Conifold Transition in Calabi–Yau Threefolds

• Conifold Transition is a topology-changing process in Calabi–Yau threefolds that replaces contracted rational curves with vanishing S³’s.
• It interweaves analytical, algebraic, and tropical methods to connect birational geometry, Hodge theory, and mirror symmetry.
• The mechanism underpins metric continuity via explicit Calabi–Yau metrics and has significant implications for string compactifications.

Conifold transition is the standard topology-changing process in complex dimension three that passes through a nodal intermediate space: one starts with a smooth threefold, contracts finitely many rational curves with normal bundle $$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$ to ordinary double points, and then smooths the resulting singularities, thereby exchanging exceptional $S^2$’s for vanishing $S^3$’s. In the Calabi–Yau setting, it lies at the intersection of birational geometry, symplectic topology, Hodge theory, mirror symmetry, and string compactification, and it is a central test case for the Gross–Siebert program, SYZ duality, and flux-vacuum dynamics (Castano-Bernard et al., 2013, Collins, 31 Aug 2025).

1. Local analytic model

A conifold singularity on a complex threefold is an isolated hypersurface singularity analytically equivalent to either

$x^2+y^2+z^2+w^2=0$

or

$z_1z_2-z_3z_4=0$

in $\mathbb{C}^4$ (Castano-Bernard et al., 2013). The affine threefold

$X_0=\{z_1z_2-z_3z_4=0\}$

is the local conifold. A smoothing deforms the equation to

$x^2+y^2+z^2+w^2=\epsilon,\qquad \epsilon\neq 0,$

or equivalently

$Y_\epsilon=\{z_1z_2-z_3z_4=\epsilon\},$

while a small resolution replaces the node by an exceptional rational curve with normal bundle $$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$ (Castano-Bernard et al., 2013, Lee et al., 2015).

One standard small resolution is

$S^2$0

with projection $S^2$1; the other small resolution interchanges $S^2$2 and $S^2$3 (Castano-Bernard et al., 2013). On the smoothing side, the vanishing cycle is a Lagrangian $S^2$4-sphere. In the local model one can write

$S^2$5

and in the differential-geometric normalization of the deformed conifold one has

$S^2$6

with $S^2$7 special Lagrangian for an appropriate Calabi–Yau structure (Castano-Bernard et al., 2013, Collins, 31 Aug 2025).

Topologically, the transition is a surgery replacing a neighborhood $S^2$8 of the vanishing sphere by a neighborhood $S^2$9 of the exceptional curve; equivalently, the unit normal sphere bundle of the exceptional $S^3$0 is diffeomorphic to $S^3$1 (Lee et al., 2015). This local exchange of $S^3$2 and $S^3$3 is the basic geometric content of every conifold transition.

2. Global transition, topology, and deformation-theoretic obstructions

A projective conifold transition $S^3$4 consists of a projective degeneration of a smooth projective threefold $S^3$5 to a singular variety $S^3$6 with ordinary double points, together with a projective small resolution $S^3$7 (Lee et al., 2015). On the smoothing side, the nodes correspond to vanishing spheres $S^3$8; on the resolution side, they correspond to exceptional curves $S^3$9 with $x^2+y^2+z^2+w^2=0$0 (Lee et al., 2015).

The classical global existence problem is governed by homological “good relations.” Friedman’s criterion states that $x^2+y^2+z^2+w^2=0$1 can be smoothed if there is a choice of small resolution for which the exceptional classes satisfy

$x^2+y^2+z^2+w^2=0$2

while the Smith–Thomas–Yau criterion gives the symplectic dual statement: a symplectic small resolution exists if the vanishing cycles satisfy

$x^2+y^2+z^2+w^2=0$3

(Castano-Bernard et al., 2013). For projective conifold transitions between Calabi–Yau threefolds, if

$x^2+y^2+z^2+w^2=0$4

then

$x^2+y^2+z^2+w^2=0$5

where $x^2+y^2+z^2+w^2=0$6 is the number of nodes (Lee et al., 2015). In the notation of small resolution $x^2+y^2+z^2+w^2=0$7 and smoothing $x^2+y^2+z^2+w^2=0$8, if $x^2+y^2+z^2+w^2=0$9 is the rank of the subgroup of $z_1z_2-z_3z_4=0$0 spanned by exceptional curves and $z_1z_2-z_3z_4=0$1 the rank of the subgroup of $z_1z_2-z_3z_4=0$2 spanned by vanishing spheres, then

$z_1z_2-z_3z_4=0$3

and

$z_1z_2-z_3z_4=0$4

(Castano-Bernard et al., 2013).

The Hodge-theoretic mechanism behind these formulas is encoded in the Picard–Lefschetz transformation

$z_1z_2-z_3z_4=0$5

with nilpotent logarithm $z_1z_2-z_3z_4=0$6 satisfying $z_1z_2-z_3z_4=0$7 (Lee et al., 2015). The same paper identifies a weight-$z_1z_2-z_3z_4=0$8 exact sequence

$z_1z_2-z_3z_4=0$9

where $\mathbb{C}^4$0 is the $\mathbb{C}^4$1-vector space of vanishing $\mathbb{C}^4$2-cycles; this exact sequence expresses the exchange between the “excess A-model” and “excess B-model” data across the transition (Lee et al., 2015).

A differential-geometric proof of the necessity direction of Friedman’s theorem was given recently using the explicit local map

$\mathbb{C}^4$3

between the singular and smoothed local conifolds, together with the expansion of the holomorphic volume form and a current identity

$\mathbb{C}^4$4

on the small resolution (Collins, 31 Aug 2025). This places the obstruction theory directly in the asymptotics of the local Calabi–Yau geometry.

3. Tropical, affine, and mirror-symmetric formulations

In the Gross–Siebert program, mirror symmetry is encoded by a discrete Legendre duality between tropical manifolds $\mathbb{C}^4$5 and $\mathbb{C}^4$6, where $\mathbb{C}^4$7 is an integral affine manifold with singularities, $\mathbb{C}^4$8 is a toric polyhedral decomposition, and $\mathbb{C}^4$9 is a strictly convex multivalued piecewise linear function (Castano-Bernard et al., 2013). A tropical conifold is the three-dimensional version in which the discriminant graph $X_0=\{z_1z_2-z_3z_4=0\}$0 is allowed to have $X_0=\{z_1z_2-z_3z_4=0\}$1-valent vertices, called tropical nodes, of two local types: positive and negative. These are exchanged by Legendre duality (Castano-Bernard et al., 2013).

Locally, tropical resolution and tropical smoothing are dual operations. Resolving a positive node corresponds to subdividing an edge in $X_0=\{z_1z_2-z_3z_4=0\}$2; on the mirror side this becomes a smoothing of a negative node by subdividing a square face $X_0=\{z_1z_2-z_3z_4=0\}$3 via a diagonal. Resolving a negative node corresponds dually to smoothing a positive node by adding a vertex and a fan structure that separates the two lines of the discriminant (Castano-Bernard et al., 2013). This gives a precise affine-geometric form of the slogan that the mirror of a resolution is a smoothing, and the mirror of a smoothing is a resolution.

The main mechanism for global compatibility is the tropical $X_0=\{z_1z_2-z_3z_4=0\}$4-cycle. A tropical $X_0=\{z_1z_2-z_3z_4=0\}$5-cycle is data $X_0=\{z_1z_2-z_3z_4=0\}$6 consisting of a compact tropical domain $X_0=\{z_1z_2-z_3z_4=0\}$7, an embedding

$X_0=\{z_1z_2-z_3z_4=0\}$8

and a primitive integral parallel vector field $X_0=\{z_1z_2-z_3z_4=0\}$9 satisfying balancing and monodromy-invariance conditions (Castano-Bernard et al., 2013). The main theorem states that if $x^2+y^2+z^2+w^2=\epsilon,\qquad \epsilon\neq 0,$0 contains nodes $x^2+y^2+z^2+w^2=\epsilon,\qquad \epsilon\neq 0,$1, then the corresponding vanishing cycles in a smoothing of $x^2+y^2+z^2+w^2=\epsilon,\qquad \epsilon\neq 0,$2 satisfy a good relation in $x^2+y^2+z^2+w^2=\epsilon,\qquad \epsilon\neq 0,$3, and the exceptional curves in a small resolution of the mirror $x^2+y^2+z^2+w^2=\epsilon,\qquad \epsilon\neq 0,$4 satisfy a good relation in $x^2+y^2+z^2+w^2=\epsilon,\qquad \epsilon\neq 0,$5; thus the obstructions to smoothing one side and resolving the mirror vanish simultaneously (Castano-Bernard et al., 2013).

This tropical picture has constructive content. The paper proves explicit simultaneous resolutions and smoothings for several classes of node configurations, including positive nodes at the corners of planar polygons and negative nodes lying on boundary faces of embedded products $x^2+y^2+z^2+w^2=\epsilon,\qquad \epsilon\neq 0,$6 (Castano-Bernard et al., 2013). In the worked $x^2+y^2+z^2+w^2=\epsilon,\qquad \epsilon\neq 0,$7-sphere example with $x^2+y^2+z^2+w^2=\epsilon,\qquad \epsilon\neq 0,$8 nodes, one side smooths to a threefold homeomorphic to a complete intersection in $x^2+y^2+z^2+w^2=\epsilon,\qquad \epsilon\neq 0,$9 of degrees $Y_\epsilon=\{z_1z_2-z_3z_4=\epsilon\},$0, $Y_\epsilon=\{z_1z_2-z_3z_4=\epsilon\},$1, $Y_\epsilon=\{z_1z_2-z_3z_4=\epsilon\},$2, while the mirror side yields a Schoen Calabi–Yau after smoothing all $Y_\epsilon=\{z_1z_2-z_3z_4=\epsilon\},$3 nodes (Castano-Bernard et al., 2013).

The same local mirror relation appears in SYZ. For smoothings of toric Gorenstein singularities arising from Minkowski decompositions, the SYZ mirror is

$Y_\epsilon=\{z_1z_2-z_3z_4=\epsilon\},$4

and this is obtained from the toric resolution mirror by analytic continuation and specialization of complex moduli (Lau, 2013). In the non-archimedean family Floer formulation of the conifold, the singular SYZ fibers match codimension-$Y_\epsilon=\{z_1z_2-z_3z_4=\epsilon\},$5 “missing points” in the mirror cluster variety, confirming the smoothing–resolution mirror relation in an explicit affinoid torus fibration (Yuan, 2022).

4. Metric geometry and analytic continuity

The local analytic geometry on both sides of the transition is governed by explicit asymptotically conical Calabi–Yau metrics. On the smoothing

$Y_\epsilon=\{z_1z_2-z_3z_4=\epsilon\},$6

an $Y_\epsilon=\{z_1z_2-z_3z_4=\epsilon\},$7-invariant Kähler potential $Y_\epsilon=\{z_1z_2-z_3z_4=\epsilon\},$8 with $Y_\epsilon=\{z_1z_2-z_3z_4=\epsilon\},$9 solves the Calabi–Yau equation, giving the Candelas–de la Ossa/Stenzel metric $$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$0; at $$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$1 the cone metric is

$$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$2

(Collins, 31 Aug 2025). On the small resolution

$$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$3

one obtains the resolved conifold metric in the class

$$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$4

again asymptotic to the same cone (Collins, 31 Aug 2025). The two local geometries converge to the conifold in the Gromov–Hausdorff sense as $$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$5 and $$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$6, respectively (Collins, 31 Aug 2025).

On compact smoothings and resolutions, Fu–Li–Yau constructed balanced metrics that glue the local conifold models to the global geometry, and Collins–Picard–Yau proved that for sufficiently small $$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$7 the tangent bundle $$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$8 admits a Hermitian–Yang–Mills metric with respect to these conformally balanced metrics (Collins et al., 2021). Near each vanishing cycle, the HYM metric is asymptotic to a multiple of the local Candelas–de la Ossa metric, so the tangent bundle geometry itself has a controlled conifold limit (Collins et al., 2021).

This analytic framework leads to a metric continuity theorem. Using balanced metrics and Hermitian–Yang–Mills metrics to geometrize the transition, it was shown that the whole operation is continuous in the Gromov–Hausdorff topology:

$$\mathcal{O}_{\mathbb{P}^1}(-1)\oplus \mathcal{O}_{\mathbb{P}^1}(-1)$$9

with the singular space $S^2$00 appearing as the common metric limit (Friedman et al., 2024). This result is specific to the non-Kähler Calabi–Yau setting of balanced/HYM geometry and extends earlier continuity statements that were purely Kähler.

5. Hodge theory, enumerative geometry, and transfer of structures

Conifold transitions do not merely change topology; they reorganize the A-model and B-model. For projective conifold transitions between Calabi–Yau threefolds $S^2$01 and $S^2$02, the all-genus Gromov–Witten theory of $S^2$03 embeds into that of $S^2$04, while the variation of Hodge structures of $S^2$05 embeds into the limiting mixed Hodge structures of the degeneration of $S^2$06 (Lee et al., 2015). More strongly, the combined information of the A-model and B-model on one side, linked along vanishing cycles or exceptional curves, determines the corresponding combined information on the other side (Lee et al., 2015).

The B-model side is controlled by the nilpotent-orbit form of the period map. If $S^2$07 are the conifold periods, then

$S^2$08

and the Yukawa couplings have simple poles

$S^2$09

along the discriminant hyperplanes (Lee et al., 2015). On the A-model side, the extremal corrections from the exceptional curves are encoded by the multiple-cover contribution

$S^2$10

and the Dubrovin connection has residue blocks determined by the relation matrix of the vanishing cycles (Lee et al., 2015).

A concrete enumerative application is the transfer of the descendent GW/PT correspondence across a projective conifold transition. If the correspondence holds for the small resolution $S^2$11, then it also holds for the smoothing $S^2$12 with stationary descendents, via two compatible degeneration formulas and local toric calculations on the exceptional pieces (Lin et al., 2023). This gives a method for propagating MNOP-type statements from toric or complete-intersection geometries to new smoothings.

The local open-string counterpart fits the same pattern. For smoothings of toric Gorenstein singularities, the generating function of open Gromov–Witten invariants is related to that of the toric resolution across the local conifold transition, and the SYZ mirror of the smoothing is recovered from wall-crossing factors determined by the Minkowski decomposition of the defining polytope (Lau, 2013). This suggests that the transition acts as a controlled passage between large-volume and conifold-limit mirror charts rather than as a breakdown of mirror geometry.

6. Variants, exceptions, and physical realizations

Several nearby notions clarify the scope of the term. Not every small geometric transition is a deformation of a conifold transition: the Namikawa cuspidal fiber product gives a “small” geometric transition whose local singularities are cusps of type $S^2$13, and the local deformation directions that split the cusp into nodes do not arise from the relevant global deformations, so the transition is not “simple” in the sense of being deformation-equivalent to a conifold transition (Rossi, 2012). Hyperconifold transitions are another distinct class: here the singularity is a finite cyclic quotient of the conifold, and the resulting transition changes the fundamental group; in the compact $S^2$14 and $S^2$15 examples constructed torically, the mirrors are ordinary conifold transitions with exactly $S^2$16 and $S^2$17 nodes, respectively (Davies, 2011).

Higher-dimensional and arithmetic generalizations also exist. In F-theory, the split/non-split transition of elliptic fibers is, except for the $S^2$18 family, a conifold transition from the resolved to the deformed side, with conifold nodes appearing at codimension-two enhancements to $S^2$19 or $S^2$20 (Kuramochi et al., 2021). In M-theory on Calabi–Yau fourfolds, the local geometry is generically a genus-$S^2$21 curve of conifold singularities, and the transition is intertwined with $S^2$22-flux quantization, Abel–Jacobi conditions, and three-dimensional $S^2$23 gauge dynamics near the intersection of Coulomb and Higgs branches (Intriligator et al., 2012). Recent work has also defined log conifold transitions for index-two Fano threefold pairs, where the geometry of the boundary del Pezzo surface forces unobstructedness and removes the global balancing conditions familiar from the Calabi–Yau case (Aguilar, 30 Jun 2026).

In string theory, conifold points are physically nontrivial rather than pathological. In Type IIB compactifications, a D3-brane wrapped on the shrinking $S^2$24 becomes massless; this is the mechanism behind the smoothing of the conifold singularity in the physical moduli space. Flux-vacuum tunneling trajectories between monodromy-related vacua were found to “graze the conifold point,” and the tunneling only occurs when warping from flux through the shrinking cycle is included (Ahlqvist et al., 2010). In heterotic compactifications, conifold transitions can be accompanied by transitions of NS5-branes and slope-stable bundles, with a coupled small-instanton description relating the cotangent bundles of the deformation and resolution branches and providing a geometric explanation of $S^2$25 target-space duality (Anderson et al., 2022). Non-perturbative topological string theory also admits geometric transitions from local $S^2$26 to the resolved conifold, with non-perturbatively corrected transition loci for the Kähler parameters (Sugimoto, 2016).

These developments show that the conifold transition is not merely a local surgery but a structural operation linking birational models, Hodge-theoretic limits, affine-mirror dualities, balanced/HYM metric geometry, and string dualities. Its basic exchange of exceptional $S^2$27 and vanishing $S^2$28 remains the organizing principle, but the modern literature makes clear that the transition is best understood simultaneously in analytic, tropical, categorical, and physical terms (Castano-Bernard et al., 2013, Lee et al., 2015).