The real differential algebra fixes the signs of integration, curvature and the divisor formula. We develop its positivity cones through their matrix representations, then prove the domination estimates used for currents. The distinction between ordinary and strong positivity is essential in middle degree.
7.1 The differential algebra
Definition 7.1.1. (Real superforms). Let with coordinates , and let be open. Write and for two copies of . The algebra of superforms is All one-form generators anticommute, including generators from different copies. Thus a form is uniquely written as where indices in a multiindex are increasing. The total degree is . In particular,
Definition. The involution is the algebra automorphism exchanging and and fixing functions. A -form is symmetric if .
Moving the primed generators to the left gives Since and have the same parity, symmetry in bidegree means precisely . This is the symmetry that will become ordinary matrix symmetry.
Definition. The differentials are the total-degree derivations determined by
For the displayed expansion of , these definitions give
NAG-900035 Proof review pending
Proposition 7.1.4. The operators satisfy For a smooth function , one has
Proof. The terms in cancel in pairs: the second derivatives are symmetric in the two new indices, whereas the primed generators are antisymmetric. The same argument applies to . In , the derivatives again agree, and exchanging the two new generators reverses the sign. The identity with can be checked on functions and the one-form generators, after which the derivation rule proves it on the algebra. Applying to gives the last formula. ◻
Remark 7.1.5 (A sign that must be corrected in the 2012 text). The displayed commutativity exponent in [CLD12] (§1.2.6, p. 14) is . Read literally, it would make commute with , contrary to (7.1.1) and to the total-degree Leibniz rule used later in that text. The displayed exponent in [CLD12] (§3.1.4, p. 36) is also incompatible with those formulas. Throughout these notes the exponent is . This is an explicit correction of the displayed formulas, not a second convention being used implicitly.
NAG-900036 Proof review pending
Proposition 7.1.6. (Conversion of the two operators). Let be the involution in Definition 7.1.1. The algebra automorphism in [CLD25] (§1.1.2) satisfies On forms of bidegree one has For a form of type , the old symmetry condition is equivalent to . The strong generators and the signed squares , with of type , are unchanged when is replaced by . Hence the three cones in Definition 7.2.2 agree under this conversion.
Proof. Every double-primed generator contributes one extra minus sign under , giving the first equality. Applying it twice gives . The symmetry assertion follows by setting . A form of type contains no double-primed factor, so . This proves the statement about the generators and, by the complementary wedge definitions, all three cones. ◻
Put and fix the positive top form Integration means . The choice is compatible with the lattice normalization when the coordinates form a lattice basis. Under an arbitrary linear change of coordinates the coefficient of this top form changes by the square of the determinant; the normalization of the integral must therefore be transformed as well.
7.2 Symmetric forms as matrices
NAG-900037 Proof review pending
Remark 7.2.1 (Matrix normalization). The factor removes the sign caused by collecting all primed generators. For a real matrix , where , define Thus is symmetric exactly when is symmetric. If , then
Definition 7.2.2. A symmetric form is positive if is positive semidefinite. It is strongly positive if is a finite sum of matrices with decomposable. It is weakly positive if for every decomposable -vector .
The zero element is permitted in all three cones. For a smooth form these conditions are imposed at every point. In the definition of strong positivity, the pointwise decomposition need not have smoothly varying summands. This chapter uses the pointwise cone condition; no smooth choice of eigenvectors is needed.
NAG-900038 Proof review pending
Proposition 7.2.3. A symmetric form is positive if and only if it is a finite sum , without a decomposability restriction. There are inclusions All three cones coincide for .
Proof. The spectral theorem gives with precisely when is positive semidefinite. Take . The converse follows from . The two cone inclusions follow from the same formula.
In the four indicated degrees every exterior vector and covector is decomposable. For degree , a nonzero exterior vector is contraction of a volume vector by a nonzero covector. A basis in which that covector is the last dual basis vector expresses the contraction as a single wedge of basis vectors. Thus the decomposability restrictions disappear. ◻
NAG-900039 Proof review pending
Proposition 7.2.4. (Hessian criterion). Let . Then This form is positive if and only if the real Hessian of is positive semidefinite at every point. If is convex, this is equivalent to convexity of .
Proof. The displayed formula follows from the definitions. A symmetric -form is represented by a real symmetric matrix. Evaluating along a real direction gives the quadratic expression . The three positivity cones in this degree agree with the cone of positive semidefinite matrices. Finally, the restriction of to the segment has second derivative ; the one-dimensional criterion for convexity proves the last assertion. ◻
NAG-900040 Proof review pending
Proposition 7.2.5. (Wedge duality). The wedge pairing between symmetric forms in complementary bidegrees is nondegenerate. Under this pairing the positive cone is self-dual across complementary degrees, and the strongly positive and weakly positive cones are dual.
Proof. Let . For increasing multiindices of size and of size , let be the sign of relative to when , and zero otherwise. The matrix is a signed permutation matrix. Expanding the wedge gives Indeed the scalar sign is . Since is invertible, nondegeneracy follows. Testing against all positive semidefinite is equivalent to being positive semidefinite: rank-one tests give all inequalities .
The complement map takes decomposable -covectors to decomposable -vectors. To see this, choose an orthonormal basis adapted to the span of the covectors in their wedge; its complementary basis gives the image under , up to a scalar and sign. Thus testing against strongly positive gives exactly weak positivity of .
For the reverse duality, the strong cone is closed. Normalize its nonzero generators by . The resulting set is compact, being the image of the closed set of unit decomposable covectors. Its convex hull is compact in finite dimension. Every nonzero element of the cone is its trace times a point of that convex hull. A convergent sequence in the cone has bounded traces, so its limit remains in the cone. The finite-dimensional separation theorem now identifies the double dual with the cone itself. ◻
7.3 A positive control form
The current estimates will require a fixed form dominating every symmetric coefficient matrix. Define
NAG-900041 Proof review pending
Lemma 7.3.1. One has , where . In particular is strongly positive and lies in the interior of the ordinary positive cone.
Proof. The degree-two factors in commute, and a repeated factor has zero wedge. Every set of distinct indices therefore occurs times. Moving the double-primed factors to the right contributes , giving (7.2.1) with the identity matrix. The result is a sum of decomposable squares. If a symmetric matrix has operator norm less than one, is positive definite; this proves the interior assertion. ◻
NAG-900042 Proof review pending
Proposition 7.3.2. (Coefficient domination). Let be a smooth symmetric form supported in a compact set . Choose with near . If , then Both inequalities mean ordinary positivity.
Proof. On , every eigenvalue of lies in , so is positive semidefinite. Off , and is positive semidefinite. This proves the assertions at every point, without choosing eigenvectors smoothly. ◻
Consequently every compactly supported symmetric form is a difference of two compactly supported positive forms. This is the decomposition needed for positive currents. Requiring both summands to be strongly positive changes the statement, and the next section shows that the stronger requirement is impossible in general.
NAG-900043 Proof review pending
Remark 7.3.3. One may replace the operator norm by a coefficient norm. If , then : for , , and summing over gives the bound. Constants in chart estimates may therefore depend on the number of coefficient directions, but not on derivatives of the test form.
7.4 The Plücker obstruction
Take and . Order the exterior coordinates by Define the symmetric matrix Its quadratic form is
NAG-900044 Proof review pending
Lemma 7.4.1. The right side vanishes for every decomposable . Moreover and .
Proof. If , then . Expanding in the volume basis gives This proves the relation. The matrix exchanges the first and sixth coordinates, exchanges the second and fifth with a minus sign, and exchanges the third and fourth. Applying it twice is the identity, and all its diagonal entries vanish. ◻
NAG-900045 Proof review pending
Theorem 7.4.2. (Failure of strong-cone spanning). In bidegree on , the strongly positive cone does not span the space of symmetric forms. The weakly positive cone contains the nonzero line . In particular it is not pointed.
Proof. The nonzero linear functional vanishes on every strong generator , because for decomposable . Hence it vanishes on the linear span of the strong cone. But , so that span is proper. The same Plücker relation on decomposable vectors gives for every . Thus every is weakly positive. ◻
Example 7.4.3. (A positive form which is not strongly positive). Let in exterior coordinate notation and put . Then is positive semidefinite, while . Since this functional vanishes on the entire strong span, is not even a difference of strongly positive forms.
This example pinpoints the interior error. The form is strongly positive, but it cannot lie in the interior of the strong cone in the full symmetric space. Every matrix in Example 7.4.3 has , regardless of , and therefore is not strongly positive. Enlarging the coefficient of a fixed strong form cannot repair this obstruction.
Remark 7.4.4 (Relation to the sources). This is the obstruction in [BGJK] (Example 2.3.6). It contradicts the real statement of [CLD12] (Lemma 5.2.3), even after symmetry is imposed. That lemma’s proof uses the interior assertion just disproved. The analytic Lemma 5.3.3 cites it through tropical charts; its printed proof therefore loses this input. An analytic counterexample would additionally require a faithful realization of the relevant forms and positivity through the analytic charts, and is not supplied by the real calculation alone.
Similarly, [GK19] (Remark 1.12(i)) asserts that weak positivity of both and implies . The form disproves this assertion in its stated real setting. The analytic uses of this remark, including Lemma 2.16, require a separate analysis of their realization and quotient relations. We do not label those analytic statements disproved solely by this example.
7.5 Pullbacks and products
The corrected coefficient estimates use ordinary positivity. Before passing to currents, we record the operations that preserve this cone. These proofs also explain which cone conditions are available when multiplying a current by a smooth form.
NAG-900046 Proof review pending
Proposition 7.5.1. Pullback by an affine map preserves each of strong, ordinary and weak positivity.
Proof. Let have linear part . On exterior covectors the pullback is . It takes a square to the square of the pulled-back covector; decomposable covectors stay decomposable or become zero. This proves the strong and ordinary assertions. For a decomposable -vector on , the exterior image is decomposable or zero. The quadratic form of evaluated at is the quadratic form of evaluated at that image, proving weak positivity. ◻
NAG-900047 Proof review pending
Proposition 7.5.2. Products of positive forms are positive. Products of strongly positive forms are strongly positive. The product of a weakly positive form and a strongly positive form is weakly positive. Products in degrees exceeding the dimension are zero.
Proof. For exterior covectors of degrees , respectively, the signs give Indeed moving past gives and . Decompose the positive matrices into squares and use this identity term by term. If both factors are strong, their generating covectors are decomposable, so their wedge is decomposable or zero.
For the last assertion it suffices to consider one decomposable square . Evaluate the product on a decomposable -vector and restrict to the subspace it spans. If restricts to zero there is nothing to prove. Otherwise choose a basis of this subspace so that is a scalar multiple of the wedge of the last dual basis vectors. The value of the product is a nonnegative square factor times the value of the first form on the wedge of the first basis vectors. Weak positivity makes the latter value nonnegative. ◻
For a smooth function, Proposition 7.1.4 identifies the matrix of with its Hessian. Thus positivity, strong positivity and weak positivity of are equivalent. On a convex open set they are equivalent to convexity of : along every line the second derivative is nonnegative, which is the one-variable criterion for convexity; conversely, second differences of a smooth convex function give a nonnegative Hessian. This special agreement in degree must not be extrapolated to all degrees.
7.6 A homotopy formula for the real double complex
The two exterior factors of a superform have different roles when one studies only one differential. This gives an elementary Poincaré lemma on an ordinary real open set, but also explains why a statement about the double complex requires more than two separate exactness statements.
Let be open and star-shaped about . Write every superform uniquely as where the first factor is an ordinary differential form in the . In this description, is the ordinary exterior derivative on each , while the form a fixed coefficient space.
NAG-900048 Definition · proof status does not apply
Definition 7.6.1. For , define coefficientwise by Set for . Define for ; for , let replace each scalar coefficient of by its value at , retaining the fixed -basis.
The semicolon separates the primed and double-primed alternating slots. Only the primed slots are scaled in this formula. In particular, is not the full geometric pullback by the constant map: that pullback would also kill positive double-primed degree.
NAG-900049 Proof review pending
Proposition 7.6.2. (Radial homotopy). The operators satisfy Consequently the -cohomology on vanishes in positive primed degree, while in primed degree zero it is the space of constant double-primed forms .
Proof. It suffices to prove the usual homotopy identity for each . Let be . Decompose the ordinary pullback of a primed form as , where contains no . The coefficient of in the identity is Integrate from to . The boundary term is , and is precisely the coefficientwise operator in Definition 7.6.1. This proves the formula. Smoothness at follows directly from the displayed pullback, or from the factor when .
If and , the identity gives . For , closedness says that every coefficient has zero differential. The star-shaped open set is connected, so those coefficients are constant. There are no forms of primed degree , giving the asserted degree-zero cohomology. ◻
There is an analogous statement for , obtained by exchanging the exterior factors with the fixed sign convention. The two statements concern separate complexes. The operator was not constructed to commute with , and no such commutation is needed for Proposition 7.6.2. This point matters when one tries to obtain a -potential by applying two homotopies successively.
7.7 The symmetric Hessian potential theorem
In bidegree the extra compatibility has a familiar form: the coefficient matrix must be symmetric. On a star-shaped open set this condition, together with closedness, suffices for a scalar potential.
NAG-900050 Proof review pending
Theorem 7.7.1. Let be open and star-shaped, and let have smooth coefficients with . If , then there is a smooth function with . The potential is unique up to addition of an affine function. If is convex, is positive if and only if is convex.
Proof. For each , the primed one-form is closed. Proposition 7.6.2, after translating the star centre to zero, gives a smooth function with . Symmetry implies . Hence the ordinary one-form is closed. Apply the same homotopy formula once more to obtain with . Its Hessian is , proving the asserted identity.
If and are two potentials, every first derivative of has zero differential and is therefore constant on the connected set . Thus is affine. Finally, positivity of a symmetric -superform is positive semidefiniteness of its matrix. On a convex domain, the second derivative of the restriction of to every line segment is . Nonnegativity of these second derivatives is equivalent to convexity. ◻
The same calculation shows that symmetry and -closedness imply -closedness here: the constructed Hessian is closed under both differentials. It also identifies the ambiguity more precisely than saying only that it is harmonic. In ordinary real coordinates the smooth Hessian kernel consists of affine functions; on an analytic space, current-harmonic functions can form a much larger class.
Example 7.7.2. (Double closedness without a scalar potential). On any open ball in , the constant form is both -closed and -closed. It cannot equal for a smooth scalar function, because that would give and . It cannot have a scalar distributional potential either, since distributional partial derivatives commute. Thus two separate Poincaré lemmas do not imply an unrestricted local -lemma for this real double complex.
The passage to a polyhedral or analytic space
The preceding results are on an open subset of one real vector space. A form on a polyhedral support is a quotient of ambient forms by tangential relations, and the tangent directions can change across faces. Before using a homotopy on that quotient, one must prove that it preserves those relations and has the required support and domain properties. An ambient homotopy formula alone does not do this.
On a Berkovich space there is a further compatibility under refinements of moment charts. A potential constructed on one tropical presentation must be independent of the presentation or be shown to glue with the permitted transition functions. These are precisely the additional issues in local-potential theorems for currents and in comparisons of tropical and analytic cohomology. The explicit Euclidean theorem is a local building block, not a theorem covering every analytic chart quotient.
7.8 A radial homotopy on a polyhedral star
The ordinary de Rham homotopy contracts all covectors of a differential form. For the complex of superforms, the primed degree must remain fixed. The correct operator contracts only the double-primed variables. This produces a local Poincaré lemma and identifies its degree-zero kernel with the multi-tangent coefficient space.
Let be the support of a finite fan, and let be an open ball centred at its vertex . Superforms on are restrictions of smooth ambient superforms, with two forms identified when their evaluations on tangent vectors of every cone agree. It is enough to work with germs at , so a common ambient representative is available. Write a representative as With the ordered-wedge convention of this book, acts on each as times the ordinary exterior derivative. This sign is essential in the homotopy formula.
For an ordinary -form with , set For set , and define The operator has bidegree . Define to be zero for , and for define it to be the constant-coefficient primed form , restricted to the fan.
NAG-900051 Proof review pending
Proposition 7.8.1. The operators descend to superforms on the fan, and satisfy Consequently every -closed germ of bidegree with is -exact.
Proof. Suppose an ambient representative vanishes on tangent vectors of every cone. For in a cone , all the points with remain in , and . Evaluating on primed and double-primed vectors in therefore evaluates only on vectors in , including the inserted vector . The integrand is zero. Its value at follows by continuity and does not affect the integral. Thus preserves the relation defining forms on .
For , if an ambient primed form vanishes on each cone, evaluate along an interior ray of a cone and let its parameter tend to zero. Its constant value at zero annihilates for every , so also preserves the relation. For there is nothing to check for .
On ambient ordinary forms, differentiation under the integral gives , where keeps only degree-zero values at the origin. One can check this on a coefficient : the surviving integrand is the derivative in of , with the evaluation on the appropriate vectors. The endpoint at gives ; the endpoint at is zero for and is for . The two factors from and cancel. Applying this identity to every proves the displayed formula on ambient forms, hence on the quotient. If and , it gives . ◻
The kernel is a multi-tangent dual
Put
NAG-900052 Proof review pending
Corollary 7.8.2. The space of -closed germs at the vertex is naturally . The local sequence is exact on finite polyhedral charts, with the natural face restrictions on .
Proof. For a closed form, Proposition 7.8.1 gives on a sufficiently small star. Thus it is represented by a constant ambient -covector. Such a covector represents zero exactly when it annihilates every , equivalently their sum . The quotient of by this annihilator is , since every linear functional on a subspace of a finite-dimensional vector space extends to the whole space. Exactness in positive double-prime degree is the proposition. Around a point in any face, translate the point to the origin and use the local star. Restriction to a smaller face star induces the indicated restrictions on the coefficient spaces. ◻
The sum in must be formed after taking exterior powers. Replacing it by changes the coefficient sheaf in general. The two-plane example in Section 24.7 makes the difference explicit. The homotopy proof identifies the correct space directly from the vanishing relation for actual forms.
Remark 7.8.3. If an integral linear map sends one fan to another, the homotopy commutes with its ordinary superform pullback: linear maps commute with radial scaling, and contraction by corresponds to contraction by its image. Translation gives the same assertion for affine charts with specified centres. This compatibility is enough for the local polyhedral argument. A global homotopy on a Berkovich space would additionally require compatible analytic domains and centres; it is not supplied by this computation.
From a local resolution to global cohomology
On a paracompact polyhedral space with finite-part charts, the sheaves admit smooth partitions of unity. They are therefore fine. The exact sequence above is a fine resolution of , so its global cohomology computes . Here we use the standard fine-resolution theorem of sheaf cohomology; its mechanism can be seen in a Čech double complex.
For a locally finite cover and subordinate partition , a Čech cochain with values in has a homotopy The sum is locally finite; extension by zero is valid because the multiplier is supported in . Thus the rows have no positive Čech cohomology. In the other direction, local exactness of the complex leaves only . Passing to sufficiently fine covers, the two computations of the total complex give the desired comparison.
This proves the finite-part local mechanism behind the tropical Dolbeault comparison. Extended tropical spaces also have sedentary boundary strata. Their form definition includes compatibility with projections to those strata, and their local Poincaré lemma requires that additional structure. The full comparison, including compact supports and extended charts, is the theorem of Jell–Shaw–Smacka stated in Chapter 24. The present radial proof does not remove the boundary hypotheses from that theorem.
References. [JSS] (Sections 2–3); the explicit homotopy above is the radial de Rham argument applied only to the double-primed variables, with the quotient and sign checks included.