Trace-Kernel Ideal in Algebraic Varieties

• Trace-Kernel Ideal is the canonical vector bundle defined as the kernel of the Frobenius trace map in smooth projective varieties over fields of positive characteristic.
• It provides a Frobenius-theoretic analog to classical theorems, with explicit decompositions that reveal key structural properties in cases such as projective spaces and Fano varieties.
• Its ampleness classification in low dimensions—being ample only for varieties like P¹, P², and certain Fano threefolds—offers actionable insights into birational and positivity theory.

The trace-kernel ideal, commonly termed the Frobenius-trace kernel, is a canonical vector bundle associated to a smooth projective variety $X$ over an algebraically closed field of positive characteristic $p > 0$. Explicitly, it is defined as the kernel of the Frobenius trace map $\Tr_F:F_*\O_X\to\O_X$, where $F=F_X:X\to X$ denotes the (absolute) Frobenius endomorphism. The study of positivity properties of this kernel yields deep links with the birational and positivity theory of algebraic varieties, providing a Frobenius-theoretic analog to classical theorems such as those of Mori and Hartshorne, and motivating the search for projective analogs of Kunz’s theorem (Carvajal-Rojas et al., 2021).

1. Algebraic Definition and Structure

Let $X$ be a smooth projective $k$-variety of dimension $d$ with $k$ algebraically closed and $\operatorname{char} k = p > 0$. The absolute Frobenius morphism induces an endomorphism $F=F_X:X \to X$. Utilizing Grothendieck duality (valid since $X$ is Gorenstein and $F$ is finite), there exists a canonical isomorphism of $F_*\O_X$-modules:

$F_*\O_X\simeq \cHom_{\O_X}(F_*\omega_X,\omega_X),$

where $\omega_X$ is the canonical bundle of $X$. The global Cartier operator $\kappa:F_*\omega_X \to \omega_X$ plays a central role, and, after twisting and application of the projection formula, induces the Frobenius trace:

$\Tr_F: F_*\O_X \xrightarrow{\cup\omega_X^{-1}} F_*(\omega_X\otimes\omega_X^{-1}) \xrightarrow{\kappa\otimes\omega_X^{-1}} \O_X.$

The trace $\Tr_F$ is surjective, yielding the Frobenius-trace kernel:

$\KX = \ker \left( \Tr_F : F_*\O_X \to \O_X \right),$

which is locally free of rank $p^d-1$. Two fundamental exact sequences are thereby established: $0 \to \KX \to F_*\O_X \xrightarrow{\Tr_F} \O_X \to 0,$ and its dual: $$0 \to \omega_X \to F_*\omega_X \xrightarrow{\kappa} \omega_X \to 0,$$ where duality is taken with respect to $\omega_X$.

2. Explicit Calculations and Positivity Phenomena

In the case of split projective bundles $P=\P(\E)\to Y$, where $\E$ splits into line bundles, the Frobenius pushforward and trace kernel admit explicit decompositions: $F_*\O_P(n)\simeq \bigoplus_{i=0}^{r-1} \bigoplus_{\substack{0\leq j_0,\dots,j_{r-1}\leq p-1\j_0+\dots+j_{r-1}\equiv n\;(\mathrm{mod}\;p)}} \O_P(\lfloor n/p\rfloor - i)\otimes\pi^*(\L_0^{j_0}\otimes\dots\otimes\L_{r-1}^{j_{r-1}}).$ For projective space $P = \P^d$, specializing $n=0$ yields: $F_*\O_{\P^d} \simeq \O_{\P^d} \oplus \bigoplus_{i=1}^{d} \O_{\P^d}(-i)^{\oplus a(i,0;d,p)},$ where the coefficients $a(i,n;d,p)$ are explicitly computable. This decomposes the trace-kernel as: $\K_{\P^d} \simeq \bigoplus_{i=1}^d \O_{\P^d}(-i)^{\oplus a(i,0;d,p)},$ implying ampleness of $\K_{\P^d}$, since $\O_{\P^d}(-i)$ is ample for $i >0$.

The following table summarizes key examples and properties as established in (Carvajal-Rojas et al., 2021):

Variety	Positivity of $\KX$	Structural Reason
$\P^d$	Ample	Decomposition into ample summands
Hirzebruch surfaces	Not ample	$\O(-1)$ summand present
Blow-up of variety	Not ample	Trivial summand upon pullback
Quadrics ($d\geq3$)	Ample if $p\neq2$	Decomposition via spinor bundles

3. Classification in Low Dimensions

The main results of Carvajal–Rojas and Patakfalvi (Carvajal-Rojas et al., 2021) provide a complete classification in dimensions 1, 2, and 3 regarding when $\KX$ is ample:

• Curves ($d=1$): $\KX$ is ample if and only if $X \cong \P^1$. For curves, this is equivalent to $\deg\KX>0$ occuring only for $\P^1$.
• Surfaces ($d=2$): Any blow-up negates ampleness. The only surface with $\KX$ ample is $\P^2$.
• Threefolds ($d=3$): Ampleness of $\KX$ forces $X$ to be Fano of Picard rank $1$. Conversely, for rank 1 Fano threefolds of index at least 2 (including projective space, quadrics with $p\ne 2$, and certain complete intersections), one checks directly that $\KX$ is ample.

The deductive strategy combines vanishing theorems, extremal contraction analysis (via Mori–Kawamata theory), and direct calculation to rule out exceptions and illuminate the geometric content behind the positivity of $\KX$.

4. Technical Tools and Vanishing Results

Several key techniques underpin these results:

• Cartier operator and duality: Identification of the Frobenius trace with the (global) Cartier operator $\kappa: F_*\omega_X \to \omega_X$ and the subsequent twist by $\omega_X^{-1}$ are fundamental to constructing the relevant sequences and interpreting their geometric meaning.
• Frobenius-splitting: The exact sequence

$0 \to \KX \to F_*\O_X \xrightarrow{\Tr_F} \O_X \to 0$

splits if and only if $X$ is $F$-split.

• Grothendieck–Lefschetz-style theorems and extremal rays: Analysis of how the presence of an extremal contraction, fibration, or exceptional divisor forces a trivial summand (and hence non-ampleness) in pullbacks or restrictions of $\KX$.
• Asymptotic invariants: For an $F$-split $X$ and ample line bundle $\L$, there is the equality

$\vol_X(\L) = \lim_{e\to\infty}\frac{h^0\bigl(X,\L\otimes\KX\bigr)}{p^{e\,d}/d!},$

establishing a relation between the positivity of the trace-kernel and the positivity invariants of line bundles.

5. Summary of Known Classification Results

The classification for ampleness of the trace-kernel is as follows:

• On $\P^d$, $\K_{\P^d}$ is ample and splits into ample line bundles.
• Among curves, only $\P^1$ has $\K_C$ ample.
• Among surfaces, only $\P^2$ has $\K_S$ ample.
• In dimension three, ampleness of $\KX$ characterizes $X$ as a rank 1 Fano threefold; for such $X$ with index at least 2, $\KX$ is ample except possibly for quadrics in characteristic 2.

These results illustrate a striking analogy with classical classification theorems and tie ampleness of the trace-kernel to strong birational and cohomological properties.

6. Open Problems and Future Directions

Several directions remain open, outlined as follows:

• Determining whether, for Fano threefolds of rank 1 and index 1, ampleness of $\KX$ persists. The answer appears to depend subtely on the characteristic $p$.
• For dimensions $d\geq4$, it is not yet resolved to what extent ampleness of $\KX$ forces $X$ to be isomorphic to $\P^d$ or another rank 1 Fano. Evidence points to a mixture of local $F$-signature bounds and global extremal-ray techniques as essential ingredients.
• Existence of a “projective-Kunz theorem” remains an open avenue—whether projective space $\P^d$ may be characterized purely by ampleness of a Frobenius-module such as the trace-kernel or higher Cartier-operator kernels.

These questions link the theory of the trace-kernel to deep conjectures in higher-dimensional algebraic geometry and the theory of singularities in positive characteristic (Carvajal-Rojas et al., 2021).