---
title: Geometric Tevelev Degrees
url: https://www.emergentmind.com/topics/geometric-tevelev-degrees
type: topic
---

# Geometric Tevelev Degrees

Searching arXiv for recent papers on geometric Tevelev degrees and related tropical/logarithmic variants.
Geometric Tevelev degrees are fixed-domain enumerative invariants attached to maps from a general pointed smooth curve to a target variety under the maximal number of point or incidence constraints compatible with a finite count. In the basic formulation, one studies the forgetful/evaluation morphism
\[
\tau:\overline M_{g,n}(X,\beta)\to \overline M_{g,n}\times X^n
\]
for a nonsingular projective target \(X\) of dimension \(r\), genus \(g\), \(n\) markings, and curve class \(\beta\), under the expected-dimension condition
\[
\int_\beta c_1(T_X)=r(n+g-1).
\]
When every irreducible component dominating \(\overline M_{g,n}\times X^n\) is generically smooth of expected dimension, the geometric Tevelev degree is the degree of \(\tau\), equivalently the number of maps from a general \(n\)-pointed genus-\(g\) curve through \(n\) general target points [2110.05520][2308.00046].

## 1. Fixed-domain definition and conceptual scope

The defining feature of a geometric Tevelev degree is that the source curve has fixed complex structure. One fixes a general pointed smooth curve \((C,p_1,\dots,p_n)\), chooses general target conditions, and counts morphisms \(f:C\to X\) in class \(\beta\) satisfying those conditions. The count is expected to be finite precisely when
\[
\int_\beta c_1(T_X)=r(n+g-1),
\]
which is the equality between the virtual dimension of \(\overline M_{g,n}(X,\beta)\) and the dimension of \(\overline M_{g,n}\times X^n\) [2110.05520][2112.14824].

This notion is distinct from the virtual Tevelev degree. The virtual invariant is defined by pushing forward the virtual fundamental class,
\[
\tau_*[\overline M_{g,n}(X,\beta)]^{\mathrm{vir}}
=
\mathrm{vTev}^X_{g,n,\beta}\,[\overline M_{g,n}\times X^n],
\]
whereas the geometric degree requires actual finiteness and transversality on a general fiber [2110.05520][2112.14824]. A persistent theme in the literature is that virtual Tevelev degrees often admit very simple formulas, but geometric and virtual counts need not agree; discrepancies come from boundary stable maps, contracted components, rational tails, or other non-enumerative contributions [2308.00046][2110.05520].

Geometric Tevelev degrees count maps, not embedded image curves modulo reparametrization. If two distinct morphisms from the fixed source have the same image, they are counted separately. This map-counting perspective is fundamental in the fixed-domain interpolation problem and aligns Tevelev degrees with Brill–Noether theory, linear series, Hurwitz theory, and the geometry of evaluation maps on moduli spaces [2110.05520][2308.00046].

## 2. Hurwitz-theoretic origins and the \(\mathbb{P}^1\) case

The original Tevelev degree arises from Hurwitz moduli. For degree-\(d\) simply branched covers \(\pi:C\to \mathbb P^1\) of genus \(g\) with \(n\) marked unramified points, the admissible-cover compactification \(\overline{\mathcal H}_{g,d,n}\) carries a natural morphism
\[
\tau_{g,d,n}:\overline{\mathcal H}_{g,d,n}\to \overline{\mathcal M}_{g,n}\times \overline{\mathcal M}_{0,n}.
\]
In Tevelev’s original situation,
\[
d=g+1,\qquad n=g+3,
\]
this map is finite, and the degree
\[
\mathrm{Tev}_g=\frac{\deg(\tau_{g,g+1,g+3})}{(4g)!}
\]
counts degree-\(g+1\) maps from a general \((g+3)\)-pointed genus-\(g\) curve to a general \((g+3)\)-pointed rational curve, up to the ordering of the \(4g\) simple ramification points. The resulting formula is
\[
\mathrm{Tev}_g=2^g
\]
[2103.14055].

The Hurwitz-space interpretation extends to the two-parameter family
\[
d[g,\ell]=g+1+\ell,\qquad n[g,\ell]=g+3+2\ell.
\]
For these parameters the source and target of the corresponding Hurwitz forgetful map still have equal dimension, and the generalized degrees \(\mathrm{Tev}_{g,\ell,r}\) count covers in which \(r\) distinguished marked points lie in one common fiber. The central recursion is
\[
\mathrm{Tev}_{g,\ell,r}
=
\mathrm{Tev}_{g-1,\ell,\max(1,r-1)}
+
\mathrm{Tev}_{g-1,\ell+1,r+1},
\]
obtained by excess intersection on boundary strata of admissible-cover spaces. In the classical sector \((\ell,r)=(0,1)\) this specializes to \(2^g\); in other regimes it interpolates with projective geometry of lines and with Castelnuovo’s count of pencils [2103.14055].

For \(\mathbb P^1\), the theory admits a further geometric generalization allowing arbitrary ramification profiles over marked target points and equal-image constraints among marked source points. If \(\mu^h=(e_{h,1},\dots,e_{h,r_h})\) are ramification profiles over \(k\) marked target points, with
\[
d=g+1+\ell,\qquad g+3+2\ell=\sum_{h=1}^k |\mu^h|,
\]
the generalized degree
\[
\operatorname{Tev}_{g,\ell,\mu^1,\dots,\mu^k}
=
\frac{\deg(\tau_{g,\ell,\mu^1,\dots,\mu^k})}{b!}
\]
counts admissible covers with prescribed grouped ramification, where
\[
b=2g+2d-2-\sum_{h=1}^k |\mu^h|+n.
\]
A basic structural simplification is that the degree depends on each \(\mu^h\) only through its total size \(|\mu^h|\). The theory admits both a genus recursion and explicit closed formulas, and it can also be expressed by Schubert-calculus formulas on \(\operatorname{Gr}(2,d+1)\) via limit linear series [2111.05880].

## 3. Projective-space geometric Tevelev degrees

For \(X=\mathbb P^r\), the fixed-domain point-interpolation problem is especially rigid. If \((C,p_1,\dots,p_n)\) is a general pointed smooth curve of genus \(g\), and \(x_1,\dots,x_n\in \mathbb P^r\) are general points, the geometric Tevelev degree
\[
\operatorname{Tev}^{\mathbb P^r}_{g,n,d}
\]
is the number of degree-\(d\) maps \(f:C\to \mathbb P^r\) with \(f(p_i)=x_i\), under the balance
\[
rn=(r+1)(d+1)-1-rg,
\qquad\text{equivalently}\qquad
n=\frac{r+1}{r}d-g+1
\]
[2308.00046][2508.19503].

The projective-space case is now known in all degrees allowed by this dimension constraint. The main formula is an explicit Grassmannian intersection:
\[
\operatorname{Tev}^{\mathbb P^r}_{g,n,d}
=
\int_{\mathrm{Gr}(r+1,d+1)}
\sigma_1^{\,g}\cdot
\left(
\sum_{\mu\subset (n-r-2)^r}\sigma_\mu\,\sigma_{\widetilde\mu}
\right)_{\lambda_0\le n-r-1},
\]
obtained by degenerating the fixed source curve and replacing naïve map spaces by spaces of complete collineations. This resolves excess intersections coming from linear dependence and base-point phenomena and yields a complete geometric answer for point conditions in \(\mathbb P^r\), as well as a complete answer in \(\mathbb P^2\) with arbitrary point and line conditions [2308.00046].

The projective-space theory sharply distinguishes geometric from virtual counts. For \(\mathbb P^r\), the virtual answer is
\[
\mathrm{vTev}^{\mathbb P^r}_{g,n,d}=(r+1)^g
\]
when the dimension constraint holds, but geometric and virtual degrees differ outside the large-degree regime. In \(\mathbb P^2\), the first non-large-degree case \((g,n,d)=(2,5,4)\) has
\[
\operatorname{Tev}^{\mathbb P^2}_{2,5,4}=4,
\qquad
\mathrm{vTev}^{\mathbb P^2}_{2,5,4}=9,
\]
and for \((g,n,d)=(4,6,6)\) one gets
\[
\operatorname{Tev}^{\mathbb P^2}_{4,6,6}=30
\]
[2308.00046].

A later combinatorial reformulation uses the RSK correspondence. Under
\[
n=\frac{r+1}{r}d-g+1\ge r+1,
\]
\(\operatorname{Tev}^{\mathbb P^r}_{g,n,d}\) equals the number of \((r+1)\)-ary words of length \(g\) satisfying three subsequence conditions: there exist at least \(g+r-d\) disjoint decreasing subsequences of length \(r+1\); there is no nondecreasing subsequence of length \(>d/r\); and there is no \((i,i+1)\)-subsequence of length \(d/r\). This converts the Schubert formula into a manifestly positive word count and explains why the large-degree regime collapses to
\[
\operatorname{Tev}^{\mathbb P^r}_{g,n,d}=(r+1)^g
\]
[2508.19503].

## 4. Virtual theory, enumerativity, and logarithmic variants

Virtual Tevelev degrees admit a uniform quantum-cohomological expression. For a nonsingular projective target \(X\), point class \([\mathrm{pt}]\), and quantum Euler class
\[
e_q(X)=\sum_j \gamma_j^\vee\star\gamma_j,
\]
the virtual invariant satisfies
\[
vTev^X_{g,d,n}
=
\operatorname{Coeff}\!\left([\mathrm{pt}]^{\star n}\star e_q(X)^{\star g},q^d\right).
\]
This reduction to small quantum cohomology yields exact formulas for projective spaces, quadrics, cominuscule homogeneous spaces, and low-degree complete intersections [2112.14824].

Enumerativity is a separate issue. A central asymptotic result proves that for fixed genus and sufficiently positive curve class, virtual Tevelev degrees are enumerative for all homogeneous varieties and for hypersurfaces of sufficiently low degree compared to the dimension. The mechanism is twofold: smooth-domain maps in a general fiber become unobstructed and transverse, while boundary strata with rational tails fail to dominate the incidence target. In genus \(0\), stronger non-asymptotic statements hold in the same classes [2110.05520].

For hypersurfaces, there is also a direct geometric computation. If \(X\subset \mathbb P^{r+1}\) is a smooth hypersurface of degree \(e\ge 3\), with
\[
r>(e+1)(e-2),
\qquad
n=\frac{r+2-e}{r}d-g+1,
\]
then
\[
\operatorname{Tev}^X_{g,d,n}
=
((e-1)!)^n\cdot (r+2-e)^g \cdot e^{(d-n)e-g+1}
\]
for all \(d\) if \(g=0\), and for sufficiently large \(d\) if \(g>0\). This reproves, by projective geometry, the asymptotic geometric count previously obtained by combining virtual formulas with enumerativity results [2203.08171].

Logarithmic variants enlarge the theory to toric targets with tangency conditions. For a toric variety \(X\), discrete data \(\Gamma=(g,n,m,c)\), and the logarithmic moduli space \(\mathcal M_\Gamma(X)\), one studies
\[
\tau:\mathcal M(\Gamma)\to \overline{\mathcal M}_{g,n}\times X^n.
\]
In genus \(0\), under the fixed-domain dimension condition
\[
m=r(n-1),
\]
the virtual logarithmic Tevelev degree equals the actual number of logarithmic stable maps from a general fixed source curve, provided boundary and toric-fixed-point degeneracies are excluded. For Hirzebruch surfaces \(\mathbb F_a\), the genus-\(0\) tropical and logarithmic counts coincide and admit explicit closed formulas in the contact partitions \(\mu_1,\dots,\mu_4\) [2309.14947].

For Fano complete intersections of dimension \(r\ge 3\), the quantum Euler class has been computed explicitly, leading to formulas for all virtual Tevelev degrees in classes \(kL\) satisfying
\[
kd=r(n+g-1).
\]
That work does not prove general geometric enumerativity, but it gives the full virtual answer and an algorithm reducing the remaining computation to genus-\(0\) two-point hyperplane invariants [2204.01151].

## 5. Tropical Tevelev degrees and correspondence theorems

The tropical counterpart is defined by replacing admissible covers with tropical admissible covers and replacing Deligne–Mumford spaces with tropical moduli spaces. In the classical Hurwitz case
\[
d=g+1,\qquad n=g+3,
\]
one has a finite tropical forgetful map
\[
F_s\times F_t:
\overline{\mathcal H}^{trop}_{g,d,n}\to M^{trop}_{g,n}\times M^{trop}_{0,n},
\]
and the tropical Tevelev degree is
\[
Tev_g^{trop}=\deg(F_s\times F_t).
\]
Because the simple branch ends are already unmarked on the tropical side, there is no \((4g)!\) denominator. The main tropical theorem is
\[
Tev_g^{trop}=Tev_g=2^g,
\]
proved by a correspondence theorem via Berkovich skeletons together with a direct combinatorial count of tropical admissible covers [2407.20025].

The tropical computation is highly explicit. One chooses a special point in
\[
M^{trop}_{g,n}\times M^{trop}_{0,n}
\]
whose source curve has a chain of \(g\) loops and whose target is a caterpillar-like tree. Each loop contributes one of exactly two allowed local fragments, \(U\) or \(D\), altering the active-path degree by \(\pm1\). The genus part is therefore encoded by lattice paths that never cross below \(y=1\), and the marked-tree part is reconstructed by cut-and-join operations along the active path. A binomial recursion counts the genus possibilities, and every resulting tropical cover has multiplicity \(1\), so the degree is literally the number of covers [2407.20025].

This tropical framework has recently been generalized by introducing the parameter \(\ell\), so that
\[
d=g+1+\ell,\qquad n=g+3+2\ell.
\]
For positive \(\ell\), the tropical degree remains unchanged:
\[
\operatorname{Tev}^{trop}_{g,\ell}=2^g.
\]
For negative \(\ell\), one obtains an explicit defect formula:
\[
\operatorname{Tev}^{trop}_{g,\ell}
=
2^g
-
\sum_{i=0}^{-\ell-1}
\bigl(g-2i+1\bigr)
\left(
{g\choose i}-{g\choose i-1}
\right),
\qquad g\ge -2\ell.
\]
The same paper also defines generalized tropical Tevelev degrees with ramification profiles \(\mu_1,\dots,\mu_k\), giving closed formulas that tropicalize the generalized algebraic degrees studied for \(\mathbb P^1\) [2601.21973].

A recurring conceptual point is that tropical Tevelev degrees are not merely analogous to the algebraic ones. In the classical case \(Tev_g^{trop}=Tev_g\) is proved by tropicalization of moduli spaces and comparison of local analytic and tropical degrees, and the later generalized work is explicitly framed as tropicalizing the algebraic theories of Cela–Pandharipande–Schmitt and Cela–Lian [2407.20025][2601.21973].

## 6. Vanishing phenomena, exceptional targets, and adjacent degree theories

Geometric Tevelev degrees are not uniformly positive. A sharp vanishing result holds for Hirzebruch surfaces
\[
\mathcal H_a=\mathbf P_{\mathbf P^1}(\mathcal O(-a)\oplus \mathcal O),
\qquad a\ge 2.
\]
If the expected-dimension condition
\[
\beta\cdot K_{\mathcal H_a}^\vee = 2(n+g-1)
\]
holds and \(n\ge 2\) (or \(g=0\) and \(n\ge 3\)), then no irreducible component of \(\overline M_{g,n}(\mathcal H_a,\beta)\) dominates \(\overline M_{g,n}\times \mathcal H_a^n\), except in the single case
\[
a=2,\qquad \beta=m[C_0].
\]
Hence, outside that case,
\[
\mathsf{Tev}^{\mathcal H_a}_{g,n,\beta}=0.
\]
The proof uses semistability of \(f^*T_X\) for general counted maps as a necessary condition for interpolation [2503.20677].

The exceptional case reduces exactly to \(\mathbb P^1\). For
\[
n=2m-g+1\ge 3,
\]
one has
\[
\mathsf{Tev}^{\mathcal H_2}_{g,n,m[C_0]}
=
\mathsf{Tev}^{\mathbf P^1}_{g,n,m}.
\]
Geometrically, a map \(u:C\to \mathcal H_2\) in class \(m[C_0]\) is equivalent to a degree-\(m\) pencil \(g:C\to\mathbb P^1\) together with a section of \(M^{\otimes 2}\), and the marked-point conditions determine that section uniquely. Thus the only nonzero Hirzebruch-surface counts for \(a\ge 2\) are inherited from the already-computed geometric Tevelev degrees of \(\mathbb P^1\) [2503.20677].

A broader but related multidegree theory is provided by Kapranov degrees on \(\overline M_{0,n}\). These are top intersections of divisor classes
\[
X_{S,i}=f_S^*\psi_i,
\]
equivalently multidegrees of the product of forgetful and Kapranov maps. They include the Castravet–Tevelev case as a special situation, and their positivity is characterized by the Cerberus condition
\[
\left|\bigcup_{j\in J} S_j\right|\ge |J|+3
\quad\text{for all nonempty }J\subseteq [n-3].
\]
In the specialization \(i_j=n\) for all \(j\), the Kapranov degree is either \(0\) or \(1\), recovering the Castravet–Tevelev phenomenon inside a larger multidegree formalism [2308.12285].

Taken together, these developments show that geometric Tevelev degrees form a moduli-theoretic family of fixed-domain interpolation invariants with several distinct faces: Hurwitz-theoretic for \(\mathbb P^1\), Schubert-theoretic for projective space, quantum-cohomological on the virtual side, logarithmic for toric targets with tangency, and tropical through correspondence theorems. The subject is marked by two complementary facts: in some regimes the answers are unexpectedly rigid, such as \(2^g\) or \((r+1)^g\); in others, boundary contributions, semistability obstructions, or ramification constraints force genuine deviations from the virtual or large-degree picture [2308.00046][2110.05520].

Source: https://www.emergentmind.com/topics/geometric-tevelev-degrees