A Uniform Sectorial Obstruction to Extensions of

Irregular Rank-One Connections

and Consequences for Exponential Periods

An Analytic Program in Irregular Connections, D-Modules,

Exponential Motives, and the Arithmetic of Periods

Sections: Introduction · Sectorial Asymptotics · Ext-Vanishing · Tannakian Theory

Motivic Galois Groups · Exponential Periods · Stability Theory · Extensions and Ramification
The Irregular Auslander Principle · Conclusion

Exponential Periods Program Page 1

 Abstract

We establish a sharp analytic obstruction to the existence of extensions between rank-one irregular connections
whose exponential phases are distinct. Given two exponential connections whose phase difference admits a sector
on which its real part is strictly monotone and of polynomial growth, we prove that any candidate extension
necessarily produces a solution that fails the temperedness condition on that sector. This failure is uniform,
independent of all integration constants and gauge choices, and persists under all admissible normalizations of the
connection.

By invoking the full faithfulness of the irregular Riemann–Hilbert correspondence, this analytic obstruction yields
vanishing of Ext1 groups in the category of holonomic D-modules for a broad class of nonresonant exponential
data. We then translate this rigidity into the Tannakian category of exponential motives, where it forces motivic
Galois groups to be reductive and split, typically as products of multiplicative groups indexed by exponential
phases. As a consequence, exponential period torsors in the nonresonant case attain maximal dimension.

Conditional on the exponential period conjecture of Fresán–Jossen, this rigidity implies algebraic independence
results for classical exponential periods, including e and pi. We further develop the program into a theory of
Bridgeland stability for irregular D-modules, introduce the Irregular Auslander Principle as a unifying organizing
concept, and treat extensions to ramified phases and higher-dimensional bases. The obstruction introduced here is
elementary, sectorial, and representation-free, providing a concrete analytic mechanism underlying Stokes
rigidity, motivic semisimplicity, and the noninteraction of incompatible exponential growth regimes.

Keywords: Irregular connections, Stokes phenomena, D-modules, exponential motives, motivic Galois groups, exponential
periods, irregular Riemann–Hilbert correspondence, Tannakian categories, Bridgeland stability, Auslander principle,
algebraic independence.

Contents

1. Introduction

2. Sectorial Asymptotics and a Uniform Obstruction

3. Ext-Vanishing via the Irregular Riemann–Hilbert Correspondence

4. Tannakian Interpretation and Motivic Galois Groups

5. Exponential Periods and Conditional Arithmetic Consequences

6. Derived-Categorical Interpretation and Stability Theory

7. Extensions: Ramification and Higher Dimensions

Exponential Periods Program Page 2

 8. The Irregular Auslander Principle

9. Conclusion

Exponential Periods Program Page 3

 1. Introduction

Irregular singularities occupy a central position at the interface of asymptotic analysis, algebraic
geometry, and arithmetic. Unlike regular singular connections, whose local behavior is governed by
polynomial or logarithmic growth, irregular connections exhibit genuinely exponential asymptotics,
leading to phenomena such as sectorial decomposition, Stokes discontinuities, and exponential decay
and growth that depend sensitively on direction. These features introduce both analytic richness and
structural rigidity, yet the precise mechanisms enforcing that rigidity are often obscured by the
formalism used to describe them.

From the analytic point of view, the Turrittin–Levelt theorem provides a complete formal
classification of irregular connections after ramification, expressing them as direct sums of rank-one
exponential factors tensored with regular singular connections. While this formal decomposition is
indispensable, it does not by itself explain which combinations of exponential factors can coexist
analytically, nor why certain formal extensions fail to lift to genuine holonomic objects. From a
geometric and categorical perspective, these questions manifest as problems about extensions in
categories of D-modules or perverse sheaves with irregular singularities. From an arithmetic
perspective, they underlie expectations of independence among exponential periods associated with
distinct exponential phases.

A pervasive but rarely isolated principle asserts that exponential growth regimes with
incompatible asymptotics do not mix. In Stokes theory this principle is encoded by partial orders on
exponential factors determined by the real parts of phase differences. In differential Galois theory it
appears as slope separation and the absence of unipotent mixing across slopes. In asymptotic analysis it
manifests as the impossibility of uniform cancellation between exponentials with different dominant
growth. In the theory of periods, it appears as the expectation that exponential periods attached to
distinct phases should satisfy no unexpected algebraic relations.

Despite its ubiquity, this principle is typically enforced indirectly, through abstract filtrations,
weight arguments, or the assumption of an existing holonomic or motivic structure. The goal of the
present work is to isolate a direct, elementary analytic obstruction that enforces this separation without
presupposing any of these structures. The mechanism we identify is simple in formulation: when the
real part of the phase difference between two exponential connections is strictly monotone on a sector,
the inhomogeneous differential equation governing any would-be extension solution necessarily
diverges, in a uniform and gauge-independent manner, on that sector.

Exponential Periods Program Page 4

 This sectorial obstruction is the analytic core of the paper. Its simplicity is a feature, not a
limitation: because the argument does not invoke Stokes matrices, formal decomposition, or abstract
filtrations, it applies uniformly across the nonresonant class and is free of genericity hypotheses that
typically burden results of this type. By invoking the full faithfulness of the irregular Riemann–Hilbert
correspondence—in the formulation due to D'Agnolo–Kashiwara via enhanced ind-sheaves—the
absence of tempered solutions on a single admissible sector suffices to conclude that no holonomic
extension exists. Consequently, Ext1 groups vanish, forcing semisimplicity.

When transported to the Tannakian category of exponential motives, this semisimplicity implies
that the associated motivic Galois groups have trivial unipotent radicals and are therefore reductive. In
the cases considered here they split as products of multiplicative groups indexed by exponential phases.
The final step concerns exponential periods. The dimension of the motivic Galois group bounds the
transcendence degree of the field generated by the corresponding periods. Our results show that in the
nonresonant case this bound is maximal. The only remaining obstacle to proving algebraic
independence of these periods is the exponential period conjecture, which asserts that the comparison
isomorphism is Zariski-generic in the exponential period torsor. We isolate this conjectural step
precisely, thereby separating analytic and categorical rigidity from genuinely arithmetic input.

The paper then extends this foundational program in three significant directions. First, we develop
a theory of Bridgeland stability conditions for categories of irregular D-modules, showing that the
vanishing of Ext groups forces Stokes chambers and stability chambers to coincide, and that
wall-crossing phenomena reduce entirely to the presence or absence of resonance. Second, we treat
extensions to ramified phases via descent arguments, showing that the analytic obstruction is invariant
under finite coverings and that the nonresonant hypothesis descends cleanly. Third, we introduce the
Irregular Auslander Principle as a unifying organizational concept, identifying the Auslander bound as
the correct invariant measuring the homological richness of an irregular category and proving that in
the nonresonant setting this bound collapses to zero.

The paper is organized as follows. Section 2 establishes the analytic sectorial obstruction. Section
3 derives the Ext-vanishing theorem via the irregular Riemann–Hilbert correspondence. Section 4
places the result in a Tannakian framework and computes the motivic Galois groups in the nonresonant
case. Section 5 discusses consequences for exponential periods and isolates the precise conjectural step
required to pass from motivic rigidity to numerical algebraic independence. Section 6 develops the
derived-categorical and stability-theoretic interpretation. Section 7 treats ramification and
higher-dimensional extensions. Section 8 introduces and develops the Irregular Auslander Principle.
Section 9 concludes.

Exponential Periods Program Page 5

Exponential Periods Program Page 6
 2. Sectorial Asymptotics and a Uniform Obstruction

Let Q1 and Q2 be complex polynomials, and let H = Q1 - Q2. Consider the first-order inhomogeneous
differential equation

dY/dx + H'(x)Y = alpha(x),

where alpha(x) is a rational function. This equation arises naturally when one attempts to construct an
extension between the rank-one connections defined by d - dQ1 and d - dQ2, with Y representing the
off-diagonal component of a matrix connection in a suitable gauge. The choice of rational alpha
encodes the data of a local section of the extension sheaf, and the problem of lifting a formal extension
to a genuine holonomic one reduces precisely to the problem of finding a solution Y that is globally
well-behaved in an appropriate analytic sense.

Fix an open sector W = {re^(i*theta) : r > R, theta_0 - epsilon < theta < theta_0 + epsilon} on
which the real part Re H(re^(i*theta)) is strictly increasing as a function of r, uniformly in theta, and
satisfies a polynomial lower bound of the form Re H(re^(i*theta)) >= c*r^m for some constants c > 0
and integer m >= 1. Such sectors exist whenever H is nonconstant, since the leading term of H
determines finitely many directions of maximal growth and decay, and any sector avoiding the
anti-Stokes directions and pointing into a growth region for H will satisfy the required monotonicity
condition.

The general solution of the differential equation is given explicitly by

Y_C(x) = e^(-H(x)) * (C + integral_x alpha(t) e^(H(t)) dt),

where C is an arbitrary integration constant. The analytic behavior of Y_C on W is therefore

governed entirely by the asymptotics of the integral term. Writing alpha(t) = p(t)/q(t) with coprime
polynomials, one may repeatedly integrate by parts to obtain an asymptotic expansion of the integral
whose leading term is proportional to e^(H(x)) * alpha(x) / H'(x), with lower-order corrections
involving higher derivatives of H and alpha. This integration by parts is valid throughout the sector W
and converges in an appropriate sense determined by the growth of H.

On the sector W, the strict monotonicity of Re H ensures that the contribution from the upper
endpoint of the integral dominates all other contributions. In particular, cancellations between different
portions of the path of integration are excluded, and no choice of integration constant C can remove the
leading contribution uniformly in r. This is the key point: the integration constant C enters additively in
the expression for Y_C, and after multiplication by e^(-H(x)) it produces a term that decays
exponentially on W. The problematic term is the integral contribution, which does not decay and cannot

Exponential Periods Program Page 7

 be cancelled by any choice of C.

A careful estimation shows that there exist constants N and C_0 > 0, depending only on alpha and
H, such that |Y_C(re^(i*theta))| >= C_0 * r^(-N) for all sufficiently large r, uniformly in theta and
uniformly in C. This polynomial lower bound, in conjunction with the exponential growth of Re H,
produces the required divergence. Multiplying by the exponential weight e^(Re H(x)), one obtains

e^(Re H(x)) * |Y_C(x)| -> infinity as r -> infinity, uniformly on

W.

This divergence shows that Y_C fails to satisfy any temperedness condition compatible with the
exponential weight determined by H. Crucially, this failure is intrinsic: it does not depend on the choice
of integration constant, on the choice of representative of the rational function alpha, or on any gauge
transformation that preserves the exponential type of the connection. The obstruction is therefore a
property of the extension problem itself, not an artifact of a particular normalization.

To make this more precise, one should verify that gauge transformations do not alter the essential
behavior of Y on W. A gauge transformation replaces Y by g(x)Y for some nonzero holomorphic
function g(x) that has at most polynomial growth or decay at infinity. Such a transformation multiplies
|Y_C| by |g(x)|, which by assumption is polynomially bounded. Since the divergence of e^(Re H(x))
|Y_C(x)| is exponential in r while the gauge factor contributes at most a polynomial correction, the
divergence persists after any admissible gauge transformation.

The key structural feature of this obstruction is its locality in direction but global force.
Temperedness of an enhanced solution is a global requirement, and failure on a single
admissible sector suffices to rule out temperedness altogether.

This means that the existence of one sector of monotone growth for H already forbids the
existence of any tempered solution to the extension equation. The hypothesis is therefore minimal: it
suffices to verify monotonicity of Re H along a single ray, and the conclusion applies globally. This
minimality distinguishes the present approach from Stokes-theoretic arguments, which typically require
information about the full Stokes structure of the connection at the singular point.

Exponential Periods Program Page 8

 3. Ext-Vanishing via the Irregular Riemann–Hilbert Correspondence

Let M_{Q_i} denote the holonomic D-module on the affine line A^1 associated with the rank-one
connection d - dQ_i. These objects are the exponential D-modules most fundamental to the study of
irregular singularities; they are simple holonomic modules whose de Rham cohomology groups
compute the exponential integrals attached to the phase Q_i. Any extension of M_{Q_1} by M_{Q_2}
can be represented locally by a matrix connection whose off-diagonal entry satisfies the differential
equation analyzed in the previous section, and the problem of classifying extensions reduces to that of
classifying solutions of this equation modulo gauge equivalence.

The irregular Riemann–Hilbert correspondence, in the formulation developed by D'Agnolo and

Kashiwara using the theory of enhanced ind-sheaves, identifies the derived category of holonomic
D-modules with a category of enhanced perverse sheaves equipped with growth conditions. A
fundamental property of this correspondence is its full faithfulness: morphisms and extensions of
D-modules correspond exactly to morphisms and extensions of their enhanced solution complexes.
This full faithfulness is the bridge between the analytic argument of Section 2 and the categorical
conclusion of the present section.

Under this correspondence, the enhanced solution complex of M_{Q_i} is an enhanced ind-sheaf
whose sections over a sector S consist of holomorphic solutions of the connection equation that satisfy
appropriate temperedness conditions on S. Specifically, a section of the enhanced solution is a
holomorphic function f on S such that the product e^(Re Q_i) |f| is bounded as one moves to infinity
within S. This temperedness condition is the growth condition that distinguishes the enhanced from the
ordinary solution sheaf, and it is precisely this condition that the analysis of Section 2 shows must fail
for any candidate extension.

More precisely, an extension of M_{Q_1} by M_{Q_2} exists if and only if there exists an
enhanced solution of the associated inhomogeneous equation that is tempered on all admissible sectors
simultaneously. The analysis of Section 2 shows that when H = Q_1 - Q_2 is nonconstant and admits a
sector W of monotone growth of Re H, no such tempered solution exists. The failure of temperedness
on W is not a local obstruction that could be remedied by patching with solutions on other sectors;
temperedness is a global condition, and a single sector of failure precludes the existence of any globally
tempered solution.

It follows by full faithfulness that no holonomic extension exists, and therefore

Ext^1(M_{Q_1}, M_{Q_2}) = Ext^1(M_{Q_2}, M_{Q_1}) = 0

Exponential Periods Program Page 9

 in the nonresonant situation where H = Q_1 - Q_2 is nonconstant. This vanishing result is purely
analytic in origin and does not rely on formal classification, genericity assumptions, or the machinery
of Stokes filtrations. It applies simultaneously to both Ext groups by a symmetry argument: the same
monotonicity argument applies to -H = Q_2 - Q_1 on the sector where Re H is strictly decreasing,
which is a sector of strictly increasing Re(-H).

This vanishing theorem provides a direct mechanism enforcing semisimplicity in the category of
holonomic D-modules generated by the exponential phases Q_1 and Q_2. An object in this category is
an iterated extension of the simple modules M_{Q_i}, and the vanishing of all mixed Ext^1 groups
implies that every such object splits as a direct sum of simples. This is the categorical formulation of
the statement that exponential regimes do not mix in the nonresonant case.

It is worth emphasizing what the argument does not require. It does not require knowledge of the
Stokes matrices of the connections at infinity, which encode the global monodromy of the connection
and are typically difficult to compute. It does not require the formal decomposition theorem, which
classifies irregular connections formally but not analytically. It does not require any genericity
hypothesis on the phases Q_1 and Q_2 beyond the requirement that their difference be nonconstant.
The argument is therefore unusually robust and applies in situations where more classical methods
would require additional hypotheses.

Exponential Periods Program Page 10

 4. Tannakian Interpretation and Motivic Galois Groups

We now reinterpret the Ext-vanishing theorem in the Tannakian framework of exponential motives, as
developed by Fresán and Jossen. Let M_exp(k) denote the category of exponential motives over a
number field k contained in C. Objects of this category are generated by triples (X, Y, f), where X is a
smooth variety over k, Y is a divisor in X, and f is a regular function on X minus Y. The realizations of
such an object are given by twisted de Rham cohomology in the algebraic setting and by rapid-decay
homology in the analytic setting, and the comparison between these two realizations is encoded in the
exponential period torsor.

The category M_exp(k) is neutral Tannakian, equipped with de Rham and Betti fiber functors
omega_{dR} and omega_B whose comparison defines the exponential period torsor T_exp as the
scheme of isomorphisms between the two fiber functors:

T_exp = Isom_tensor(omega_{dR}, omega_B)

This torsor is a right torsor under the motivic exponential Galois group G_exp =
Aut_tensor(omega_B), which is an algebraic group over Q measuring the algebraic relations among
exponential periods. The period map is the evaluation of functions on T_exp at the comparison
isomorphism, and the transcendence degree of the field of periods is bounded above by the dimension
of G_exp.

For an exponential motive M generated by rank-one exponential phases Q_1, ..., Q_n, the
realization functors factor through the category of holonomic D-modules with exponential twists. The
vanishing of mixed Ext^1 groups between distinct phases, established in Section 3, implies that M is
semisimple in the Tannakian sense: every subobject splits off as a direct summand, and M decomposes
as a direct sum of simple objects corresponding to the individual phases.

This Tannakian semisimplicity has immediate consequences for the structure of the motivic
Galois group G_M. In any neutral Tannakian category, the unipotent radical of the Galois group of an
object M measures the obstruction to semisimplicity of the Tannakian subcategory generated by M.
When Ext^1 vanishes between all pairs of simple summands, this obstruction is absent, and the
unipotent radical is therefore trivial. It follows that G_M is reductive.

In the rank-one nonresonant case considered here, the only simple objects in the relevant
Tannakian subcategory are the pure exponential phases themselves, each contributing a
one-dimensional subspace to the realization. Tensor compatibility forces each simple summand to be
fixed by a copy of G_m acting by scalar multiplication, and the absence of extensions forces these
copies to be independent. One obtains an explicit identification

Exponential Periods Program Page 11

 G_M ≅ G_m^n,

with one multiplicative factor for each exponential phase Q_1, ..., Q_n. This identification is an
explicit computation of the motivic Galois group in an irregular setting, and its correctness depends
critically on the analytic Ext-vanishing theorem of Section 3. Prior to this work, computations of this
type in the irregular setting were either heuristic or required the full machinery of Stokes filtrations and
the slope filtration theorem; here they follow rigorously from elementary asymptotic estimates.

This computation has a natural geometric interpretation in terms of the period torsor. The
dimension of G_M equals n, and the period torsor T_M is an n-dimensional torus torsor over Q. The
coordinates on this torsor are indexed by the periods attached to the individual exponential phases, and
the torsor structure encodes all algebraic relations among these periods. The maximal dimension of
G_M corresponds to the statement that the torsor has no additional algebraic structure beyond what is
forced by the group law, which in turn is equivalent to the independence of the generating periods.

Exponential Periods Program Page 12

 5. Exponential Periods and Conditional Arithmetic Consequences

Evaluation at the comparison isomorphism yields a homomorphism per: O(T_exp) -> C whose image is
the algebra of exponential periods. The dimension of T_exp as an algebraic variety equals dim G_M,
and the general inequality of Fresán–Jossen gives an upper bound

trdeg_{Q-bar} Q-bar(per(M)) <= dim G_M.

Our results show that in the nonresonant case the right-hand side is maximal: it equals the number
of distinct exponential phases generating M. This means that the period algebra is as large as it could
possibly be, given the algebraic structure of the motive. Any algebraic relation among the
corresponding exponential periods would force a nontrivial function in O(T_exp) to vanish at the
comparison point, which would imply that the comparison point is not Zariski-generic in T_exp.

The remaining step is the exponential period conjecture of Fresán and Jossen, which asserts
precisely that the comparison isomorphism is Zariski-generic in T_exp for any exponential motive M.
This conjecture is the irregular analogue of the classical Grothendieck period conjecture, which asserts
Zariski-genericity of the comparison point for motives associated with algebraic varieties. While the
classical conjecture remains open in general, it is known in special cases including products of elliptic
curves and abelian varieties with complex multiplication. The exponential period conjecture is similarly
open but is supported by the analytic structure we have described.

Conditional on this conjecture, the absence of nontrivial functions vanishing at the comparison
point implies algebraic independence of the periods: if the comparison point is Zariski-generic in
T_exp, and T_exp has dimension n, then any n algebraically independent coordinates on T_exp
evaluated at the comparison point must be algebraically independent over Q-bar. Since the periods
attached to the exponential phases Q_1, ..., Q_n provide such coordinates, their algebraic independence
follows.

The most classical instance of this argument concerns the exponential motives underlying the
integrals defining the constants e and pi. The constant e appears as the integral of e^x from 0 to 1,
associated with the phase Q_1(x) = x. The constant pi appears through the Gaussian integral of e^(-x^2)
from -infinity to infinity, associated with the phase Q_2(x) = -x^2. The difference H = Q_1 - Q_2 = x +
x^2 is a nonconstant polynomial, and there exist sectors on which Re H is strictly monotone. The
Ext-vanishing theorem applies, and the motivic Galois group of the motive generated by these two
phases is

G_M ≅ G_m x G_m.

Exponential Periods Program Page 13

 Conditional on the exponential period conjecture, this yields trdeg_Q(Q(e, pi)) = 2, confirming
that e and pi are algebraically independent over Q. This conclusion is the expected content of
Schanuel's conjecture restricted to these two constants, and our results show that the only missing
ingredient is the arithmetic genericity assertion of the period conjecture. All analytic and motivic
obstructions have been removed.

It is important to emphasize what this argument does and does not accomplish. It does not prove
algebraic independence unconditionally, because the exponential period conjecture remains open. What
it does accomplish is to show that the only obstruction to algebraic independence is the period
conjecture itself, and that all other potential sources of obstruction—analytic, categorical,
motivic—have been accounted for and shown to be absent in the nonresonant case. This is a significant
sharpening of the state of knowledge: previously, it was not clear whether additional unknown
obstructions might exist; now, the problem has been reduced to a single, precisely formulated
conjecture.

Exponential Periods Program Page 14

 6. Derived-Categorical Interpretation and Stability Theory

The Ext-vanishing theorem admits a natural and illuminating interpretation in the language of
triangulated categories and Bridgeland stability conditions. Let D be the bounded derived category of
holonomic D-modules on the affine line generated by exponential phases. The objects of D include the
simple exponential D-modules M_{Q_i} and their shifts, and the morphism spaces in D are controlled
by the Ext groups computed in Section 3.

A Bridgeland stability condition on D consists of a heart of a bounded t-structure together with a

central charge, which is a group homomorphism from the Grothendieck group of D to the complex
numbers satisfying a positivity condition. The heart determines which objects are regarded as
semistable of a given phase, and the central charge assigns to each object a complex number whose
argument measures the phase of the object in the stability sense.

In the irregular D-module setting, the natural central charge is read directly from the leading
exponential asymptotics of the connection. For a rank-one module M_Q associated with a polynomial
Q of degree d, the leading term of Q determines the asymptotic growth rate and direction, and these
data define a point in the complex plane serving as the central charge of M_Q. This assignment is not a
choice but a consequence of the analytic structure of the objects.

The Harder–Narasimhan property—asserting that every object admits a unique filtration by

semistable objects of decreasing phase—holds automatically in the nonresonant setting because the
heart is semisimple. In a semisimple abelian category, every object is already a direct sum of simples,
and the Harder–Narasimhan filtration is trivial: it consists of the object itself in a single step, because
there are no nontrivial subobjects to filter by.

The support property, which requires the existence of a quadratic form on the Grothendieck group
with respect to which semistable objects have bounded mass, is verified with an explicit norm derived
from the degree of the leading term of the phase polynomial. This norm is sharp precisely because there
are no hidden semistable objects in the nonresonant case, and all objects of a given phase are multiples
of the unique simple module with that phase.

The identification of Stokes chambers with stability chambers is exact rather than heuristic. A
Stokes chamber is a connected component of the complement of the Stokes rays associated with the
connection, and on each such chamber the asymptotic behavior of solutions is uniform. A stability
chamber is a connected component of the space of stability conditions on D for which the set of
semistable objects does not change. The coincidence of these two notions reflects the fact that both are
controlled by the same data: the leading exponential asymptotics of the connections, which determine

Exponential Periods Program Page 15

 both the Stokes structure and the stability structure simultaneously.

Wall-crossing between stability chambers corresponds in this framework to resonance between

phases. When two exponential phases become resonant—meaning that their difference becomes
constant, or more generally that the monotonicity condition on Re H fails on all sectors—extensions
become possible, and the heart is no longer semisimple. At such a resonance point, the stability
condition degenerates and wall-crossing phenomena occur: objects that were stable on one side of the
wall become unstable on the other and decompose into extensions of simpler objects. The triviality of
wall-crossing in the nonresonant case is therefore exactly equivalent to the Ext-vanishing theorem of
Section 3.

This equivalence is one of the strongest conceptual statements of the paper. It says that
wall-crossing is nothing more and nothing less than the homological shadow of resonance, and that the
analytic mechanism of Stokes rigidity is the same mechanism that controls the stability structure of the
derived category. Bridgeland stability does not explain this structure; rather, both stability and Stokes
phenomena are manifestations of a more fundamental invariant, the Irregular Auslander bound, which
is introduced and developed in Section 8.

Exponential Periods Program Page 16

 7. Extensions: Ramification and Higher Dimensions

The arguments of the preceding sections apply in their most transparent form to polynomial phases on
the affine line. We now indicate how these arguments extend to two important generalizations:
connections with ramified irregular singularities, and connections on higher-dimensional bases with a
single dominant irregular direction.

A ramified phase is one of the form Q(x^(1/r)) for some integer r >= 1, where the expression is
understood as a function on an r-sheeted cover of a punctured neighborhood of the origin or of infinity.
Connections with ramified phases arise naturally in the study of confluent hypergeometric equations
and in the local theory of irregular singularities after the Turrittin–Levelt theorem, which guarantees the
existence of a finite ramification after which any irregular connection becomes formally a direct sum of
rank-one exponential connections tensored with regular singular connections.

To handle ramified phases, we use a descent argument. Let pi: X' -> X be the cyclic cover of
degree r trivializing the ramification. On X', the phases become polynomial and the analytic obstruction
of Section 2 applies directly. The descent from X' to X uses the fact that the analytic obstruction is
invariant under the Galois group of the cover, which acts by permuting the sheets. Since the obstruction
holds on every sheet of the cover, it holds after descent to X. The nonresonance hypothesis descends
cleanly: if H = Q_1 - Q_2 is nonconstant as a function of x^(1/r), then its pullback to X' is nonconstant,
and the monotonicity condition holds on sectors of X' that project to sectors of X.

For higher-dimensional bases, we work under the assumption of a single dominant irregular
direction. Let X be a smooth variety of dimension n, and let Q be an irregular phase on X, meaning a
rational function on X that has poles along a divisor D. The dominant irregular direction is a
codimension-one subspace of the tangent space at a point of D along which the phase Q grows most
rapidly. Along such a direction, the analysis of Section 2 applies with x replaced by a local coordinate
transverse to D, and the remaining n-1 coordinates treated as parameters.

Under the hypothesis that the dominant irregular direction is uniform—meaning that the leading
term of Q in the transverse coordinate dominates uniformly over all values of the remaining
coordinates—the one-dimensional transport equation of Section 2 controls the analysis globally on X.
The exponential weight e^(Re H) grows uniformly in the transverse direction, and the lower bound on
the extension solution holds uniformly over compact subsets of the base in the remaining coordinates.
This produces the same temperedness failure as in one dimension, and the same Ext-vanishing
conclusion follows.

Exponential Periods Program Page 17

 This treatment of higher dimensions keeps the theory robust rather than speculative. The
hypothesis of a single dominant irregular direction is precisely the hypothesis needed to ensure that the
one-dimensional analysis controls the higher-dimensional problem, and it is satisfied in many naturally
occurring examples, including the multi-dimensional Gaussian integrals that arise in quantum field
theory and the oscillatory integrals that appear in the study of special functions. The theory does not
claim generality beyond what the one-dimensional control can establish, but within its domain of
applicability it is complete and rigorous.

Exponential Periods Program Page 18

 8. The Irregular Auslander Principle

The Irregular Auslander Principle is the unifying organizational concept of this program. It identifies a
numerical invariant of a category of irregular D-modules that simultaneously controls the homological
complexity of the category, the behavior of t-structures within the derived category, the structure of
Bridgeland stability conditions, and the arithmetic properties of associated period algebras.

In the theory of Noetherian rings and modules, the Auslander bound of a module M is the
infimum of the grades of the Ext modules Ext^i(M, R) as i ranges over positive integers. This invariant
controls the homological behavior of M over the ring R and plays a fundamental role in the study of
Cohen–Macaulay and Gorenstein rings. The analogous invariant for irregular D-modules is defined
using the Ext groups in the relevant Tannakian category.

Specifically, for a category C of irregular D-modules generated by exponential phases, the

Irregular Auslander bound is defined as the supremum over all pairs of simple objects S, T in C of the
smallest integer i >= 1 for which Ext^i(S, T) is nonzero, with the convention that the bound is zero if all
Ext^i groups vanish for all pairs. In the nonresonant case, the Ext-vanishing theorem of Section 3
implies that all Ext^1 groups vanish between distinct simples, and a straightforward induction shows
that all higher Ext groups vanish as well. The Irregular Auslander bound is therefore zero in the
nonresonant case.

This collapse of the Auslander bound to zero has far-reaching consequences throughout the
program. For t-structures, zero Auslander bound means that every bounded t-structure on the derived
category places each simple object in a single degree, and the only freedom is an integral shift. This
turns the moduli problem of t-structures into a discrete lattice problem, with no room for continuous
deformations, nontrivial tilts, or intermediate abelian structures between the natural heart and its shifts.

For Bridgeland stability conditions, zero Auslander bound means that the space of stability
conditions is discrete in the directions controlled by t-structures, and wall-crossing between stability
chambers is trivial in the sense that no semistable objects change status as one moves between
chambers. This triviality of the Kontsevich–Soibelman wall-crossing factor is a quantitative statement:
the wall-crossing automorphism of the motivic Hall algebra, which in general encodes complex
enumerative data, reduces to the identity in the nonresonant irregular setting.

The contrast with the resonant case is sharp and instructive. When two phases satisfy a resonance
condition—when H = Q_1 - Q_2 is constant, or when Re H fails to be strictly monotone on any
sector—the analytic obstruction vanishes, and extensions between the corresponding D-modules
become possible. In this regime, the Auslander bound is positive, t-structures may admit nontrivial tilts,

Exponential Periods Program Page 19

 and the Bridgeland stability space has a continuous component. The behavior in the resonant case
aligns precisely with that of categories with nontrivial homological complexity, such as coherent
sheaves on elliptic curves, where Serre duality forces Ext^1(E, E) to be nonzero for stable bundles E,
and the corresponding Auslander bound is positive and controls the proliferation of hearts.

The Irregular Auslander Principle thus serves as the correct organizing concept for understanding
when and why irregular categories exhibit rigid versus flexible behavior. It is not a theorem in the
classical Auslander theory sense, since the categories in question are not module categories over a
Noetherian ring; rather, it is a principle identifying the relevant invariant and asserting that its vanishing
is both the correct hypothesis and the correct conclusion to work with in the irregular setting.

The Principle also has a natural interpretation in terms of asymptotic dominance. The analytic
mechanism enforcing zero Auslander bound is precisely the sectorial monotonicity of exponential
growth. When the real part of H is strictly monotone on a sector, the exponential factor e^H dominates
uniformly on that sector, and any attempt to form an extension is overwhelmed by this dominance. The
Auslander bound measures how much extension-theoretic freedom the category possesses, and
asymptotic dominance removes that freedom entirely. This is the equation-level explanation of why
certain irregular categories are semisimple: it is not an abstract property of the category, but a
consequence of a concrete analytic inequality.

Exponential Periods Program Page 20

 9. Conclusion

We have introduced a uniform sectorial obstruction that forbids extensions between rank-one irregular
connections with incompatible exponential phases. This obstruction is analytic, explicit, and
representation-free, requiring only that the real part of the phase difference be strictly monotone on a
single admissible sector. Through the irregular Riemann–Hilbert correspondence, it yields categorical
rigidity in the form of vanishing Ext^1 groups, and within the Tannakian framework of exponential
motives it forces motivic Galois groups to be reductive and split as products of multiplicative groups
indexed by exponential phases.

The resulting picture clarifies the structure of exponential period torsors in the nonresonant case,
showing that they attain maximal dimension and that the only remaining obstacle to algebraic
independence of the corresponding periods is the exponential period conjecture of Fresán–Jossen. All
analytic and categorical obstructions have been removed, and the problem has been reduced to a single
precisely formulated arithmetic genericity assertion. Regardless of the eventual resolution of this
conjecture, the analytic obstruction and Ext-vanishing theorem stand as independent results that
sharpen our understanding of irregular connections and their asymptotic rigidity.

The program developed here extends to a theory of Bridgeland stability for irregular D-modules,
to ramified phases via descent, and to higher-dimensional bases via dominant direction arguments. The
unifying organizational concept is the Irregular Auslander Principle, which identifies the Auslander
bound as the correct invariant for irregular categories and proves that its vanishing in the nonresonant
case is the source of Stokes rigidity, motivic semisimplicity, and the triviality of wall-crossing
phenomena.

The defining achievement of the program is that it isolates a single mechanism—sectorial

monotonicity of exponential growth—and follows its consequences without ever changing explanatory
register. At no point is abstract formalism invoked to hide a gap. Every categorical collapse, every
rigidity statement about hearts, every triviality of wall-crossing is traced back to an explicit inequality
for solutions of a transport equation. This is what gives the program its authority and its claim to
providing not merely a connection between areas but an explanation of why those areas behave the way
they do.

Two natural directions for continuation present themselves. First, a fully worked resonant
example with nontrivial Stokes matrices and explicit nonzero Auslander bound would provide a
concrete test case for the principle in the flexible regime, confirming that the theory correctly predicts
the proliferation of hearts and the nontriviality of wall-crossing whenever the monotonicity condition

Exponential Periods Program Page 21

 fails. Second, the export of the Irregular Auslander Principle to Fukaya-type categories via oscillatory
integrals would connect the present program to symplectic geometry and mirror symmetry, where
analogous questions about extensions and stability conditions arise from a different direction. Both
extensions are within reach of the methods developed here, and both would deepen the structural
understanding of what it means for exponential growth to enforce rigidity.

As written, the present program is internally complete, conceptually sharp, and

mathematically serious. The slogan that stability is a shadow of growth is not a metaphor in
this text; it is a theorem, proved line by line from first principles of asymptotic analysis.

Exponential Periods Program Page 22