in
108: B}b_i^{\dag}b_i-B\sum_{i\in A}a_i^{\dag}a_i \\ \nonumber %
109: =Js\sum_k \gamma_k(a_kb_{-k}+a_k^{\dag}b_{-k}^{\dag})+\alpha\sum_k
110: b_k^{\dag}b_k+\beta\sum_k a_k^{\dag}a_k \; . \hspace{1.8cm}
111: \end{eqnarray}
112: Where $\alpha\equiv Js\lambda z+B$ and $\beta\equiv Js\lambda
113: z-B$ and we have Fourier expanded the operators in the last
114: equality . Now to diagonalize (3)I apply the (canonical)
115: Bogoliubov transformation :
116: \begin{eqnarray}
117: c_k=u_ka_k-v_kb_{-k}^{\dag} \\ \nonumber %
118: d_k=u_kb_k-v_ka_{-k}^{\dag}
119: \end{eqnarray}
120: For the transformation to be canonical we must have:$u^2-v^2=1$
121: and for (3) to be diagonal we must have:
122: $Js\gamma(u^2+v^2)+(\alpha+\beta)uv=0$. No problem arises if we
123: assume $u$ and $v$ are real. Finally we obtain:
124: \begin{eqnarray}
125: \hat{\mathcal{H}}\approx\sum_k {\mathcal{E}}^{(1)}_k
126: c_k^{\dag}c_k+\sum_k {\mathcal{E}}^{(2)}_k d_k^{\dag}d_k \\
127: \nonumber {\mathcal{E}}^{(1,2)}_k\equiv
128: 2Js(\lambda^2-\cos^2ak)^{1\over 2}\pm B .
129: \end{eqnarray}
130: One of The branches is always gapfull which has to be the
131: spectrum of fluctuations trying to align spins against the
132: magnetic field direction and the other one is also gapfull which
133: tries to saturate the system, but in this branch the gap vanishes
134: at: $B_c=2Js(\lambda^2-1)^{1\over 2}$. This result shows that for
135: $\lambda\geq 1$ the correlation length $\xi\propto
136: (\lambda-1)^{-{1\over 2}}$ that is $\nu ={1\over 2}$ above the
137: critical point $\lambda =1$. This result is more reliable for the
138: following reason:
139: \newline
140: we can compute the site magnetization reduction due to quantum
141: fluctuations in our spin wave approximation:
142: \begin{eqnarray}
143: \mid\Delta <S_z>\mid = \int_{-{\pi\over a}}^{\pi\over a} v_k^2 dk
144: =
145: \int_{-{\pi\over a}}^{\pi\over a} (\frac{\lambda z/2}{\sqrt
146: \lambda^2z^2-\gamma_k^2}-{1\over 2})dk \\ \nonumber
147: =-{2\pi\over a}+{4{\sqrt 2}\lambda\over a}\frac{F({-1\over
148: \lambda^2-1})}{\sqrt{\lambda^2-1}} \; .
149: \end{eqnarray}
150: Where $F$ is the complete elliptic integral of first kind.We can
151: see in Fig.I that this reduction is finite at $\lambda > 1$ and
152: vanishes very fast as $\lambda$ grows. %
153: \par
154: Now we can sketch the line of transition from saturation to IAF
155: and other phases as follows(fig.II):
156: \begin{eqnarray}
157: B_c(\lambda)=2Js\sqrt{\lambda^2-1} \hspace{1cm} \lambda\geq 1
158: \\ \nonumber
159: B_c(\lambda)=4Js(1-\lambda) \hspace{1.2cm} \lambda\leq 1 .
160: \end{eqnarray}
161: \section{Field Theoretic Discussion}
162: In this section I am going to analyze the anisotropic spin chain
163: by means of nonlinear sigma model transformation. The Minkowskian
164: action for anisotropic Heisenberg chain
165: ${\hat{\mathcal{H}}}=\sum_{j=1}^{N}\alpha_{\mu\nu}S_j^{\mu}S_{j+1}^{\nu}-\sum_{j=1}^{N}{\bf
166: B}\cdot {\bf S}$ with the symmetric rank two tensor $\alpha$ is :
167: \begin{equation}
168: S_M=s\sum_{j=1}^{N} S_{WZ}[{\bf n}_j]-\int_{0}^{\beta\hbar} dx_0
169: \sum_{j=1}^{N}
170: \alpha_{\mu\nu}n_j^{\mu}n_{j+1}^{\nu}-\int_{0}^{\beta\hbar}dx_0
171: \sum_{j=1}^{N} {\bf B}\cdot {\bf n} \; .
172: \end{equation}
173: Where $x_0$ is the real time and $\beta=1/k_BT$ that in this
174: paper we discuss the zero temperature situation
175: $\beta\hbar\rightarrow\infty$.The first term known as Wess-Zumino
176: term is a Berry phase for each spin and represents the area of a
177: cap of the unit sphere formed by the closed path passed by the
178: unit vector ${\bf n}(x_0)$ from $x_0=0$ to $x_0=\beta\hbar$[10]
179: .Under the critical magnetic field $B_c$ we can separate out
180: rapidly varying fluctuations as follows:
181: \newline
182: First of all we stagger the field: ${\bf n}_j\rightarrow
183: (-1)^j{\bf n}_j$ then we write ${\bf n}_j$ as a short range
184: $Ne\grave{e}l$ field ${\bf l}_j$ plus fluctuations around it
185: ${\bf m}_j$ : ${\bf n}_j={\bf m}_j+a_0(-1)^j{\bf l}_j$ where
186: ${\bf m}^2=1$, ${\bf m}\cdot {\bf l}=0$ and $a_0$ is the lattice
187: constant. This method works only for two sublattice systems close
188: to $Ne\grave{e}l$ state which is not the eigenstate of
189: $\hat{\mathcal{H}}$ of course!
190: \newline
191: Expanding the action up to second order in ${\bf l}$ and
192: integrating it out we obtain $ Z=\int {\mathcal{D}}{\bf m}\delta
193: ({\bf m}^2-1) \; e^{iS_{eff}}$ where :
194: \begin{eqnarray}
195: S_{eff}=\int d^2x [{1\over
196: a_0}\alpha_{\mu\nu}m^{\mu}m^{\nu}+{a_0\over
197: 2}\alpha_{\mu\nu}m^{\mu}\partial_1^2m^{\nu}]+
198: {s\over 2}\int d^2x \; {\bf m}\cdot (\partial_0{\bf m}\times
199: \partial_1{\bf m})+ \\ \nonumber
200: +{s^2\over 4a_0}\int\alpha_{\mu\nu}^{-1}[({\bf m}\times\partial_0
201: {\bf m})_{\mu}({\bf m}\times\partial_0 {\bf m})_{\nu}]d^2x
202: -{1\over Js\lambda a_0}{\bf B}\cdot ({\bf m\times\partial_0 m}) .
203: \hspace{0.8cm}
204: \end{eqnarray}
205: For the XXZ model $\alpha_{\mu\nu}=Js^2 \; diag(1,1,\lambda)$ by
206: which we find:
207: \begin{eqnarray}
208: {\mathcal{L}}_{eff}=\frac{s}{4a_0 J}(\partial_0
209: m_z)^2-{a_0J\lambda\over 2}(\partial_1 m_z)^2-{J\over
210: a_0}(1-\lambda)m_z^2-{a_0J\over 2}[(\partial_1 m_x)^2+ \\
211: \nonumber +(\partial_1 m_y)^2]+\frac{1}{4a_0J}\{(\partial_0
212: m_y)^2[m_z^2+\lambda^{-1}m_x^2]+(\partial_0
213: m_x)^2[m_z^2+\lambda^{-1}m_y^2]+ \\ \nonumber
214: -2(m_ym_z\partial_0m_z\partial_0m_y+m_xm_z\partial_0m_x\partial_0m_z+\lambda^{-1}m_xm_y\partial_0m_y\partial_0m_x)\}+
215: \\ \nonumber
216: -{B\over Js\lambda a_0}({\bf m\times\partial_0 m})\cdot\hat{z}
217: +{\theta\over 8\pi}\epsilon_{\mu\nu}{\bf
218: m}\cdot(\partial_{\mu}{\bf m}\times\partial_{\nu}{\bf m})\; .
219: \hspace{3cm}
220: \end{eqnarray}
221: where for the last term $\mu ,\nu =0,1$.The last term is the
222: Pontryagin index (or winding number) of field ${\bf m}$ as a map
223: from $S^2$ (real space-time) to $S^2$ (configuration space) and
224: is a topological term and $\theta=2\pi s$[10]. This is the well
225: known $\theta$-term that for isotropic Heisenberg chain resulted
226: in Haldane's conjecture, however in XXZ chain as we can see in
227: (10) due to massive fluctuations of $m_z$ $(\Delta\propto
228: 1-\lambda)$ the $z$ component will be negligible in ground state,
229: in other words this component is short range and so at long
230: wavelengths(low energies)is irrelevant.To find the low energy
231: behavior we can use the mean-field approximation $m_z\approx 0$.
232: In this approximation the winding number will be ineffective.
233: \section{Topological Phase Transition at T=0}
234: In our mean-field approximation removing the $z$ component in the
235: Lagrangian density results in:
236: \begin{eqnarray}
237: {\mathcal{L}}_{eff}=-{a_0J\over 2}[(\partial_1 m_x)^2+(\partial_1
238: m_y)^2]+\frac{1}{4a_0 J\lambda}[(\partial_0 m_y)^2m_x^2+ \hspace{2cm} \\
239: \nonumber
240: +(\partial_0 m_x)^2m_y^2 -2m_xm_y\partial_0 m_x\partial_0 m_y]-\frac{B}{J\lambda sa_0}(m_x\partial_0m_y-m_y\partial_0m_x) \\
241: \nonumber
242: \rightarrow\frac{1}{2g}[(\partial_0\varphi)^2-(\partial_1\varphi)^2]-{2B\over
243: g s}(\partial_0 \varphi) \; . \hspace{4.3cm}
244: \end{eqnarray}
245: Where in the last line we have parametrized ${\bf m}$ as
246: $(\cos\varphi,\sin\varphi,0)$ and scaled the space-time so that
247: the spin wave velocity is one and $g=(2\lambda)^{1\over 2}$ is
248: the coupling constant.This is the Lagrangian for massless
249: fluctuations of XXZ model in 1+1 dimensions.(The term coupled to
250: magnetic field is a total derivative, i.e. the magnetic field has
251: no effect on in plane excitations in mean-field approximation.)
252: The Lagrangian in (11) has topological defects for the field
253: $\varphi(x_0,x)$ is an angle.These defects drastically alter the
254: behavior of system when they are present.In this
255: scenario(Kosterlitz-Thouless), at weak couplings $(g<g_c)$ the
256: ground state has bound vortices(defects in space-time) and is
257: partially ordered with algebraically falling correlations.At
258: strong couplings the vortices become unbound and the partial order
259: completely vanishes and the correlations fall off exponentially
260: with correlation length $\xi\propto 1/\ln(2g))$.This drastic
261: change is a second order quantum phase transition at zero
262: temperature without symmetry breaking.
263: \section{Summary}
264: The low energy phase diagram of antiferromagnetic XXZ chain in
265: magnetic field is shown in fig.II . At $0\leq\lambda<1$ and under
266: $B_c\approx 4Js(1-\lambda)$ we have two kind of ground states: at
267: $\lambda<\lambda_{KT}$ the ground state is partially ordered while
268: at $\lambda>\lambda_{KT}$ the ground state has free vortices and
269: is disordered.Above the critical magnetic field the $z$ component
270: is no longer ignored and we have saturated ferromagnetic(SF)
271: order.At $\lambda>1$ and $B<B_c$ obviously the system has
272: antiferromagnetic Ising (IAF) behaviour at low energies while in
273: this regime and above $B_c$ we have again saturated phase. For
274: $B=0$ it is inferred that we have a quantum phase transition at
275: $\lambda=1$ from disordered phase to ordered IAF phase which has
276: already been shown in continum limit for spin half case by
277: Bosonization method[10].
278: \section{Acknowledgement}
279: The author is grateful to IPM for funding this research as a
280: student research work .
281: \section{References}
282: [1]S.M. Girvin, Cond-Mat/9711233 and references therein. \newline
283: [2]P.W. Anderson, Science{\bf 235},1196(1987), S.Chakravarty,B.I.
284: Halperin,D.R. Nelson, Phys. Rev. B,{\bf 39},2344(1989) and
285: references therein. \newline [3]F.D.M. Haldane, Phys. Rev.
286: Lett.,{\bf 50},1153(1983). \newline [4]E. Fradkin,{\it Field
287: theories of Condensed Matter Systems}(Addison-Wesely,1991),Chp.4
288: \newline [5]I. Affleck, J. Phys. :Cond. Matt. {\bf
289: 1},3047(1989). \newline [6] A. Auerbach, {\it Interacting
290: Electrons and Quantum Magnetism}(Springer-Verlag,New York,1995) .
291: \newline
292: [7]R. Abolfath, H. Hamidian,A. Langari, Cond-Matt/9901063 .
293: \newline
294: [8] B. Normand,J. Kyriakidis,D. Loss, Cond-Matt/9902104 .
295: \newline
296: [9]A. Langari, Phys. Rev.B,{\bf 58},14467(1998) .
297: \newline
298: [10]Ref.4 charter 5.
299: \newline
300: [11]A.Lopez,A.G.Rojo,E.Fradkin, Phys. Rev. B,{\bf 49},15139(1994).
301: \vspace{2cm}
302: \begin{center}
303: Figure caption
304: \end{center}
305: Fig. I)Site magnetization reduction due to quantum fluctuations
306: at ground state in the regime of higher anisotropy $\lambda > 1$ .
307: \newline
308: \newline
309: Fig. II)The low energy phase diagram of the XXZ antiferromagnetic
310: chain.The straight line shows the threshold of saturation.The
311: curve $\sqrt{B^2+1}$ is the threshold of Ising
312: Antiferromagnetism(IAF).
313: \end{document}
314: