Beta Decay Modulation Analysis

ISEF Category: Physics and Astronomy

Subcategory: Nuclear and Particle Physics  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

Some physics questions hide in plain data. A tiny yearly wobble in a decay curve could point to noise, a seasonal effect, or something deeper. You can test that claim with public datasets, careful statistics, and a wavelet plot that acts like a spotlight for changing patterns.

What Is It?

Nuclear decay usually looks random for each atom, but a huge sample follows a smooth trend. Think of it like coin flips in bulk. You cannot predict one flip, but you can predict the average number of heads. This project asks whether beta-decay counts show a small yearly pattern that should not be there.

The big idea comes from published reports about possible annual modulation in long-term decay measurements. A modulation is just a repeating up-and-down pattern. Your job is not to assume the effect is real. Your job is to test the public data for periodic signals, compare them to noise, and estimate how large any solar-neutrino-linked effect could be before the data would have shown it.

Why This Is a Good Topic

This makes a strong science fair topic because the data already exist, the question is sharp, and the analysis can be tested step by step. You can learn time-series analysis, signal detection, uncertainty, and how to separate real trends from artifacts like drift or seasonality. The real-world connection is basic nuclear physics and a live scientific debate about whether decay rates are perfectly constant.

Research Questions

• How does applying a wavelet decomposition change the visibility of annual power in the BNL and PTB decay datasets?
• What is the effect of detrending choice on the estimated amplitude of any yearly modulation?
• Does the apparent annual signal persist after removing known systematic drifts in detector response?
• To what extent do BNL and PTB datasets agree on the phase of any yearly modulation?
• Which wavelet scales show the strongest excess power near a one-year period?
• What is the effect of using bootstrap resampling on the upper limit for a solar-neutrino-linked modulation?

Basic Materials

• Computer with internet access
• Spreadsheet software or Python installed
• Public BNL and PTB decay datasets
• Text editor or notebook for documenting analysis choices
• Plotting software such as Python matplotlib or Google Sheets charts
• Calculator for quick sanity checks
• Reference notes on time-series analysis and Fourier ideas.

Advanced Materials

• Computer with internet access
• Python with NumPy, SciPy, pandas, matplotlib, and PyWavelets
• Jupyter Notebook or similar notebook environment
• Public BNL and PTB decay datasets
• Statistical analysis package for bootstrap and hypothesis testing
• Version control tool such as Git for tracking analysis changes
• ImageJ if you also inspect scanned plots from published papers.

Software & Tools

• Python: Runs the time-series analysis, wavelet decomposition, and uncertainty calculations.
• Jupyter Notebook: Keeps code, plots, and written reasoning together in one place.
• pandas: Organizes the decay datasets into clean time series.
• SciPy: Helps fit trends, test models, and estimate confidence limits.
• PyWavelets: Performs the wavelet decomposition that can reveal time-localized periodic signals.

Experiment Steps

1. Define the exact datasets, time span, and decay channels you will compare.
2. Clean the data so you know how to handle missing points, outliers, and normalization.
3. Choose a baseline model that removes the main decay trend before you search for yearly structure.
4. Build a standard signal search using wavelets, periodograms, or both, so you can compare methods.
5. Plan controls that separate a true annual effect from detector drift, sampling gaps, and seasonal lab artifacts.
6. Set an upper-limit method that converts your null result or weak signal into a quantitative bound.

Common Pitfalls

• Mixing datasets with different sampling intervals, which can create fake periodic structure.
• Skipping detrending, which leaves the main decay curve in the data and hides small modulations.
• Treating one spike in the wavelet plot as proof, which ignores how noisy time-series data can be.
• Ignoring phase differences between BNL and PTB, which can make unrelated patterns look like one effect.
• Using too many analysis settings until one happens to look significant, which inflates false positives.

What Makes This Competitive

A strong version of this project does more than point at a plot. You can compare multiple analysis methods, justify every preprocessing choice, and test whether the result survives resampling, detrending changes, and dataset splits. The best projects also turn the analysis into a limit, not just a yes-or-no claim, so your conclusion has a real numerical boundary. That kind of careful work looks much closer to research than a class demo.

Project Variations

• Analyze alpha-decay datasets instead of beta-decay datasets and compare whether any annual pattern changes by decay mode.
• Compare wavelet results with Lomb-Scargle periodograms to see whether both methods pick out the same seasonal structure.
• Test whether the effect size changes when you analyze only one detector, one isotope, or one contiguous time window.

Learn More

• PubMed: Search for review articles on nuclear decay rate variability and solar influence hypotheses.
• NASA: Use mission and heliophysics background material to understand solar cycles and neutrino-related context.
• NIH National Library of Medicine Bookshelf: Look for free textbook chapters on statistics, uncertainty, and signal analysis.
• MIT OpenCourseWare: Search for course notes on probability, statistics, and data analysis for physics.
• USGS Earthquake Catalog methods pages: Read about time-series filtering, event detection, and noise handling in long records.
• Physical Review Letters and Physical Review C: Search for peer-reviewed papers on reported decay anomalies and follow-up tests.