The Hard-Hexagon Model and Rogers--Ramanujan Type Identities

George E. Andrews
PNAS 1981;78;5290-5292
doi:10.1073/pnas.78.9.5290
This information is current as of March 2007.

This article has been cited by other articles:

www.pnas.org#otherarticles
E-mail Alerts Receive free email alerts when new articles cite this article - sign up in the box
at the top right corner of the article or click here.
Rights & Permissions To reproduce this article in part (figures, tables) or in entirety, see:
www.pnas.org/misc/rightperm.shtml
Reprints To order reprints, see:
www.pnas.org/misc/reprints.shtml

Notes:
Proc. Natl Acad. Sci. USA
Vol. 78, No.. 9, pp. 5290-5292, September 1981
Mathematics

The hard-hexagon model and Rogers.-Ramanujan type identities

(statistical mechanics/combinatorial identities/basic hypergeometric series)
GEORGE E. ANDREWS
410 McAllister Building, Pennsylvania State University) University Park, Pennsylvania 16802
Communicated by Donald E. Knuth, May 29, 1981

ABSTRACT In regimefI ofBaxter's solution of the hard-hexa- 00

gon model [Baxter, R. J. (1980)J. Phys. A 13, L61-L70], he pre- (a). (a;q). = fl(1 - aq);
= [1.10]
sents six conjectures.identifying certain one-dimensional.partition j=o
functions with infinite products. An outline of the proof of these
conjectures is given here. HI[al,a2,.-,a,;q] = (a1),,(a9J,,(.. [1.11]
1. Introduction These results either are given explicitly by Rogers (2, 3) or are
In 1980, Baxter (1) found his beautiful solution to the hard-hexa- immediate -consequences of his work: Eq. 1. 1 is equation 1 on
gon model of statistical mechanics.. His treatment ofthis model p. 328 of ref. 2; Eq. 1.2 is equation 2 on p. 329 of ref. 2; Eq.
is naturally divided into four regimes that depend on values 1.3 is equation 2, line 3, on p. 330 of ref. 3; Eq. 1.4 is equation
taken by various parameters associated with the model. Then 2, line 2, on p. 330 of ref. 3; Eq. 1.5 is implicit in the identity
in truly astounding fashion it turns out that eight of equations 2 and 3 on p. 330 of ref. 2 when A = 1 and q is
Rogers-Ramanujan type identities [all essentially. known to replaced by -q (explicitly given by Slater (5), equation 94); Eq.
Rogers (2, 3)] are thefundamental keys for finding infinite prod- 1.6 is equation 13 on p. 332 of ref. 2; Eq. 1.7 is the second
uct representations ofthe related statistical mechanics partition equation of p. 331 ofref. 2; Eq. 1.8 is equation 3, line 2, on page
functions in regimes I, III, and IV. Indeed, the required iden- 330 of ref. 3.
tities are these: For regime II, however, it turns out that one must consider
Regime I the following rather complicated one-dimensional partition
function:
E
n=O
q2/(q)n =
l/l(q,q4;q); [1.1]
Fk(a1) = lim > q i~Fi+2-Tt+1+0+1) [1.12]
mA2°2,3, -- -am
E q 2/(q)n l/H(q2,q3;q5);
n=O
=
[1.2] in which the summation runs over all possible (m - 1)-tuples
Regime III (oT2,...,m) subject to the conditions: (1) or; = 0 or 1 for 1 ' j
'm; (2) o + Sr' 1 for 1 c j S m -1; (3) arm = om+1 =
> qn(3n-l)/2/(q)n(q;q2) = II(qq6,ql0;q10/(q)x; [1.3] °; (4) ai = 1 if i-k(mod 3), otherwise Bri = 0. Baxter obtains
n=O recurrence relations for refinements of these functions Fk(al);
however, the techniques that he applies successfully to solve
> q3n(n+1)/2/(q) (q;q2) +1 = IH(q2,q8,q1';q10/(q),; [1.4] the recurrence relations in the other three regimes fail here.
n=O
For this reason he is unable to find counterparts of the infinite
series in Eqs. 1.1-1.8. By direct expansion he obtains over-
Regime IV whelming evidence to conjecture that each of F1(0), F1(1), F2(0),
F2(1), F3(0), and F3(1) are identical with elegant combinations
2
qn(n+l)/(q)2n+l fl(q3 q7 qlO;ql1&I`(q4 qlS;q2)/(q),;
= [1.5] of infinite products.
n=O In Section 2 we shall give double series expansions'for the
n00 Fk(orl) that indeed establish all six of Baxter's conjectures. Apart
>
n=O
qf( +)/(q)_ I(qq9,q'O;q10IH(q8,q12;q20)/(q) [ from their contribution to Baxter's solution ofthe hard-hexagon
model, these results are also surprising mathematically. They
yafl/()2n = 0)q 17 are not apparently limiting cases of known basic hypergeome-
tric series identities; this is in contradistinction to the fact that
Eqnf/(qq)21= 1/H(q4,q6;qn)(q;q2)
n=O
[1.7] the place of Eqs. 1.1-1.8 in the hierarchy of basic hypergeo-
oo
metric series is well known [cf. Slater (5), Bailey (6, 7)]. In Sec-
tion 3, we shall describe the results and techniques required
n=1
qn2/(9;9q2- q/fI(q9 xql;q )(q;q2)x. [1.8] to establish these theorems.
We are herein utilizing the following standard' notation of
Slater (4) 2. The identities for Baxter's conjectures
n-I THEOREM 1.
(a)n (a;q)n
= =
Hl (1--a); [1.9]
J=o
F1(0) = q(3n2+3n)/2-
The publication costs ofthis article were defrayed in part by page charge n0O=r(3 n.+ 1)/2 (q;q)r(q)n-2r+1
payment. This article must therefore be hereby marked ''advertise-
ment" in accordance with 18 U. S. C. §1734 solely to indicate this fact. ={H1[q4 q 11,q 15 ;q 15 ] + qHl[q.,q 14 q 1 5;q5]}/(q),.-
15

5290
 Mathematics: Andrews Proc. NatL Acad. Sci. USA 78 (1981) 5291

THEOREM 2.
00 (3n2+3n)/2 - r > q 3[[N ] [3.2]
n=O Or53n/2 (q2sq2X(q)32r
= {H[q q ,q ;q'51 - qH[q ,q ,q ;q 5]}/(q)0. in which
(1-.qN)(I1 qN1) ...( 4N-M+l) fo
_
THEOREM 3. [N] f (1 - qM)(1 _qMl).(1 - q) forM_0
co
n(3n-1)/2 - r
M , for M < 0, NO_0,
F2(0) = l y- (q 'q%(qhn-2,1
[3.3]
= ii[q6,q9,q'5;q'5}/(q). and [x] = the largest integer not exceeding x. If in identity 3.1
we let N a0, the first result required for regime III (namely
THEOREM 4. Eq. 1.3) is obtained. Similarly. ifN -X in identity 3.2, we ob-
qn(3n+5)/2 + 1 - r tain Eq. 1.4.
F2(1) =>E To obtain Theorems 1-6, one merely replaces N by 3N + a
n =0 Osr5(3n + 1)/2 (q2;q2)r(qn-2r+1 (a = 0 +t1) in identities 3.1 and 3.2, then-replaces q by q-', next
= qI[q3,q 2,q5;q'5]/(q). multiplies by the minimal power of q necessary to produce poly-
nomials in q, and then lets N ---* 00. This process produces the
identity of series and products described in these theorems, and
THEOREM 5. the relationship between regimes II and III that follows from
00 n(3n+ 1)/2 - r the.replacement of q by q' provides the identity with the var-
F3(0) =
I -0 r3W2 (q2;q )r(q)3n-2 ious Fk(9l)
.4. Conclusion
=
Hi[q6,q9,q'5;q'5]/(q). Of course the immediate interest of the results described here
lies in the fact that the Rogers-Ramanujan type identities for
THEOREM 6. regime II of the hard-hexagon model are now rigorously estab-
co
qn(3nf+1)/2 - r lished. On the other hand, there are numerous interesting long-
F3(1)= > range questions more of interest in the theory of partitions and
n=l O'r'(3n- 1)/2 (q2;q2)r(q)3n-2r-1 q-series that will be. explored in our complete exposition:
(i) Suppose the Rogers-Ramanujan partition ideal [see An-
= qH[q ,q ,q ;q'5]/(q)0. drews (ref. 8, chapter 8) for detailed discussion of partition
ideals] is replaced by another classical partition ideal; what hap-
Baxter had conjectured the identity ofeach ofthe Fk(ol) with pens in regimes II, III, and IV appropriately modified?
the corresponding infinite products given above. Two major (ii) The q - q-' duality of regimes II and III also exists be-
steps are needed to prove these theorems. First, one must de- tween regimes I and IV. Indeed, the relevant polynomial iden-
velop methods for the treatment ofthe expressions given in Eq. tities for this latter relationship are
1.12 so that the double series representations given above can
be found. Second, a set of transformations is required to allow E qJ [N i] = A (-1)AqA(5A+l)/2 [ N j; [4. 1]
identification of the double series with the appropriate infinite
product expression.
3. Outline of'the; proofs of the theorems
Our attack diverges from that of Baxter immediately. Baxter's > e+j [N j]
development of series-product identities relies on the taking of
the limit as m tends to 00 in Eq. 1.12. We instead find repre-
sentations for. the partition functions arising in regime III with
m remaining fixed and finite. We then utilize the powerful fact
that when m is finite one may effectively pass from regime III
to regime II by the transformation q -* q'. Our solution of
regime III (and consequently our proof of Theorems 1-6) relies
on the two following polynomial identities:
=0 (_1)AqA(5A-3)/2 [
E
Lr 2 ]
N + 15j +]

q
These identities were completely stated.in ref. 9 and have their
[4.2]

origin in the work of Schur (10). The arguments used to obtain

Theorems 1-6 from identities 3.1 and 3.2 may now be turned
on identities 4.1 and 4.2 to obtain Eqs. 1.5-1.8, a relationship
qn(3n+1)/2 [N -2n - 2r] q [r + n q
qr previously unnoticed. Thus a "duality theory" between various
sets of identities of the Rogers-Ramanujan type deserves
exploration.
(iii) The analytic duality described above has a corresponding
- A0 ( )Aq5A2+A [[ N i {3.1] manifestation in the pattition-theoretic interpretations of the
various identities considered. Thus the well-known combina-
torial interpretations of Eqs. 1. I -and 1.2 are "dual" to the com-
binatorial interpretations of Eqs. 1.4-4.8 (see refs. 11-13). The
3n(n+)/ [N - 2n-2r-1] [r + n qr possible scope of this duality will also be explored.
5292 Mathematics: Andrews Proc. Natd Acad. Sci. USA 78 (1981)

(iv) One referee has pointed out that if the signs joining the and studying the function that arises by changing the + sign on
two products in Theorems 1 and 2 are reversed, one obtains the left to a - sign.
expressions that'can be converted to products. by using the quin-
tuple product identity (equation 7.4.7 on p. 205 of ref. 4. For This work was partially supported by National Science Foundation
Grant, MCS-75-19162.
example,
1. Baxter, R. J. (1980). J. Phys. A 13, 161-170.
n~q4qll ql5 _ qH[q q14 1q5;q'5] 2. Rogers, J J. (1894) Proc. Lond. Math. Soc. First Ser. 25, 318 -343.
[4.3]
,9 q q ;q1q. 3. Rogers, L. J. (1917) Proc. Lond. Math. Soc. Second Ser. 16,
=
H[q,-q-q q-q
315-336.
This suggests taking the general quintuple product identity, say 4. Slater, L. J. (1966) Generalized Hypergeometric Functions (Cam-
in the form bridge Univ., Canbridge, England).
5. Slater, L. J. (1952) Proc; Lond. Math. Soc. Second. Ser. 54,
147-167.
(1 - a3x3n2)(1 - a3x3n-1)(1 x3n)-
6. Bailey, W. N. (1947) Proc. Lond. Math. Soc. Second Ser. 49,
nnl
421-435.
7. Bailey, W. N. (1949) Proc. Lond. Math. Soc. Second Ser. 50, 1-10.
8. Andrews, G. E. (1976) The Theory of Partitions, Encyclopedia of
Mathematics and Its Applications (Addison-Wesley, Reading,
+ a [j (1 -
a3X3n-1)(1 - a-3x3n2)(1 x3n) -
MA), Vol. 2.
n_1 9. Andrews, G. E. (1970) Scr. Math. 28, 297-305.
10. Schur, I. (1917) Sitzungsber. Akad. Wiss., Bert KL, Math. Phys.
Tech. 302-321.
= [l (1 - a2X2n -1)(1 -
a-2X2n-1) 11. Gordon, B. (1965) Duke Math. J. 31, 741-748.
nal 12. Andrews, T. E. & Askey, R. (1977) in Higher Combinatorics, ed.
Aigner, M. (Reidel, Dordrecht, The Netherlands), pp. 3-25.
13. Hirschhorn, M. D. (1979) J. Combinatorial Theory Ser. A 27,
(1 + ax ')(1 + a lx')(1 -x') 33-37.