Resonances sets of Schrödinger operators Thanks: The work of PG in Sections 2, 3 and 6 was performed at the Saint Petersburg Leonhard Euler International Mathematical Institute and supported by the Ministry of Science and Higher Education of the Russian Federation (agreement no. 075–15–2025–343). The work of YB of Sections 4 and 5 is supported by the Ministry of Science and Higher Education of the Russian Federation in the framework of a scientific project under agreement No. 075-15-2025-013. YB and PG are winners of the BASIS competitions and would like to thank its sponsors and jury. Yurii Belov and Pavel Gubkin Abstract . We prove that resonances of the Schrödinger operator with compactly supported potential can contain arbitrary subset of the angle { z : − Im z > C | Re z | } \{z:-\mathop{\rm Im}z>C|\mathop{\rm Re}z|\} that satisfies Blaschke condition. We also establish sufficient conditions for the subsets of wider domains. 1. Introduction In the present paper we consider one-dimensional Schrödinger operators \mathcal{H}_{q} on the half-line \mathbb{R}_{+}=[0,+\infty) of the form \displaystyle\mathcal{H}_{q}=-\frac{d^{2}}{dx^{2}}+q(x), with real-valued compactly supported potentials q\in L^{1}(\mathbb{R}_{+}) . The Jost solution \psi(k,x) of \mathcal{H}_{q} is defined as the unique solution of \displaystyle-\psi^{\prime\prime}(k,x)+q(x)\psi(k,x)=k^{2}\psi(k,x),\qquad k\in\mathbb{C},\qquad x\geqslant 0, that satisfies \psi(k,x)=e^{ikx} for all large x . The corresponding Jost function is defined by \displaystyle w(k)=\psi(k,0),\qquad k\in\mathbb{C}. This function is entire, its zeroes in the lower half-plane \{z\colon\mathop{\rm Im}z<0\} are called resonances of the operator \mathcal{H}_{q} , we denote the set of resonances by \mathop{\mathrm{Res}}\nolimits(q) . Strictly speaking \mathop{\mathrm{Res}}\nolimits(q) is not a set but a multiset, where the multiplicity of each resonance equals the multiplicity of the corresponding zero. In the free case q\equiv 0 we have \psi(k,x)=e^{ikx} , w(k)=1 and \mathop{\mathrm{Res}}\nolimits(q)=\emptyset . When q is nontrivial, the set \mathop{\mathrm{Res}}\nolimits(q) is infinite; together with the corresponding bound states (zeroes of w with positive imaginary part) resonances uniquely determine the potential, see Theorem 1.1 in [ 10 ] and the papers [ 11 ] , [ 15 ] , [ 14 ] devoted to the stability of the inverse spectral problem. Resonances can also be defined as the poles of the meromorphic continuation of the resolvent of \mathcal{H}_{q} , see Chapter 2 in the book [ 5 ] . The techniques from complex analysis that relate the growth of entire function to the distribution of its zeroes provide valuable information about the location of resonances. For example, the classical Blaschke condition (1.1) \displaystyle\sum_{\lambda\in\mathop{\mathrm{Res}}\nolimits(q)}\frac{|\mathop{\rm Im}\lambda|}{1+|\lambda|^{2}}<\infty holds for every q with compact support. The counting function \displaystyle N(r)=\#\Big\{z\colon w(z)=0,\;|z|0 we have the asymptotic relation (1.2) \displaystyle\#\Big\{z\colon w(z)=0,\;|z|0 , there is only a finite number of resonances in the horizontal strip \{z\colon-h<\mathop{\rm Im}z<0\} . Another point of interest is the characterization problem of resonances sets. The results mentioned above give natural necessary conditions for a set to be a resonances set. Jost functions that correspond to compactly supported potentials are described in [ 10 ] , [ 2 ] , see Theorem 2.1 by E. Korotyaev below, however the explicit sufficient conditions for the resonances sets are not present in the literature. Resonances in every bounded domain are free parameters, see Theorem 1.2 in [ 10 ] . Theorem 3 in [ 2 ] states that there exists a potential with infinitely many resonances in the logarithmic strip of the form ( 1.3 ), in Proposition 7 in the paper [ 17 ] M. Zworski constructed a smooth potential on the real line with infinitely many resonances on the imaginary line. In the present paper we study the following question: (1.4) Let \Lambda be a set of points in \{z\colon\mathop{\rm Im}z<0\} . Does there exist q such that \Lambda\subset\mathop{\mathrm{Res}}\nolimits(q) ? Theorem 1.1 . If \Lambda satisfies the classical Blaschke condition \sum_{\lambda\in\Lambda}\frac{|\mathop{\rm Im}\lambda|}{1+|\lambda|^{2}}<\infty and for some C>0 we have \Lambda\subset\{z:-\mathop{\rm Im}z\geq C|\mathop{\rm Re}z|\} then there exists q\in L^{1}(\mathbb{R}_{+}) with compact support such that \Lambda\subset\mathop{\mathrm{Res}}\nolimits(q) . Theorem 1.2 . Assume that \Lambda is contained in the domain \{z:-\mathop{\rm Im}z\geq C|z|^{\beta}\} for some C>0 and \beta>0 . If the counting function n of \Lambda satisfies the relation n(r)\lesssim r^{\alpha} for some \alpha<\beta then there exists compactly supported q\in L^{1}(\mathbb{R}_{+}) such that \Lambda\subset\mathop{\mathrm{Res}}\nolimits(q) . In the next theorem we show that that the previous theorems are sharp in the following sense: the conclusion of Theorem 1.1 no longer holds if we replace the angle with any wider domain; the relation from Theorem 1.2 between the counting function and the width of the allowed domain is essential. Theorem 1.3 . Let \tau,\rho\colon\mathbb{R}_{+}\to\mathbb{R}_{+} be two positive unbounded increasing continuous functions such that \tau(r)=o(\rho(r)) as r\to\infty . There exists \Lambda\subset\{z:-\mathop{\rm Im}z\geq\tau(|z|)\} with n(r)\leqslant\rho(r) that is not a subset of \mathop{\mathrm{Res}}\nolimits(q) for any q\in L^{1}(\mathbb{R}_{+}) with compact support. Let us finish the introduction with some concluding remarks. • Theorem 1.2 can be formulated for the domains \{z:-\mathop{\rm Im}z\geq\tau(|z|)\} when the function \tau satisfies \tau(r)\gtrsim\log^{5}r . In this case there exists an unbounded \rho such that the assertion n(t)\leqslant\rho(t) implies the existence of the required potential q . • Assertions on the given set \Lambda from Theorem 1.1 imply n(r)=o(r) as r\to\infty , which is consistent with ( 1.2 ); in Theorem 1.2 we also have n(r)\lesssim r^{\alpha}=o(r) as r\to\infty . On the other hand, we have N(r)=O(r) as r\to\infty for the counting function of \mathop{\mathrm{Res}}\nolimits(q) hence in both theorems \Lambda\neq\mathop{\mathrm{Res}}\nolimits(q) and, even more, \Lambda is very small part of \mathop{\mathrm{Res}}\nolimits(q) . 2. Characterization of Jost functions Jost functions corresponding to the Schrödinger operators with compactly supported potentials are entire functions of exponential type. Let \mathcal{PW}_{\sigma} be the classical Paley-Wiener space of entire functions with spectrum in [-\sigma,\sigma] and \mathcal{P}_{\sigma}=e^{iz\sigma}\mathcal{PW}_{\sigma} . We also fix the notation \displaystyle\mathbb{C}_{+}=\{z\colon\mathop{\rm Im}z>0\},\qquad\overline{\mathbb{C}_{+}}=\{z\colon\mathop{\rm Im}z\geqslant 0\},\qquad\mathbb{C}_{-}=\{z\colon\mathop{\rm Im}z<0\}. In the present section we use the characterization of Jost functions from the paper [ 10 ] to reduce question ( 1.4 ) to the construction of the suitable entire function. Theorem 2.1 (E. Korotyaev, Theorem 1.1, [ 10 ] ) . Let f be an entire function without zeroes in the closed half-plane \overline{\mathbb{C}_{+}} such that (2.1) \displaystyle z(f(z)-1)\in\mathop{\mathrm{Const}}\nolimits+\mathcal{P}_{\sigma}. Then f is a Jost function of some Schrödinger operator with q\in L^{1}([0,\sigma]) . Corollary 2.2 . Consider the set \Lambda\subset\mathbb{C}_{-} . If there exists r\in\mathcal{P}_{\sigma} such that r(0)=0 and r(\lambda)=\lambda for all \lambda\in\Lambda then \Lambda\subset\mathop{\mathrm{Res}}\nolimits(q) for some potential q\in L^{1}(\mathbb{R}_{+}) with \mathop{\mathrm{supp}}\nolimits q\subset[0,\sigma] . Proof. Let f be an entire function defined by z(f-1)=-r . If z\neq 0 the equality f(z)=0 is equivalent to r(z)=z hence for every \lambda\in\Lambda we have f(\lambda)=0 . Furthermore, r\in\mathcal{P}_{\sigma} is bounded in \overline{\mathbb{C}_{+}} hence f has only a finite number of zeros in \overline{\mathbb{C}_{+}} , denote them by z_{1},\ldots,z_{N} . Fix N arbitrary points w_{1},\ldots,w_{N}\in\mathbb{C}_{-} and define polynomials \displaystyle P(z)=(z-z_{1})\cdot\ldots\cdot(z-z_{n}),\qquad Q(z)=(z-w_{1})\cdot\ldots\cdot(z-w_{n}). The function f_{1}=f\cdot\frac{Q}{P} is entire, has no zeroes in \overline{\mathbb{C}_{+}} and it satisfies f_{1}(\lambda)=f(\lambda)=0 for every \lambda\in\Lambda . We claim that f_{1} is a Jost functions corresponding to some potential supported on [0,\sigma] . According to Theorem 2.1 , it suffices to verify (2.2) \displaystyle z(f_{1}-1)\in\mathop{\mathrm{Const}}\nolimits+\mathcal{P}_{\sigma}. Both P and Q are monic with \deg P=\deg Q=N hence there exists c\in\mathbb{C} such that (2.3) \displaystyle\deg(z(Q-P)-cP)0 such that \Lambda\subset K_{C} , where \displaystyle K_{C}=\{z:-\mathop{\rm Im}z\geq C|\mathop{\rm Re}z|\} is the angle in the lower half-plane \mathbb{C}_{-} . Fix arbitrary \sigma>0 and denote \gamma=\sigma/3 . There exists H\in\mathcal{PW}_{\gamma} such that H(0)\neq 0 and H(\lambda_{n})=0 for every n\geqslant 0 , see Theorem 2.1.5 in [ 8 ] . Without loss of generality, we may also assume that \big|[-\gamma,-\gamma+\varepsilon]\cap\mathop{\mathrm{supp}}\nolimits\hat{H}\big|>0 for all \varepsilon>0 . The Hayman theorem, see Lecture 15 in [ 12 ] , states that (4.1) \displaystyle\log|H(z)|=-\gamma|\mathop{\rm Im}z|+o(|z|),\qquad|z|\to\infty, holds everywhere in \mathbb{C}_{-} outside of the union of disks of finite view. In other words, there exists a family of circles \displaystyle B_{n}=\{z\colon|z-z_{n}|\leqslant r_{n}\},\qquad n\geqslant 1, such that \sum_{n\geqslant 1}r_{j}/|z_{j}|<\infty and ( 4.1 ) holds in the domain \mathcal{U}=\mathbb{C}_{-}\setminus\cup_{n\geqslant 1}B_{n} . In particular, for some A>C we have \displaystyle\{z\colon-\mathop{\rm Im}z=A|\mathop{\rm Re}z|\}\subset\mathcal{U}. Let K_{A}=\{z:-\mathop{\rm Im}z\geq A|\mathop{\rm Re}z|\} be the corresponding angle. Relation ( 4.1 ) implies the inequality (4.2) \displaystyle\log|H(z)|\geqslant-2\gamma|\mathop{\rm Im}z| when z\in\mathcal{U}\cap K_{A} and |\mathop{\rm Im}z| is large enough (say \mathop{\rm Im}z<-h<0 ). Furthermore, there exists an unbounded sequence h\leqslant h_{0}0 and for every k\geqslant 1 define \varepsilon_{k} as the solution of \psi(1/\varepsilon_{k})=k , where (5.3) \displaystyle\psi(t)=C\int_{a}^{t}\frac{w(u)}{u}du,\qquad t\geqslant a. Here C>0 is a constant chosen such that \cos(e^{-C})=1/e . Theorem 5.1 (Luxemburg, Korevaar, Theorem 5.2 [ 13 ] ) . Fix \sigma>0 . If the parameter a is large enough then the product \displaystyle H(z)=\prod_{n\geq 1}\left(1-\frac{z^{2}}{\lambda_{n}^{2}}\right)\prod_{k\geq 1}\cos\left(\varepsilon_{k}z\right) converges for every z\in\mathbb{C} and defines an entire function H\in\mathcal{PW}_{\sigma} . There exist different ways to construct function from \mathcal{PW}_{\sigma} with prescribed smallness (or even two-sided estimates) on the real line, see, e.g., Chapters X and XI in the book [ 9 ] and the papers [ 3 , 16 ] . The following result states that H(z) cannot be very small when z is separated from the zeroes of H . We prove it in the next section. Theorem 5.2 . If n(t)\lesssim t^{\alpha} for some 0<\alpha<1 and a>0 is large then the estimate \displaystyle\log|H(z)|\gtrsim-|z|^{\alpha}\log|z| holds for all z\in\mathbb{C} that satisfy \displaystyle\mathop{\mathrm{dist}}\nolimits(z,\Lambda)\geq 1,\qquad\mathop{\mathrm{dist}}\nolimits(-z,\Lambda)\geq 1,\qquad|\mathop{\rm Im}z|\geqslant 1. 5.2. Proof of Theorem 5.2 Recall that n is a counting function of \Lambda , \psi is given by ( 5.3 ) and each \varepsilon_{k} is the solution of \psi(1/\varepsilon_{k})=k . The function \theta is defined in ( 5.2 ) and w satisfies w\geqslant\theta+2\log t . Let us introduce two more functions (5.4) \displaystyle\xi(t)=\int_{t}^{\infty}\frac{n(u)}{u^{2}}\,du=\sum_{|\lambda_{k}|\geqslant t}\frac{1}{|\lambda_{k}|}+\frac{n(t)}{t},\qquad t\geqslant 0, \displaystyle\gamma(t)=\int_{t}^{\infty}\frac{w(u)}{u^{2}}du,\qquad t\geqslant 0. Lemma 5.3 . For every nonzero z\in\mathbb{C} we have \displaystyle|\cos(z)|\geq\frac{|\mathop{\rm Im}z|}{2|z|}. Proof. Let z=re^{i\theta}=r(\cos\theta+i\sin\theta) . The required inequality is equivalent to |\cos(z)|\geqslant|\sin\theta|/2 . We have iz=r(-\sin\theta+i\cos\theta) and \displaystyle e^{iz}=e^{r(-\sin\theta+i\cos\theta)}=e^{-r\sin\theta}\left(\cos(r(\cos\theta))+i\sin(r(\cos\theta))\right), \displaystyle e^{-iz}=e^{r(\sin\theta-i\cos\theta)}=e^{r\sin\theta}\left(\cos(r(\cos\theta))-i\sin(r(\cos\theta))\right), \displaystyle 2\cos z=\cos(r\cos\theta)\left(e^{-r\sin\theta}+e^{r\sin\theta}\right)+i\sin(r\cos\theta)\left(e^{-r\sin\theta}-e^{r\sin\theta}\right). If |\cos(r\cos\theta)|\geqslant\sqrt{2}/2 then we write \displaystyle|\cos z|\geqslant|\mathop{\rm Re}\cos z|=\bigg|\frac{\cos(r\cos\theta)\left(e^{-r\sin\theta}+e^{r\sin\theta}\right)}{2}\bigg|\geqslant|\cos(r\cos\theta)|\geqslant\frac{\sqrt{2}}{2}\geqslant\frac{|\sin\theta|}{2}. Otherwise the inequality |\sin(r\cos\theta)|\geqslant\sqrt{2}/2 holds. This implies r\geqslant r|\cos\theta|\geqslant\pi/4 and \displaystyle|\cos z|\geqslant|\mathop{\rm Im}\cos z| \displaystyle=\bigg|\frac{\sin(r\cos\theta)\left(e^{-r\sin\theta}-e^{r\sin\theta}\right)}{2}\bigg| \displaystyle\geqslant|\sin(r\cos\theta)|\cdot|r\sin\theta|\geqslant\frac{\sqrt{2}}{2}\cdot\frac{\pi|\sin\theta|}{4}\geqslant\frac{|\sin\theta|}{2}. The proof is concluded. ∎ Lemma 5.4 . For every z\in\mathbb{C} with |\mathop{\rm Im}z|\geqslant 1 , we have \displaystyle\log\bigg|\prod_{k\geq 1}\cos\left(\varepsilon_{k}z\right)\bigg|\gtrsim-|z|\cdot\gamma(2|z|)-\psi(2|z|)\cdot\log|z|. Proof. Consider the following disjoint sets \displaystyle\mathcal{A}=\left\{k\colon\varepsilon_{k}\leq\frac{1}{2|z|}\right\},\qquad\mathcal{B}=\left\{k\colon\frac{1}{2|z|}<\varepsilon_{k}\right\}. The set \mathcal{B} is finite, we have \#\mathcal{B}=\psi(2|z|) . The previous lemma gives \displaystyle\log\left|\prod_{k\in\mathcal{B}}\cos(\varepsilon_{k}z)\right|\geq\#\mathcal{B}\cdot\log\left(\frac{\operatorname{Im}z}{2|z|}\right)=-\log\frac{2|z|}{\mathop{\rm Im}z}\cdot\psi(2|z|)\gtrsim-\log|z|\cdot\psi(2|z|). For every k\in\mathcal{A} we have |\varepsilon_{k}z|<1/2 hence \log|\cos(\varepsilon_{k}z)|\gtrsim-|\varepsilon_{k}z| . This implies \displaystyle\log\left|\prod_{k\in\mathcal{A}}\cos(\varepsilon_{k}z)\right|\gtrsim-|z|\sum_{k\in\mathcal{A}}\varepsilon_{k}. To conclude the proof of the lemma we notice \displaystyle\sum_{k\in\mathcal{A}}\varepsilon_{k}=\sum_{k\geqslant\psi(2|z|)}\frac{1}{\psi^{-1}(k)}<\int_{\psi(2|z|)}^{\infty}\frac{dx}{\psi^{-1}(x)}=\int_{2|z|}^{\infty}\frac{d\psi(x)}{x}\asymp\int_{2|z|}^{\infty}\frac{w(u)}{u^{2}}du\lesssim\gamma(2|z|). ∎ Lemma 5.5 . If z satisfies \mathop{\mathrm{dist}}\nolimits(z,\Lambda)\geq 1 and \mathop{\mathrm{dist}}\nolimits(-z,\Lambda)\geq 1 then \displaystyle\log\bigg|\prod_{k\geq 1}\left(1-\frac{z^{2}}{\lambda_{k}^{2}}\right)\bigg|\gtrsim-|z|\cdot\xi(2|z|)-\log|z|\cdot n(2|z|). Proof. As in the previous lemma we partition positive integers into the two following sets: \displaystyle\mathcal{C}=\left\{k\colon|\lambda_{k}|\leq 2|z|\right\},\qquad\mathcal{D}=\left\{k\colon 2|z|<|\lambda_{k}|\right\}. The uniform estimate \log|1-w|\gtrsim-|w| for w with |w|<1/2 gives \displaystyle\log|1-z^{2}/\lambda_{k}^{2}|\gtrsim-|z/\lambda_{k}|^{2}\geqslant-|z/\lambda_{k}|,\qquad k\in\mathcal{D}. The latter implies \displaystyle\log\left|\prod_{k\in\mathcal{D}}\left(1-\frac{z^{2}}{\lambda_{k}^{2}}\right)\right|\gtrsim-\sum_{k\in\mathcal{D}}\frac{|z|}{|\lambda_{k}|}\geqslant-|z|\xi(2|\lambda|). On the other hand, for every k\in\mathcal{C} we can use the estimate \displaystyle\left|1-\frac{z^{2}}{\lambda_{k}^{2}}\right|=\frac{|(z-\lambda_{k})(z+\lambda_{k})|}{|\lambda_{k}|^{2}}\geqslant\frac{1}{|\lambda_{k}|^{2}}\geqslant\frac{1}{4|z|^{2}}. The set \mathcal{C} is finite, we have \#\mathcal{C}\leqslant n(2|z|) therefore \displaystyle\log\left|\prod_{k\in\mathcal{C}}\left(1-\frac{z^{2}}{\lambda_{k}^{2}}\right)\right|\gtrsim-\log(4|z|^{2})\cdot\#\mathcal{C}\gtrsim-\log|z|\cdot n(2|z|). To finish the proof we combine the obtained estimates for products over \mathcal{C} and \mathcal{D} . ∎ Proof of Theorem 5.2 . From the definitions of \theta and \psi we know \displaystyle\theta(t)=\int_{0}^{\infty}\frac{n(u)}{u}\frac{t^{2}}{t^{2}+u^{2}}\,du\lesssim\int_{0}^{\infty}u^{\alpha-1}\frac{t^{2}}{t^{2}+u^{2}}\,du=t^{\alpha}\int_{0}^{\infty}\frac{x^{\alpha-1}}{1+x^{2}}\,dx\lesssim t^{\alpha}, \displaystyle\xi(t)=\int_{t}^{\infty}\frac{n(u)}{u^{2}}\,du\lesssim\int_{t}^{\infty}u^{\alpha-2}\,du\lesssim t^{\alpha-1}. We can also choose w(t)\lesssim t^{\alpha} so that \psi and \gamma satisfy \displaystyle\psi(t)\lesssim\int_{1}^{t}\frac{w(u)}{u}dt\lesssim\int_{1}^{t}u^{\alpha-1}dt\lesssim t^{\alpha}, \displaystyle\gamma(t)=\int_{t}^{\infty}\frac{w(u)}{u^{2}}du\lesssim\int_{t}^{\infty}u^{\alpha-2}\,du\lesssim t^{\alpha-1}. The required inequality for H now follows from Lemmas 5.4 and 5.5 : \displaystyle\log|H(z)| \displaystyle=\log\bigg|\prod_{k\geq 1}\cos\left(\varepsilon_{k}z\right)\bigg|+\log\bigg|\prod_{n\geq 1}\left(1-\frac{z^{2}}{\lambda_{n}^{2}}\right)\bigg| \displaystyle\gtrsim-|z|\cdot\gamma(2|z|)-\psi(2|z|)\cdot\log|z|-|z|\cdot\xi(2|z|)-\log|z|\cdot n(2|z|)\gtrsim-|z|^{\alpha}\log|z|. ∎ 5.3. Points Clustering In this section we describe the construction of contours that will be used in the proof of Theorem 1.2 , recall Figure 1 . Fix a number r>0 and let (5.5) \displaystyle V_{r}=\big\{z\colon\mathop{\mathrm{dist}}\nolimits(z,\Lambda)\leqslant r\big\}=\bigcup_{n\geq 1}\big\{z:|\lambda_{n}-z|\leqslant r\big\} be the neighborhood of \Lambda of the radius r . Denote the connected components of V_{r} by U_{r,1},U_{r,2},\ldots and let \Gamma_{r,n} be the contours that run along the boundary of U_{r,n} , see Figure 5 . We remark that \Gamma_{r,n} may be disconnected. Let us show that each U_{r,n} is bounded and the length of \Gamma_{n} is not very large, provided that the set \Lambda satisfies assertion ( 5.1 ). x y \Gamma_{2} U_{2} \Gamma_{1} U_{1} \Gamma_{0} U_{0} Figure 5. Clusters visualization By the construction we have \Lambda\subset V_{r}=\cup_{n\geqslant 1}U_{r,n} . For every n\geqslant 1 denote \Lambda_{r,n}=\Lambda\cap U_{r,n} and let \{\lambda_{r,n,0},\lambda_{r,n,1},\ldots\} be the elements of \Lambda_{r,n} enumerated such that |\lambda_{r,n,0}|\leqslant|\lambda_{r,n,1}|\leqslant\ldots holds. Assume that the size of \Lambda_{r,n} is at least N . Then for every k0 for all w\in\Gamma_{n} . Substituting the two latter inequalities into ( 5.9 ), we get \displaystyle|F_{n}(z)|\lesssim\frac{\mathop{\mathrm{len}}\nolimits(\Gamma_{n})}{\mathop{\mathrm{dist}}\nolimits(z,\Gamma_{n})}\sup_{w\in\Gamma_{n}}e^{-C_{2}|w|^{\beta}}. From estimates ( 5.6 ) and ( 5.7 ) we know the relations \mathop{\mathrm{len}}\nolimits(\Gamma_{n})\lesssim|\lambda_{n,0}| and |w|\asymp|\lambda_{n,0}| for all w\in\Gamma_{n} . It follows that the series \displaystyle g(z)=\sum_{n\geqslant 1}H(z)F_{n}(z),\qquad z\in\mathbb{C}, converges for all z\in\mathbb{C} and defines an entire function g that satisfies |g(z)|\lesssim|H(z)| when \mathop{\mathrm{dist}}\nolimits(z,\Gamma_{n})>1 for all n\geqslant 0 . Notice that \displaystyle\bigcup_{n\geqslant 0}\big\{z\colon\mathop{\mathrm{dist}}\nolimits(z,\Gamma_{n})\leqslant 1\big\}=\big\{z\colon\mathop{\mathrm{dist}}\nolimits(z,\Lambda)\leqslant 2\big\}=V_{2}, where V_{2} is defined in ( 5.5 ). In the previous section we have shown that V_{2} has a sparse structure, recall Figure 5 . Maximum modulus principle and the Paley-Wiener theorem give g\in\mathcal{PW}_{\sigma} . To finish the proof of the theorem we need to fulfill the assertion g(0)=0 and drop ( 5.8 ). At this point we proceed as in the proof of Theorem 1.1 : construct \tilde{g} for the set \displaystyle\tilde{\Lambda}=\{-2i\}\cup\{\lambda_{n}-2i,\,n\geqslant 1\}. as described above and take g=\tilde{g}(\cdot+2i) . 6. Sharpness of the results. Proof of Theorem 1.3 Resonances of the Schrödinger operator can be defined as the zeroes of the corresponding de Branges function E=A+iB , see [ 2 ] . The following lemma is stated in Lemma 3 in the preprint [ 1 ] . We reproduce its proof here for the reader’s convenience. Lemma 6.1 . For every compactly supported potential q\in L^{1}(\mathbb{R}_{+}) we have \displaystyle\sup_{t\in\mathbb{R}}\sum_{s\in\mathop{\mathrm{Res}}\nolimits(q)}\frac{|\mathop{\rm Im}s|}{|t-s|^{2}}<\infty. Proof. The phase function of the Schrödinger operator is defined by the relation E/\overline{E}=e^{-2i\varphi} , see [ 2 ] , it satisfies \displaystyle\varphi^{\prime}(t)=\sum_{s\in\mathop{\mathrm{Res}}\nolimits(q)}\frac{|\mathop{\rm Im}s|}{|t-s|^{2}}+a, where a is a real-valued constant. The calculation shows \displaystyle-2i\varphi^{\prime} \displaystyle=\frac{E^{\prime}}{E}+\frac{\overline{E^{\prime}}}{\overline{E}}=\frac{E^{\prime}\overline{E}-E\overline{E}^{\prime}}{|E|^{2}} (6.1) \displaystyle=\frac{AB^{\prime}-A^{\prime}B}{A^{2}+B^{2}}=\frac{AB^{\prime}-A^{\prime}B}{B^{2}}\cdot\frac{1}{(A/B)^{2}+1}=\left(\frac{A}{B}\right)^{\prime}\frac{1}{(A/B)^{2}+1}. Let \{\lambda_{n}\}_{n\in\mathbb{Z}} and \{\mu_{n}\}_{n\in\mathbb{Z}} be the zeroes of A and B respectively. Lemma 1 in [ 2 ] states that for q\in L^{2}(\mathbb{R}_{+}) the functions A and B satisfy (6.2) \displaystyle\left|\frac{A(z)}{\sin z}\right|\asymp\frac{\mathop{\mathrm{dist}}\nolimits(z,\{\lambda_{n}\})}{\mathop{\mathrm{dist}}\nolimits(z,\pi\mathbb{Z})},\qquad\left|\frac{B(z)}{\cos z}\right|\asymp\frac{\mathop{\mathrm{dist}}\nolimits(z,\{\mu_{n}\})}{\mathop{\mathrm{dist}}\nolimits(z,\pi\mathbb{Z}+\pi/2)}. We claim that these relations hold for all q\in L^{1}(\mathbb{R}_{+}) . In [ 2 ] the L^{2} assertion is used in the asymptotic formula (3.3) for \lambda_{n} and \mu_{n} , \displaystyle\lambda_{n}=\pi n+\frac{C}{n}+\frac{a_{n}}{n},\qquad\mu_{n}=\pi_{n}-\frac{\pi}{2}+\frac{C}{n}+\frac{b_{n}}{n},\qquad\{a_{n}\},\{b_{n}\}\in\ell^{2}(\mathbb{Z}_{+})\qquad n\geqslant 1. For q\in L^{1}(\mathbb{R}_{+}) the latter holds with \{a_{n}\},\{b_{n}\}\in\ell^{\infty}(\mathbb{Z}_{+}) , see, e.g., [ 4 ] , which is sufficient for the rest of the proof. Substitute z=\mu_{k} into ( 6.2 ). We have \displaystyle|A(\mu_{k})|\asymp|\sin(\mu_{k})|\frac{\mathop{\mathrm{dist}}\nolimits(\mu_{k},\{\lambda_{n}\})}{\mathop{\mathrm{dist}}\nolimits(\mu_{k},\pi\mathbb{Z})}\asymp 1,\qquad|B^{\prime}(\mu_{k})|\asymp 1. For every n\in\mathbb{Z} let v_{n}=A(\mu_{n})/B^{\prime}(\mu_{n}) . This sequence is bounded and we can write \displaystyle\frac{A}{B}=\sum_{n\in\mathbb{Z}}\frac{A(\mu_{n})}{B^{\prime}(\mu_{n})}\left(\frac{1}{z-\mu_{n}}+\frac{1}{\mu_{n}}\right)=\sum_{n\in\mathbb{Z}}v_{n}\left(\frac{1}{z-\mu_{n}}+\frac{1}{\mu_{n}}\right), \displaystyle\left(\frac{A}{B}\right)^{\prime}=\sum_{n\in\mathbb{Z}}v_{n}\frac{1}{(z-\mu_{n})^{2}}. Inequality ( 6.1 ) implies \displaystyle|\varphi^{\prime}(x)|\lesssim\sum_{n\in\mathbb{Z}}\frac{1}{(x-\mu_{n})^{2}}\cdot\Bigg|1+\bigg(\sum_{n\in\mathbb{Z}}v_{n}\left(\frac{1}{z-\mu_{n}}+\frac{1}{\mu_{n}}\right)\bigg)^{2}\Bigg|^{-1}. The second multiplier is always less or equal to 1 hence we get \displaystyle|\varphi^{\prime}(x)|\leqslant\sum_{n\in\mathbb{Z}}\frac{1}{(x-\mu_{n})^{2}}\lesssim 1. Fix small \varepsilon>0 . If \mathop{\mathrm{dist}}\nolimits(x,\{\mu_{n}\})\geqslant\varepsilon this gives |\varphi^{\prime}(x)|\lesssim 1 . Otherwise there exists \mu_{k} such that |x-\mu_{k}|=\mathop{\mathrm{dist}}\nolimits(x,\{\mu_{n}\})\leqslant\varepsilon . In this case we have \displaystyle\sum_{n\in\mathbb{Z}}\frac{1}{(x-\mu_{n})^{2}}\asymp\frac{1}{(x-\mu_{k})^{2}},\qquad\left|\sum_{n\in\mathbb{Z}}v_{n}\left(\frac{1}{z-\mu_{n}}+\frac{1}{\mu_{n}}\right)\right|\asymp\frac{1}{|z-\mu_{k}|}, which again implies |\varphi^{\prime}(x)|\lesssim 1 . The proof is concluded. ∎ Proof of Theorem 1.3 . First of all we notice that it is sufficient to prove theorem only when \rho satisfies \rho(r)=o(r) as r\to\infty . Indeed, for every compactly supported q\in L^{1}(\mathbb{R}_{+}) , the counting function N of \mathop{\mathrm{Res}}\nolimits(q) satisfies N(r)=O(r) as r\to\infty hence we may assume \rho(r)=O(r) and \tau(r)=o(\rho(r))=o(r) as r\to\infty . Next, if \rho(r)=o(r) does not hold we can prove the theorem with \tilde{\rho}(r)=\sqrt{\rho(r)\tau(r)} instead of \rho . Consider another function \kappa(r)=\frac{1}{2}(\rho(r)+\tau(r)) and define r_{n} as the solution of \kappa(r_{n})=n for all n\geqslant\kappa(0) . The assertion \kappa(r)=o(r) as r\to\infty implies (6.3) \displaystyle\lim_{n\to\infty}\frac{n}{r_{n}}=\lim_{n\to\infty}\frac{\kappa(r_{n})}{r_{n}}=0. We also have \tau(r)=o(r) as r\to\infty hence for all large n there exists a unique z_{n}\in\mathbb{C} that satisfies \displaystyle|z_{n}|=r_{n},\qquad\mathop{\rm Im}z_{n}=-\tau(r_{n}),\qquad\mathop{\rm Re}z_{n}>0. We construct an increasing sequence \{n_{0},n_{1},\ldots\} inductively. Choose n_{0} for which z_{n_{0}} is well-defined. Next, for every k\geq 1 select n_{k}>n_{k-1} such that 2^{k}n_{k}\leq r_{n_{k}} and \displaystyle n_{0}+\cdots+n_{k-1}\leqslant\rho(r_{n_{k}})-\kappa(r_{n_{k}}). Such n_{k} exists for every k\geqslant 1 because of ( 6.3 ) and \displaystyle\lim_{n\to\infty}\left(\rho(r_{n})-\kappa(r_{n})\right)=\frac{1}{2}\lim_{n\to\infty}\left(\rho(r_{n})-\tau(r_{n})\right)=+\infty. We form the multiset \Lambda=\{z_{n_{k}}\colon k\geq 0\} , where z_{n_{k}} appears with multiplicity n_{k} . Equivalently, we can replace each instance of z_{n_{k}} with n_{k} distinct points arbitrarily close to z_{n_{k}} . We have \displaystyle\sum_{\lambda\in\Lambda}\frac{1}{|\lambda|}=\sum_{k=0}^{\infty}\frac{n_{k}}{|z_{n_{k}}|}=\sum_{k=0}^{\infty}\frac{n_{k}}{r_{n_{k}}}\leq\sum_{k=0}^{\infty}\frac{1}{2^{k}}<\infty. Furthermore, we can estimate the counting function as follows: \displaystyle n(r_{n_{k}})=n_{0}+\cdots+n_{k}\leqslant\rho(r_{n_{k}})-\kappa(r_{n_{k}})+n_{k}=\rho(r_{n_{k}}). The inequality n(r)\leq\rho(r) for all r\geq 0 follows. Assume that \Lambda\subset\mathop{\mathrm{Res}}\nolimits(q) for some compactly supported q\in L^{1}(\mathbb{R}_{+}) . By Lemma 6.1 the function \displaystyle p(t)=\sum_{s\in\Lambda}\frac{|\mathop{\rm Im}s|}{|t-s|^{2}} is uniformly bounded for t\in\mathbb{R} . On the other hand, we have \displaystyle p(\mathop{\rm Re}z_{n_{k}})=\sum_{s\in\Lambda}\frac{|\mathop{\rm Im}s|}{|\mathop{\rm Re}z_{n_{k}}-s|^{2}}\geq\frac{n_{k}|\mathop{\rm Im}z_{n_{k}}|}{|\mathop{\rm Re}z_{n_{k}}-z_{n_{k}}|^{2}}=\frac{n_{k}}{|\mathop{\rm Im}z_{n_{k}}|}=\frac{\kappa(r_{n_{k}})}{\tau(r_{n_{k}})}\to\infty,\qquad n\to\infty. This contradiction shows that \Lambda is not a subset of \mathop{\mathrm{Res}}\nolimits(q) . ∎