Rutherford Gold Foil Experiment

What Actually Happened in That Lab at Manchester

Geiger and Marsden were running the measurements under Rutherford's direction around 1909. They fired alpha particles at an extremely thin gold foil and watched where the scattered particles landed on a zinc sulfide screen. Most of the time the alpha particles went straight through with barely any deflection. That was the expected result based on the Thomson plum pudding model. The surprise came when a small fraction bounced back at angles greater than ninety degrees. Rutherford later said it was almost as unbelievable as if you fired a fifteen-inch shell at tissue paper and it came back and hit you. If you are trying to recreate this for a teaching demonstration or just want to actually see the physics work instead of reading about it, the realistic route is to use a computational simulation. There are several freely available ones online. The PhET simulation from the University of Colorado is probably the most reliable starting point. You can find it by searching for "PhET Rutherford scattering." The raw download isn't always necessary since it runs in a browser, but if you want it offline you can grab the zip from their GitHub mirror and unpack it locally. I spent a couple of afternoons last year trying to get a custom Python implementation working so I could tweak parameters for an advanced undergraduate lab. The standard approach uses a Monte Carlo algorithm where you sample impact parameters from a uniform distribution and calculate the scattering angle using the Rutherford formula. The differential cross section is given by:

d/d = (ZZe² / 16E)² × 1/sin(/2) Here Z is the atomic number of the alpha particle (which is 2), Z is the atomic number of gold (79), e is the elementary charge, is the vacuum permittivity, E is the kinetic energy of the alpha particle, and is the scattering angle. The formula blows up as approaches zero, which sounds problematic but is actually just the mathematical reflection of the fact that most particles scatter at very small angles. You need to introduce a physical cutoff angle or an impact parameter cutoff in your simulation to avoid numerical overflow. The problem I hit was that the simulation was generating too many large-angle scattering events for the default parameter set. What was happening was that my random number generator for the impact parameter wasn't properly weighted. I was sampling b uniformly from zero to some maximum value, but the proper distribution for a scattering experiment requires sampling proportional to b db because the number of particles hitting an annulus of width db at radius b is proportional to the area of that annulus. Once I switched to sampling b with a square root distribution — basically taking the square root of a uniform random number and scaling it — the angular distribution matched the theoretical prediction almost exactly. This took me about three hours to track down because the output looked qualitatively reasonable at first glance. The total count rates were correct, just the angular spread was wrong.

Another detail that nobody emphasizes enough is the foil thickness. In the original experiment, the gold foil was roughly four hundred micrometers thick, which is about two thousand atoms thick. If you are simulating this, you need to account for multiple scattering events. A single alpha particle doesn't just interact with one nucleus and stop. It passes through many atomic layers and accumulates small-angle deflections. The original Geiger-Marsden setup was sensitive enough to detect the large-angle events that resulted from single close encounters precisely because the foil was thin enough to make multiple large-angle scattering statistically unlikely. For a realistic simulation, you should apply a Gaussian smearing to the final scattering angle to approximate the effect of multiple small-angle collisions, using a width determined by the material's radiation length or by tabulated multiple scattering distributions from something like the Highland formula.

Why the Results Broke the Plum Pudding Model

The Thomson model predicted that positive charge was spread diffusely throughout the atom. Under that model, the maximum electric field an alpha particle would encounter is relatively weak, and the cumulative effect of many small deflections should produce a narrow angular distribution centered near zero. The observed backward scattering was completely incompatible with that prediction. The only way to get an alpha particle to reverse direction is if it encounters a very strong electric field over a very short distance, which requires the positive charge and most of the mass to be concentrated in a tiny region. That region is the nucleus. The nucleus itself is on the order of femtometers across while the atom is about angstroms across. That means the nucleus occupies roughly one ten-trillionth of the atom's volume. Most alpha particles pass through the empty space between nuclei and experience minimal deflection. A few pass close enough to a nucleus to be strongly scattered. The probability of a large-angle scatter scales with the square of the nuclear charge and inversely with the fourth power of the alpha particle's kinetic energy. That is why Rutherford's formula contains the Z² term and the 1/E² dependence.

Practical Limitations You Should Know About

The Rutherford scattering model works beautifully for alpha particles and light ions at energies below about ten MeV. Above that threshold, you start running into problems because the alpha particle can get close enough to the nucleus that the strong nuclear force becomes relevant, and the purely Coulombic cross section formula breaks down. For gold with Z = 79, the Coulomb barrier is roughly thirty MeV for an alpha particle, so this isn't usually an issue at typical lab energies of four to eight MeV, but it matters if you are pushing the simulation into higher energy regimes. Another limitation is that Rutherford scattering assumes a point-like nucleus and a pure Coulomb potential. At very small impact parameters, electron screening becomes negligible, but at larger impact parameters the atomic electrons partially shield the nuclear charge. This reduces the effective Z experienced by the alpha particle at grazing angles and causes a measurable deviation from the pure Rutherford formula at scattering angles below about five degrees. If your experiment or simulation requires accuracy in that forward angle region, you need to incorporate a screening correction, typically using a Thomas-Fermi or Molière screening function. The model also completely ignores quantum mechanical effects like diffraction and interference, which become relevant when the de Broglie wavelength of the incident particle is comparable to the nuclear size. For a five MeV alpha particle, the de Broglie wavelength is roughly 6.4 femtometers, which is indeed in the same ballpark as a gold nucleus radius of about 7 femtometers. In practice this doesn't ruin the classical description for most scattering angles, but it does mean the classical formula is an approximation, not an exact result. Full quantum mechanical treatments using partial wave analysis give essentially the same answer for this energy range, which is a nice consistency check but doesn't change how you should present or teach the experiment.

What to Actually Watch For in the Data

If you are running a real lab with a radioactive source and a Geiger-Müller tube or silicon detector, the first thing you need to worry about is background radiation. A typical alpha source for this kind of experiment might be Americium-241 at about one microcurie, which gives you a count rate of a few thousand counts per second at close range. Your background is usually around twenty to thirty counts per minute from cosmic rays and environmental gamma radiation. That background is negligible for the direct beam measurement but becomes relevant when you are measuring very small scattering angles where the signal is already low because most particles go straight through with minimal deflection. The detector geometry matters a lot. The solid angle subtended by your detector determines your counting rate, and it goes as the detector area divided by the square of the distance from the foil. If you move the detector farther away to reduce background from the unscattered beam, your count rate drops with the square of the distance. A practical rule of thumb is to keep the detector about ten centimeters from the foil and use a detector aperture of roughly one square centimeter. This gives you a solid angle of about one percent of four pi steradians, which is enough for reasonable statistics without being overwhelmed by the primary beam. One edge case that catches people off guard: the gold foil is not a single crystal. It is a polycrystalline film with randomly oriented grains. If you were to use a single crystal, you would get channeling effects where alpha particles traveling along certain crystallographic directions experience much less scattering than expected. The polycrystalline nature of the foil effectively averages out these directional effects, which is why the original experiment used evaporated gold foil rather than a crystal. If you ever source your own foil and it turns out to be crystalline enough to show channeling, your angular distribution will look nothing like the Rutherford prediction. I learned this the hard way when a supplier sent me a batch of gold foil that was annealed and grain-sized enough to produce slightly anisotropic scattering patterns. Switching to a different supplier who guaranteed a randomly oriented microcrystalline structure fixed the problem immediately.

The experiment also assumes that the alpha particles lose a negligible amount of energy passing through the foil. For a four MeV alpha in four hundred micrometers of gold, the energy loss is roughly fifty to one hundred keV according to the Bethe-Bloch formula. This is small enough that you can treat the energy as approximately constant for the purpose of calculating the scattering angle, but if you need higher precision, you should integrate the energy loss continuously through the foil and recalculate the scattering at each layer. For a teaching lab this correction is unnecessary. For a research-quality measurement, it changes the inferred nuclear charge by less than one percent.

Common Misconceptions

People often say that Rutherford "discovered the nucleus" in this experiment. That is technically inaccurate. The experiment demonstrated that the positive charge must be concentrated in a small region, but Rutherford did not immediately calculate the nuclear size or propose a complete atomic model until 1911, two years after the experiment was conducted. The initial paper by Geiger and Marsden was purely experimental. Rutherford's theoretical interpretation came afterward when he worked through the mathematics of Coulomb scattering and realized what the data implied about atomic structure. Another frequent mistake is the claim that this experiment directly measured the size of the nucleus. It did not. The experiment established that the nucleus must be small, but the actual size determination came later from higher energy scattering experiments and from analyzing the deviations from the Rutherford formula at very large scattering angles where the alpha particle gets close enough to the nucleus for the finite nuclear size to matter. The original gold foil experiment could only set an upper bound on the nuclear radius, which Rutherford estimated to be less than ten femtometers. The experiment is also sometimes confused with the later cloud chamber work by Wilson or the spectroscopic studies of atomic spectra. Those are entirely different investigations, though they all contributed to the development of atomic physics in the same general period. The Rutherford Gold Foil Experiment is specifically about particle scattering and the determination of atomic structure through direct mechanical interaction rather than through electromagnetic absorption or emission.

Where to Go From Here

If you want to dig deeper into the actual historical data, Geiger and Marsden's original papers are available through the Royal Society's document delivery service or through scanned copies in various university library archives. The data tables are surprisingly clean. For a modern treatment that covers both the classical and quantum mechanical aspects, the textbook "Introductory Nuclear Physics" by Krane has a thorough chapter on scattering theory that derives the Rutherford formula from both approaches and discusses the limitations in detail. For a hands-on computational exercise, writing your own Monte Carlo simulator in Python or Julia and comparing it against the analytical cross section is a genuinely useful way to understand what is actually happening in the experiment rather than just memorizing the formula.