Complex singularities in the specific heat of the SU(2) lattice gauge model

Complex singularities in the specific heat of the SU(2) lattice gauge model

Volume 102B, number 4 PHYSICS LETTERS 18 June 1981 COMPLEX SINGULARITIES IN THE SPECIFIC HEAT OF THE SU(2) LATTICE GAUGE MODEL M. FALCIONI a, E. MA...

186KB Sizes 9 Downloads 50 Views

Volume 102B, number 4

PHYSICS LETTERS

18 June 1981

COMPLEX SINGULARITIES IN THE SPECIFIC HEAT OF THE SU(2) LATTICE GAUGE MODEL M. FALCIONI a, E. MARINARI

a,b,

M.L. PACIELLO a, G. PARISI c and B. TAGLIENTI a

a Istituto Nazionale di Fisica Nucleate, Sezione di Roma, Rome, Italy b lstituto di Fisica dell'Universitd di Roma, Rome, Italy c Istituto Nazionale di Fisica Nucleate, Laboratorio Nazionali di Frascati, Frascati (Rome), Italy Received 19 March 1981

A recent Monte Carlo simulation showing a sharp peak versus/3 in the specific heat for an SU(2) lattice gauge theory is compared with the high temperature expansion. An interpretation based on complex ~-plane singularities is proposed.

In the last years, Monte Carlo computer simulations have provided a powerful tool to study lattice gauge theories in the non-perturbative region [1,2]. These methods were introduced to study the properties of statistical systems [3] and quite recently Wilson [4] suggested an extension to field theory. Computer simulations of lattice gauge theories for different gauge groups have been performed by several authors [5,6] to investigate the occurrence of phase transitions in the intermediate region lying between the high-temperature (strong coupling) regime and the low-temperature (weak coupling) one. In particular, Creutz [5] analyzed the average plaquette energy as a function of/3 - 1/kT = 4/g 2 and found a considerable variation of the slope near/3 2.2. To analyze the nature of the transition between the strong and weak coupling it seems more suitable to study-the specific heat rather than the energY. Recently, Lautrup and Nauenberg have performed a Monte Carlo simulation of the specific heat C(/3) for U(1) [7] and SU(2) [8] gauge groups. In both cases a sharp peak shows up, with a completely different behaviour under the increasing of the lattice size: the peak height increases for U(1), as it is expected in a phase transition, and remains nearly constant for SU(2). While for the U(1) group one expects a transition from a confining phase to a deconfining one, for the SU(2) group such an interpretation seems to be unlikely. 270

The purpose of this letter is to provide an explanation of the peak of ref. [8] in terms of complex/3-singularities rather than a phase transition. In our analysis, we will assume that the available terms of the high-temperature expansion are enough to get a rough picture of the nearby singularities. The high-temperature expansion of the average plaquett e energy is: 10 N=2 the E N are positive numbers given by Wilson [9] * 1, fsu(2) dU l t r U exp(1/3 tr U) j(/3) =

fsu(2) dU exp(½/3 tr U) (2)

= -2//3 +I0(/3)/I1(~) = ~ o Ak/32k+l' 10(/3) and 11(/3) being Bessel functions. From (1) the specific heat is obtained:

C(/3) = _(32 dE/d~3

(3) lO

=/32 dJ (l +N~=2 (2N+ l ) E N ( - 1 ) N j 2 N + o ( j 2 2 ) ) *1 The values ofE N can be found in ref. [6]. Up to the order jlS the EN coefficients can be found in ref. [10].

0 031-9163/81/0000-0000/$ 02.50 © North-Holland Publishing Company

Volume 102B, number 4

PHYSICS LETTERS

18 June 1981

10(

The ratio test applied to the first terms suggests a singularity on the negative real axis at j 2 ~ - 1 / 6 ; as we are trying to reach a singularity at fl ~ 2.2 (corresponding to j 2 ~. 0.22), we push the j 2 ~ - 1 / 6 singularity far away by a conformal transformation. We choose:

a ( z ) = N =~ 2

z = 6 j 2 / ( 6 J 2 + 1).

F(z) contains two complex singularities at z = Zc(1

(4)

The specific heat reads now as: 10 CO3)=j32d-~(1 + ~ r N z N + O ( z l l ) ) N=2

,

(5)

where the coefficients rN are easily derived from E N. As expected the rN are all positive, suggesting a singularity on the real axis. Under this hypothesis, the ratio test gives now a singularity at z -~ 0.59, corresponding to/3 ~ 2.4. This is an important point to remark: the peak of Lautrup and Nauenberg [8] occurs at a value of 13near the point where the high-temperature expansion suggests a singularity, however the precise values of the singularity points do not coincide as it should be. It is possible that this discrepancy is due to the finite length of the high-temperature expansion or to finite size effects in the Monte Carlo simulations. In this letter we explore an alternative point of view i.e.: both the asymptotic behaviour of the hightemperature expansion and the "experimental" peak in the specific heat originate from a pair of complex singularities near the real axis, as suggested in ref. [ 11 ]. The reader may find it interesting that the specific heat of a two-dimensional superconductor in the Random Phase Approximation shows both a marked peak and a pair of complex singularities near the real axis [ 12]. In the RPA approximation this two-dimensional system is similar to gauge theories as far as it has both asymptotic freedom and absence of phase transition. I n t h e hypothesis of the existence of complex singularities a more refined analysis is needed for the SU(2) specific heat. We proceed by looking for a function with the following properties: (i) It has the right high temperature expansion, (ii) It reproduces the simulation data near fl ~ 2.2. We have studied the following simple f o r m :

CO3) = fl2(dJ/d~)[1 + F(z) + a ( z ) ] ,

(6)

where:

Y(z) = [A(1 - zlzc) + B ] / [ ( i

-

zlzJ

+

~>21.r12,

(7)

rN

N! dzNz=o (8)

+ i5); by construction eq. (6) automatically satisfies condition (i). The adjustable parameters A, B, z c, 5, 3' must be chosen in such a way that (i) CO3) =/32(dJ/d13)[1 + F(z) + G(z)] fits the simulation data near fl ~ 2.2. (ii) G(z) must be well approximated by a low-order polynomial in the region of interest near z ~ 0.57; in practice only its first 4 ~ 5 terms must be sensibly 4= 0. (This amount to that the singularities of G(z) are further away.) Two reasonable sets of values we get for the parameters are:

(a): A = 1.52,

B =0.657,

zc

0.570,

:

(9) 6 = 0.0507,

3' = 0.530.

1.5

t

\

\

i

I

I

1

2

3

\

,~

Fig. 1. Specific heat as function of/3; the dots indicate the simulation data, the solid curve is the fit (a), the dashed curve is the fit (b). 271

Volume 102B, number 4

PHYSICS LETTERS

Table 1 CoefficientsgN defined in (8) for the two fits (a) and (b) [see (9)] compared with the rN as defined in (5).

N

rN

gN (fit a)

gN (fit b)

2 3 4 5 6 7 8 9 10

0.5556 0.8519 1.3241 2.1287 3.5196 5.9055 9.9365 16.6354 27.5977

0.330 0.188 0.861 -0.735 -0.805 -0.109 -0.726 0.676 -0.519

0.210 0.109 0.264 -0.530 -0.112 -0.125 -0.731 0.167 -0.631

× 10 -1 X 10 -2 × 10 -1 × 10 -1 10 -2 x 10 -3

×

X 10 -1 × 10 -1 × 10 -3

(b): A =1.04,

B =0.623,

6 --- 0.0532,

zc--0.565,

3' = 0.517.

The reader m a y w o n d e r h o w m a n y terms o f the high-temperature expansion are needed to detect complex singularities. The answer is given in fig. 2, where the ratios

R N _ CN CN+ 1 ' × 10 -1 × 10 -1

(9 c o n ' d )

In fig. 1 we compare the results o f the fit with the simulation data, in table 1 we give the coefficients gN o f G(z) c o m p a r e d w i t h rN.

18 June 1981

CN

_ 1 d _ ~ , F(z)! N ! d(ZlZc)~V Iz=O

(10)

are p l o t t e d versus N . We get R N ~ 1 (as for a real singularity at z = Zc) up to N ~ 30; after this order we have a fast decrease. We conclude that the c o m p l e x singularities hypothesis, although it has m a n y attractive features, is very hard to test by looking at the high-temperature expansion only (unless very long series w o u l d be available). We are i n d e b t e d to K. Wilson for having provided us w i t h the coefficients o f the high-temperature expansion. We thank also C. R e b b i for having provided us w i t h a c o p y o f ref. [6] prior to publication and N. Sourlas for useful discussions.

References

Cli CN+I

Q0 0000gqIQIIQOOOOQIO QOOOBQQQ

I

I

8

I

I

16

I

I

24

I

I

32

N

Fig. 2. Ratios R N = CN/CN+ 1 [see eq. (10)] versus N for the fit (a), a very similax behaviour is obtained for the fit (b). 272

[1] K. Wilson, Corneil preprint CLNS/80/442 (January 1980). [2] For a recent review see: G. Parisi, talk XXth Conf. on High energy physics (Madison, 1980), Frascati preprint LNF-80/52 (P) (September 1980). [3] K. Binder, ed., Monte Carlo methods (Springer, Berlin, 1979). [4] K. Wilson, Carg6se Summer School (1977) (Plenum, New York, 1978). [5] M. Creutz, L. Jacobs and C. Rebbi, Phys. Rev. D20 (1979) 1915; M. Creutz, Phys. Rev. Lett. 43 (1979) 553; M. Creutz, Phys. Rev. Lett. 45 (1980) 313; M. Creutz, Phys. Rev. D21 (1980) 2308; C. Rebbi, Phys. Rev. D21 (1980) 3350. [6] G. Bhanot and C. Rebbi, SU(2) string, gluebail mass and interquark potential by Monte Carlo computations •(1980), to be published. [7] B. Lautrup and M. Nauenberg, Phys. Lett. 95B (1980) 63. [8] B. Lautrup and M. Nauenberg, Phys. Rev. Lett. 45 (1980) 1755. [9] K. Wilson, private communication. [10] R. Balian, J.M. Drouffe and C. Itzykson, Phys. Rev. D l l (1975) 2104; D19 (1979) 2514 (E). [11] D.J. Gross, Les Houches (1980), to be published in Phys. Rep. [12] W.E. Masker, S. Mar~elja and R.D. Parks, Phys. Rev. 188 (1969) 745.