---
title: Mukai's Syzygies Conjecture Explained
url: https://www.emergentmind.com/topics/mukai-s-syzygies-conjecture
type: topic
---

# Mukai's Syzygies Conjecture Explained

Mukai’s Syzygies Conjecture is a syzygy-theoretic program that appears in two adjacent senses in the literature summarized here. In the adjoint-line-bundle form, the conjecture attributed to Mukai states that if \(X\) is smooth projective, \(L\) is ample, and \(m \ge \dim X + 2 + k\), then \(K_X + mL\) satisfies \((N_k)\) [2509.06629]. In the canonical-curve form, Mukai’s work on canonical curves and homogeneous varieties suggests that the syzygies of a general canonical curve should admit an explicit geometric description through special linear sections of homogeneous varieties, especially Grassmannians and related constructions; later work formulates this as a refinement of Green/Mukai-type syzygy statements for canonical curves [1907.07553]. Taken together, these strands indicate that higher syzygies are expected to reflect intrinsic geometry with substantial rigidity.

## 1. Terminological scope and core formulations

In the adjoint setting, the conjectural statement is explicit: if \(X\) is smooth projective, \(L\) is ample, and \(m \ge \dim X + 2 + k\), then \(K_X + mL\) satisfies \((N_k)\) [2509.06629]. This is a strong generalization of Fujita-type adjoint positivity conjectures, and it predicts not only projective normality or quadratic generation, but linearity of the first \(k\) steps of the minimal free resolution. The same source emphasizes that the conjecture remains widely open in general, and that even for surfaces the \(k=0\) case is not fully known in general [2509.06629].

In the canonical-curve literature, the emphasis shifts from adjoint bundles on arbitrary varieties to the geometry of the extremal linear strand. Mukai’s ideas connect canonical curves to linear sections of homogeneous varieties and to explicit geometric models for their moduli, while Green-style conjectures predict vanishing and nonvanishing patterns in the linear strand. Kemeny’s theorem in even genus is presented as a refined geometric version of Mukai/Green-style syzygy statements: not merely a vanishing theorem, but a statement that the last linear syzygy space is generated by syzygies of concrete geometric origin [1907.07553].

| Context | Central syzygy statement | Representative source |
|---|---|---|
| Adjoint bundles | \(m \ge \dim X + 2 + k \Rightarrow K_X+mL\) satisfies \((N_k)\) | [2509.06629] |
| Canonical curves | Extremal linear syzygies should be geometrically generated by scrolls, Grassmannians, or related models | [1907.07553] |

This suggests that “Mukai’s Syzygies Conjecture” is best understood not as a single isolated sentence, but as a coherent expectation that syzygies should be governed by geometry in a particularly explicit way.

## 2. Algebraic framework: \((N_k)\), Koszul cohomology, and linear strands

The adjoint formulation is expressed through the minimal graded free resolution of the section ring
\[
R(L)=\bigoplus_{n\ge 0} H^0(Y,nL)
\]
over
\[
S_L=\operatorname{Sym} H^0(Y,L).
\]
If
\[
0 \to E_d(L)\to \cdots \to E_1(L)\to E_0(L)\to R(L)\to 0
\]
is the minimal graded free resolution, then \(L\) satisfies \((N_k)\) if
\[
E_0(L)=S_L,\qquad E_i(L)=\bigoplus S_L(-(i+1)) \quad \text{for }1\le i\le k.
\]
Accordingly, \((N_0)\) is projective normality, \((N_1)\) is projective normality plus generation of the ideal by quadrics, and higher \((N_k)\) encode linearity of the first \(k\) syzygy modules [2509.06629].

A standard cohomological criterion uses the kernel bundle
\[
0\to M_L \to H^0(Y,L)\otimes \mathcal O_Y \to L \to 0.
\]
If
\[
H^1\!\left(Y, M_L^{\otimes(i+1)}\otimes L^h\right)=0 \quad\text{for all }0\le i\le k,\ h\ge 1,
\]
then \(L\) satisfies \((N_k)\) [2509.06629]. For quotients of abelian varieties, this criterion is combined with \(IT(0)\), \(GV\), Castelnuovo–Mumford regularity, and the structure of the abelian cover [2509.06629].

The broader syzygy formalism is Koszul cohomology. For a projective scheme \(X\), a globally generated line bundle \(L\), and a coherent sheaf \(B\), the module \(R(X,B;L)\) has a minimal free resolution whose summands are controlled by
\[
E_p=\bigoplus_q K_{p,q}(X,B;L)\otimes_k S(-p-q),
\]
where \(K_{p,q}(X,B;L)\) is the cohomology of the standard Koszul-type complex [2405.18022]. In particular, for canonical curves the groups \(K_{p,1}(C,\omega_C)\) are the linear syzygies, and their dimensions are the linear-strand Betti numbers [1907.07553].

## 3. Canonical curves, rank, and Mukai’s geometric philosophy

A canonical curve is a smooth non-hyperelliptic curve embedded by the canonical linear system,
\[
C \hookrightarrow \mathbb P(H^0(C,\omega_C)^\vee),
\]
and for a genus \(g\) curve this embedding lies in \(\mathbb P^{g-1}\) [1907.07553]. The syzygies of the canonical ideal are measured by the groups \(K_{p,1}(C,\omega_C)\), and the linear strand carries a notion of rank: a class
\[
a\in K_{p,1}(X,L)
\]
has rank equal to the dimension of the smallest linear subspace \(V\subset H^0(X,L)\) such that
\[
a\in K_{p,1}(X,L,V).
\]
A nondegenerate variety satisfies \(\operatorname{rank}(a)\ge p+1\); rank \(p+1\) syzygies give rational normal scrolls, rank \(p+2\) syzygies arise from linear sections of Grassmannians, and a syzygy is called geometric when \(\operatorname{rank}(a)\le p+2\) [1907.07553].

For canonical curves of even genus \(g=2k\), Green’s conjecture predicts
\[
K_{k-1,1}(C,\omega_C)\neq 0,\qquad K_{k,1}(C,\omega_C)=0,
\]
so \(K_{k-1,1}\) is the highest linear syzygy group [1907.07553]. In this range, the Mukai program is not only about the existence of nonzero syzygies but about explaining them geometrically. Mukai’s work suggests that the syzygies of general canonical curves should be describable through special linear sections of homogeneous varieties, especially Grassmannians and related constructions [1907.07553].

This philosophy sharpens Green’s classical result. For a non-hyperelliptic curve \(C\) of genus \(g\ge 5\), Green proved that
\[
K_{1,1}(C,\omega_C)
\]
is generated by rank two syzygies coming from line bundles \(L\in W^1_d(C)\); under the identification
\[
K_{1,1}(C,\omega_C)\cong (I_C)_2,
\]
this means that the quadrics containing the canonical curve are generated by quadrics of rank four [1907.07553]. The later even-genus results extend this pattern from the first linear syzygy group to the last one.

## 4. The geometric syzygy theorem in even genus

The most explicit realization of the canonical-curve version is Kemeny’s theorem on the Geometric Syzygy Conjecture in even genus. The conjectural statement is that for a general curve \(C\) of genus \(g=2k\), the last linear syzygy space
\[
K_{k-1,1}(C,\omega_C)
\]
is spanned by geometric syzygies [1907.07553]. The theorem proves a stronger form: if \(C\) is a general canonical curve of genus \(g=2k\), then
\[
K_{k-1,1}(C,\omega_C)
\]
is generated by the rank \(k\) syzygies
\[
a \in K_{k-1,1}\!\left(C,\omega_C, H^0(\omega_C\otimes A^{-1})\right), \qquad A\in W^1_{k+1}(C).
\]
Thus the last linear syzygy group is generated by syzygies arising from minimal pencils \(A\) of degree \(k+1\), equivalently from the subspaces
\[
H^0(\omega_C\otimes A^{-1}) \subset H^0(\omega_C).
\]
The corresponding syzygy scheme is the rational normal scroll defined by the \(2\times 2\) minors of the Petri matrix
\[
H^0(A)\otimes H^0(\omega_C\otimes A^{-1}) \to H^0(\omega_C)
\]
[1907.07553].

This theorem is presented as an extension of Green’s theorem on rank-four quadrics. Green’s theorem concerns the first linear syzygy group \(K_{1,1}\); Kemeny’s theorem concerns the last linear syzygy group \(K_{k-1,1}\) for genus \(2k\). In the paper’s formulation,
\[
\text{rank 4 quadrics generating } K_{1,1} \quad \leadsto \quad \text{rank }k\text{ syzygies generating } K_{k-1,1}
\]
[1907.07553].

The proof passes through K3 geometry. For a K3 surface \(X\) with
\[
\operatorname{Pic}(X)=\mathbb Z[L],\qquad L^2=4k-2,
\]
let \(E\) be the Lazarsfeld–Mukai bundle on \(X\). The paper proves that the morphism
\[
\mathbb P(H^0(X,E)) \longrightarrow \mathbb P\big(K_{k-1,1}(X,L)\big),\qquad s\mapsto a(s),
\]
is the Veronese embedding of degree \(k-2\), hence there is a natural isomorphism
\[
\operatorname{Sym}^{k-2} H^0(X,E)\cong K_{k-1,1}(X,L)
\]
[1907.07553]. Rathmann later gave a substantially shorter proof of Voisin’s result for K3 surfaces of even sectional genus by constructing geometric Koszul complexes on Grassmann varieties; in the even case \(L^2=4k-2\), the paper proves
\[
K_{k-2,2}(X,L)=0
\]
and again identifies
\[
K_{k-1,1}(X,L)\cong \operatorname{Sym}^{k-2}H^0(E)
\]
[2205.00266].

## 5. Related constructions: scrolls, projections, symmetric products, and K3 reconstruction

A major refinement of the Mukai–Schreyer picture is the assertion that extremal linear syzygies actually come from the scroll attached to a minimal pencil. For a general \(k\)-gonal curve \(C\) of genus \(g\ge 2k-1\), \(k\ge 4\), and a line bundle \(L\) with
\[
\deg(L)\ge 2g+k,
\]
the extremal group
\[
K_{r(L)-k,1}(C,L)
\]
has the scroll-predicted dimension \(r(L)-k\), and the restriction map from the associated rational normal scroll \(X_L\) is an isomorphism:
\[
K_{r(L)-k,1}(X_L,\mathcal O_{X_L}(1)) \to K_{r(L)-k,1}(C,L).
\]
In this sense, the extremal syzygies of \(C\) are exactly those coming from the scroll [1811.01105]. The same paper proves that the syzygy scheme of a nonzero linear syzygy can be recovered from the syzygy schemes of its pointwise projections:
\[
\mathrm{Syz}(a) = \bigcap_{x \in Z} \mathrm{Cone}_x\!\big(\mathrm{Syz}(\mathrm{pr}_x(a))\big)
\]
under explicit projective-normality and spanning hypotheses [1811.01105].

Another line of development uses symmetric products of curves. A survey of this method explains that weight-one syzygies can be recast on \(C_{p+1}\), and that vanishing of \(K_{p,1}\) reduces to
\[
H^1(C_{p+1}, M_{p+1,B}\otimes N_{p+1,L})=0.
\]
This framework yields the effective gonality theorem, Green’s \((2g+1+p)\)-theorem, results on secant varieties, and the tangent-developable approach to generic Green’s conjecture [2405.18022]. The survey does not isolate a theorem formally named “Mukai’s Syzygies Conjecture,” but it places Mukai’s philosophy squarely within the modern language of Koszul cohomology, \(N_p\)-properties, and intrinsic geometric control of Betti tables [2405.18022].

Mukai’s program also operates through rank-two Brill–Noether loci on curves lying on K3 surfaces. For \(g=2s+1\), the locus
\[
M_C(2,K_C,s) = \left\{ [F]\in M_C(2,K_C)\;:\; h^0(C,F)\ge s+2 \right\}
\]
has tangent space controlled by the Petri map
\[
\mu:S^2H^0(F)\longrightarrow H^0(S^2F),
\]
and in the rank-two K3 case the paper proves that the association
\[
x\in S \longmapsto E_x
\]
gives an isomorphism
\[
S \xrightarrow{\sim} M_C(2,K_C,s).
\]
Under the congruence assumption \(g\equiv 3 \pmod 4\), the resulting Brill–Noether K3 surface is then used in a Fourier–Mukai reconstruction of the original K3 surface [1309.0496]. This is not an adjoint \((N_k)\)-statement, but it is a precise realization of the claim that canonical syzygies, Petri geometry, and K3 geometry are manifestations of the same structure.

## 6. Confirmed cases, obstructions, and disambiguations

In the adjoint formulation, several cases are now known. For a bielliptic surface \(S\) over an algebraically closed field of characteristic \(\neq 2,3\), if \(L\) is ample and
\[
m \ge \max\{4,\,2k+2\},
\]
then
\[
K_S+mL \text{ satisfies } (N_k).
\]
For a smooth projective variety \(Y\) of dimension \(d\) in characteristic \(0\) with \(K_Y\equiv 0\), if \(P\) is ample and globally generated, \(N\) is numerically trivial, and \(N+P\) is globally generated, then
\[
N+mP \text{ satisfies } (N_k)\quad \text{if } m\ge \max\{d+1,\ k+1\}.
\]
For a hyperelliptic variety \(X=A/G\) of dimension \(d\) with \(\operatorname{char}(k)\nmid |G|\), if \(L\) is ample and \(N\) is numerically trivial, then
\[
N+mL \text{ satisfies } (N_k)\quad \text{if } m\ge \max\{2d+1,\ 2k+2\}.
\]
When \(G\) is commutative, \(\operatorname{char}(k)=0\), and \(h^1(X,\mathcal O_X)>0\), the paper improves this to
\[
N+mL \text{ satisfies } (N_1)\ \text{for } m\ge 2d,
\]
and
\[
N+mL \text{ satisfies } (N_k)\ \text{for all } k \text{ with } 2d>2k+2.
\]
These results confirm Mukai’s conjecture for bielliptic surfaces when \(k\le 2\), for hyperelliptic varieties in the range \(\dim X-1\le k\le \dim X\), and for complex Bagnera–de Franchis varieties in the range \(\dim X-2\le k\le \dim X\) [2509.06629].

The current bounds are not presented as final. The same paper explicitly states that the stronger statement
\[
mL \text{ satisfies } (N_k)\text{ for } m\ge k+3
\]
for general hyperelliptic varieties remains open, because a certain splitting of kernel bundles on the abelian cover does not hold [2509.06629].

At the level of canonical-curve heuristics, there are also structural obstructions to overly naive determinantal expectations. Eisenbud’s conjectural picture had suggested that if \(I_2\) is generated by quadrics of rank at most four, then the last nonvanishing linear syzygy should come from a determinantal construction. Counterexamples show that rank \(\le 4\) quadrics do not by themselves control the rank of the top linear syzygies: for any odd \(n\), there exists an arithmetically Cohen–Macaulay toric ideal generated by \(n\) quadrics of rank \(\le 4\), with only one linear first syzygy, of rank \(n\); analogous even-\(n\) Gorenstein examples also exist [1012.0933]. The same paper proves a modified theorem in which the rank of the top linear syzygy determines whether the ideal must contain \(2\times 2\) minors of a \(1\)-generic matrix or \(4\times 4\) Pfaffians of a skew-symmetric \(1\)-generic matrix [1012.0933]. This does not refute Mukai’s geometric philosophy; it refines it by showing that low-rank quadrics alone are insufficient without control of the rank of the corresponding syzygy.

A final disambiguation is necessary. The name “Mukai conjecture” is also used for numerical inequalities in Fano geometry, such as
\[
\dim X \ge \rho_X\,(i_X-1),
\]
with equality only in the projective-space case for the quiver-moduli setting [2310.15927], and for the total-index variant considered in effective non-vanishing theory [2306.08841]. Those conjectures are distinct from Mukai’s Syzygies Conjecture. The latter concerns minimal free resolutions, Koszul cohomology, and the geometric origin of syzygies; the Fano inequalities concern dimension, Picard number, and index.

Source: https://www.emergentmind.com/topics/mukai-s-syzygies-conjecture