MEAN STAIRCASE OF THE RIEMANN ZEROS: A COMMENT ON THE LAMBERT W FUNCTION AND AN ALGEBRAIC ASPECT 9 DAVIDEAMARCA,STEFANOBELTRAMINELLI,ANDDANILOMERLINI 0 0 2 Abstract. In this note we discuss explicitly the structure of two simple set ofzeroswhichareassociatedwiththemeanstaircaseemergingfromthezeta n functionandwespecifyasolutionusingtheLambertWfunction. Theargu- a mentofitmaythenbesetequaltoaspecialN×N classicalmatrix(forevery J N) related to the Hamiltonian of the Mehta-Dyson model. In this way we 1 specifyafunctionofanhermiteanoperatorwhoseeigenvaluesarethe“trivial 2 zeros”onthecriticalline. Thefirstsetoftrivialzerosisdefinedbytherela- ] stieotn.s(ITmo`dζi`st21in+guii·sht´´fr=om0∧thReeu`sζu`al12t+rivii·atl´´ze(cid:54)=ro0s asn=dvρic+evie·rsta=for−t2hne,snec(cid:62)ond1 T integer) N . h t This (heuristic, non rigorous) research note is dedicated to the international a m Swiss-Italian mathematician and physicist Professor Dr. Sergio Albeverio on the occasion of his seventieth birthday; a friend and for years the scientific director [ of Cerfim (Research Center for Mathematics and Physics of Locarno), situated 1 opposite the ”Rivellino”1. v 7 7 3 1. Introduction: a search for an hermitean operator associated 3 with the Riemann Zeta Function . 1 There is much interest in understanding the complexity related to the Riemann 0 9 Hypothesis and concerned with the location and the structure of the non trivial 0 zeros of the Riemann zeta function ζ(s) where s=ρ+i·t is the complex variable. : FollowingasuggestionofHilbertandPolya,inrecentyearsmanyeffortshavebeen v i devoted to a possible construction of an hermitean operator having as eigenvalues X theimaginarypartst ofthenontrivialzerosofζ (ζ beingmeromorphic,thezeros n r are countable). These are given by the solutions of the equation ζ(ρ +i·t )=0, a n n n = 1,2,...,. If ρ = 1 for all n, then all the zeros lie on the critical line (the n 2 Riemann Hypothesis is true); the program is then to find an hermitean “operator” T such that T · ϕ = t · ϕ in some appropriate space (ϕ would be the nth n n n n eigenvector of T). There are today many strategies in the direction of constructing such anoperatorandin thesequel wewill commenton some(among manyothers) Date:17.1.2009. 1991 Mathematics Subject Classification. 11M26. Keywordsandphrases. Riemannzetafunction,LambertWfunction,Riemannzeros,harmonic oscillator,RiemannHypothesis. 1The Bastion ”Il Rivellino”, situated opposite Cerfim, is 99.9% attribuable to Leonardo da Vinci(1507). 1 2 DAVIDEAMARCA,STEFANOBELTRAMINELLI,ANDDANILOMERLINI verystimulating worksonthesubject. In[1], Pitk¨anen’s heuristic workgoesinthe direction of constructing orthogonality relations between eigenfunctions of a non hermitean operator related to the superconformal symmetries; a different operator than the one just mentioned has also been proposed in [2] by Castro, Granik and Mahecha in terms of the Jacobi Teta series and an orthogonal relation among its eigenfunctions has also been found. In the rigorous work by Elizalde et.al [3] some problemswiththoseapprocheshavebeenpointedout. Inaworkofsomeyearsago Julia [4] proposed a fermionic version of the zeta function which should be related tothepartitionfunctionofasystemofp-adicoscillatorsinthermalequilibrium. In two others pioneering works of these years, Berry and Keating [5, 6] proposed an interesting heuristic operator to study the energy levels t (the imaginary parts of n the non trivial zeros of the zeta function). The proposed Hamiltonian has a very simple form given, on a dense domain, by: H =p·x+ 1, where 2 (cid:18) (cid:19) 1 ∂ (1) p= i ∂x in one dimension. As explained by the authors, the difficulty is then to define appropriatespacesandboundaryconditionstoproperlydeterminepandH asher- mitean operators. In such an approach the heuristic appearance of “instantons” is also discussed. In another important work Bump et al. [7] introduced a local Rie- mannHypothesisandprovedinparticularthattheMellintransformoftheHermite polynomials (associated with the usual quantum mechanical harmonic oscillator) contain as a factor a polynomial p (s), corresponding to the n-energy eigenstate n of the oscillator, whose zeros are exactly located on the critical line σ = 1. The 2 relation of the polynomials p (s) with some truncated approximation of the en- n tire funcion ξ(s) (the Xi function), related to the Riemann zeta function seems to be still lacking. Others important mathematical results concerning the non trivial Riemann zeros, have been obtained by many leading specialist (see among others the work by Connes [8] and the work by Albeverio and Cebulla [9]). For a recent work on the xp hamiltonian see G. Sierra [10]. Let us also mention that for the nontrivial zeros of zeta an interesting equation hasbeenproposedoriginallybyBerryandKeatingin[5]. Infact,rememberingthe definition of ξ(s) where ζ(s) is the Riemann zeta function, given by: (2) ξ(s)= 1 ·s·(s−1)·π−2s ·Γ(cid:16)s(cid:17)·ζ(s) 2 2 the equation for possible zeros of ξ proposed in [5] is given by: π2s π1−2s (3) + =0 Γ(cid:0)s(cid:1) Γ(cid:0)1−s(cid:1) 2 2 Asstatedbytheauthors,Eq. (3)couldbeconsideredasa“quantizationcondition”. Unfortunately, as mentioned in [5], Eq. (3) possesses complex zeros and so can not be used to provide an hermitean operator which would generate the non trivial zeros of ζ. The content of our note is concerned with the “mean staircase” of the Riemannzeros: wefirstspecifythetwosetoftrivialzerosonthecriticallinerelated toitandpointoutanexplicitconstructionusingtheLambertWfunction; thenwe introduce a specific argument (a n×n hermitean matrix H, describing a discrete harmonic oscillator with creation and annihilation “operators” a and a∗ such that MEAN STAIRCASE OF THE RIEMANN ZEROS 3 [a,a∗] = −2) into the Lambert W function. We obtain then, for the trivial zeros, the goal that the “Polya-Hilbert program” has for the non trivial zeros. 2. The mean staircaise of the Riemann zeros and the trivial zeros on the critical line associated with it Letξ(s)bethexifunctiongivenby(2). IfN(t)denotesthenumberofzerosofξ in the critical strip of height smaller or equal to t, and if S(t)≡ 1 arg(cid:0)ζ(cid:0)1 +it(cid:1)(cid:1), π 2 then [11] (cid:18) (cid:19) 1 (4) N(t)=(cid:104)N(t)(cid:105)+S(t)+O , t where (cid:18) (cid:18) (cid:19) (cid:19) t t 7 (5) (cid:104)N(t)(cid:105)= · ln −1 + 2π 2π 8 (cid:104)N(t)(cid:105), the “bulk contribution” to N, is called the “mean staircase of the zeros” (cfr. [11]). The fluctuations of the number of zeros around the mean staircase, are given by the function S(t). It is known [11] that S(t) = O(lnt) without assuming (cid:16) (cid:17) RH while, assuming RH is true, it is known that S(t)=O lnt . At this point, ln(lnt) since our remark has mainly to do with (cid:104)N(t)(cid:105), we will set S(t) = 0 in Eq.(4). We shall study the relation N(t) = (cid:104)N(t)(cid:105)+O(cid:0)1(cid:1). The two sets (which we call t here “trivial zeros on the critical line”) of interest are defined by the above mean staircase as follows. The first set is given by the zeros of Imζ(cid:0)1 +it(cid:1) alone i.e. 2 such that Reζ(cid:0)1 +it(cid:1) (cid:54)= 0 (the first set of trivial zeros on the critical line). The 2 second set is given by the zeros of Reζ(cid:0)1 +it(cid:1), such that Imζ(cid:0)1 +it(cid:1) (cid:54)= 0. For 2 2 the first set: (cid:18) (cid:19) (cid:18) (cid:19) 1 1 (6) Imζ +it∗ =0∧Reζ +it∗ (cid:54)=0 2 n 2 n Then since Imζ(cid:0)1 +it∗(cid:1) = 0 we have that π−1argζ(cid:0)1 +it∗(cid:1) = π−1(−πn) and 2 n 2 n π−1argξ =0 is given by those t∗ such that n t∗ (cid:18) (cid:18)t∗ (cid:19) (cid:19) 7 (7) N(t∗)∼=(cid:104)N(t∗)(cid:105)= n · ln n −1 + =n,ninteger(cid:62)1 n n 2π 2π 8 at large values of t or n. So the nonlinear equations to be solved which should give the values where only Imζ(cid:0)1 +it(cid:1) vanishes, i.e {t∗}, is given by: 2 n (8) (cid:104)N(t∗)(cid:105)=n,ninteger(cid:62)1 n while for the second set 1 (9) (cid:104)N(t∗∗)(cid:105)=n− ,ninteger(cid:62)1 n 2 (The first set has been known for a long time and constitutes the Gram points, sin(θ)=0, where θ is the phase of the ζ function, while for the second set one has cos(θ)=0). We note, the values of interest are given by the abscissa of the intersection points between the staircase (Eq.(5)) and the two functions π−1argξ(cid:0)1 +it(cid:1) and 2 π−1argξ(cid:0)1 +it(cid:1)−1 . TheplotofFig1illustratesthesituationforsomelowlying 2 2 zeros. Thevaluesfort∗ liemostlyinbetweentheexactvalueoftheRiemannzeros n t and t , but it is known that the Gram law fails for the first time at t=282.4 n−1 n 4 DAVIDEAMARCA,STEFANOBELTRAMINELLI,ANDDANILOMERLINI (“first istanton” according to [5]). The solution of the above equation which gives t∗, t∗∗ using a very special function (the LambertW function, see [12]) is given n n below. 5 4 3 2 1 10 20 30 40 -1 Figure1. Theplotof(cid:104)N(t)(cid:105)(continuoscurve),ofN(t)(fullstair) and N(t)− 1 (intermittent stair) 2 3. An exact solution for the sequence t∗ and t∗∗ n n The equation corresponding to (8), may be written in the form (cid:18) (cid:19) t (10) t 2πe =en−e78 2πe and the equation corresponding to (9) in the form (cid:18) (cid:19) t (11) t 2πe =en−12e−78 2πe (cid:16)n−7(cid:17) (cid:16)n−1−7(cid:17) so that introducing the new variables x = exp 8 resp. x = exp 2 8 we e e obtain the equation (from (10) and (11), x>0) (12) W(x)·exp(W(x))=x The function W(x) is called the Lambert W function and has been studied exten- sively in these recent years. In fact such an equation appears in many fields of science. In particular the use of such a function has appeared in the study of the wave equation in the double-well Dirac delta function model or in the solution of a jet fuel problem. See [12] for an important work on the subject. Moreover the Lambert W function appears also in combinatorics as the generating function of trees and as explained in [12] the W function has many applications, even if the presence of the W function often goes unrecognized. The Lambert W function has many complex branches; of interest here is the principal branch of W which is analytic at x=0. So, the solution of (8, 9) is given MEAN STAIRCASE OF THE RIEMANN ZEROS 5 by (cid:18) (cid:18)n− 7(cid:19)(cid:19) (13) t∗ =2πe·exp W 8 n e (cid:18) (cid:18)n− 1 − 7(cid:19)(cid:19) (14) t∗∗ =2πe·exp W 2 8 n e Wehavethusspecified,withthehelpoftheLambertWfunction,thesequences{t∗} n resp. {t∗∗},whicharethezerosof Im(cid:0)ζ(cid:0)1 +i·t(cid:1)(cid:1)suchthatRe(cid:0)ζ(cid:0)1 +i·t(cid:1)(cid:1)(cid:54)=0 n 2 2 resp. Re(cid:0)ζ(cid:0)1 +i·t(cid:1)(cid:1)=0 and Im(cid:0)ζ(cid:0)1 +i·t(cid:1)(cid:1)(cid:54)=0. 2 2 It should be noted here that in Eq.(8), n, which would correspond to the exact value of a true zero value t (non trivial zero) of the ζ function would not be an n integer n or n− 1 since we have replaced in Eq. (5) S(t) by zero. For the first few 2 low zeros (the true zeros), it may be observed numerically that the corresponding values,letsayn∗,arerandomlydistribuedmostlybetweentwoconsecutiveintegers, but the mean values are nearby the integers plus 1. A calculation with some zeros 2 gives a mean value of 0.49 instead of 0.5. So, in average it seems that the behavior of the true zeroes t “follows” more the pattern of the set t∗∗. In the similar way n n thezerosofthefirstset, i.e. t∗, liemostlyinbetweentwonontrivialzerosofζ but n of course it is known that there are very complicated phenomena associated with the chaotic behavior of the non trivial zeros of the Riemann ζ function. Asan example, thefirstof theistantons, correspondingto n=126, citedabove, is located at the value of t=282.4.... On the table below we give the values of t∗ n and of t of a true zero around t=280. n t =279.22925 126 t∗ =280.80246 126 t∗ =282.4547596 127 t =282.4651147 127 t =283.211185 128 t∗ =284.1045158 128 t =284.8359639 129 Table 1. From those numerical computations we see that two consecutive zeros of Im(ζ) alone are followed by two consecutive true zeros, that is t∗ anticipates t . 127 127 The difference between the two subsequent t values is very small and given by ∆t = 0.0103. The phase change is given by iπ as illustrated on the plot of Im(cid:0)ln(cid:0)ζ(cid:0)1 +it(cid:1)(cid:1)(cid:1) (step curve) and that of Im(cid:0)ζ(cid:0)1 +it(cid:1)(cid:1). 2 2 For the first 500 energy levels, that is for values of t from 0 to t = 811.184... (levelnumbern=500), itmaybeseenthatthereare13istantons(inthelanguage of [5]), all with a Maslov phase change of +iπ or of −iπ. The width is usally small but it is larger for the istanton located at t = 650.66 (n corresponding to 379), where this time ∆t=0.31.... Returning now to the trivial zeros (the two sets t∗, n t∗∗ defined above), we note the elementar relation which follow from (8) and (9), n and given by: t∗ +t∗ (15) n n+1 =t∗∗ 2 n+12 6 DAVIDEAMARCA,STEFANOBELTRAMINELLI,ANDDANILOMERLINI Figure 2. The first istanton and t∗∗ +t∗∗ (16) n−21 n+12 =t∗ 2 n Eq.(15, 16) say that the zeros of the real part alone are obtained by those of the imaginary part alone by simple average and viceversa. The two sequences are regularly spaced and the mean distance between two trivial zeros at the height t, as the mean staircase indicates (Eq.(5)), is given approximatively by: t 2π 2π (17) = = (cid:104)N(t)(cid:105) log(cid:0) t (cid:1) log(n) 2π for t and n large Before proposing an hermitean operator for the sequences of the trivial zeros it is important to investigate a possible “quantization condition” for the non trivial zeros. For this we start with the Riemann symmetry of the ζ function. From the exact relation for the ξ-function given by: 1 (cid:16)s(cid:17) (18) ξ(s)= π−2sΓ ζ(s)s(s−1) 2 2 (cid:18) (cid:19) 1 1−s =ξ(1−s)= π−1−2sΓ ζ(1−s)(1−s)(1−s−1) 2 2 s∈C, we have that (cid:16)s(cid:17) (cid:18)1−s(cid:19) (19) π−2sΓ ζ(s)=π−1−2sΓ ζ(1−s) 2 2 In equation (19) we limit ourselves to consider the values s = ρ+it = 1 ↓ +it, 2 t ∈ R, and thus 1−s = 1 ↑ −it; moreover we are interested in high values of t so 2 MEAN STAIRCASE OF THE RIEMANN ZEROS 7 that we may use the Stirling’s formula for the Gamma function given by: (20) Γ(x)∼=(2π)12 xx−12e−x as x→∞. From (19) and (20) we then obtain (asymptotically for t→∞) (cid:18) (cid:18)(cid:18) (cid:19)(cid:18) (cid:18) (cid:19) (cid:19) (cid:19) (cid:18) (cid:18) (cid:19)(cid:19)(cid:19) t t 1 1 (21) exp iπ ln −1 − +iarg ζ +it = 2π 2π 8 2+ (cid:18) (cid:18)(cid:18) (cid:19)(cid:18) (cid:18) (cid:19) (cid:19) (cid:19) (cid:18) (cid:18) (cid:19)(cid:19)(cid:19) t t 1 1 exp −iπ ln −1 − +iarg ζ −it 2π 2π 8 2− Since (cid:18) (cid:18) (cid:18) (cid:19)(cid:19)(cid:19) (cid:18) (cid:18) (cid:18) (cid:19) (cid:19)(cid:19) 1 1 exp iarg ζ ↑−it =exp iarg ζ ↓+it +iπ 2 2 (cid:18) (cid:18) (cid:18) (cid:19)(cid:19)(cid:19) 1 =−exp iarg ζ ↓+it 2 we then have, taking the limit ρ= 1 ↓= 1, that: 2 2 (cid:18) (cid:18) (cid:19) (cid:19) (cid:18) (cid:18) (cid:19)(cid:19) t t π 1 (22) cos(Ψ)=0whereΨ= ln −1 − +arg ζ +it 2 2π 8 2 Thus Ψ=π(cid:0)n+ 1(cid:1). We then obtain: 2 (cid:18) (cid:18) (cid:19) (cid:19) (cid:18) (cid:18) (cid:19)(cid:19) t t 1 1 1 1 ln −1 − + arg ζ +it =n− 2π 2π 8 π 2 2 hence (cid:18) (cid:18) (cid:19) (cid:19) (cid:18) (cid:18) (cid:19)(cid:19) t t 7 1 1 1 (23) ln −1 + + arg ζ +it =n+ 2π 2π 8 π 2 2 Eq.(22)maybeseenasanapproximate“quantumcondition”forthetrueRiemann zeros, but it is only a consequence of the Riemann symmetry (Eq. 19). In fact, if in equation (22) we neglect the last term arg(ζ), then (22) has as a solution the second set of trivial zeros {t∗∗}. It is true, as remarked by Berry and Keating, n that their Eq.(3) has complex zeros wich are not the Riemann zeros, but it should be remarked that if in Eq.(3) we set Re(s) = 1 then Eq.(3) reduces to Eq.(22) 2 without the fluctuation term arg(ζ); so the solution of Berry and Keating Eq.(3) for Re(s)= 1 is the same as the second set of trivial zeros {t∗∗} we have specified. 2 n Below the plots of the left hand side of Eq.(22), with and without the term argζ(cid:0)1 +it(cid:1). As an illustration, we may observe on the plot the first istanton 2 discussed above and the second one. In fact the maximum of the function which gives t∗ (Eq. 22 without the term argζ) is outside the plot of the step function given by (22) (the true function). This is visible on the plot near t=282 and near t = 296 (the second istanton). This conclude our remark on (3) and Eq.(22). In thenextsection,weshallcostructanhermiteanoperatorwhoseeigenvaluesarethe trivial zeros of the zeta function on the critical line. Now in (13) and (14) the value of a trivial zero (t∗ or t∗∗) is given through his n n index n by means of the Lambert W function so that such zeros are related in a non linear way to the integers n, i.e. in principle to the spectrum of an harmonic oscillator. So, for the trivial zeros, no boundary condition is needed here, since they are obtained by means of (13) and (14) in the large t limit. At this moment we are free to introduce a hermitean matrix which may generates the trivial zeros. 8 DAVIDEAMARCA,STEFANOBELTRAMINELLI,ANDDANILOMERLINI Figure 3. Plot of the function cos(Ψ) of Eq.(22) with the term arg(cid:0)ζ(cid:0)1 +it(cid:1)(cid:1) (step function) and without that term. 2 4. An Hermitean operator (matrix) associated with the mean staircase (trivial zeros) of the Riemann Zeta function As remarked above, in (13) and (14) the only “quantal number” is the index n of the trivial zeros and the construction may be given using a hermitean n×n matrix H, for any n, at our disposal and related to the classical one dimentional many body system whose fluctuation spectrum around the equilibrium positions is that of the harmonic oscillator. In fact, the one dimensional Mehta-Dyson model of random matrices (which may be seen as a classical Coulomb system with n particles) has, at low temperature an energy fluctuation spectrum given by the integersanditispossibletointroduceclassicalannihilationandcreationoperators, as studied in [13] (a short discussion is presented in the Appendix). The matrix elements of the associated hermitean matrix are then functions of the zeros of the Hermitepolynomials;inthiscasewedonothaveaHilbertspaceandnoSchr¨odinger Equation will be associated with the Lambert W function. Another direction, i.e. that of introducing a Schr¨odinger Equation to describe the trivial zeros may in principle be obtained as an application of the results given by G. Nash [14]; this because for large n, as it is known, (13) and (14) give the behavior ([11], pag. 214) related to the asymptotic behavior of the LambertW function: 2πn (24) t = ,n→∞ n ln(n) and thus the spectrum appears in fact as a one where the associated Schr¨odinger Equation contains a Gaussian type of potential [14]. Here we will consider the matrixformulation: thepointmayseemtobesomewhatartificialbutthehermitean matrixwewilluse(specifiedintheAppendix)isrelatedtotheMehta-Dysonmodel, the “starting point” of the random matrix theory. To do this, we begin to write MEAN STAIRCASE OF THE RIEMANN ZEROS 9 (10) in a slighly different form using the Stirling formula for the Gamma function of real argument given by: Γ(x)=(2π)12 xx−12e−xasx→∞ We then have that, as t→∞, (cid:18) (cid:18) (cid:19)(cid:19) (cid:18) (cid:19) t 1 t t 7 1 7 (25) ln Γ + = ln −1 + + ln(2π)− 2πe 2 2πe 2πe 8 2 8 1 7 =n∗+ ln(2π)− =n∗+θ 2 8 where θ = 1ln(2π)− 7. 2 8 ThusintroducingtheoperatorT =T(H)whoseeigenvaluesshouldbethetrivial energy levels (for the first as well as for the second set defined by (13) and (14) as well as H, the hermitean matrix given in the Appendix and related to the Mehta- Dyson model, we may write the following heuristic matrix equation: (cid:18) (cid:19) T I (26) Γ + =eH+θ 2π 2 where I is the unit matrix. Eq.(26) is the equation for T, giving the trivial zeros. The inversion of this formula (if it is possible to take it) yields heuristically: (cid:18) (cid:19) (27) T =T(H)=2π Γ−1(cid:0)eH+θ(cid:1)− I 2 To conclude, if Hϕ =(cid:0)n+ 1(cid:1)ϕ , where ϕ is the nth eigenfunction of H, then n 2 n n (cid:18) (cid:19) (cid:18) (cid:19) (28) Tϕ =2π Γ−1(cid:0)eH+θ(cid:1)− I ϕ =2π Γ−1(cid:0)en+θ(cid:1)− 1 ϕ =t ϕ n 2 n 2 n n n where t =t∗ resp. t∗∗. (in thats latter case with θ lowered by 1/2) n n n OfcourseEq.(26)fortheoperatorT ismoreappealingthan(13,14)(wherenis replaced by H and t∗ resp. t∗∗ are replaced by T) due to the combinatorial nature n n of the Gamma function, but the eigenvalues of the operators are the same in the “termodynamic limit”, t→∞. Remark: Ifoneconsidertheusualmapz →1−1 thenthecriticallines= 1+it s 2 (t∈R)ismappedontotheunitcircle|z|=1; thetwosetsoftrivialzeros{t∗}and n {t∗∗}haveasaccumulationpointz =1(asn→∞),whichisthesameaccumulation n pointfortherealzerosoftheζ functiongivenbyz¯ =1− 1 =1+ 1 ,asn→∞ n −2n 2n (see Fig .4). Neglecting the real zeros {z }, Fig. 4 illustrate by means of two sets of trivial n zeros {t∗} and {t∗∗} the Lee-Yang Theorem for the zeros of the partition function n n for some general spin lattice system studied in statistical mechanics. If RH is true, then all non trivial zeros of ζ(s) shall be located at the same circle |z| = 1, with z =1 as accumulation point. 5. Conclusion InthisnotewehavespecifiedanoperatorequationfortheoperatorT,havingas eigenvalues the trivial zeros of the ζ function on the critical line (13 and 14); such zeros are the so called trivial zeros (mean staircases), and are related in a strong non linear way to the eigenvalues of a discrete harmonic oscillator described by H in the large t limit. The introduction of H may seem to be artificial: nevertheless 10 DAVIDEAMARCA,STEFANOBELTRAMINELLI,ANDDANILOMERLINI †z§á1 Σá0 zá0 zá1 Σá1 1 ΣáÄ 2 Figure 4. z-plane the two sets of trivial zeros ({t∗} and {t∗∗}) are the eigenvalues of T(H). In a n n subsequent note we will present a study of another sequence of zeros possibly more connected with the true Riemann zeros. Update Very recently G. Sierra and P.K. Townsend [15] introduced and studied an in- terestingphysicalmodel(achargedparticleintheplaneinpresenceofanelectrical andamagneticpotential). Inparticular, thelowestLandaulevelisconnectedwith the smoothed counting function that gives the average number of zeros, i.e. the staircasewhichherewehavestudied,bymeansofaclassicalone-dimentionalmodel of N interacting charged particles. Note added Very recently Schumayer et al [16] constructed (in particular) the Quantum mechanicalpotentialforξ(s)zeros,withthefirst200energyeigenvalues(nontrivial zeros). Itisexpectedthatthesameformofaquantummechanicalpotentialwould appear using only the two sets of zeroes we have discussed in this note. For the constructionofanHamiltonianwhosespectrumcoincideswiththeprimes, seealso the recent work of S. Sekatskii [17].