Birch Rank in Elliptic Curves and Forms

• Birch rank is a dual concept that defines the analytic and algebraic rank for elliptic curves under BSD and the codimension of the singular locus for forms.
• In elliptic curve theory, Birch rank plays a pivotal role in understanding regulator comparisons, twist families, and stability results in low analytic rank cases.
• For homogeneous forms, Birch rank quantifies the codimension of singularities, linking it to Schmidt rank and establishing criteria for complete intersections.

Searching arXiv for the cited works and closely related papers to ground the article in current arXiv records. “Birch rank” is used in two distinct but mathematically connected senses. In the arithmetic of elliptic curves, what one might colloquially call the Birch rank is the rank predicted by the Birch and Swinnerton–Dyer conjecture: the common value of the analytic rank $\operatorname{ord}_{s=1}L(E,s)$ and the algebraic rank $\operatorname{rank}E(\mathbb{Q})$. In the geometry of forms, the Birch rank of a homogeneous polynomial or system is the codimension of the singular locus cut out by the gradient or Jacobian rank-deficiency equations. The first meaning is central to the modern theory of elliptic curves and $L$-functions; the second is central to Birch’s work on forms in many variables and to later comparisons with Schmidt rank (Barrios et al., 1 Oct 2025, Lampert et al., 2022).

1. Birch rank in the Birch–Swinnerton-Dyer conjecture

For an elliptic curve $E/\mathbb{Q}$, the algebraic rank is the Mordell–Weil rank

$$r_{\mathrm{alg}}(E/\mathbb{Q})=\operatorname{rank}E(\mathbb{Q}),$$

while the analytic rank is

$$r_{\mathrm{an}}(E/\mathbb{Q})=\operatorname{ord}_{s=1}L(E,s).$$

The weak Birch–Swinnerton-Dyer conjecture predicts

$$\operatorname{ord}_{s=1}L(E,s)=\operatorname{rank}E(\mathbb{Q}).$$

In this sense, the Birch rank is simply the rank appearing in BSD.

The strong BSD formula identifies the leading Taylor coefficient at $s=1$ with arithmetic invariants. In the normalization used for elliptic curves over $\mathbb{Q}$,

$\frac{L^{(r)}(E,1)}{r!\,\Omega_E} = \frac{\#\Sha(E/\mathbb{Q})\;\prod_{\ell} c_\ell(E/\mathbb{Q})\;\operatorname{Reg}(E/\mathbb{Q})}{\#E(\mathbb{Q})_{\mathrm{tors}}^{\,2}},$

where $\operatorname{rank}E(\mathbb{Q})$0, $\operatorname{rank}E(\mathbb{Q})$1 is the Néron period, $\operatorname{rank}E(\mathbb{Q})$2 are Tamagawa numbers, $\operatorname{rank}E(\mathbb{Q})$3 is the Néron–Tate regulator, and $\operatorname{rank}E(\mathbb{Q})$4 is the Tate–Shafarevich group. In rank $\operatorname{rank}E(\mathbb{Q})$5, the regulator is the Néron–Tate height of a generator of the free part of $\operatorname{rank}E(\mathbb{Q})$6. In the analytic rank $\operatorname{rank}E(\mathbb{Q})$7 range, Gross–Zagier and Kolyvagin imply equality of analytic and algebraic rank and finiteness of $\operatorname{rank}E(\mathbb{Q})$8, so the main remaining issue is the precise leading-term identity (Jetchev et al., 2015).

2. Low-rank twist families and BSD modulo squares

A recent use of the elliptic-curve Birch rank studies how the BSD leading term behaves in quadratic twist families. Let $\operatorname{rank}E(\mathbb{Q})$9 have conductor $L$0 with $L$1 and $L$2 coprime and $L$3 squarefree, and let $L$4 be a fundamental discriminant satisfying the modified Heegner hypothesis: primes dividing $L$5 split and primes dividing $L$6 are inert in $L$7. For the quadratic twist $L$8, the comparison is carried out in $L$9, because regulators, periods, and $E/\mathbb{Q}$0 are naturally only controlled modulo squares.

The parity of the twist rank is governed by the root number. If $E/\mathbb{Q}$1 is a squarefree product of an odd number of primes, then $E/\mathbb{Q}$2, so $E/\mathbb{Q}$3 and $E/\mathbb{Q}$4 have opposite analytic-rank parity. If $E/\mathbb{Q}$5 is a squarefree product of an even number of primes, then $E/\mathbb{Q}$6, so the parities agree. This leads to two comparison regimes: rank patterns $E/\mathbb{Q}$7 or $E/\mathbb{Q}$8 in the odd-$E/\mathbb{Q}$9 case, and $$r_{\mathrm{alg}}(E/\mathbb{Q})=\operatorname{rank}E(\mathbb{Q}),$$0 or $$r_{\mathrm{alg}}(E/\mathbb{Q})=\operatorname{rank}E(\mathbb{Q}),$$1 in the even-$$r_{\mathrm{alg}}(E/\mathbb{Q})=\operatorname{rank}E(\mathbb{Q}),$$2 case.

The main conclusion is a stability statement for the BSD package in low rank. If $$r_{\mathrm{alg}}(E/\mathbb{Q})=\operatorname{rank}E(\mathbb{Q}),$$3 is semistable with analytic rank at most one, and $$r_{\mathrm{alg}}(E/\mathbb{Q})=\operatorname{rank}E(\mathbb{Q}),$$4 is a positive fundamental discriminant coprime to the conductor such that $$r_{\mathrm{alg}}(E/\mathbb{Q})=\operatorname{rank}E(\mathbb{Q}),$$5 again has analytic rank at most one, then the BSD formula modulo $$r_{\mathrm{alg}}(E/\mathbb{Q})=\operatorname{rank}E(\mathbb{Q}),$$6 holds for $$r_{\mathrm{alg}}(E/\mathbb{Q})=\operatorname{rank}E(\mathbb{Q}),$$7 if and only if it holds for $$r_{\mathrm{alg}}(E/\mathbb{Q})=\operatorname{rank}E(\mathbb{Q}),$$8. A central ingredient is a Gross–Zagier-type formula modulo squares for $$r_{\mathrm{alg}}(E/\mathbb{Q})=\operatorname{rank}E(\mathbb{Q}),$$9, with $$r_{\mathrm{an}}(E/\mathbb{Q})=\operatorname{ord}_{s=1}L(E,s).$$0, together with a delicate comparison of periods and Tamagawa numbers under twisting, especially when $$r_{\mathrm{an}}(E/\mathbb{Q})=\operatorname{ord}_{s=1}L(E,s).$$1 is even (Barrios et al., 1 Oct 2025).

3. $$r_{\mathrm{an}}(E/\mathbb{Q})=\operatorname{ord}_{s=1}L(E,s).$$2-adic and refined notions of Birch rank

In the $$r_{\mathrm{an}}(E/\mathbb{Q})=\operatorname{ord}_{s=1}L(E,s).$$3-adic setting, the relevant rank is not always the classical Mordell–Weil rank. For a $$r_{\mathrm{an}}(E/\mathbb{Q})=\operatorname{ord}_{s=1}L(E,s).$$4-ordinary elliptic curve $$r_{\mathrm{an}}(E/\mathbb{Q})=\operatorname{ord}_{s=1}L(E,s).$$5 over a number field, the multi-variable $$r_{\mathrm{an}}(E/\mathbb{Q})=\operatorname{ord}_{s=1}L(E,s).$$6-adic BSD conjecture introduces the extended rank

$$r_{\mathrm{an}}(E/\mathbb{Q})=\operatorname{ord}_{s=1}L(E,s).$$7

where $$r_{\mathrm{an}}(E/\mathbb{Q})=\operatorname{ord}_{s=1}L(E,s).$$8 and $$r_{\mathrm{an}}(E/\mathbb{Q})=\operatorname{ord}_{s=1}L(E,s).$$9 is the number of primes above $$\operatorname{ord}_{s=1}L(E,s)=\operatorname{rank}E(\mathbb{Q}).$$0 at which $$\operatorname{ord}_{s=1}L(E,s)=\operatorname{rank}E(\mathbb{Q}).$$1 has split multiplicative reduction. The extra term records exceptional zeros. The conjecture predicts that a multi-variable $$\operatorname{ord}_{s=1}L(E,s)=\operatorname{rank}E(\mathbb{Q}).$$2-adic $$\operatorname{ord}_{s=1}L(E,s)=\operatorname{rank}E(\mathbb{Q}).$$3-function $$\operatorname{ord}_{s=1}L(E,s)=\operatorname{rank}E(\mathbb{Q}).$$4 vanishes at the trivial character to order at least $$\operatorname{ord}_{s=1}L(E,s)=\operatorname{rank}E(\mathbb{Q}).$$5, and that its $$\operatorname{ord}_{s=1}L(E,s)=\operatorname{rank}E(\mathbb{Q}).$$6-th leading term is governed by an extended $$\operatorname{ord}_{s=1}L(E,s)=\operatorname{rank}E(\mathbb{Q}).$$7-adic regulator on an extended Mordell–Weil group. In low rank this is proved in several cases, including the genuinely multi-variable case where $$\operatorname{ord}_{s=1}L(E,s)=\operatorname{rank}E(\mathbb{Q}).$$8 and there are two exceptional zeros, so that a mixed third derivative appears (Disegni, 2016).

A different refinement arises in the conjectures of Mazur and Tate. Their Mazur–Tate elements $$\operatorname{ord}_{s=1}L(E,s)=\operatorname{rank}E(\mathbb{Q}).$$9, defined by modular symbols for finite abelian extensions $s=1$0, are conjectured to lie in augmentation-ideal powers measuring rank and split multiplicative contributions. Recent work proves the $s=1$1-part of a weak main conjecture for these elements, proves an order-of-vanishing statement of the form

$s=1$2

with $s=1$3 the $s=1$4-rank of the dual Selmer group, and establishes refined leading-term congruences via the rank-zero component of the equivariant Tamagawa Number Conjecture. This suggests that, in refined BSD settings, Birch rank is best understood as a Galois-equivariant or Selmer-theoretic order of vanishing rather than merely an integer-valued Mordell–Weil rank (Bullach et al., 10 Nov 2025).

4. Geometric Birch rank for forms

For a form $s=1$5 of degree $s=1$6, the Birch rank is defined geometrically by

$s=1$7

Equivalently, it is the codimension of the singular locus of the affine hypersurface $s=1$8. For a collection $s=1$9 of forms of common degree $\mathbb{Q}$0, if

$\mathbb{Q}$1

then

$\mathbb{Q}$2

Over an algebraically closed field of characteristic not dividing $\mathbb{Q}$3, Birch rank is essentially equivalent to Schmidt rank. One always has

$\mathbb{Q}$4

and Kazhdan–Lampert–Polishchuk give the reverse control

$\mathbb{Q}$5

Consequently, large Schmidt rank forces large Birch rank and, once the threshold $\mathbb{Q}$6 is crossed, the common zero set is a complete intersection of codimension $\mathbb{Q}$7. For admissible fields such as number fields, finite fields, and finite separable extensions of $\mathbb{Q}$8, polynomial bounds compare Schmidt rank over the ground field with Schmidt rank over the algebraic closure, and hence indirectly with geometric Birch rank. These comparisons transfer circle-method results about integer points, prime points, and rational points from Birch-rank hypotheses to Schmidt-rank hypotheses (Lampert et al., 2022).

5. Computational, heuristic, and statistical viewpoints

One computational interpretation of Birch rank is the analytic rank itself. Under BSD and GRH, explicit-formula methods give sharp upper bounds for $\mathbb{Q}$9. A particularly useful test function is

$\frac{L^{(r)}(E,1)}{r!\,\Omega_E} = \frac{\#\Sha(E/\mathbb{Q})\;\prod_{\ell} c_\ell(E/\mathbb{Q})\;\operatorname{Reg}(E/\mathbb{Q})}{\#E(\mathbb{Q})_{\mathrm{tors}}^{\,2}},$0

whose Fourier transform is compactly supported, truncating the prime-power side of the explicit formula. This method gives exact upper bounds for curves known to have rank at least $\frac{L^{(r)}(E,1)}{r!\,\Omega_E} = \frac{\#\Sha(E/\mathbb{Q})\;\prod_{\ell} c_\ell(E/\mathbb{Q})\;\operatorname{Reg}(E/\mathbb{Q})}{\#E(\mathbb{Q})_{\mathrm{tors}}^{\,2}},$1 and $\frac{L^{(r)}(E,1)}{r!\,\Omega_E} = \frac{\#\Sha(E/\mathbb{Q})\;\prod_{\ell} c_\ell(E/\mathbb{Q})\;\operatorname{Reg}(E/\mathbb{Q})}{\#E(\mathbb{Q})_{\mathrm{tors}}^{\,2}},$2, and for the known curve of rank at least $\frac{L^{(r)}(E,1)}{r!\,\Omega_E} = \frac{\#\Sha(E/\mathbb{Q})\;\prod_{\ell} c_\ell(E/\mathbb{Q})\;\operatorname{Reg}(E/\mathbb{Q})}{\#E(\mathbb{Q})_{\mathrm{tors}}^{\,2}},$3 it gives a bound of $\frac{L^{(r)}(E,1)}{r!\,\Omega_E} = \frac{\#\Sha(E/\mathbb{Q})\;\prod_{\ell} c_\ell(E/\mathbb{Q})\;\operatorname{Reg}(E/\mathbb{Q})}{\#E(\mathbb{Q})_{\mathrm{tors}}^{\,2}},$4 (Bober, 2011). A complementary GRH-based approach proves that Elkies’ curve has Mordell–Weil rank $\frac{L^{(r)}(E,1)}{r!\,\Omega_E} = \frac{\#\Sha(E/\mathbb{Q})\;\prod_{\ell} c_\ell(E/\mathbb{Q})\;\operatorname{Reg}(E/\mathbb{Q})}{\#E(\mathbb{Q})_{\mathrm{tors}}^{\,2}},$5 and analytic rank at most $\frac{L^{(r)}(E,1)}{r!\,\Omega_E} = \frac{\#\Sha(E/\mathbb{Q})\;\prod_{\ell} c_\ell(E/\mathbb{Q})\;\operatorname{Reg}(E/\mathbb{Q})}{\#E(\mathbb{Q})_{\mathrm{tors}}^{\,2}},$6, and similarly treats a curve of rank $\frac{L^{(r)}(E,1)}{r!\,\Omega_E} = \frac{\#\Sha(E/\mathbb{Q})\;\prod_{\ell} c_\ell(E/\mathbb{Q})\;\operatorname{Reg}(E/\mathbb{Q})}{\#E(\mathbb{Q})_{\mathrm{tors}}^{\,2}},$7 and the classical examples of ranks $\frac{L^{(r)}(E,1)}{r!\,\Omega_E} = \frac{\#\Sha(E/\mathbb{Q})\;\prod_{\ell} c_\ell(E/\mathbb{Q})\;\operatorname{Reg}(E/\mathbb{Q})}{\#E(\mathbb{Q})_{\mathrm{tors}}^{\,2}},$8 through $\frac{L^{(r)}(E,1)}{r!\,\Omega_E} = \frac{\#\Sha(E/\mathbb{Q})\;\prod_{\ell} c_\ell(E/\mathbb{Q})\;\operatorname{Reg}(E/\mathbb{Q})}{\#E(\mathbb{Q})_{\mathrm{tors}}^{\,2}},$9 (Klagsbrun et al., 2016).

A heuristic use of Birch rank starts from Birch and Swinnerton-Dyer’s original product

$\operatorname{rank}E(\mathbb{Q})$00

together with Sato–Tate statistics. Matching the asymptotic $\operatorname{rank}E(\mathbb{Q})$01 against the Sato–Tate expectation for $\operatorname{rank}E(\mathbb{Q})$02 leads to an expected Mordell–Weil rank of $\operatorname{rank}E(\mathbb{Q})$03 for non-CM curves, and also $\operatorname{rank}E(\mathbb{Q})$04 for CM curves over fields not containing the CM field, while the expected rank is $\operatorname{rank}E(\mathbb{Q})$05 for CM curves over fields containing the CM field (Thakur, 2022).

Large-scale data analysis gives a further perspective. A study of $\operatorname{rank}E(\mathbb{Q})$06 isomorphism classes of elliptic curves with conductor $\operatorname{rank}E(\mathbb{Q})$07 treats rank as a statistical variable alongside conductor, period, Tamagawa product, regulator, torsion size, and $\operatorname{rank}E(\mathbb{Q})$08. In that dataset, $\operatorname{rank}E(\mathbb{Q})$09, with only a single example of rank $\operatorname{rank}E(\mathbb{Q})$10. Gradient-boosted trees show that rank is difficult to infer from Weierstrass coefficients alone but is strongly correlated with the other BSD invariants, and the normalized right-hand side of the BSD formula exhibits a Beta-type distribution after rescaling (Alessandretti et al., 2019).

6. Established cases, explicit families, and broader scope

A large body of work concerns cases where BSD is known in low rank, prime by prime or family by family. For semistable elliptic curves over $\operatorname{rank}E(\mathbb{Q})$11 of analytic rank one, the $\operatorname{rank}E(\mathbb{Q})$12-part of the BSD formula is proved for each prime $\operatorname{rank}E(\mathbb{Q})$13 of good reduction such that $\operatorname{rank}E(\mathbb{Q})$14 is irreducible, and for $\operatorname{rank}E(\mathbb{Q})$15 under an additional supersingular condition (Jetchev et al., 2015). This extends to primes of multiplicative reduction in the semistable case, again in analytic rank one (Castella, 2017). In the CM setting, for analytic rank one and any odd potentially good ordinary prime $\operatorname{rank}E(\mathbb{Q})$16 with $\operatorname{rank}E(\mathbb{Q})$17, the $\operatorname{rank}E(\mathbb{Q})$18-primary part of $\operatorname{rank}E(\mathbb{Q})$19 is shown to have the order predicted by BSD (Li et al., 2016).

Other results make Birch rank explicit in families. For the CM family

$\operatorname{rank}E(\mathbb{Q})$20

with $\operatorname{rank}E(\mathbb{Q})$21, the rank is $\operatorname{rank}E(\mathbb{Q})$22 or $\operatorname{rank}E(\mathbb{Q})$23, and under BSD one has

$\operatorname{rank}E(\mathbb{Q})$24

where $\operatorname{rank}E(\mathbb{Q})$25 is defined by a specific recurrence formula (Nomoto, 2021). A generalized Birch lemma produces explicit infinite families of quadratic twists $\operatorname{rank}E(\mathbb{Q})$26 and $\operatorname{rank}E(\mathbb{Q})$27 with analytic and algebraic ranks $\operatorname{rank}E(\mathbb{Q})$28 and $\operatorname{rank}E(\mathbb{Q})$29, respectively, and also proves the $\operatorname{rank}E(\mathbb{Q})$30-part of BSD for those families when the original curve satisfies it (Shu et al., 2021). For a large class of elliptic curves over $\operatorname{rank}E(\mathbb{Q})$31, there are also explicit infinite families of quadratic twists of analytic rank $\operatorname{rank}E(\mathbb{Q})$32 for which the $\operatorname{rank}E(\mathbb{Q})$33-part of BSD is established; these results were subsequently used by Xin Wan to prove the full BSD conjecture for some explicit infinite families of non-CM elliptic curves (Cai et al., 2017).

At a global level, a majority of elliptic curves over $\operatorname{rank}E(\mathbb{Q})$34, when ordered by height, satisfy the BSD rank conjecture: at least $\operatorname{rank}E(\mathbb{Q})$35 have algebraic and analytic rank $\operatorname{rank}E(\mathbb{Q})$36 or $\operatorname{rank}E(\mathbb{Q})$37, with at least $\operatorname{rank}E(\mathbb{Q})$38 of curves having both ranks equal to $\operatorname{rank}E(\mathbb{Q})$39 and at least $\operatorname{rank}E(\mathbb{Q})$40 having both ranks equal to $\operatorname{rank}E(\mathbb{Q})$41 (Bhargava et al., 2014). For individual curves of analytic rank $\operatorname{rank}E(\mathbb{Q})$42 or $\operatorname{rank}E(\mathbb{Q})$43, there is an algorithm to prove the full BSD conjectural formula, and with computer assistance it was applied to $\operatorname{rank}E(\mathbb{Q})$44 of the $\operatorname{rank}E(\mathbb{Q})$45 such curves of conductor less than $\operatorname{rank}E(\mathbb{Q})$46 (Miller, 2010). Taken together, these results suggest a bifurcated picture: as an elliptic-curve invariant, Birch rank is most tractable in analytic rank $\operatorname{rank}E(\mathbb{Q})$47 and $\operatorname{rank}E(\mathbb{Q})$48, while as a geometric invariant of forms it is a codimension theory intimately tied to singular loci, Schmidt rank, and arithmetic applications.