Where not to find the spectrum of the partial theta function
Yousra Gati, Vladimir Petrov Kostov · 2026
Read Where not to find the spectrum of the partial theta function by Yousra Gati, Vladimir Petrov Kostov (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-29903.txt.
Introduction
Yousra Gati and Vladimir Petrov Kostov map where the spectrum of Ramanujan's partial theta function cannot lie. They show no spectral value falls in a wide sector plus a small disk around zero, of radius just over a fifth. One spectral value sits in the same sector plus a slightly larger disk, at about three tenths. Inside the safe zone, zero moduli split apart by negative half-integer powers of the parameter size. The result fences off most of the sector and pins the lone value inside it.
Catalog record
| CREATOR | Yousra Gati, Vladimir Petrov Kostov |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 25426 |
| LICENSE | http://creativecommons.org/publicdomain/zero/1.0/ |
| RIGHTS | CC0 |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.29903
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Where not to find the spectrum of the partial theta function Yousra Gati, Vladimir Petrov Kostov Abstract . The spectrum of Ramanujan’s partial theta function θ ( q , x ) := ∑ j = 0 ∞ q j ( j + 1 ) / 2 x j \theta(q,x):=\sum_{j=0}^{\infty}q^{j(j+1)/2}x^{j} , q ∈ 𝔻 1 q\in\mathbb{D}_{1} (the unit disk centered at the origin), x ∈ ℂ x\in\mathbb{C} , is the set of values of the parameter q q for which θ ( q , . ) \theta(q,.) has a multiple zero. We show that there is no spectral value in the set 𝕊 ∪ 𝔻 c 0 \mathbb{S}\cup\mathbb{D}_{c_{0}} , c 0 = 0.2078750206 … c_{0}=0.2078750206\ldots , where 𝕊 \mathbb{S} is the sector { 0 < | z | < 0.6 , arg ( z ) ∈ [ π / 4 , 7 π / 4 ] } \{0<|z|<0.6,{\rm arg}(z)\in[\pi/4,7\pi/4]\} . There is a single spectral value in the set 𝕊 ∪ 𝔻 0.31 \mathbb{S}\cup\mathbb{D}_{0.31} which equals 0.309249 … 0.309249\ldots . For q ∈ 𝕊 ∪ 𝔻 c 0 q\in\mathbb{S}\cup\mathbb{D}_{c_{0}} , the moduli of the zeros of θ \theta are separated by the negative half-integer powers of | q | |q| . Key words: partial theta function, Jacobi theta function, Jacobi triple product AMS classification: 26A06 1. Introduction We consider S. Ramanujan’s partial theta function \sum_{j=0}^{\infty}q^{j(j+1)/2}x^{j} (see [ 1 ] and [ 32 ] ) regarding x\in\mathbb{C} as a variable and q\in\mathbb{D}_{1} as a parameter; \mathbb{D}_{r}\subset\mathbb{C} stands for the open disk centered at the origin and of radius r . The function’s name is chosen because of its resemblance with the Jacobi theta function \Theta(q,x)=\sum_{j=-\infty}^{\infty}q^{j^{2}}x^{j} , as \theta(q^{2},x/q)=\sum_{j=0}^{\infty}q^{j^{2}}x^{j} ; “partial” refers to the summation being performed only for non-negative indices. We remind that \Theta(q,x^{2})=\prod_{m=1}^{\infty}(1-q^{2m})(1+x^{2}q^{2m-1})(1+x^{-2}q^{2m-1})~~~\,{\rm(Jacobi~triple~product)} which implies the equalities (1.1) \Theta^{*}(q,x):=\Theta(\sqrt{q},\sqrt{q}x)=\sum_{j=-\infty}^{\infty}q^{j(j+1)/2}x^{j}=\prod_{m=1}^{\infty}(1-q^{m})(1+xq^{m})(1+q^{m-1}/x)~. Thus one can write (1.2) \theta(q,x)=\Theta^{*}(q,x)-G(q,x)~,~~~\,G(q,x):=\sum_{j=-\infty}^{-1}q^{j(j+1)/2}x^{j}=\sum_{j=1}^{\infty}q^{j(j-1)/2}x^{-j}~. The partial theta function is of interest in many domains: Ramanujan type q -series ( [ 32 ] ), the theory of (mock) modular forms ( [ 4 ] ), statistical physics and combinatorics ( [ 28 ] ), asymptotic analysis ( [ 2 ] ), asymptotics and modularity of partial and false theta functions and their interaction with representation theory and conformal field theory ( [ 3 ] , [ 5 ] ), relationship between Appell-Lerch sums and mock theta functions ( [ 24 ] ), Artin-Tits monoids ( [ 6 ] ), quantum many-body systems ( [ 30 ] ). Andrews-Warnaar identities for the partial theta function can be found in [ 31 ] , [ 33 ] and [ 29 ] . The explicit combinatorial interpretation of the coefficients of the leading root of \theta as a series in q is given in [ 27 ] . Padé approximants of \theta are studied in [ 23 ] . For each q fixed, the function \theta is an entire function of order 0 , see [ 12 , formula (5)] . It has countably-many zeros (denoted by \xi_{j} ) which are distinct for |q|<c_{0} , see Notation 1 and part (3) of Remarks 1 . To continue the list of applications of \theta we define its spectrum as the set of values of the parameter q for which \theta(q,.) has a multiple zero; its elements are called spectral numbers or spectral values . Recently, the role played by \theta in the theory of section-hyperbolic polynomials (i. e. real univariate polynomials with all roots real negative and with all their truncations having only real negative roots) was revealed in [ 25 ] , [ 9 ] and [ 22 ] , based on the classical works of Hardy, Petrovitch and Hutchinson (see [ 7 ] , [ 26 ] and [ 8 ] ). The relationship with the spectrum of \theta was mentioned in [ 22 ] . This stimulated the study of the analytic properties of \theta , see [ 14 ] – [ 21 ] . Remark 1 . The positive spectral numbers 0<\tilde{q}_{1}=0.309249\ldots<\tilde{q}_{2}=0.516959\ldots<\tilde{q}_{3}=0.630628\ldots<\cdots form an increasing sequence tending to 1^{-} , see the list of the 6-digit truncations of the first 25 of them in [ 22 ] . The spectral number \tilde{q}_{1} is the closest to 0 . The negative spectral numbers \cdots<\bar{q}_{2}<\bar{q}_{1}=-0.72713332\ldots<0 form a decreasing sequence tending to -1^{+} , see [ 16 , Table 1] . The existence of the complex spectral numbers v_{\pm}:=0.4353184958\ldots\pm i\cdot 0.1230440086\ldots is justified in [ 11 , Proposition 8] . Notation 1 . For q\in\mathbb{D}_{1} , we denote by \mathbb{A}_{a,b}(|q|)\subset\mathbb{C} the open annulus \{|q|^{-a}<|x|<|q|^{-b}\} , 0<a<b , and by \mathbb{S}_{r;\alpha,\beta}\subset\mathbb{C} the sector \{0<|x|<r,{\rm arg}(x)\in[\alpha,\beta]\} , 0\leq\alpha<\beta\leq 2\pi . We denote by c_{0}:=0.2078750206\ldots the solution to the equation 2\sum_{\nu=1}^{\infty}|q|^{\nu^{2}/2}=1 . We set \theta_{k}:=\sum_{j=0}^{k}q^{j(j+1)/2}x^{j} (the k th truncation of \theta ) and \theta_{k}^{\bullet}:=\sum_{j=k+1}^{\infty}q^{j(j+1)/2}x^{j} . Definition 1 . We say that the zeros \xi_{j} of the partial theta function are strongly separated in modulus if (1.3) \xi_{1}\in(\mathbb{D}_{|q|^{-3/2}}\setminus\{0\})~~~\,{\rm and}~~~\,\xi_{k}\in\mathbb{A}_{k-1/2,k+1/2}(|q|)~,~~~\,k\geq 2~. It is clear that if for q=q_{*} , the zeros of \theta are strongly separated in modulus, then q_{*} is not a spectral number. We prove in Section 3 the following theorem: Theorem 1 . (1) For q\in\mathbb{S}_{0.6,\pi/4,7\pi/4}\cup\mathbb{D}_{c_{0}} , the zeros of \theta are strongly separated in modulus. (2) The only spectral number of \theta in the set \mathbb{S}_{0.6,\pi/4,7\pi/4}\cup\mathbb{D}_{0.31} is \tilde{q}_{1} . Remarks 1 . (1) Theorem 1 improves the main result in [ 19 ] which says that for q\in\mathbb{S}_{0.55,\pi/2,3\pi/2} , the zeros of \theta are strongly separated in modulus and that for q\in\mathbb{S}_{0.6,\pi/2,3\pi/2} and k\neq 2 , 3 , conditions ( 1.3 ) hold true. (2) The existence of the following spectral numbers of modulus <0.6 explains why the sector \mathbb{S}_{0.6,\pi/4,7\pi/4} cannot be replaced by the disk \mathbb{D}_{0.6} : \tilde{q}_{1} , \tilde{q}_{2} and v_{\pm} (see Remark 1 ). The radius 0.6 cannot be extended much due to the presence of the spectral number \bar{q}_{1} , see Remark 1 . (3) For q\in\overline{\mathbb{D}_{c_{0}}} , the zeros of \theta are strongly separated in modulus, see [ 11 , Lemma 1] . Acknowledgement. The authors thank Boris Shapiro from the University of Stockholm for the figure included in this paper. 2. Comments 2.1. The spectral values of \theta in \mathbb{D}_{0.6} A natural question to ask is which spectral values of \theta except \tilde{q}_{1} , \tilde{q}_{2} and v_{\pm} belong to the disk \mathbb{D}_{0.6} , because 0.6 is the radius of the sector \mathbb{S}_{0.6,\pi/2,3\pi/2} , see Theorem 1 . To this end one can consider the truncations \theta_{k} of \theta and the spectral values (numbers) of \theta_{k} , i. e. the values of q for which \theta_{k} has a multiple zero. The truncations \theta_{8} , \theta_{9} and \theta_{10} have multiple zeros for (but not only for) the following values of q : \begin{array}[]{cc}0.5374009225\ldots\pm i\cdot 0.1800191987\ldots~,&0.5373319887\ldots\pm i\cdot 0.1803221641\ldots\\ \\ {\rm and}&0.5373387872\ldots\pm i\cdot 0.1803273624\ldots~.\end{array} We list here the truncations up to 4 digits of their 5 respective zeros of least modulus: \begin{array}[]{lll}-2.0471-i\cdot 1.4799~,&-4.7152+i\cdot 1.7057~,&-2.4631+i\cdot 7.6623~,\\ \\ -2.4639+i\cdot 7.6621~,&1.6159+i\cdot 20.5308~,&\\ \\ -2.0467-i\cdot 1.4773~,&-4.7087+i\cdot 1.7093~,&-2.4678+i\cdot 7.6608~,\\ \\ -2.4684+i\cdot 7.6612~,&0.8219+i\cdot 17.0772~,&\\ \\ -2.0467-i\cdot 1.4772~,&-4.7085+i\cdot 1.7091~,&-2.4680+i\cdot 7.6613~,\\ \\ -2.4685+i\cdot 7.6611~,&0.7472+i\cdot 17.0808~.\end{array} Definition 2 . We say “the j th and (j+1) st zero of \theta (or its truncation \theta_{r} ) in terms of the modulus” in the sense that there are exactly