From taming to compatibility in four dimensions

← All digests

Family 342 · Differential geometry
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-26

1. The question

Complex geometry begins with an operation on tangent vectors that behaves like multiplication by $i$. Symplectic geometry begins with a closed, nondegenerate two-form that measures oriented area. These structures can cooperate in two different ways: the area form may be positive on complex lines, or it may also respect the complex structure exactly.

The first relation is called taming; the stronger relation is compatibility. Donaldson asked whether, on a closed four-manifold, the existence of a taming symplectic form guarantees the existence of a compatible one.

Family 342 claims an affirmative answer. The almost-complex structure is held fixed. The compatible form is allowed to change its cohomology class, an essential qualification: the manuscript explicitly gives an example where some taming classes have no compatible representative.

The difficulty is easy to state. Averaging a taming form with its complex transform makes it compatible pointwise, but generally destroys the differential equation saying that the form is closed. The problem is to achieve both conditions simultaneously across the entire manifold.

2. Background and essential definitions

Two structures on a four-manifold

Let $X$ be a smooth real four-manifold. An almost-complex structure is a smooth linear map $J$ on each tangent space satisfying

$$J^2=-I.$$

This is the algebra of a quarter-turn. It need not arise from complex coordinates; when it does, $J$ is called integrable. No integrability is assumed in the main theorem.

A real two-form $\omega$ is symplectic if it is nondegenerate and closed, with $d\omega=0$. It tames $J$ if

$$\omega(v,Jv)>0\qquad(v\ne0).$$

Thus every complex line has positive oriented area. Compatibility additionally requires

$$\omega(Ju,Jv)=\omega(u,v).$$

A compatible pair defines a Riemannian metric by $g(u,v)=\omega(u,Jv)$. Together they form an almost-Kähler structure. It becomes Kähler only if $J$ is integrable.

Why projection is insufficient

For any two-form, its invariant part is

$$\omega^{\mathrm{inv}}(u,v) =\tfrac12\bigl(\omega(u,v)+\omega(Ju,Jv)\bigr).$$

If $\omega$ tames $J$, this projection remains positive on complex lines and is invariant. But $J$ varies with position, and differentiation of the projected form introduces derivatives of $J$. Therefore $d\omega=0$ does not generally imply $d\omega^{\mathrm{inv}}=0$.

The missing condition is global and differential, not a failure of pointwise linear algebra.

Classes and cones

A closed form determines a cohomology class $[\omega]\in H^2(X;\mathbb R)$. The class records its integrals over closed surfaces. Write $\mathcal K_J^t$ for the set of taming classes and $\mathcal K_J^c$ for the compatible classes. They are cones: positive scaling and addition preserve the corresponding positivity conditions.

A two-form $\alpha$ is anti-invariant when $\alpha(Ju,Jv)=-\alpha(u,v)$. Such a form vanishes on every complex line. Let $H_J^-$ be the subspace of cohomology represented by closed anti-invariant forms. Adding one of these to a compatible form preserves taming, though not compatibility.

The intersection pairing of two classes is

$$a\cdot b=\int_X\alpha\wedge\beta,$$

where $\alpha,\beta$ are closed representatives and $X$ has the orientation determined by $J$. It depends only on the classes.

3. Development history and previous partial results

Donaldson formulated the existence question in his 2006 paper on two-forms and elliptic equations. He also proposed a related a priori estimate problem for prescribed volume. These are distinct problems: existence of some compatible form does not solve every prescribed-volume equation. Donaldson.

For integrable structures, the question belongs to the theory of compact complex surfaces. Buchdahl and Lamari established Kähler criteria without relying on the classification of surfaces. Positive currents supplied both existence tools and obstructions. Li–Zhang proved the equivalence between the existence of taming and compatible symplectic forms on complex surfaces and developed a cohomological framework for comparing the two cones. Li–Zhang, 2009.

Pseudoholomorphic curves offered another route. Gromov constructed compatible forms for almost-complex structures on the complex projective plane tamed by its standard symplectic form, using averages of currents associated with pseudoholomorphic lines. Taubes developed integration over curve moduli spaces to establish compatibility for a residual set of structures in the case $b_2^+=1$, with prescribed-class conclusions under stated additional assumptions. Here residual means a countable intersection of open dense sets; it does not mean every structure. His later correction addresses the relevant first-Betti-number case. Taubes’s original paper, the correction.

Li–Zhang extended curve-based methods to all tamed structures on certain rational surfaces and to further cases under hypotheses involving embedded spheres. Other approaches used restrictions on the anti-invariant cohomology dimension. Li–Zhang, rational four-manifolds.

The new manuscript claims to remove these integrability, genericity, and cohomological-dimension restrictions for the existence question. It uses positive-current duality rather than assuming that a suitable family of pseudoholomorphic curves is already available.

4. What the new manuscript claims

The main theorem states: if a smooth almost-complex structure $J$ on a closed connected smooth real four-manifold is tamed by a symplectic form, then a smooth symplectic form compatible with that same $J$ exists.

The source is the October 6, 2026 manuscript. Its conclusion places no restriction on the compatible form’s cohomology class.

The second theorem describes the compatible cone more precisely. Define

$$V=(H_J^-)^\perp, \qquad C=\mathcal K_J^t\cap V.$$

Orthogonality is with respect to the intersection pairing. The manuscript asserts

$$\mathcal K_J^t=C+H_J^-,$$

and

$$\mathcal K_J^c =\{a\in C:a\cdot y>0 \text{ for all }0\ne y\in\overline C\}.$$

The closure $\overline C$ is taken inside $V$. The sum denotes all sums of one element from each set. Crucially, the strict positivity test includes every nonzero boundary point of the closed cone, not just its interior.

Write $h_J^-=\dim H_J^-$, and let $b_2^+$ be the positive index of the intersection form. Under the additional condition

$$h_J^-=b_2^+-1,$$

the paper obtains $\mathcal K_J^t=\mathcal K_J^c+H_J^-$. In particular, if $b_2^+=1$, every taming class has a compatible representative.

Without that dimension condition, the manuscript explicitly asserts that the inclusion can be strict, even when $H_J^-=0$. Its four-torus example has two taming classes whose intersection is zero. A compatible class must pair strictly positively with every taming class, so neither of those two classes can be compatible. Nevertheless, the example has another compatible form. This makes the distinction between existence and prescribed class concrete.

5. Why the advance matters

Importance: 5/5. If correct, the result settles a central existence question at the interface of symplectic and almost-complex geometry in dimension four. It says that positivity on complex lines can always be upgraded to a closed compatible geometry, even though the obvious pointwise projection fails.

Keeping $J$ fixed is important. Choosing a new almost-complex structure compatible with a given symplectic form is standard; finding a compatible form for a prescribed tamed structure is the difficult direction.

The cone theorem adds information about which global periods can occur. It would replace a bare existence statement with an intersection-theoretic test, while explaining why arbitrary taming classes need not survive the upgrade.

The manuscript’s current estimates also feed into the hypersymplectic deformation construction in family 341. That application adds its own prescribed-class hypotheses and deformation arguments. The rating concerns the significance of these stated claims, not their verification.

6. The main idea

Translate failure of existence into an obstruction

The proof begins by supposing that no compatible form exists. Positive invariant forms make an open convex cone, while closed invariant forms make a linear space. A separation theorem then provides a nonzero positive functional separating them.

This functional is a current $P$. One can think of a current as generalized integration over oriented surfaces, allowing both concentrated and diffuse mass. Positivity means that it evaluates nonnegatively on forms positive on complex lines. By construction, $P$ vanishes on all closed invariant forms.

An elliptic correction supplies an anti-invariant distribution $Q$ such that

$$T=P+Q$$

annihilates every closed two-form. If one could prove $Q=0$, the contradiction would be immediate: a nonzero positive current must evaluate positively on a taming form, but $T$ is required to vanish on that closed form.

The analytic task is therefore to remove the anti-invariant correction without losing control of the positive current.

Control mass and changing directions

Represent $P$ as a weighted average of oriented complex-line elements. Let $\mu$ be the resulting measure on $X$. Corrected radial tests produce a quadratic local mass bound

$$\mu(B_r(x))\leq Cr^2.$$

This is the scaling expected for something surface-like inside a four-dimensional ambient space. It does not by itself say that $P$ is integration over a smooth surface.

Regularization using the Hodge operator gives an $L^2$ bound for $Q$, meaning that its squared magnitude is integrable, and an estimate on how the complex-line directions vary between nearby points. The first regularized pairing gives boundedness, but a remaining negative error prevents it from directly proving $Q=0$.

Separate the part with positive density

The manuscript establishes existence of the two-dimensional density

$$\theta(x)=\lim_{r\downarrow0}r^{-2}\mu(B_r(x))$$

and considers the set $E$ where this density is positive. Its central step claims that the restriction $P_E$ is closed. This is subtle: restricting a closed object to an arbitrary measurable set can create a boundary.

The argument uses smooth cutoffs of density and combines two kinds of error. An angular estimate controls variation of the tangent directions; a transverse estimate controls how nearby base points depart from the approximate tangent plane. After the appropriate normalization, their product tends to zero. The $L^2$ control of $Q$ handles its contribution to the cutoff boundary.

This produces closedness of the positive-density part without requiring a specified rate of convergence of tangent measures. The remaining positive part is concentrated at zero-density centers, precisely where the troublesome error in the regularized pairing can be made to vanish.

Finish the contradiction and refine it by class

Pairing the residual current with $T$, using the closedness just obtained, forces $Q=0$. The resulting positive current cannot annihilate the original closed taming form. This contradicts the assumed absence of a compatible form.

For the cone theorem, the manuscript repeats separation with a cohomology class prescribed. The resulting obstruction is expressed through the intersection pairing and the closure of the invariant part of the taming cone. This yields the strict positivity condition in the theorem rather than a universal equality of taming and compatible classes.

A specifically four-dimensional identity supports the estimates: anti-invariant two-forms are self-dual for a compatible auxiliary metric. The proof therefore has a structural dependence on dimension, not merely a convenient choice of notation.

7. Scope, limitations, and open questions

The theorem assumes compactness without boundary and real dimension four. It does not assert an analogous implication in arbitrary dimension, where the two-form decompositions and available obstructions differ.

Compatibility does not make the prescribed $J$ integrable. The output is almost-Kähler geometry, with Kähler geometry only in the integrable case.

The result also does not preserve the original taming form, its class in general, or a prescribed volume density. It gives existence for fixed $J$, plus the stated cone criterion. Donaldson’s stronger analytic estimate questions require separate treatment.

The cone statement is particularly sensitive to quantifiers: positivity against all nonzero elements of $\overline C$ is stronger than positivity against a selected collection of classes. The special dimension condition gives a simpler equality, but the paper does not claim that equality without it.

These are the manuscript’s stated conclusions and mechanisms. The digest does not verify the density-splitting argument, the example, or the claimed resolution independently.

8. Sources and further reading

The primary sources linked above cover Donaldson’s question, cone comparisons, and curve-based partial results. Optional background appears under almost complex manifold, symplectic manifold, and current in mathematics.

Source manuscripts

• M1. OpenAI, Taming implies compatibility on four-manifolds (October 6, 2026).

Sources consulted on 26 October 2026 at repository revision 1581d4c. Download the manuscript citations.