p q = E/2π

E/2πΛ

FIGURE 2

As explained in details in Chapter II, §3.2 of [12], one restricts to even functions to handle the Riemann zeta function, and to the first quadrant in order to estimate the semiclassical count of the zeros of ζ with positive imaginary part less than

E. This procedure is depicted in Figure 3. In order to compare this semiclassical picture with white light, one considers the model of white light given by the regular representation of the additive group R. We adjust the cutoff so that this corresponds, using the isomorphism u 7→ exp(u) of R with the multiplicative group R^∗₊, to the subgraph of the hyperbola pq =₂^E_π restricted to the interval [1/Λ, Λ] as shown in Figure 4. The exponential u 7→ exp(u) is obviously an isomorphism exp: R → R^∗₊ and the following map is a symplectic isomorphism

since its Jacobian is given by d exp(u) ∧ d(v exp(−u)) = exp(u)du ∧ (exp(−u)dv − v exp(−u)du) = du ∧ dv.

THE SCALING HAMILTONIAN105

p q = E/2π

E/2πΛ

FIGURE 3

The symplectic isomorphism Φ transforms the phase space of the regular representation of R cutoff in the interval [− log Λ, log Λ] and in a range of energy (the dual variable to log(q)) in the interval [0, E/2π], into the portion of the phase space (q, p) shown in Figure 4. We refer to [10] and Chapter II, §3.2 of [12] for the understanding of careful normalizations.

106ALAIN CONNES AND CATERINA CONSANI

Λ E/2π

E/2π

-LogΛ LogΛ

u -u

Φ(u,v)=(e,v e)

p q ≤ E/2π

-1

Λ Λ

FIGURE 4

THE SCALING HAMILTONIAN107

There are two main differences between the cutoff phase space of Figure 3 and the white light of Figure 4. Firstly, there is a piece of Figure 4 which is not covered by Figure 3, it extends from p = Λ to p = ΛE/2π and is represented in dark blue in

ΛE/2π

p q = E/2π

1/Λ Λ

FIGURE 5

Then there is also the rectangle, with vertices at (0, 0), (1/Λ, 0),(1/Λ, Λ),(0, Λ), in Figure 3, which does not fit inside Figure 4. It is of area equal to 1 and thus corresponds to a single quantum state.

108ALAIN CONNES AND CATERINA CONSANI

In order to compare this set-up with the Berry-Keating framework we perform the symplectic transformation of the (q, p) plane given by

This transformation preserves the area and the hyperbola pq =₂^E_π. It transforms Figure 5 into the following Figure 6:

E/2π

pq=E/2π

1 ->Λ

FIGURE 6

One easily obtains that the missing (absorbed) part of the white light is given by the semiclassical picture of the “Berry-Keating" Hamiltonian as in Figure 7. This discussion shows that all the semiclassical computations for the Riemann zeros viewed as an absorption spectrum (as in [10,12]), are identical to those performed with the “Berry-Keating" Hamiltonian viewed as an emission spectrum ([1, 2]). The presence of the opposite signs involved in absorption and emission spectra does not create a conflict at this point. However, one should not conclude that the two models are equivalent.

THE SCALING HAMILTONIAN109

FIGURE 7. Figure taken from [2], there is a different normalization of the Fourier transform which accounts for the discrepancy in the factors 2π.

In fact, the minus sign coming from the absorption spectrum strikes back when one investigates the quantum fluctuations. In [1], Berry and Keating are forced to set the Maslov phases all equal to π in order to account for this minus sign. They write explicitly:

2. SEMI-LOCAL TRACE FORMULA

After the initial work [10] on the semi-local trace formula a better proof was obtained based on an idea of [6] in the case of a single archimedean place. This development was explained in the class of the first author at Collège de France in

110ALAIN CONNES AND CATERINA CONSANI

1999 [11]. We refer to Chapter 2 of [12] for a detailed exposition. A key tool here is the quantized calculus of Chapter IV of [9]. 2.1. QUANTIZED CALCULUS. The main idea of the quantized calculus is to give an operator-theoretic version of the calculus rules, based on the operator-theoretic differential (2.1) d⁻f:= [F, f], where f is an element in an involutive algebra A represented as bounded oper-ators on a Hilbert space H. The right-hand side of (2.1) is the commutator with a self-adjoint operator F on H with F² = 1. One then defines the analogue of differential forms of degree n as linear combinations of monomials of the form

1 2

The differential of an element of Ω is defined as the graded variant of (2.1) (2.2) d ω:= Fω − (−1)ⁿωF.

One checks, using F² = 1, that the square of the differential is 0, d²ω = 0.

Next, we briefly recall the framework for the quantized calculus in one variable, as in Chapter IV of [9]. We let functions f of one real variable act as multiplication operators in L²(R, ds), by (2.3) (f ξ)(s):= f (s)ξ(s), ∀s ∈ R, ξ ∈ L²(R, ds).

We let F_eR be the Fourier transform wi∫th respect to the character e_R(x):= e^−2πix

1 ξ(t)

THE SCALING HAMILTONIAN 111 its Haar measure. The elements of L^∞(Ĉ) act as multiplication operators on the

Hilbert space H:= L²(Ĉ). The operator H on H is given by (2.9) H:= 2F_C 1_P F_C⁻¹ − 1, where F_C: L²(C) → H is the Fourier transform and 1 is the multiplication by P the characteristic function of the set P = {u ∈ C | |u| ≥ 1}. Such H plays the role of the Hilbert transform. In analogy with (2.7), we define the “quantized" differential of f as the operator (2.10) d⁻f:= [H, f] = H f − f H.

We also use the following notation whose normalization fits with (2.5): (2.11) Π_b]:= F_C 1_Y F_C⁻¹, for Y = {u ∈ C | e^2πa ≤ |u| ≤ e^2πb}. [a, 2.2. THE HILBERT SPACE L2(XS) FOR A FINITE SET OF PLACES. In the following we consider the global field Q of rational numbers. In the adèle ring of Q the additive Haar measure and the Haar measure of the multiplicative group of idèles are mutually singular. This problem does not arise w∞hen one restricts the attention to a finite set S of places of Q. We assume that ∈ S. Let us consider the locally compact ring (2.12) A_S = Q_v. ∏ v∈S

It contains Q as a subring, using the diagonal embedding. We let Q_S denote the subring of Q given by all rational numbers whose denominator only involves powers of primes p ∈ S. In other words, (2.13) Q_S = {q ∈ Q | |q|_v ≤ 1, ∀v ∈/ S}. The group Q^∗_S of invertible elements of the ring Q_S is of the form (2.14) Q^∗_S = GL₁(Q_S) = {±pⁿ₁¹ · · · pⁿ_k ^k: p_j ∈ S \ {}, n_j ∈ Z}. ∞

We approximate the adèle class space by the semi-local adèle class space X S which is the quotient (2.15) X_S:= A_S/Q^∗_S.

The group (2.16) A^∗_S = GL₁(A_S) = GL₁(Q_p) CQ,S = A∗S/Q∗S, ∏ p∈S acts naturally by multiplication on the quotient X. S We recall from [10,12] that the Fourier transform F on the Bruhat-Schwartz space S(A_S) induces a unitary operator on the Hilbert space L²(X_S). For each place v ∈ S we choose a basic character α of Q_v. This makes it possible to identify the v locally compact abelian group Q_v with its Pontrjagin dual by the pairing (2.17) 〈a, b〉:= α_v(ab), ∀a, b ∈ Q_v.

112ALAIN CONNES AND CATERINA CONSANI

We normalize the additive Haar measure so that it is self-dual. Then α =_v α is ∏ v a basic character of the additive group A_S, and we let F_α denote the corresponding Fourier transform, acting on the space S(A_S). We refer to [10,12] for the proof of the following result LEMMA 2.2. Consider the character α = α and the corresponding Fourier trans-∏v v form F_α as above. The map f 7→ F_α(f), for f ∈ S(A_S), extends uniquely to a unitary operator F_α on the Hilbert space L²(X_S).

The scaling operator ϑ_a(λ), for λ ∈ GL₁(A_S), is defined as (2.18) (ϑ_a(λ)ξ)(a) = ξ(λ⁻¹a), ∀a ∈ A_S, ∀ξ ∈ S(A_S).

It is not unitary but its module is given by the module of the modulated group Q,S (2.19) ϑ_a(λ)^∗ϑ_a(λ) = ϑ_a(λ)ϑ_a(λ)^∗ = |λ|_S.

We let w be the unitary identification of L²(X_S) with L²(C_Q,S) (see [12] Proposition 2.30). It intertwines the representation ϑ with the following regular repre-a

Fourier transform F_αv relative to the character α, as a composition of the inver-v sion I(f)(s):= f (s⁻¹) with the multiplicative convolution operator associated to u. The function u is given by the ratio of the local factors of L-functions v v (see [17]). This classical formula extends to our context as follows. We denote by u the function (2.22) u:= u_v ◦ π_v. ∏ where π: Ĉ_,S → Q̂^∗_v is the projection dual to the embedding Q^∗_v ↪→ C_Q,S, v Q x 7→ (1, 1,..., x,..., 1) which puts x at the place v. We now have LEMMA 2.3. (see [12]) On L²(X_S) one has (2.23) F_α = w⁻¹ ◦ I ◦ F_C⁻¹ ◦ u ◦ F_C ◦ w, where F_C is the Fourier transform of C and I(f)(s):= f (s⁻¹) is the inversion. Q,S

THE SCALING HAMILTONIAN113

2.3. THE SEMI-LOCAL TRACE FORMULA. As in the case of the single place, we introduce infrared and ultraviolet cutoffs. For the infrared, we use the orthogonal projection P onto the subspace Λ (2.24) {ξ ∈ L²(X_S) |ξ(x) = 0, ∀x with |x| > Λ}.

Thus, P is the operator acting as multiplication by the function ρ, with ρ_Λ (x) = 1 for |x|^Λ≤ Λ and ρ_Λ (x) = 0 for |x| > Λ. Λ

We define an ultraviolet cutoff as P̂_Λ = F_αP_ΛF_α⁻¹, where F_α is the Fourier transform of Lemma 2.2, which depends upon the choice of the basic character α = ∏v v α. Then, we let (2.25) R_Λ = P̂_Λ P. Λ

We recall from Chapter 2 of [12] the following result.

LEMMA 2.4. Let S 3 be as above. For any Λ > 0 there is a unitary operator (2.26) W = W_Λ: L²(X_S) → L²(Ĉ_Q,S) such that, for any h_j ∈ S(C_Q,S), j = 1, 2, one has (2.27) W ϑ_a(h̃₁)P̂_Λ(P_Λϑ_a(h̃₂)W =) ĥ u d u Π_2oπgΛ] + Π_[0,2l ĥ. 1 2 − ^{2 l} 2 1 −1 − ∞ [, 2oπg Λ]

Here h̃_j(λ) = |λ|^−1/2h_j(λ), the operator d u is the quantized differential of the function − u:=_S u_v ◦ π_v of (2.22) and ĥ is the multiplication operator by the Fourier transform ∏ j

114 ALAIN CONNES AND CATERINA CONSANI ∫ ∫

Note that ϑ_a(f) = f (λ)ϑ_a(λ)d^∗λ corresponds by (2.21) to ϑ_m(h) = h(λ)ϑ_m(λ)d^∗λ, where one has h(λ) = |λ|^1/2 f (λ). In terms of the test function h the terms in the

In [7], J. F. Burnol found a spectral realization of the zeros of zeta, closely related to the spectral realization of [10], but using the Sonine spaces implemented by L. de Brange in his approach to RH. Motivated by the role of Sonine spaces, as well as his own work on the semi-local trace formula [18], X.-J. Li made, in early 2019, a brave attempt at proving the Weil positivity which is equivalent to RH (see [5]). Rather than introducing a cutoff for large q and p as in [10], he prescribes a cutoff near zero, i.e. for small q and p. In fact he does this in a balanced way i.e. say requiring q > Λ and p > Λ⁻¹. The coincidence between this prescription and the above relation between the “absorption" and "emission pictures" is very striking since the proposed cutoff near zero corresponds exactly to Figures 5, 6, 7. This attempt though fails. Nonetheless it is worthwhile to review it in some details and also to understand what is still missing, while retaining the interesting part. There are two basic facts on which Li’s attempt is based:

FACT 3.1. Let A, B be positive bounded operators in a Hilbert space H and assume that AB is of trace class, then Tr(AB) ≥ 0. Proof. Let A be the square root of A. Then, since in general the spectrum of a 1/2 product ST is the same as the spectrum of TS except possibly for 0, one obtains, denoting Spec the non-zero part of the spectrum

This shows that all eigenvalues of the trace class operator AB are positive, and one concludes using Lidskii’s theorem.

FACT 3.2. The Riemann-Weil explicit formula applied to a test function h ∈ S(C_Q) with compact support only involves finitely many places.

Proof. By hypothesis the function h only depends on the modulus. The contribution of each prime p only involves the value of h on non-zero powers of p so that only finitely many primes contribute.

The natural idea then, is to use the semi-local framework of [10] recalled in Section 2 and mo∫re precisely formula (2.28) to show an inequality of the form ′ |w|1/2 (3.1) h(w)d^∗u ≤ 0, h = h₁? h₂, h₂(w):= h₁(w⁻¹). ∑ v _∈SQ∗v |1 − w|

THE SCALING HAMILTONIAN115

Note that by construction the operators ĥ given as the multiplication operators j by the Fourier transform F_C(h_j) fulfill, when h₂ = h^∗₁ as in (3.1), the following

FACT 3.3. When h₂ = h^∗₁, the operator ĥ ĥ is positive. 1 2

Now, using the cyclic property of the trace and the commutativity of the algebra of multiplication by functions, one gets from ((2.28) the followin)g equality ∫ ′ |w|1/2 1 −1 (3.2) h(w)d^∗w = Tr ĥ₁ĥ₂(u d⁻u). ∑ v _∈SQ^∗_v |1 − w| 2

The idea of X.-J. Li is to replace the operator R_Λ = P̂_Λ P by the positive operator Λ T = PP̂P, where P is the projection corresponding to the cutoff near zero, i.e. restricting to |w| ≥ 1. One gets, using the notations of (2.28), that

One sees that Fact 3.1 would imply (3.1) provided one could prove that

As we shall see this is equivalent to controlling the sign of the operator u d u. −1 − The quantized calculus gives in fact the following general criterion to control the sign of the “logarithmic derivative" u d u of a unitary u. −1 −

LEMMA 3.4. Let H be a Hilbert space, u a unitary operator and F = 2P − 1, where P is an orthogonal projection. Then, with d⁻u:= [F, u], the following three conditions are equivalent (i) Pu^∗(1 − P)uP = − ¹₂ u d u −1 − (ii) u d⁻u ≤ 0 −1 (iii) Pu = PuP Proof. (i) ⇒ (ii) since Pu^∗(1 − P)uP is a positive operator. (ii) ⇒ (iii). If u d⁻u ≤ 0 one has u^∗[P, u] ≤ 0 and u Pu ≤ P, thus Pu^∗P = u^∗P −1 ∗ since the range of u P is inside P. Thus, taking adjoints, one obtains Pu = PuP. (iii) ⇒ (i). Assume Pu = PuP. Taking adjoints one derives Pu^∗P = u^∗P, and thus the projection Q = u^∗Pu is less than P, one has PQP = Q and

Pu^∗(1 − P)uP = P(1 − Q)P = P − Q = −u^∗[P, u] = − u d u −1 − which proves (i).

When working with the quantized calculus in one variable one gets the following COROLLARY 3.5. Let H and F be as in Definition 2.1, and u a unitary function. Then

116ALAIN CONNES AND CATERINA CONSANI

Proof. The function u is inner if and only if it preserves the subspace which is the range of the projection P, with F = 2P − 1. Then Lemma 3.4 gives the required equivalence.

Equipped with Lemma 3.4 and Corollary 3.5, we can now understand why the direct approach proposed by X.-J. Li fails to prove Weil’s positivity. Indeed, in our case and taking a single place to simplify the discussion, the issue is whether the functions u, where u is given by the ratio of local factors on the critical line, v v spaces H (C^±₁₂) of the half plane happen to_∞be inner functions. In fact we can ask if u belongs to one of the Hardy v

At a finite prime p the function u is given by the ratio of local factors p 1 − p^−(1−z) 1 (3.3) u_p(s):=, z = + is. 1 − p^−z 2

It is a function of modulus 1 for s ∈ R. The zeros of the denominator 1 − p^−z give poles,

∈ 2πi ∈ log p which show that in the left half-plane the function is not holomorphic. On the other hand in the right half-plane the denominator does not vanish and the function is holomorphic. However it is not bounded since for real values of z it be-haves as follows 1 − p−(1−z 1 − p _−z⁾ ∼ −p^z−1, z →.

We now look at the archimedean place. There, the ratio of local factors is () (3.4) u^∞(s):= φ(+ is), φ(z) =. 1 π Γ 1−2 z(− 2z)z 2 2 1−z Γ

The function φ(z) is of modulus 1 for z = ¹₂ + is, and has poles in the left halfplane. It is holomorphic in the right half-plane <(z) ≥₂¹, however it is not bounded there. In fact one has using the complement formula π π Γ (x)Γ (1 − x) = ⇒ = Γ (x)sin(πx) sin(πx) Γ (1 − x)

THE SCALING HAMILTONIAN117

Thus even on the real axis the function φ(z) that vanishes at odd integers takes very large values on even integers such as

1856156927625

Thus we conclude that none of the functions u_v(s)^∗ fulfills the requirement of Corollary 3.5. We finally remark that, a priori, the inequality (3.1) could still hold even though the above described attempt to prove it fails. In fact one has the following

FACT 3.6. The inequality (3.1) does not hold in general.

Proof. We consider the simplest case when S is reduced to the single archimedean place. Inequality (3.1) would then imply, using the expression of the left hand side in terms of the logarithmic derivative of u^∞(s), that this logarithmic derivative (multiplied by i) has a constant sign. But u^∞(s) = e²ⁱ where θ(s) is the θ(s)

Riemann-Siegel angular function (see [12] Chapter II, §5.1, Lemma 2.20). Thus the inequality (3.1) would imply that the Riemann-Siegel angular function θ(s) is a monotonic function, but by looking at its graph (Figure 8) one sees that it is not the case.

20

10

M p∈S S Thus we see, heuristically as above, that the Poisson Formula gives the conceptual explanation for Lemma 2.3. There is in general a choice involved for basic additive characters α of A_S such that Q_S is a self-dual lattice, but this ambiguity S disappears when working in the quotient X_S = A_S/Q^∗_S. In particular, the Fourier transform F_α is independent of such choices. Indeed, the Fourier transforms of

120 ALAIN CONNES AND CATERINA CONSANI the same function for different choices of characters for which Q_S is a self-dual lattice are related by the action (by scaling) of the group Q^∗_S and thus their difference vanishes in the space L²(X_S). The Weil inequality needs the above normalization of additive characters and this indicates that the Poisson Formula should play a key role in the solution of the Conjecture 4.1. In fact, there is another striking reason why the Poisson Formula is intimately related to Conjecture 4.1. It is the construction, due to J.P. Kahane (see [8]), of non-trivial elements of the Sonine space, i.e. of functions which vanish identically near the origin as well as their Fourier transform. Let Π:=_n δ be the Pois-∑ ∈Z n son distribution, which is the sum of Dirac distributions on the integers. Then the tempered distribution D(Π) has the above vanishing property, when D is a differential operator whose range D(S(R)) consists of functions which fulfill the two conditions (4.7) f (0) = f̂ (0) = 0.

The simplest choice for D is D(f)(x):= x f ^′′(x), since one checks by integration by parts that the integral of D(f) vanishes. For any such choice of D the distribution D(Π) has the required Sonine support property. One then obtains a solution in Schwartz space by applying to D(Π) the smoothing by convolution with a smooth function φ with small support near zero, and multiplication by the Fourier transform of φ. In fact the above choice of the differential operator D, though simple, is not symmetric with respect to Fourier transform. One checks that the simple function D:= D_u² + D_u of our scaling Hamiltonian D_u(f)x):= x f ^′(x) does indeed work, commutes with Fourier transform, and one gets (4.8) D(f)(x) = x² f ^′′(x) + 2x f ^′(x).

One checks directly that the range D(S(R)) consists of functions which fulfill the two conditions (4.7). All this construction extends verbatim to the semi-local case.

REMARK 4.2. In an intriguing manner the adjoint expression ∆₂:= D_u² − D_u provides the tropical structure of the Scaling Site as explained in [14], equation (1).

4.2. TIME VERSUS ENERGY. In the approach to RH advocated by M. Berry and

J. Keating [1, 2], one looks for an interpretation of the zeros of the Riemann zeta function as energy levels of a quantum system. In our own approach however the zeros appear naturally as “times" i.e. in a dual manner (see [10]). While this fact might look perplexing at first sight, we shall briefly explain here below why this dual point of view is in fact more natural in view of the generalization of the Riemann zeta function in the framework of global fields. Indeed, in this set-up one meets much simpler avatars of the Riemann zeta function. They are associated to a curve C over a finite field F_q. It turns out that these analogues of the Riemann zeta function ζ(s) are in fact functions of the form L(q^−s), where q is the

THE SCALING HAMILTONIAN 121 cardinality of the finite field over which the curve is defined. Moreover, L(z) is a rational fraction and its zeros are determined by the zeros z of the polynomial j numerator P(z) whose degree is twice the genus g of the curve. By a famous theorem of A. Weil, all z are on the circle of radius q. Thus the zeros s of L(q^−s) j −1/2 are all of real part <(s) =₂¹ and their imaginary parts =(s) are of the form () =(s) = α_j + 2πk, j ∈ {1,... 2g}, k ∈ Z, log q where the −α_j are the arguments of the z and are determined only modulo 2π. j This type of distribution of numbers, being periodic, is very natural as a distribution in “time", not in “energy".

4.3. ANALYSIS AND GEOMETRY. The above discussion does not dismiss the nagging thought we started with. In fact there is a real possibility that one could find a way to solve Conjecture 4.1, by first concentrating on the simplest case of a single place and finding a way to make use of the support condition on the test function. A scary aspect of this strategy is that Weil’s positivity is very much dependent on the careful choices of principal values and it is difficult to avoid computational mistakes. This analytical approach is in this respect, by its vulnerability, quite different from the geometric approach that we have gradually developed in [13, 15]. While the adèle class space is clearly implicit in the semilocal analytic approach described above, its geometric structure is only unveiled at the global level and it reveals its conceptual meaning as the points of the Scaling Site, where the action of the scaling group reveals itself as the action of the Frobenius.

REFERENCES

[1] M. Berry and J. Keating, H = qp and the Riemann zeros, “Supersymmetry and Trace Formulae: Chaos and Disorder”, edited by J.P. Keating, D.E. Khmelnitskii and I.V. Lerner, Plenum Press. Link to download the paper [2] M. Berry and J. Keating, The Riemann zeros and eigenvalue asymptotics. SIAM Rev. 41 (1999), no. 2, 236–266. [3] A. Beurling, On two problems concerning linear transformations in Hilbert space. Acta Mathematica 81, (1948) 239–255. [4] E. Bombieri, J. Lagarias, Complements to Li’s criterion for the Riemann hypothesis. J. Number Theory 77 (1999), no. 2, 274–287. [5] E. Bombieri, The Riemann hypothesis. The millennium prize problems, 107–124, Clay Math. Inst., Cambridge, MA, 2006. [6] J. F. Burnol, Sur les formules explicites. I. Analyse invariante [On explicit formulae. I. Invariant analysis] C. R. Acad. Sci. Paris Sér. I Math. 331 (2000), no. 6, 423–428.

122ALAIN CONNES AND CATERINA CONSANI

[7] J. F. Burnol, Sur certains espaces de Hilbert de fonctions entières, liés à la transformation de Fourier et aux fonctions L de Dirichlet et de Riemann. (French) [Some Hilbert spaces of entire functions associated with the Fourier transform and Dirichlet and Riemann L-functions] C. R. Acad. Sci. Paris Sér. I Math. 333 (2001), no. 3, 201–206. [8] J. F. Burnol, Sur les espaces de Sonine associés par de Branges à la transformation de Fourier.

C. R. Acad. Sci. Paris, Ser. I 335 (2002) 689–692. [9] A. Connes, Noncommutative geometry, Academic Press (1994). [10] A. Connes, Trace formula in noncommutative geometry and the zeros of the Riemann zeta function. Selecta Math. (N.S.) 5 (1999), no. 1, 29–106. [11] A. Connes, Formules explicites, formules de trace et réalisation spectrale des zéros de la fonction zéta, Course at Collège de France, 1999. [12] A. Connes, M. Marcolli, Noncommutative Geometry, Quantum Fields, and Motives, American Mathematical Society, 2008. [13] A. Connes, An essay on the Riemann Hypothesis. In “Open problems in mathematics", Springer (2016), volume edited by Michael Rassias and John Nash. [14] A. Connes, C. Consani, Geometry of the Scaling Site. Selecta Math. (N.S.) 23 (2017), no. 3, 1803–1850. [15] A. Connes, C. Consani, The Riemann-Roch strategy, complex lift of the Scaling Site, To appear on “Advances in Noncommutative Geometry, On the Occasion of Alain Connes’ 70th Birthday”, Chamseddine, A., Consani, C., Higson, N., Khalkhali, M., Moscovici, H., Yu, G. (Eds.), Springer (2019). Available at http://arxiv.org/abs/1805.10501. ISBN 978-3-030-29596-7. [16] A. Connes, C. Consani, On Absolute Algebraic Geometry preprint (2019). Available at https://arxiv.org/abs/1909.09796 [17] J. Tate, Fourier analysis in number fields and Hecke’s zeta-function, Ph.D. Thesis, Princeton, 1950. Reprinted in J.W.S. Cassels and A. Frölich (Eds.) “Algebraic Number Theory”, Academic Press, 1967. [18] Li, Xian-Jin, A generalization of A. Connes’ trace formula. J. Number Theory 130 (2010), no. 2, 386–430.

ALAIN CONNES, COLLÈGE DE FRANCE, IHES AND OHIO STATE UNIVERSITY E-mail address: alain@connes.org

CATERINA CONSANI, DEPARTMENT OF MATHEMATICS, THE JOHNS HOPKINS UNIVER-SITY, BALTIMORE MD, 21218, USA E-mail address: kc@math.jhu.edu