--- title: Tamagawa Ratio in Arithmetic Geometry url: https://www.emergentmind.com/topics/tamagawa-ratio type: topic --- # Tamagawa Ratio in Arithmetic Geometry Searching arXiv for recent papers on Tamagawa ratio and related Tamagawa-number formulations. The **Tamagawa ratio** is not a single standardized invariant across arithmetic geometry. In the literature, the phrase most often denotes either a ratio of Selmer-group sizes attached to an isogeny, a quotient of products of Tamagawa-type local factors across intermediate fields, or a cohomological quotient that computes a Tamagawa number. In each case, the underlying theme is the comparison of global arithmetic objects by means of local correction terms. Recent work has clarified that these apparently different uses are connected by local-to-global product formulas, regulator constants, root numbers, and Galois-cohomological defect groups [2504.17962], [1406.6745], [2606.31649]. ## 1. Terminological scope and basic meanings The phrase **“Tamagawa ratio”** is used in several distinct but related senses. In the setting of elliptic curves with a rational \(2\)-isogeny, the Tamagawa ratio is defined as \[ \mathcal{T}(E/E') := \frac{|\mathrm{Sel}_\phi(E/\mathbb{Q})|}{|\mathrm{Sel}_{\hat\phi}(E'/\mathbb{Q})|}, \] where \(\phi:E\to E'\) is the isogeny and \(\hat\phi\) its dual [1406.6745]. Its \(2\)-adic valuation is written \[ t(A,B) := \operatorname{ord}_2 \mathcal{T}(E_{A,B}/E'_{A,B}), \] for curves \(E_{A,B}: y^2=x^3+Ax^2+Bx\) with rational \(2\)-torsion [1406.6745]. In the norm-relations framework for rank prediction, the term refers not to a single invariant with fixed notation, but to a quotient of products of Tamagawa-type local factors. The relevant global factor is \[ C_{E/F}=\prod_v C_v(E/F,\omega), \] with \[ C_v(E/L,\omega)= \begin{cases} c(E/L_v) & \text{if } E/L_v \text{ has semistable reduction},\\[4pt] c(E/L_v)\cdot \left|\omega/\omega^0\right|_{L_v} & \text{if } E/L_v \text{ has additive reduction}, \end{cases} \] and the corresponding “ratio/product of Tamagawa numbers” is typically a quotient of the form \[ \prod_i (C_{E/F^{H_i}})^{n_i} \] attached to a formal sum of subgroups \(\Theta=\sum_i n_iH_i\) [2504.17962]. Other papers do not define a separate invariant called a Tamagawa ratio, but instead study the **global Tamagawa product** \[ \operatorname{Tam}(E):=\prod_p c_p(E) \] for elliptic curves over \(\mathbb{Q}\) [2105.03513], or \[ \operatorname{Tam}(E/K):=\prod_{\mathfrak p} c_{\mathfrak p} \] over a number field \(K\) [2108.13625]. In BSD-motivated discussions, the quotient \[ \frac{c(E)}{|E(\mathbb{Q})_{\mathrm{tors}}|} \] or more generally \[ \frac{c(E/K)}{|E(K)_{\mathrm{tors}}|} \] plays the closest analogous role [2505.20479], [2202.06235]. For algebraic tori, the “ratio” language appears in Ono-type formulas, where the Tamagawa number itself is expressed as a quotient of cohomological invariants, such as \[ \tau_k(T)=\frac{| H^1(\Gamma,X(T)) |}{| \Sha^2(\Gamma,X(T))|}, \] or variants involving \(\Sha^1\) [2109.04121], [2009.04431]. This suggests that the term has a family resemblance rather than a universal definition. ## 2. Isogenies, Selmer groups, and Cassels-type product formulas The classical arithmetic context for the Tamagawa ratio is the comparison of Selmer groups associated to an isogeny and its dual. For elliptic curves with a rational point of order \(2\), one has a degree-\(2\) isogeny \[ \phi : E_{A,B}\to E'_{A,B}, \] with dual \(\hat\phi:E'_{A,B}\to E_{A,B}\), and the Tamagawa ratio \[ \mathcal{T}(E/E') := \frac{|\mathrm{Sel}_\phi(E/\mathbb{Q})|}{|\mathrm{Sel}_{\hat\phi}(E'/\mathbb{Q})|} \] measures the imbalance between the \(\phi\)-Selmer group and the dual \(\hat\phi\)-Selmer group [1406.6745]. A key structural fact is Cassels’ product formula \[ \mathcal{T}(E/E') = \prod_{v} \frac{|H^1_\phi(\mathbb{Q}_v, C)|}{2}, \] where \(C=\langle(0,0)\rangle\) is the kernel of \(\phi\), and the product is effectively over places of bad reduction and \(v=2,\infty\) [1406.6745]. For odd primes \(p\neq 2\), \[ |H^1_\phi(\mathbb{Q}_p, C)| = \frac{c_p'}{c_p}, \] so the local contribution is a relative Tamagawa factor for the isogenous pair [1406.6745]. In this sense, the Tamagawa ratio is literally a product of local Tamagawa-number quotients. For the family \[ E_{A,B}: y^2 = x^3 + Ax^2 + Bx, \] the local factors are controlled by reduction type. At primes \(p\neq 2\), the curve has good reduction if \(p\nmid B(A^2-4B)\), additive reduction if \(p\mid B\) and \(p\mid A^2-4B\), and multiplicative reduction if \(p\) divides exactly one of \(B\) and \(A^2-4B\) [1406.6745]. In the multiplicative case, \[ |H^1_\phi(\mathbb{Q}_p,C)|=\frac{c_p'}{c_p} = \begin{cases} 4 & \text{if } v_p(A^2-4B)\text{ is odd or } \left(\frac{-2AB}{p}\right)=1,\\[4pt] 1 & \text{if } v_p(B)\text{ is odd or } \left(\frac{B}{p}\right)=1,\\[4pt] 2 & \text{otherwise.} \end{cases} \] This local classification is the source of the asymptotic behavior of the global ratio [1406.6745]. The paper “The distribution of the Tamagawa ratio in the family of elliptic curves with a two-torsion point” proves that if \(\mathcal{E}(X)\) denotes the set of minimal models with \(|A|,B^2\le X\), then \[ \#\mathcal{E}(X) \sim \frac{4X^{3/2}}{\zeta(6)}, \] and \[ \{t(A,B):(A,B)\in\mathcal{E}(X)\} \] becomes normally distributed with mean \(0\) and variance \(2\log\log X\) as \(X\to\infty\) [1406.6745]. More precisely, the ratio is governed by the difference of two additive functions, \[ g_1(A,B) := \sum_{p \mid A^2-4B} 1, \qquad g_2(A,B) := \sum_{p\mid B} 1, \] so that, away from exceptional sets, \[ t(A,B)\approx g_1(A,B)-g_2(A,B) \] [1406.6745]. The same paper deduces that the average size of the \(2\)-Selmer group in this family is unbounded [1406.6745]. A highly rigid special case occurs for the family \[ E_{\sigma 2D}: y^2=x^3+\sigma 2Dx, \qquad E'_{\sigma 2D}: y^2=x^3-\sigma 8Dx, \] with \(\sigma=\pm1\) and \(D=\prod p_i^{e_i}\), \(e_i\in\{1,3\}\). There the paper defines \[ T_A=\frac{\#S^{(\phi)}(E_A)}{\#S^{(\phi)}(E_{-4A})}=2^{-t_A} \] and proves \[ T_{2D}=\frac12 \quad (\sigma=-1), \qquad T_{-2D}=1 \quad (\sigma=+1), \] equivalently \[ t_{2D}=1,\qquad t_{-2D}=0 \] [2106.00340]. The proof reduces local solubility conditions for homogeneous spaces to linear algebra over \(\mathbb F_2\), using matrices built from Legendre symbols [2106.00340]. A broader formulation appears in the Greenberg–Wiles framework. For a finite \(G_F\)-module \(M\) with local conditions \(L_v\), the Tamagawa ratio is defined by \[ T(M):=\frac{\#H^0(G_F,M)}{\#H^0(G_F,M^\vee)}\cdot \prod_v T_v(M), \qquad T_v(M):=\frac{\#L_v}{\#H^0(G_v,M)}, \] and the Greenberg–Wiles formula gives \[ \frac{\#Sel\,M}{\#Sel\,M^\vee}=T(M) \] [2606.31649]. In geometric isogeny settings, the local factors are often ratios of Tamagawa numbers, so the classical isogeny-Selmer ratio is a special case of this general formalism [2606.31649]. ## 3. Norm relations, products of Tamagawa numbers, and rank prediction A distinct recent use of the term concerns the prediction of positive Mordell–Weil rank by means of products of Tamagawa numbers across intermediate fields. For a Galois extension \(F/\mathbb{Q}\) with group \(G\), a representation \(\rho\), and a virtual permutation decomposition \[ \bigg( \bigoplus_{\sigma \in \Gal(\mathbb{Q}(\rho)/\mathbb{Q})} \rho^\sigma \bigg)^{\oplus m} = \bigg( \bigoplus_i \mathbb{C}[G/H_i] \bigg)\ominus \bigg( \bigoplus_j \mathbb{C}[G/H_j'] \bigg), \] the norm relations test examines the quotient \[ \frac{\prod_i C_{E/F^{H_i}}}{\prod_j C_{E/F^{H_j'}}} \] [2504.17962]. The criterion is that if this quantity is **not** a norm from a quadratic field \(\mathbb{Q}(\sqrt D)\subset \mathbb{Q}(\rho)\), or is not a square in the even-\(m\) case, then \(E(F)\) has positive rank [2504.17962]. In this setting the “Tamagawa ratio” is thus a quotient of global Tamagawa-type factors \(C_{E/F^{H}}\), not merely a quotient of local Tamagawa numbers. The paper “Tamagawa numbers and positive rank of elliptic curves” proves that this Tamagawa-number criterion is contained in the parity-conjecture framework [2504.17962]. The central compatibility theorem states that for a quadratic extension \(K/\mathbb{Q}\), a \(K\)-relation \(\Theta=\sum_i n_iH_i\), and an elliptic curve \(E/L\) semistable at primes above \(2,3\), \[ \prod_i (C_{E/F^{H_i}})^{n_i} \equiv \prod_{\tau \in \Irr_{\mathbb{Q}}(G)} \mathcal{C}_\Theta(\tau)^{u(E/L,\chi_\tau)} \pmod{N_{K/\mathbb{Q}}(K^\times)}, \] where \(\mathcal{C}_\Theta(\tau)\) is the regulator constant and \[ w(E/L,\chi_\tau)=(-1)^{u(E/L,\chi_\tau)} \] [2504.17962]. This theorem turns the Tamagawa-product expression into a product of regulator constants with exponents determined by twisted root numbers [2504.17962]. The paper then proves that, under semistability at \(2,3\), \[ \text{If the norm relations test predicts } \rk E/F>0,\text{ then there exists an irreducible orthogonal }\rho\text{ with } w(E/\mathbb{Q},\rho)=-1 \] [2504.17962]. Its conceptual conclusion is that \[ \boxed{\text{Norm-relations/Tamagawa test} \subseteq \text{Parity-conjecture method}.} \] Hence the Tamagawa-ratio criterion does not detect positive rank beyond what parity for twists already detects [2504.17962]. The local mechanism is expressed through a theorem asserting that for local data over \(\mathcal K\) there is a \(\mathbb{Q}D\)-module \(\mathcal V\) such that for all self-dual \(\tau\), \[ \frac{w(E/\mathcal K,\tau)}{w(\tau)^2} = (-1)^{\langle \tau,\mathcal V\rangle}, \] and for a quadratic \(K/\mathbb{Q}\) and a \(K\)-relation \(\Theta\), \[ \mathcal C_\Theta(\mathcal V) \equiv \prod_i C_v(H_i)^{n_i} \pmod{N_{K/\mathbb{Q}}(K^\times)} \] [2504.17962]. This provides the bridge from local Tamagawa factors to root-number exponents. The same paper also compares Tamagawa products with BSD quotients. Assuming parity for twists and finiteness of \(\Sha\), it shows \[ \prod_i \BSD(E/F^{H_i})^{n_i}\in N_{K/\mathbb Q}(K^\times), \] and in particular \[ \prod_i (C_{E/F^{H_i}})^{n_i} \equiv \prod_i (\Reg_{E/F^{H_i}})^{n_i} \pmod{N_{K/\mathbb Q}(K^\times)} \] [2504.17962]. This identifies the Tamagawa part and regulator part modulo norms under parity hypotheses. ## 4. Local Tamagawa numbers, global products, and variation phenomena A large part of the literature studies not a ratio in the strict sense, but the local or global Tamagawa factors that underlie such ratios. For elliptic curves over \(\mathbb Q\) in short Weierstrass form \[ E(a_4,a_6): y^2=x^3+a_4x+a_6, \] the local Tamagawa number is \[ c_p(E):=[E(\mathbb Q_p):E_0(\mathbb Q_p)], \] and the global Tamagawa product is \[ \operatorname{Tam}(E):=\prod_p c_p(E) \] [2105.03513]. The paper “Tamagawa products of elliptic curves over \(\mathbb{Q}\)” constructs the Dirichlet series \[ L_{\mathrm{Tam}}(s)=\sum_{m=1}^\infty \frac{P_{\mathrm{Tam}}(m)}{m^s}, \] where \(P_{\mathrm{Tam}}(m)\) is the limiting proportion of curves with \(\operatorname{Tam}(E)=m\) [2105.03513]. It proves \[ P_{\mathrm{Tam}}(1)=0.5053\ldots \] and \[ L_{\mathrm{Tam}}(-1)=1.8193\ldots, \] so the average Tamagawa product is \(1.8193\ldots\) [2105.03513]. Over a general number field \(K\), the analogous product is \[ \operatorname{Tam}(E/K)=\prod_{\mathfrak p} c_{\mathfrak p}, \] and the associated Tamagawa \(L\)-series is \[ L_{\mathrm{Tam}(K,s)}=\sum_{m=1}^{\infty}\frac{P_{\mathrm{Tam}(K,m)}}{m^s} \] [2108.13625]. The paper proves \[ P_{\mathrm{Tam}(K,1)}=\prod_{\mathfrak p}\delta_{K,\mathfrak p}(1) \] and \[ L_{\mathrm{Tam}(K,-1)}=\prod_{\mathfrak p}\left(\sum_{c\ge1} c\,\delta_{K,\mathfrak p}(c)\right) \] [2108.13625]. It also shows that there exist sequences of number fields for which \(P_{\mathrm{Tam}(K,1)\!}\to 0\) and \(L_{\mathrm{Tam}(K,-1)}\to\infty\), as well as sequences for which both tend to \(1\) [2108.13625]. This suggests that the statistics of Tamagawa products are highly sensitive to local splitting, especially at \(2\) and \(3\). For Jacobians of semistable hyperelliptic curves, the relevant local factor is \[ c_{Jac(X)/K}:=\#\Phi_{Jac(X)/K}(k), \] written \(c_{X/K}\) in the paper [1808.05479]. A crucial theorem identifies the component group with the Jacobian of the dual graph, equivariantly for Frobenius, giving \[ c_{X/K}=c_{\Gamma,\mathrm{Frob}} \] [1808.05479]. For semistable hyperelliptic curves this is refined via the BY tree \(T=(T,S)\), and one has \[ c_{X/K}=c_{T,\epsilon F}:=\#\Phi_T^{\epsilon F} \] [1808.05479]. The paper’s strongest ratio-type statement concerns base change. If \(L/K\) is a finite extension with ramification degree \(e\) and residue degree \(f\), then \[ c_{X/L} = \prod_{d\mid f}\left(a_d\cdot e^{r_d}\cdot\gcd(e,2)^{s_d}\right)^{\varphi(d)} \] for constants \((a_d,r_d,s_d)\) with \((a_d,r_d,s_d)=(1,0,0)\) for almost all \(d\) [1808.05479]. In particular, for an unramified extension \(K_q/K\) of prime degree \(q\), \[ c_{X/K_q}/c_{X/K} \] is a \((q-1)\)st power [1808.05479]. The paper stresses that this is special to hyperelliptic curves and fails for general semistable curves, giving a non-hyperelliptic genus \(5\) counterexample with \[ c_{X/K}=1 \quad\text{but}\quad c_{X/K_5}=121 \] [1808.05479]. Recent work also gives a four-factor product formula for Jacobians over a discrete valuation field with perfect residue field: \[ |\phi_J(k)|= \underbrace{\prod_{v\in \VG} m_v^{\deg v - 2}}_{\text{unipotent part}} \cdot \underbrace{\det\!\bigl(\langle\cdot,\cdot\rangle|_{H_1(\DG,\mathbb Z)}\bigr)}_{\text{toric part}} \cdot \underbrace{\frac{\prod_{e\in \EG} r_e}{\prod_{v\in \VG} r_v}}_{\text{arithmetic part}} \cdot \underbrace{\gcd_{v\in \VG}(r_vm_v)\,|\Psi_J|}_{\text{cohomological part}} \] [2606.06713]. Although not cast as a Tamagawa ratio, this factorization makes explicit how local geometry, descent, and cohomology combine to produce the Tamagawa number of the Jacobian. ## 5. Tori, groups, and cohomological ratio formulas For algebraic tori and linear algebraic groups, Tamagawa ratios appear as genuine cohomological quotients computing Tamagawa numbers. For a torus \(T\) over a global field \(k\), the Tamagawa number \(\tau_k(T)\) is defined adelically, but Ono’s formula expresses it as \[ \tau_k (T) = \frac{| H^1 ( \Gamma , X(T) ) |}{| \Sha^2(\Gamma,X(T))|}, \] or equivalently \[ \tau_k (T) = \frac{| H^1 ( \Gamma , X(T) ) |}{| \Sha^1(\Gamma,T)|} \] [2109.04121]. The numerator is global Galois cohomology of the character lattice, and the denominator is a local-global defect group. In this setting the Tamagawa number is itself a cohomological “ratio.” For CM tori \(T=T^{K,\mathbb Q}\), one has the explicit formula \[ \tau(T)=\frac{2^r}{n_K}, \qquad n_K=[A^\times:N(T(A))\cdot \mathbb Q^\times], \] where \(r\) is the number of CM factors in the CM algebra \(K=\prod_{i=1}^r K_i\) [2109.04121]. The same paper proves that for any integer \(n\), there exists a CM torus \(T\) over \(\mathbb Q\) such that \[ \tau(T)=2^n \] [2109.04121]. Thus every positive or negative power of \(2\) occurs as the Tamagawa number of a CM torus. A related study of norm-condition tori considers \[ T(k)=\{x\in K^\times : N_{K/K^+}(x)\in k^\times\} \] for extensions \(k\subset K^+\subset K\) with \(K/K^+\) fixed by a central subgroup of prime order \(p\) [2009.04431]. There the Tamagawa number satisfies \[ \tau(T)=\frac{|\widehat H^1(k,\mathbf X^\star(T))|}{|\Sha^2(\mathbf X^\star(T))|}, \] and in the Galois case the numerator is explicitly \[ \widehat H^1(G,\mathbf X^\star(T))= \begin{cases} 0,& G_p \text{ cyclic},\\ \mathbb Z/p\mathbb Z,& G_p \text{ noncyclic}. \end{cases} \] [2009.04431]. The paper proves general bounds \[ \frac{p}{|G^{\mathrm{ab}}[p]|}\le \tau(T)\le p \] and shows, for the auxiliary torus \(T_1\), that \[ \tau(T_1)=p \] [2009.04431]. For quasi-split semisimple simply connected groups over function fields, the Tamagawa number is defined as the volume of the adelic quotient \[ T(G)=\operatorname{vol}(G(F)\backslash G(\mathbb A)) \] with respect to the Tamagawa measure [2309.10076]. The main theorem is \[ T(G)=1 \] for quasi-split semisimple simply connected groups over the function field of a smooth projective curve over \(\mathbb F_q\), \(q\neq 2\) [2309.10076]. The proof reduces the group Tamagawa number to that of a maximal torus \(A\), with the comparison mediated by Eisenstein series and the pole of the longest intertwining operator [2309.10076]. While this setting does not use the phrase “Tamagawa ratio” explicitly, the argument is organized around a ratio-like comparison between global volume and torus volume. ## 6. BSD, Bloch–Kato, and broader arithmetic interpretations In BSD-type formulas for elliptic curves, the ratio \[ \frac{c(E)}{|E(\mathbb Q)_{\mathrm{tors}}|} \] is a standard global arithmetic factor [2505.20479], and the related divisibility problem \[ p\mid c(E/K) \] for curves with rational \(p\)-torsion has been studied systematically [2202.06235]. For a global field \(K\) and an elliptic curve \(E/K\), the local factor is \[ c_v(E/K):=[E(K_v):E_0(K_v)], \] and the global Tamagawa number is \[ c(E/K):=\prod_v c_v(E/K) \] [2202.06235]. The paper proves that for every number field \(K/\mathbb Q\) there exists a constant \(n_K\) such that for every prime \(p\ge 7\) and every elliptic curve \(E/K\) with a \(K\)-rational point of order \(p\), one has \(p\mid c(E/K)\) with at most \(n_K\) exceptions [2202.06235]. Over \(\mathbb F_q(t)\), if \(q\) is a power of \(p\ge5\) and \(E/\mathbb F_q(t)\) is non-isotrivial with an \(\mathbb F_q(t)\)-rational point of order \(p\), then \[ p\mid c(E/\mathbb F_q(t)) \] [2202.06235]. For modular forms at Eisenstein primes, the rank-\(0\) Tamagawa number formula takes the form \[ \ord_p\!\left(\frac{L(f,r)}{\Omega_f}\right) = \ord_p\!\left(\#\Sha_{\Nek}(f/\mathbf Q)\cdot \Tam(A_f/\mathbf Q)\right), \] where \[ \Tam(f/\mathbf Q)=\prod_{\ell\mid N} c_\ell(A_f/\mathbf Q) \] [2410.24193]. In the anticyclotomic rank-\(1\) setting, the paper proves a control formula \[ \#\mathcal O/f^\Sigma_{ac}(0) = \frac{\#\Sha_\BK(f/K)\cdot C^\Sigma(A_f)}{(\#\H^0(K,A_f))^2} (\#\delta_v)^2, \] which is explicitly described there as the closest analogue of a Tamagawa ratio [2410.24193]. For motives, the Bloch–Kato Tamagawa number conjecture again presents Tamagawa numbers as ratios between arithmetic and analytic invariants. One paper recalls the formula \[ \operatorname{Tam}(M) = \frac{\#H^0(\mathbb Q,\mathcal O^*\otimes\mathbb Q/\mathbb Z(1))}{\#\,\Sha(M)} \] and proves that for a pure motive \(M\), \[ \lim_{B\to\infty} \frac{\#\{x\in B(\mathbb Q)\mid H_{o,d}(x)\le B\}} {\mu\bigl(\{x\in \prod_p B(\mathbb Q_p)\mid H_{o,d}(x)