The Shape Behind the Infinities: Renormalization as Slope Extraction and the Anti-Geometric Axiom

J. Rogers, SE Ohio

Abstract

The infinities of quantum electrodynamics are the unbounded growth of the 1/r profile toward small scales, the signature of a shape that continues at every scale. Renormalization stands as a procedure, and this paper supplies the reason it works. It is the subtraction of two divergent evaluations of the same 1/r tower to find a finite quantity on its slope. The coupling α is the overlap of the ratio of two physical extents, r_e/ƛ_e. The beta function is the first slope of that shape. The infinities cancel in the subtraction because both are the same growth of the same shape, which is why a finite point can be extracted. The electron is not a point: it has the extent r_e, and the effects keep growing below it. The shape was not seen because the standard framework contains an explicit axiom that prohibits it, the point particle, and that axiom is protected by six mechanisms: the gauge smokescreen, the unit Jacobian confusion, the positivist ethos, the historical inversion, the identity threat, and the double standard. Results are reported as named concerns, not as simplified magic numbers: ratio, shape, projection, and chart, each shown by name. The math does not change. The referents are restored.

1. Introduction

The standard framework treats renormalization as necessary and meaningful. Loop integrals diverge. Counterterms are introduced. The divergences are subtracted. The finite remainder is identified with physical quantities, and the running of the coupling is derived from the renormalization group equations. This is presented as a deep feature of quantum field theory.

We argue that it is a slope extraction applied to a series whose divergence is the unbounded growth of the shape. The physical content is the slope of the shared 1/r profile. Renormalization is the framework's way of computing a derivative without knowing the shape it is differentiating.

Because renormalization has no accepted referent in the standard framework (the textbook position is "bookkeeping"), this paper does not need to refute an existing explanation. It supplies what the symbols refer to. It also reports every result as named concerns, ratio, shape, projection and chart, with the product given last as the compiler's output. A coefficient written as one simplified number cannot be checked, because many different structures reduce to it.

2. The Two Sectors and Their Shared Profile

Both the charge sector and the kinematic sector have the same exterior radial profile. In three spatial dimensions the static Green's function for a point disturbance is

G(r) = 1/(4πr)

This shape is not owned by either sector. It is the shape of any long-range static influence in 3D. The charge sector has

V_C(r) = (e²/ε₀) · (1/4π) · (1/r)

and the kinematic sector has

V_K(r) = −GM · (1/r)

Both are the same 1/r shape. The difference is only the coupling and the sign. The derivative tower is identical. Each term is named by concern: the coupling with its unit chart (e²/ε₀), the spherical projection (1/4π), and the shape (1/r).

3. The Derivative Tower

The n-th radial derivative of 1/r is

dⁿ/drⁿ (1/r) = (−1)ⁿ n! / rⁿ⁺¹

So

• d/dr (1/r) = −1/r²
• d²/dr² (1/r) = 2/r³
• d³/dr³ (1/r) = −6/r⁴
• d⁴/dr⁴ (1/r) = 24/r⁵

The coefficients are n!. The force is the first slope. The tide is the second. The jerk is the third. The tower is the same for both sectors.

4. The Overlap Ratio

The coupling α is the overlap of two physical extents:

α = r_e / ƛ_e

where r_e is the charge extent and ƛ_e is the kinematic extent. The same two extents bound an integral of the slope:

∫ from r_e to ƛ_e of dr/r² = 1/r_e − 1/ƛ_e = (1 − α)/r_e

Since r_e ≪ ƛ_e, this is dominated by 1/r_e, and it diverges as r_e → 0. That is the point limit. The divergence is the point-limit singularity of 1/r. The dimensionless content, the ratio r_e/ƛ_e = α, is finite and physical. The bounds are the two extents because they are the two points at which the two sectors are read, the same two-point evaluation that renormalization subtracts (Section 6.1). The integral between them is the finite difference of the slope between those two points, and nothing outside them is a second evaluation.

5. The Origin of the Divergence

5.1 The Point-Particle Idealization

In QED the electron is treated as a point. The self-energy loop has no minimum radius, and the integral over loop momentum runs to infinity:

Σ(p) ~ ∫ d⁴k/(2π)⁴ · 1/(k² − m²) ⋯

The divergence is the unbounded growth of the shape as the loop shrinks. A point has no scale, so a point could not grow. The integral diverges because the 1/r tower keeps growing at every smaller scale.

A generic one-loop correction has the form

I(μ) = ∫₀^∞ (dk/k) f(k/μ)

which diverges logarithmically as k → ∞ (r → 0). Regulated with a cutoff Λ:

I(Λ) = A + B ln(Λ/μ) + ⋯

The divergence sits in the ln Λ term. It comes from the r → 0 limit of the 1/r profile. The shape keeps growing at every smaller scale, so the integral runs to infinity.

5.2 The Finite Extent and the Continuing Shape

The electron is a finite transducer, the intersection of the charge sector and the kinematic sector with extent r_e, known since Thomson (1881). The extent is the reference scale that fixes α, and it is not a wall. Below r_e the tower continues, and the effects grow without bound at every smaller scale. The infinities of the standard framework are that growth, read off a point that was assumed to have no scale.

6. Renormalization as Slope Extraction

6.1 Subtraction at Two Scales

Compute the same integral at two scales μ₁ and μ₂:

I(μ₁) = A + B ln(Λ/μ₁) + ⋯

I(μ₂) = A + B ln(Λ/μ₂) + ⋯

Subtract:

I(μ₁) − I(μ₂) = B ln(μ₂/μ₁)

The divergent ln Λ terms cancel. They are identical because they come from the same r → 0 limit of the same 1/r shape. The finite remainder is the difference between two points on the shape.

In the geometric picture, the logarithm is the integral of the 1/r profile itself:

∫ from r₁ to r₂ of dr/r = ln(r₂/r₁)

The subtraction is the area under the 1/r shape between two scales. This is also why the beta function is a derivative with respect to ln μ: the coupling is read off as a slope against the logarithm of scale, and the divergence common to both evaluations drops out.

6.2 Subtraction of Divergent Series

The perturbative expansion of a physical quantity is

Q = Σₙ cₙ αⁿ

The coefficients cₙ grow like n! at high order. The series is asymptotic and diverges. Renormalization subtracts one divergent series from another:

Q_ren = Σₙ cₙ αⁿ − Σₙ dₙ αⁿ

The divergences cancel order by order and the remainder is finite. The two series are the same 1/r tower evaluated at two different scales. Their divergent parts are identical and cancel. The finite remainder is the difference between two points on the shape, and to first order

Q_ren ~ (d/dr)(1/r) · Δr

Renormalization is the finite difference of a derivative.

The map from cₙ to the tower is explicit. The n-th derivative of 1/r carries n!, and a Taylor expansion would divide it by n!. The series of derivative values without the 1/n! is

Σₙ [dⁿ/drⁿ (1/r)] Δⁿ = (1/r) Σₙ n! (−x)ⁿ, with x = Δ/r

This is the Euler series. It is asymptotic, and its Borel sum is the finite quantity (1/r) ∫₀^∞ e^(−t)/(1 + x t) dt. Identifying x with the expansion parameter α, the step Δ read against r in units of the shared ruler, matches cₙ to the tower term by term, cₙ ↔ (−1)ⁿ n!/rⁿ⁺¹, and the Borel sum is the finite point on the slope. The identification of x with α is the paper's claim. The sign structure of the actual QED coefficients is not shown here.

6.3 Why the Subtraction Requires a Shape

Subtracting two infinities gives a meaningful result only if both are evaluations of the same thing. A point has no shape, so there is nothing for the two evaluations to be evaluations of, and nothing that guarantees their divergent parts are identical. On the point-particle axiom, the cancellation is a coincidence.

The procedure also borrows the shape it denies. The scale μ, the regulator Λ, the running with scale, and the logarithm in the remainder all presuppose that something changes as you move through scales. A scale-free point has none of these. Renormalization uses structure at every scale while the axiom says there is none, so the method contradicts the foundation it sits on. With the shape restored, the contradiction is gone: the two divergences cancel because they are the same growth of the same 1/r tower.

6.4 Separation of Concerns in the Coefficient

The coefficient of the log, B, is one number made of several concerns: α × (1/π) × 2 × 2 × (1!·1!/3!). The reduced form is the compiler's output and is not used as a result. Matching the product to data confirms nothing, because many different splits give the same product, as with 6π⁴/90 reduced to π⁴/15 in the Stefan-Boltzmann constant (Rogers, The Illusion of the Final Formula). The coefficient is tested one concern at a time.

Concern	Factor	Owner
Overlap ratio	α	r_e/ƛ_e
Shape	1!·1!/3!	The tower: the Beta function B(2,2) = Γ(2)Γ(2)/Γ(4), the 3! of rung 3
Projection	1/π	Spherical geometry of the transducer
Chart	2	The log is in Q² and the scale is Q, so ln Q² = 2 ln Q
Unassigned	2	Spinor trace, no owner yet

The product contains a factorial. It is not the factorial. The 3! in 4/(3!·π) cancels against the 4 to give 2/3, the same cancellation that hides a 3! inside the 15 of the Stefan-Boltzmann constant. Counted by derivative order the factor is n! with n = 3. Counted by the power of 1/r it is (n−1)! with n = 4. It is the same 3!.

The criterion is that every factor has a single owner: shape, projection, or chart. A factor no concern claims, here the 2 from the spinor trace, fails the decomposition until it is assigned.

6.5 Numerical Check Against the Standard Coefficient

The one-loop electron vacuum polarization was evaluated exactly, Δ ln α(Q) = (2α/π) ∫₀¹ x(1−x) ln(1 + x(1−x)Q²/m²) dx, with each concern kept as its own value.

Concern	Name	Value
Overlap ratio	r_e/ƛ_e = 2.8179×10⁻¹⁵ m / 3.8616×10⁻¹³ m	0.0072974
Projection	1/π	0.31831
Trace	spin factor	2 (unowned)
Shape moment	M0 = 1!·1!/3!	0.166667
Chart	d ln Q² / d ln Q	2
Split moment	M1 = ∫ w ln[x(1−x)] dx = −5/18	−0.277778

The slope of the exact curve against ln Q is 1.548544×10⁻³ at Q = 1 GeV and 1.548546×10⁻³ at 100 GeV and at 10⁴ GeV. The named factors give 1.548546×10⁻³, the same as the standard reduced coefficient 2α/(3π). The match is an identity. It shows that the named parts reconstruct the standard coefficient, and it tests nothing more.

What is tested is the form. The running is logarithmic in the scale ratio above a few multiples of 1/ƛ_e. Below that the threshold remainder, which is the finite extent of the mass sector, falls by about a factor of 100 for each factor of 10 in Q: 2.6×10⁻³ at 10 MeV and 2.6×10⁻⁵ at 100 MeV. The split moment M1 is the constant term. It moves with the reference scale and the scheme, so only differences between two scales are physical.

7. The Beta Function as the First Slope

The running of the coupling is described by

β(α) = μ dα/dμ

This is a derivative: the slope of the coupling with respect to the energy scale. Since α = r_e/ƛ_e and ƛ_e = ℏ/(m_e c) is a fixed kinematic ruler, the running of α is the running of r_e against that ruler as the probe scale changes:

β(α) ~ d/d ln μ ( r_e(μ) / ƛ_e )

The beta function is the first slope of the overlap ratio. The higher coefficients of the beta function are the higher derivatives of the tower, and the n! growth is the factorial tower.

In named concerns, the one-loop relative running per unit-charge fermion is

\frac{d\ln\alpha}{d\ln Q} = \underbrace{\alpha}_{\text{ratio}} \times \underbrace{\tfrac{1}{\pi}}_{\text{projection}} \times \underbrace{2}_{\text{chart}} \times \underbrace{2}_{\text{unowned}} \times \underbrace{\tfrac{1!\,1!}{3!}}_{\text{shape}}

The product of the named parts is the number QED reports as one reduced coefficient. It is shown last, as the compiler's output, and the test is on the named parts (Section 6.4).

The running is not a quantum correction from vacuum polarization. It is the resolution-dependent overlap of the charge extent and the kinematic extent. At higher energies you probe smaller r and see more of the tower. The overlap ratio changes. The coupling runs.

8. The Extent as the Reference Scale

If the loop were cut off at k ~ 1/a, the self-energy would be finite:

Σ ~ ∫₀^(1/a) dk k³ · 1/(k² + m²) ~ 1/a²

That finite value is what a hard cutoff at the extent would give. It is not what the shape does. The extent r_e fixes the scale of α, and the tower continues below it, so the integral still grows and renormalization still extracts the finite point. The cutoff Λ in Section 5.1 is the regulator, and the shape's growth is what it regulates. The finite self-energy is the renormalized one, fixed by the measured mass, and this paper does not compute its value.

9. The Anti-Geometric Axiom

If the shape is this plainly present, with 1/r as the profile of both sectors, n!/rⁿ⁺¹ as the tower, and the overlap ratio as the coupling, why was it not seen? The answer is not that physicists were not smart enough. The paradigm contained an explicit axiom that prohibited the shape from being seen. That axiom is the point particle.

A point has no extent, no interior, no shape, and no derivative structure. To say "the electron is a point" is not a neutral empirical claim. It is a geometric prohibition: there is no extent to geometrize, and the charge is a naked label attached to a mathematical zero. The axiom is not derived. It is assumed. Once assumed, it forces what follows:

Consequence	Origin
α is an orphan ratio	No extents to compare
Loop integrals diverge	Growth of the shape read as a pathology
Renormalization is needed	Infinities must be subtracted
Particle spectrum is free parameters	No harmonics of a shape
Force is primitive	No shape to be the slope of
Virtual particles are reified	No extent to store stress
Dark energy is a substance	No rate change to explain acceleration
Measurement problem persists	No ontology of the transducer

The point particle is not a simplification. It is a deletion. It deletes the geometry, and the framework then spends eighty years calculating the consequences of the deletion and running the procedure of renormalization without knowing why it works.

10. Six Protections of the Axiom

The point particle alone might have been exposed. It was protected by six mechanisms.

10.1 The Gauge Smokescreen

In gauge theory the potential is declared not real. It is gauge-dependent, only the field strength is physical, and the zero point is arbitrary. This is true as far as it goes. The zero point is arbitrary, but the shape is not. You can shift the zero. You cannot shift the hill. The gauge principle collapsed the two: the potential is gauge, therefore the hill is not real; the metric is coordinate-dependent, therefore curvature is all there is; α is a coupling, therefore the extents are not physical. The shape is not the gauge. The shape is what the gauge describes.

10.2 The Unit Jacobian Confusion

The standard framework puts c, ℏ, G, and α into one bucket labeled "fundamental constants." This is a category error. c is a unit Jacobian, meters per second, and its numerical value is convention. ℏ is a hybrid: a unit Jacobian (h) divided by a geometric ratio (2π). α is a pure dimensionless ratio, r_e/ƛ_e, and its value is physical. By lumping them together the framework could no longer ask "ratio of what?" and α became an orphan because its referents were buried in the same bucket as c.

10.3 The Positivist Ethos

"Shut up and calculate" is not an epistemic position. It is a defense. It says do not ask what the symbol refers to, what shape the slope is on, or what the ratio is between. But "the math works" is an admission: the paradigm has become a calculator that does not know what its numbers are numbers of. The ethos is enforced socially. Asking "what is α a ratio of?" is treated as philosophy, and asking "why does renormalization work?" gets the technical answer "it's just effective field theory." The foundational question is quarantined.

10.4 The Historical Inversion

Thomson calculated the electromagnetic mass and its relation to the radius of a charged sphere in 1881. That is a physical extent, from mechanics and electromagnetism, before quantum theory and before the fine-structure constant existed. Sommerfeld introduced α in 1916 to explain the fine structure of hydrogen. It was noticed in the spectrum, not derived from r_e. QED then wrote

r_e = α · ℏ/(m_e c)

presenting the classical electron radius as derived, as if it were α times the Compton wavelength. Historically and logically the order is the reverse. The extent came first and the ratio second, and the ratio was only ever one extent divided by another. QED took the later, orphaned ratio and made it the parent, and took the earlier physical extent and made it the child. That is a historical erasure of the priority of the extent, and with it the priority of the shape.

10.5 The Identity Threat

A paradigm is not just a set of equations. It is a professional self. The equations are internalized as "how a competent physicist thinks," and seeing the shape means seeing that the gauge was a description, not the thing. That changes who you are. The defenses are the familiar ones: "It's just philosophy." "The math works." "Not even wrong." "No real physicist talks like that." Resistance is strongest when the math does not change. If the math changes, it is a technical dispute and normal science. If only the meaning changes, the threat is total, because it means the old guard did not understand the meaning of their own theory, though they understood the calculations.

10.6 The Double Standard

The standard framework assumes a fixed time dimension with uniform steps, a smooth spacetime manifold, a fundamental metric, locality and unitarity as axioms, real point particles, fundamental gauge symmetry, a potential that is a fiction whose effects are real, and dark energy as a substance. None of it is derived, and when asked why, the answer is "that's the framework." A new idea is held to a different standard. It must derive α from first principles, reproduce the particle spectrum, make testable predictions, have an action and a Lagrangian, and account for every number. That standard is never applied to the foundations of the standard framework. It is a social standard, not a scientific one. It prevents the shape from being seen by demanding that any alternative explain everything while the paradigm explains nothing.

11. Predictions

1. High-order coefficients. The n-th order coefficient in the perturbative expansion should grow like n!, matching the derivative tower of 1/r. This is already observed in QED and post-Newtonian gravity.
2. Structure at every scale. At distances comparable to r_e and below, the 1/r tower should show its growing derivatives rather than flatten. Deviations from Coulomb and Newton should grow as the tower, with no scale at which they stop.
3. Running of α. The running should be derivable from the overlap geometry, not from loop integrals. The beta function should match the derivative of the ratio r_e/ƛ_e, reported as named concerns (ratio, shape, projection, chart), not as a single coefficient.
4. Unbounded growth below the extent. Below r_e the effects should keep growing without bound as the derivative tower. The direct test is to check that the logarithmic coefficients follow from the geometry rather than being fitted, and that the coupling does not stop running at any scale.
5. Particle spectrum. The stable modes of the transducer should be the harmonics of the n! tower. Mass ratios should be related to the roots of the derivative series. No roots are computed in this paper, and the prediction is open.

12. Falsification Criterion

The claim is falsified if the one-radius geometry fails to produce the observables, or if a different shape produces the same observables.

The geometry is a circle, so it has a single radius. There is no interior profile to tune and no free shape parameter. The ratio r_e/ƛ_e = α is an identity, so it is the premise and is not what is tested. What is tested is the extraction: one measured point on the shape fixes the finite remainder at any other scale, as renormalization does, and it works only because both divergences have the same shape. The criterion has three horns, and any one falsifies:

1. Failure to reproduce. The divergent parts of the two evaluations must cancel identically, leaving the integral of the 1/r shape between the two scales: logarithmic in the scale ratio. One measured point must then fix α at every other scale on that form, with no further parameter, with the beta function coefficients following from the derivative tower rather than from loop integrals. It must also show the effects growing without bound below r_e; if they flatten at any scale, only a point does that. The coefficient is tested by separate concerns, not as one number: shape, projection and chart must each be derived on their own, and every factor must have a single owner (Section 6.4). A match of the product alone confirms nothing. If any of these fails, the shape is not the physics.
2. Equivalence by a different shape. If a different shape, a different one-parameter geometry or any other profile, gives the same observables, then the observables do not select this shape, and the claim that this shape is the physics is falsified.

3. Per-order effects. Each derivative order is a distinct physical effect with a fixed coefficient, n!/rⁿ⁺¹: the force at order 1, the tide at order 2, the jerk at order 3. Shrinking the scale extends the known effect by adding the next order, and the orders add. Effects may compose across the two sectors, so what is observed as one effect can be a sum of several tower terms, and at smaller scales the terms may become harder to untangle. The test is then on the sum: with the coefficients fixed by the shape, the composed result must equal the observed effect. The claim is falsified if a remainder is left that no tower term accounts for, if the scaling of the sum departs from the tower, or if the orders do not add.

The first horn is a calculation. The second is a comparison: run the same derivation for a rival shape and check whether the numbers differ. Only a geometry that reproduces the observables and that no other shape reproduces survives. The third horn gives a separate chance to fail at every order.

Status. Owned: the overlap ratio, the shape moment, and the chart factor. Open: the spin trace factor 2, and the origin of the weight x(1−x) from the geometry. The weight enters Section 6.5 from the standard loop calculation and is not yet generated by the tower. The derivation of each named factor from the geometry alone is the work still to be done, and Prediction 5 is stated and not yet computed.

13. Conclusion

Renormalization stands as a procedure, and it is a slope extraction: the subtraction of two divergent evaluations of one shape. The divergence is the unbounded growth of that shape, and it cancels because both evaluations share it. The infinities are the signature of structure at every scale. The physical content is the slope of the shared 1/r profile. The beta function is the first derivative, and the running coupling is the geometry.

The shape was always there: in the 1/r potential, in the n!/rⁿ⁺¹ tower, in the overlap ratio r_e/ƛ_e, in the running of α, in the electron shell lobes, in the n of Planck's quantum, in the beta function, and in the finite extent of the transducer. It could not be seen because the paradigm contained an axiom that prohibited it, and that axiom was protected by six layers of defense. To see the shape you have to break the axiom: restore the extent, geometrize charge and kinematics, treat time as a rate rather than a dimension, separate unit Jacobians from physical ratios, ask "ratio of what?" and "slope of what?", and report every result as named concerns, not as a simplified number.

The math does not change. The referents are restored. The point is the fiction, the shape is the physics, and the exit is through the geometry that was always there.

References

• Newton, I. Philosophiæ Naturalis Principia Mathematica, General Scholium, 1713.
• Planck, M. Scientific Autobiography and Other Papers, 1949.
• Kuhn, T. The Structure of Scientific Revolutions, 1962.
• Thomson, J.J. "On the Electric and Magnetic Effects produced by the Motion of Electrified Bodies," Philosophical Magazine, 1881.
• Einstein, A. "Über einen die Erzeugung und Verwandlung des Lichtes betreffenden heuristischen Gesichtspunkt," Annalen der Physik, 1905.
• Dyson, F. "The Radiation Theories of Tomonaga, Schwinger, and Feynman," Physical Review, 1949.