Completing Kodaira's generalized potential theory on open Riemannian manifolds: a Riemann-Roch dimension formula, vanishing third-kind fields of cycles, and essential singularities

In 1949 Kunihiko Kodaira constructed a generalized potential theory of harmonic fields (differential forms that are simultaneously closed and co-closed) on orientable analytic Riemannian manifolds, assigning to every finite chain $C$ a harmonic field $e[C]$ 'of the third kind' with singularities on $\partial C$, and proving, for closed (compact) manifolds, a generalized Riemann-Roch dimension formula (Theorem 27) that counts such fields with prescribed poles and periods. For open (noncompact) manifolds the theory was left incomplete: no counting theorem exists, and Kodaira's cylinder example $\Gamma\times\mathbb{R}$ exhibits non-bounding cycles $Z$ with $e[Z]=0$, so finite-energy harmonic fields fail to detect the homology of the manifold. The open problem consists of three linked parts: (1) a normalizability condition at infinity under which the spaces of harmonic fields with prescribed poles and periods on an open analytic manifold are finite-dimensional, together with a dimension formula specializing to Kodaira's closed-manifold formula; (2) an intrinsic characterization of the cycles $Z$ with $e[Z]=0$, equivalent to characterizing which classes admit square-integrable harmonic representatives; (3) the classification of harmonic fields with isolated essential (infinite-order) singularities, including the exceptional ranks. All known Riemann-Roch-type theorems impose structural hypotheses at infinity (compactness, coverings, conical or fibred ends); none covers the general open analytic case, so the problem remains open.

Background

Harmonic fields. On an $n$-dimensional Riemannian manifold $\mathfrak{M}$, the exterior derivative $d$ maps differential $k$-forms to $(k+1)$-forms, and the codifferential $\delta$ (its metric adjoint, expressible via the Hodge star operator $*$ as $\delta = \pm{*}{d}{*}$) maps $k$-forms to $(k-1)$-forms. A harmonic field of rank $\rho$ is a $\rho$-form $\omega$ with $d\omega = 0$ and $\delta\omega = 0$ (closed and co-closed). On a closed (compact, boundaryless) $\mathfrak{M}$, Hodge's theorem identifies the space of harmonic $\rho$-fields with the $\rho$-th de Rham cohomology, so its dimension is the Betti number $b_\rho$; this finite-dimensional counting underlies the classical theory of Abelian integrals and, on Riemann surfaces, the Riemann-Roch theorem, which computes the dimension of the space of meromorphic functions with prescribed zeros and poles (a divisor).

Kodaira's 1949 theory. In his Annals paper, Kodaira developed 'generalized potential theory' for harmonic fields on orientable analytic Riemannian manifolds that may be open (noncompact), extending classical function theory on Riemann surfaces to arbitrary rank and dimension: fields of the first kind are the globally regular harmonic fields; fields of the second kind $e[p]$ carry prescribed point singularities (poles) at points $p$; fields of the third kind $e[C]$ are attached to finite $\rho$-chains $C$ (finite singular chains, with boundary $\partial C$) and carry period-type singularities along $\partial C$ (a period is the integral of a field over a cycle). Existence and uniqueness of $e[C]$ (Theorem 15) and convergence of the associated expansions (Theorem 20) were proved for general manifolds, including open ones; $e[C]$ is characterized by square-integrability ($L^2$: finite norm $\int |\omega|^2$) on compact subdomains and outside every neighborhood of $\partial C$, together with normalization conditions fixing its periods against an explicit auxiliary class of fields, so that its singular set is exactly $\partial C$. For a cycle $Z$ (a chain with $\partial Z = \emptyset$), $e[Z]$ is a globally square-integrable harmonic field, expected to detect the homology class of $Z$ through its periods.

Two theorems left behind. Theorem 25 is a Laurent-type structure theorem: near an isolated singular point of finite order (the singular part of the field is a finite combination of explicit model singular fields — the analog of a pole), a harmonic field admits an explicit asymptotic expansion; in the exceptional ranks $\rho = 1$ and $\rho = n-1$ this requires auxiliary flux conditions, denoted (11.1)/(11.1)* in the paper (supplementary integral conditions on the field near the singular point), and without them the expansion is recovered only after adding a specific line-integral term. Theorem 27 is a generalized Riemann-Roch theorem for closed manifolds: with $b$ the $\rho$-th Betti number, $r$ and $s$ the numbers of prescribed poles $p_j$ and $q_k$ (for the problem and its adjoint), and $t$, $u$ the numbers of prescribed $\rho$-chains $C_k$ and $(n-\rho)$-chains $C_k^*$ — all assumed linearly independent modulo the regular harmonic fields — the dimensions $A$ and $B$ of the two associated spaces of harmonic fields with those prescribed singularities and vanishing periods satisfy $A = B - b + r - s - t - u$.

The open-manifold gap. For open $\mathfrak{M}$, the two central questions of the program were left open. First, no Riemann-Roch-type counting theorem was proved: the compactness (complete-continuity) arguments behind Theorem 27 fail on open manifolds, and without a condition at infinity the admissible spaces of harmonic fields with prescribed poles are typically infinite-dimensional (for instance, on an open Riemann surface the meromorphic functions with a single prescribed pole already form an infinite-dimensional space), so a meaningful theorem must first identify a normalizability class at infinity. Second, the passage from cycles to fields breaks: in Kodaira's example $\mathfrak{M} = \Gamma \times \mathbb{R}$ (a cylinder over a cross-section $\Gamma$), the cycles $Z_y$, carried by the slices $\Gamma\times\{y\}$ over the line factor, are not boundaries of finite chains, yet $e[Z_y] = 0$; a nontrivial homology class on an open manifold need not be detected by any finite-energy harmonic field. In modern language, the question is which cohomology classes on a noncompact manifold are realized by square-integrable harmonic forms (reduced $L^2$-cohomology: the quotient of closed $L^2$ forms by the closure of the exact ones).

Modern context. (i) The closed-manifold counting program was completed: Kodaira proved Riemann-Roch for compact analytic surfaces in 1951, and Hirzebruch's Riemann-Roch theorem and the Atiyah-Singer index theorem gave the general finite-dimensional counts of which Theorem 27 is an early instance. (ii) On open manifolds, Gaffney extended the Hilbert-space theory of harmonic integrals (Hodge's classical term for harmonic forms); Atiyah-Patodi-Singer introduced boundary correction terms ($\eta$-invariants); Cheeger-Gromov established bounds on the von Neumann dimension (a dimension theory for infinite-dimensional $L^2$ spaces) of $L^2$-cohomology; Roe built an index theory for open manifolds. (iii) Closest to Kodaira's question, Gromov and Shubin (1992-1994) proved a Riemann-Roch theorem for elliptic operators with rigged divisors on compact manifolds — a divisor augmented with finite-dimensional spaces of prescribed local singular and vanishing data: the dimension $r(\mu, A)$ of the solution space of an elliptic equation with rigged divisor $\mu$ satisfies $r(\mu, A) = \operatorname{ind} A + \deg(\mu) + r(\mu^{\wedge}, A^*)$ — the modern operator-theoretic descendant of Theorem 27. Their exposition states explicitly that the noncompact case is treated only with proper boundary conditions and 'conditions at infinity'; subsequent variants (Shubin's $L^2$ version; Schulze-Tarkhanov for conical singularities; Kha-Kuchment for abelian coverings, where infinity is adjoined to the divisor) all impose structural hypotheses on the end. (iv) Realizability by $L^2$ harmonic forms is known only case-by-case: Zucker computed it for warped products (metrics of the form $dt^2 + f(t)^2\,g_\Gamma$ on $\Gamma\times\mathbb{R}$, a class containing the product cylinders of Kodaira's example), Hausel-Hunsicker-Mazzeo for fibred gravitational-instanton ends (gravitational instantons are the complete noncompact hyperkähler four-manifolds arising in physics) via intersection cohomology (a cohomology theory for singular spaces that restores Poincaré duality) of the boundary; no intrinsic characterization in terms of the geometry of the ends exists. (v) For singularities of infinite order (essential singularities, where no finite combination of model singular fields captures the singular behavior — the analog of an essential singularity of a holomorphic function), no extension of Theorem 25 was found in the literature: the analog of the Weierstrass theory for harmonic fields on analytic Riemannian manifolds appears undeveloped.

Problem Statement

Complete Kodaira's generalized potential theory of harmonic fields of the third kind from closed to open (noncompact) orientable analytic Riemannian manifolds. The problem has three linked components.

Component 1 (global counting). Let $\mathfrak{M}$ be an open, orientable, analytic Riemannian $n$-manifold, and let $e[\cdot]$ denote Kodaira's third-kind construction (harmonic fields with prescribed singularities on finite chains, square-integrable away from those chains, with Kodaira's period normalization). Identify a normalizability condition at infinity under which: (a) the two spaces of admissible harmonic fields associated, as in Kodaira's Theorem 27, with the prescribed poles $p_1,\dots,p_r$ and $q_1,\dots,q_s$ (for the problem and its adjoint, respectively) and the prescribed vanishing periods over the finite $\rho$-chains $C_1,\dots,C_t$ and $(n-\rho)$-chains $C_1^*,\dots,C_u^*$ are finite-dimensional; and (b) their dimensions, denoted $A$ and $B$ as in Theorem 27, satisfy an explicit formula — the expected shape is $A = B + (\text{correction terms})$, with correction terms computable from topological data (Betti numbers, counts of prescribed poles and chains) together with geometric data of the ends, but the problem does not prescribe the exact algebraic form of the count or the nature of the condition. The formula and the class must satisfy three mandatory consistency checks: (i) when $\mathfrak{M}$ is closed they reduce to Kodaira's Theorem 27, $A = B - b + r - s - t - u$; (ii) they reproduce the vanishing $e[Z_y] = 0$ for the slice cycles of the cylinder $\Gamma\times\mathbb{R}$; (iii) they agree with the known special cases — the Gromov-Shubin rigged-divisor theorem on compact manifolds and its extensions to abelian coverings, conical singularities, and $L^2$ settings — on the manifolds where those apply. The class must be nontrivial: it must contain the harmonic fields occurring in all the listed special cases.

Component 2 (homological realizability). Characterize, in intrinsic geometric terms of the ends, the finite $\rho$-cycles $Z$ on an open $\mathfrak{M}$ for which $e[Z] = 0$; in modern language, characterize which classes are realized by globally square-integrable harmonic fields (reduced $L^2$-cohomology realizability). The characterization must reduce on closed manifolds to the classical statement (no such vanishing for nonzero classes), must predict the vanishing for the cylinder slices $Z_y$, and must be effectively checkable at least on product/cylindrical ends, warped products, fibred ends, and hyperbolic cusp ends, where the answer is known from Zucker and Hausel-Hunsicker-Mazzeo.

Component 3 (local singularity theory). Extend Theorem 25 to isolated singular points of infinite order: for a harmonic field with an essential isolated singularity (no finite combination of model singular fields captures the singular behavior) at a point of an analytic Riemannian manifold, determine which singular behaviors occur and what normal form the field takes near the point (the analog of the Laurent/Weierstrass theory), and settle uniqueness and classification in the exceptional ranks $\rho = 1$ and $\rho = n-1$ when the flux conditions (11.1)/(11.1)* of the 1949 paper fail, beyond the addition of the specific line-integral term identified there.

References

1. Kunihiko Kodaira, 'Harmonic fields in Riemannian manifolds (generalized potential theory)', Annals of Mathematics (2) 50(3) (1949), 587-665. DOI: 10.2307/1969552. URL: https://doi.org/10.2307/1969552. Reprinted as the chapter '( 17 ) Harmonic fields in Riemannian manifolds (generalized potential theory)', pp. 172-250, in Kunihiko Kodaira: Collected Works, Volume I, Princeton University Press (2015). DOI: 10.1515/9781400869855-018. URL: https://doi.org/10.1515/9781400869855-018
2. Kunihiko Kodaira, 'Relations between harmonic fields in Riemannian manifolds', Mathematica Japonicae 1 (1948), 6-23. zbMATH: 0041.29503. URL: https://zbmath.org/?q=an:0041.29503
3. Kunihiko Kodaira, 'The theorem of Riemann-Roch on compact analytic surfaces', American Journal of Mathematics 73(4) (1951), 813-875. DOI: 10.2307/2372120. URL: https://doi.org/10.2307/2372120
4. Matthew P. Gaffney, 'The heat equation method of Milgram and Rosenbloom for open Riemannian manifolds', Annals of Mathematics (2) 60(3) (1954), 458-466. DOI: 10.2307/1969846. URL: https://doi.org/10.2307/1969846
5. Matthew P. Gaffney, 'Hilbert space methods in the theory of harmonic integrals', Transactions of the American Mathematical Society 78 (1955), 426-444. DOI: 10.2307/1993072. URL: https://doi.org/10.2307/1993072
6. Mitsuru Nakai and Leo Sario, 'Harmonic fields with given boundary behavior in Riemannian spaces', Journal d'Analyse Mathematique 18 (1967), 245-257. DOI: 10.1007/BF02798047. URL: https://doi.org/10.1007/BF02798047
7. I. V. Skripnik, 'A-harmonic fields with singularities', Ukrainian Mathematical Journal 17(4) (1965), 130-133 (Russian). zbMATH: 0156.42402. URL: https://zbmath.org/?q=an:0156.42402
8. Friedrich Hirzebruch, 'Topological Methods in Algebraic Geometry', Grundlehren der mathematischen Wissenschaften 131, Springer (1966; English translation of 'Neue topologische Methoden in der algebraischen Geometrie', Ergebnisse der Mathematik und ihrer Grenzgebiete 9, 1956). DOI: 10.1007/978-3-642-62018-8. URL: https://doi.org/10.1007/978-3-642-62018-8
9. Michael F. Atiyah and Isadore M. Singer, 'The index of elliptic operators on compact manifolds', Bulletin of the American Mathematical Society 69 (1963), 422-433. DOI: 10.1090/S0002-9904-1963-10957-X. URL: https://doi.org/10.1090/S0002-9904-1963-10957-X
10. Michael F. Atiyah, Vijay K. Patodi, and Isadore M. Singer, 'Spectral asymmetry and Riemannian geometry. I', Mathematical Proceedings of the Cambridge Philosophical Society 77 (1975), 43-69. DOI: 10.1017/S0305004100049410. URL: https://doi.org/10.1017/S0305004100049410
11. Steven Zucker, 'L2 cohomology of warped products and arithmetic groups', Inventiones Mathematicae 70 (1982), 169-218. DOI: 10.1007/BF01390727. URL: https://doi.org/10.1007/BF01390727
12. Jeff Cheeger and Mikhael Gromov, 'Bounds on the von Neumann dimension of L2-cohomology and the Gauss-Bonnet theorem for open manifolds', Journal of Differential Geometry 21 (1985), 1-34. DOI: 10.4310/jdg/1214439461. URL: https://doi.org/10.4310/jdg/1214439461
13. John Roe, 'An index theorem on open manifolds. I', Journal of Differential Geometry 27 (1988), 87-113. DOI: 10.4310/jdg/1214441652. URL: https://doi.org/10.4310/jdg/1214441652 (continued as Part II, ibid. 115-136)
14. Mikhael Gromov and Mikhail A. Shubin, 'The Riemann-Roch theorem for elliptic operators and solvability of elliptic equations with additional conditions on compact subsets', Inventiones Mathematicae 117(1) (1994), 165-180. DOI: 10.1007/BF01232238. URL: https://doi.org/10.1007/BF01232238
15. Mikhael Gromov and Mikhail A. Shubin, 'The Riemann-Roch theorem for elliptic operators', in: I. M. Gelfand Seminar, Advances in Soviet Mathematics 16, Part 1, American Mathematical Society (1993), 211-241. DOI: 10.1090/advsov/016.1/07. URL: https://doi.org/10.1090/advsov/016.1/07
16. Mikhael Gromov and Mikhail A. Shubin, 'The Riemann-Roch theorem for elliptic operators and solvability of elliptic equations with additional conditions on compact subsets', Journées équations aux dérivées partielles (1993), 1-13. DOI: 10.5802/jedp.455. URL: https://doi.org/10.5802/jedp.455
17. Mikhail A. Shubin, 'L2 Riemann-Roch theorem for elliptic operators', Geometric and Functional Analysis 5(2) (1995), 482-527. DOI: 10.1007/BF01895677. URL: https://doi.org/10.1007/BF01895677
18. Bert-Wolfgang Schulze and Nikolai Tarkhanov, 'The Riemann-Roch theorem for manifolds with conical singularities', Osaka Journal of Mathematics 36(4) (1999), 1011-1047. zbMATH: 0964.58021. URL: https://zbmath.org/?q=an:0964.58021
19. Tamás Hausel, Eugenie Hunsicker, and Rafe Mazzeo, 'Hodge cohomology of gravitational instantons', Duke Mathematical Journal 122(3) (2004), 485-548. DOI: 10.1215/S0012-7094-04-12233-X. URL: https://doi.org/10.1215/S0012-7094-04-12233-X (preprint: arXiv:math/0207169, https://arxiv.org/abs/math/0207169)
20. Minh Kha and Peter Kuchment, 'Liouville-Riemann-Roch theorems on abelian coverings', Lecture Notes in Mathematics 2245, Springer (2021). DOI: 10.1007/978-3-030-67428-1. URL: https://doi.org/10.1007/978-3-030-67428-1 (preprint: arXiv:1909.13386, https://arxiv.org/abs/1909.13386)