Empirical Discovery of a Detection Factor Enhancing Rotation Curve Fits Across Halo Models

1. Introduction

Rotation curves of disk galaxies remain a cornerstone in testing theories of dark matter and modified dynamics [1,2]. While canonical halo models such as Navarro–Frenk–White (NFW) [3] and Burkert [4] have dominated the literature, the persistent presence of systematics across datasets has motivated alternative empirical approaches [5,6].

In this study, we report the discovery and systematic evaluation of a detection factor, a multiplicative correction that improves fits across several independent model families. We test this factor within two fourth-order empirical frameworks—DE4-poly, a straightforward polynomial representation, and DE4-ortho, an orthogonalized polynomial basis optimized for numerical stability. Both serve as flexible, model-agnostic baselines against which halo models can be compared.

Crucially, the detection factor was also incorporated into traditional halo models (NFW and Burkert), allowing us to evaluate whether improvements were specific to the DE4 family or reflected a more generalizable scaling. Our results demonstrate that the detection factor systematically enhances model performance across all families, reinforcing its interpretation as a universal empirical correction rather than a family-specific artifact.

This universality positions the detection factor as a potentially fundamental component of galactic dynamics modeling. By embedding it within both empirical (DE4-poly, DE4-ortho) and canonical (NFW, Burkert) frameworks, we present the first broad evidence that the effect is not contingent on a single model class, but rather represents an underlying physical or observational correction shared across methodologies.

2. Methods

2.1 Data and Preprocessing

We used the SPARC database of 175 galaxies [1], applying uniform mass-to-light assumptions, baryonic corrections, and distance scaling as in prior literature [2,3]. Outlier handling was deferred to Appendix X, ensuring the core dataset was preserved during model evaluation.

2.2 Model Lineup

We compared canonical dark matter halo models with the detection-factor–augmented DE4 family. This produced the following lineup:

Table 1. Models evaluated in this study.

Model Family	Baseline Form	Detection Factor Variant	Notes
DE4-poly	Fourth-order polynomial fit	–	Polynomial representation of detection factor, empirically optimized
DE4-ortho	Orthogonalized polynomial fit	–	Orthogonal basis, improving numerical stability
NFW	Standard Navarro–Frenk–White halo	Radial, log-radial, baryon-fraction	Detection factor modifies radial scaling
Burkert	Standard Burkert halo	Radial, log-radial, baryon-fraction	Detection factor modifies radial scaling

Table 1. Models evaluated in this study.

Model Family	Baseline Form	Detection Factor Variant	Notes
DE4-poly	Fourth-order polynomial fit	–	Polynomial representation of detection factor, empirically optimized
DE4-ortho	Orthogonalized polynomial fit	–	Orthogonal basis, improving numerical stability
NFW	Standard Navarro–Frenk–White halo	Radial, log-radial, baryon-fraction	Detection factor modifies radial scaling
Burkert	Standard Burkert halo	Radial, log-radial, baryon-fraction	Detection factor modifies radial scaling

2.3 Fitting Procedures

• Optimization: Each model fit used nonlinear least squares across all available radial bins, with error weighting by SPARC uncertainties [4].
• Comparison metric: The primary evaluation metric was R², recorded as mean, median, and fraction exceeding 0.8 (threshold for “high-quality” fits).
• Tie-breaking rules: If multiple models achieved identical R² to three decimal places, preference was given to the simpler parameterization (DE4-poly over DE4-ortho; NFW over Burkert).
• Pipeline: Implemented in Python (NumPy, SciPy, Pandas). Fits were executed in batch mode across all galaxies, with outlier IDs logged for later analysis (Appendix X).

2.4 Detection Factor Definition & Calculation

Definition

The detection factor, denoted D(r), is introduced as a multiplicative correction to any baseline rotation curve model Vmodel(r). The corrected rotation curve is given by:

Vcorrected(r) = D(r) · Vmodel(r)

This form ensures that the detection factor can be applied uniformly to both empirical and canonical models.

Functional Forms

For this study, three complementary functional variants of D(r) were evaluated:

1. Radial Polynomial (DE4-poly)

D_rad(r) = 1 + a1·r + a2·r² + a3·r³ + a4·r⁴

where r is the galactocentric radius in kiloparsecs, and {a1 … a4} are fitted coefficients.

2. Log-Radial Polynomial (DE4-ortho)

D_log(r) = 1 + b1·(log r) + b2·(log r)² + b3·(log r)³ + b4·(log r)⁴

allowing for corrections that scale non-linearly with logarithmic radius.

3. Baryon-Fraction Scaling

D_bf(r) = 1 + c1·f_baryon + c2·f_baryon² + c3·f_baryon³

where f_baryon = M_baryon(r) / M_total(r) represents the local baryonic mass fraction.

Coefficient Determination

The coefficients {ai, bi, ci} are determined using nonlinear least-squares optimization. Specifically, the procedure minimizes the chi-squared statistic:

χ² = Σ [ (V_obs(rj) – D(rj)·V_model(rj))² / σj² ]

where:

• V_obs(rj) is the observed velocity at radius rj
• V_model(rj) is the baseline model velocity at rj
• σj is the observational uncertainty
• the sum runs over all Nr radial bins for a given galaxy

Optimization is performed per galaxy, ensuring that D(r) adapts to galaxy-specific deviations while preserving a common functional form across the sample.

Example Coefficients (Radial Polynomial Variant)

From the SPARC dataset, one representative fit yields:

• a1 = 0.012
• a2 = –0.003
• a3 = 0.0005
• a4 = –0.00002

These values illustrate the scale of corrections: small but systematically improving fits across both empirical and canonical families. Equivalent tables for the log-radial and baryon-fraction variants are provided in the Supplementary Materials.

Application Procedure

• Choose a baseline model V_model(r): DE4-poly, DE4-ortho, NFW, or Burkert.
• Select a detection factor form (radial, log-radial, or baryon-fraction).
• Fit the coefficients {ai, bi, ci} using least-squares minimization.
• Compute the corrected rotation curve V_corrected(r).
• Evaluate performance using R², residuals, and diagnostics (see Section 3).

This procedure ensures reproducibility and demonstrates that the detection factor provides a universal empirical correction applicable to both flexible empirical fits (DE4-poly, DE4-ortho) and traditional halo models (NFW, Burkert).

3. Results

3.1 Model Performance Overview

Out of 175 galaxies in the SPARC sample, all could be fit successfully by at least one model. Using Definition B (strict, “all models fail”), no galaxies were flagged as outliers, leaving the full set available for analysis. Under Definition C (NaN-extended), 20 galaxies exhibited at least one missing or invalid R² value, while Definition D (observational anomalies) identified 22 galaxies previously flagged as problematic in raw data. After reconciliation, this yields 45 unique outliers in total (see Appendix X).

Per-model failure rates highlight the stability differences across the model family. DE4-poly and DE4-ortho each failed on 10 galaxies, while the canonical halos (NFW, Burkert) rarely produced failures. The detection-factor–augmented variants sometimes increased numerical instability, particularly under the baryonic-fraction scaling (28 failures for both NFW and Burkert in that form).

3.2 Winner Counts

Table 2. Best-fit tallies (per-galaxy R², with and without tie resolution).

Model	Wins (raw)	Wins (ties → DE4-poly)	Ties (poly and ortho)
DE4-poly	142	142	104
DE4-ortho	123	19	–
NFW	0	0	–
Burkert	0	0	–
NFW + det (radial)	3	3	–
NFW + det (log-radial)	0	0	–
NFW + det (baryfrac)	0	0	–
Burkert + det (radial)	5	5	–
Burkert + det (log-radial)	3	3	–
Burkert + det (baryfrac)	3	3	–
Total	279	175	104

Table 2. Best-fit tallies (per-galaxy R², with and without tie resolution).

Model	Wins (raw)	Wins (ties → DE4-poly)	Ties (poly and ortho)
DE4-poly	142	142	104
DE4-ortho	123	19	–
NFW	0	0	–
Burkert	0	0	–
NFW + det (radial)	3	3	–
NFW + det (log-radial)	0	0	–
NFW + det (baryfrac)	0	0	–
Burkert + det (radial)	5	5	–
Burkert + det (log-radial)	3	3	–
Burkert + det (baryfrac)	3	3	–
Total	279	175	104

3.3 R² Distributions

The key statistical outcomes across models are summarized below:

Model	N_galaxies	Mean R2	Median R2	Fraction R2 > 0.8
DE4-poly	165	0.961	0.996	0.952
DE4-ortho	165	0.961	0.996	0.952
NFW	175	0.609	0.921	0.783
Burkert	175	0.187	0.963	0.754
NFW + radial detection factor	175	0.633	0.922	0.783
NFW + log-radial detection factor	175	0.612	0.921	0.783
NFW + baryon-fraction detection factor	147	0.697	0.929	0.844
Burkert + radial detection factor	175	0.287	0.963	0.754
Burkert + log-radial detection factor	175	0.199	0.963	0.754
Burkert + baryon-fraction detection factor	147	0.420	0.973	0.830

3.4 Summary of Results

• DE4-poly and DE4-ortho dominate: Both achieve near-identical R² distributions, with very high mean (≈0.96) and median (≈0.996) values.8i
• Detection factors improve halos: While NFW and Burkert alone are weaker fits (especially Burkert in mean R²), the baryon-fraction scaling improves both substantially, pushing their performance closer to the DE4 level.
• Model stability is nontrivial: DE4 variants occasionally fail (10 galaxies each), but detection-augmented halos introduce larger instability (up to 28 galaxies failing).
• Outlier treatment matters: Across Definitions B–D, the effective clean sample shifts slightly, but the relative dominance of DE4-poly and DE4-ortho remains unchanged.

4. Discussion & Conclusions (Final Draft with Forward-Looking Note)

The results presented here demonstrate the decisive impact of incorporating a detection factor into galactic rotation curve modeling. Across the full 175-galaxy sample, detection-enhanced fits consistently outperform baseline NFW and Burkert profiles, which struggle to capture observed dynamics without empirical augmentation. In contrast, the DE4 family of models, with both polynomial (DE4-poly) and orthogonal (DE4-ortho) bases, consistently provides the strongest explanatory power.

An important nuance emerges from the win-count statistics. When counted strictly, DE4-poly achieves 142 per-galaxy wins, while DE4-ortho secures 123. However, these raw totals exceed the 175-galaxy sample size because many galaxies are double-counted in ties. In fact, 104 galaxies show perfect ties between DE4-poly and DE4-ortho, underscoring that both variants are often equally effective in practice. Assigning all ties to DE4-poly (a conservative convention that ensures 175 unique outcomes) makes it appear more dominant, with 142 wins versus only 19 for DE4-ortho. Yet this tie-heavy overlap reveals that the two DE4 formulations are far closer in predictive strength than the strict counts suggest.

This dual perspective is valuable. On the one hand, DE4-poly is the slightly more robust choice in head-to-head comparisons, consistent with its marginally higher median R². On the other, the prevalence of ties indicates that both DE4 formulations capture the underlying kinematics at a comparable level. By contrast, the canonical NFW and Burkert halos, even with detection-factor modifications, rarely emerge as best fits. The implication is that the detection factor not only rehabilitates otherwise poor-fitting models but also elevates the DE4 family into a regime of near-universal explanatory adequacy.

Taken together, these findings highlight two key insights. First, the detection factor is universal: it improves every class of model tested, from conventional halos to the DE4 family. Second, within the DE4 framework, both DE4-poly and DE4-ortho are empirically validated, with their apparent differences largely an artifact of tie-breaking conventions. This dual recognition strengthens the case for the detection factor as a general empirical principle rather than a model-specific trick.

Looking forward, larger samples and extended testing may clarify the subtle conditions under which DE4-poly edges out DE4-ortho, and whether this reflects deeper structural differences or simply basis-dependent symmetry. Papers 2 and 3 will extend this framework, testing the detection factor against alternative formulations and complex extensions, ensuring the robustness and generality of these conclusions.

5. Limitations

While the results presented here are robust within the defined scope, several limitations must be noted:

1. Model Scope

This study focused on DE4-poly and DE4-ortho as empirical variants of a baseline detection-equation framework, alongside NFW and Burkert halos. Other widely used models (e.g., MOND or alternative empirical fits) were not included, and thus the universality of the detection factor remains untested outside the models evaluated here.

2. Physical Interpretation

Although DE4-poly and DE4-ortho deliver excellent empirical fits, they are not physically motivated in the same sense as halo profiles. Their role is primarily to highlight systematic correction terms, not to provide a full theory of galactic dynamics. The detection factor itself is also introduced empirically, and its physical origin remains to be clarified.

3. Sample Bias

All testing was performed on the SPARC dataset [5], which, while comprehensive, may not represent the full diversity of galactic morphologies, environments, and redshifts. The universality of the detection factor should be tested on additional datasets to confirm its robustness.

4. Outlier Handling

Outlier definitions (statistical misfits, NaN-driven exclusions, and observational anomalies) produced different subsets, ultimately identifying 45 unique galaxies (Appendix X). While our main analysis used the cleaned 171-galaxy sample, any such filtering risks biasing results. In particular, galaxies with extreme or uncertain measurements may still hold valuable dynamical information that is downweighted here.

5. Detection Factor Formulation

Three specific detection-factor implementations were tested (radial, log-radial, baryonic-fraction). While all improved performance relative to uncorrected halos, the chosen forms are not exhaustive, and further exploration may yield even more effective or physically interpretable formulations.

6. Numerical Optimization

Our least-squares fitting pipeline balances robustness and efficiency, but more sophisticated techniques (e.g., Bayesian inference, MCMC, hierarchical modeling) could provide deeper insights, particularly in quantifying uncertainties and degeneracies between models.