For some time now I've been working on a system that can produce an analysis of elements that are present in an unknown sample (within limits, explained later). It's based on a lot of excellent work by an Italian group called Theremino. They use a combination of parts that can be acquired on ebay, such as a type of x-ray detector called a scintillator, a photomultiplier tube, high voltage power supply, some pulse-shaping circuitry and a computer's audio board (one that can digitize audio).
The basic idea is to use a variation on fluorescence, where some materials will emit visible light when illuminated by ultraviolet light. The exact term is XRF, for X-Ray Fluorescence. For ultraviolet illumination, only a few materials will emit light: and they, as far as I am aware, are compounds, doped or otherwise. The emission wavelengths are more dependent on the dopant than the compound. While this might sound OK, the dopants typically are exotic elements like Yttrium, Eutropium, Erbium and the like -- not of much interest if you want to know if your jewelry really is gold or silver.
On the other hand, x-ray illumination can excite fluorescence of all elements except hydrogen and helium. The key is that the emitted light is characteristic of the element, not a dopant. So if we had a way to (1) illuminate a sample with x-rays and (2) determine the wavelength of light emitted by the sample's fluorescence, we can determine the elements present and get an approximate idea of their concentration.
The first problem to solve is a hobbyist-available X-Ray source. The Theremino group uses radioactive capsules from ionization-type smoke detectors. They have a small amount of Americium-241 in them, representing about 1 micro-curie of radioactive decay (about 37000 decays per second). Along with the desired alpha particles used by the smoke detector circuitry, the capsule emits gamma rays with an energy of about 60,000 electron-volts, 60 Kev (visible light is in the 2-4 electron-volt range). Before anyone freaks out, the gamma ray flux is quite low, and, since smoke detectors typically are installed on the ceiling, a long way away from people. This reduces the radiation dose even more. So in normal use they are safe. The Theremino application uses several of these to produce a higher count rate for better analysis results. I don't think the total x-ray flux is much of an issue, but, even so, my system uses lead shielding to completely remove any misgivings on my part.
The other issue with regard to the Americium241 (Am241) is the 60Kev photon. It can only excite fluorescence in elements if their characteristic fluorescence is at a lower wavelength. That makes sense, right? The emitted light's energy per photon CANNOT exceed the incident photons that are stimulating the fluorescence. Otherwise you have a variation on the perpetual-motion machine, where the energy-out exceeds the energy-in. So Am241 can't excite the primary fluorescence line for any element past Tungsten, whose k-alpha line is about 58Kev. Elements with higher atomic numbers won't get "tickled" enough to emit X-rays. It gets a bit more complicated because inner-shell electrons can get involved and emit much lower energy x-rays, but their energies are much more difficult to detect, out of the range of commonly-available x-ray detectors.
So. We have a fairly readily-available x-ray source to cause elements up to Tungsten to fluoresce. Now, how do we tell them apart? The key is a particular type of radiation detector called a proportional detector. For each incoming x-ray photon its output is proportional to the photon energy, which is directly related to its wavelength. Planck's constant at work. Look it up! Anyway, there are a few hobbyist-available proportional detectors. The first is a scintillator. It's a special crystal that outputs a burst of light whose intensity is proportional to the x-ray photon that generated it. The crystal typically is optically coupled to a high-speed detector called a photomultiplier tube that can turn the burst of light into an electrical signal, along with a substantial gain (that's the 'multiplier' part of the name). The photomultiplier tube (PMT) output also is proportional to the light input, so the pulse height coming out of the PMT is proportional to the x-ray photon that created it. Each element has a characteristic x-ray fluorescence wavelength, so by sorting the pulse height into bins, where each bin represents a small range of pulse heights, we obtain a spectrum that is a fingerprint of the elements present in the sample.
This sounds pretty good, except for a few things I haven't mentioned yet. The first is that the PMT requires a fairly high voltage to operate, about 800-200 volts. It doesn't require much current, but that's still a challenge. The voltage determines the gain of the PMT, so it also is necessary to regulate that voltage, while avoiding instability (oscillations) and noise. Photomultipliers also are sensitive to magnetic fields so they need to be magnetically shielded, using mu-metal surrounds. They are very light-sensitive so they also need to be completely shielded from visible light (and if exposed to light when their supply voltage is still on, they can be destroyed!). Finally, the best scintillator crystals are hygroscopic -- they absorb moisture -- so ones you find on ebay may be seriously degraded.
PMT's can be found on Ebay for not a lot of money, but then you need to add in the power supply and scintillator.
An in-between solution is to use a silicon photomultiplier, also called an SiPM. It is a type of avalanche photodetector and is more robust compared to a PMT, plus it usually only needs 20-50V to operate. It still requires a scintillator crystal with all the attendant issues related to the types that offer the best energy resolution (i.e., very hygroscopic).
Finally, there are PIN (P-type/Intrinsic/N-type) diode detectors. The name is related to their structure, with the important part being the relatively wide lightly doped Intrinsic region that can capture lots of x-rays. This type of diode can be all over the map, depending on the range of elements you want to analyze, plus the sensitivity of the system. The most sensitive detectors have an energy range that covers Boron, with a primary x-ray wavelength of 185 electron-volts on up to lead and beyond, in the 10,000 electron-volt range. These are operated in a liquid nitrogen dewar at 77 degrees Kelvin, operate at hundreds of volts, and would destroy themselves if for some reason they rose to room temperature. These systems are very expensive and very unforgiving if you screw up.
However, if we're willing to give up something on the lighter-atomic-weight range, we can get by with something that runs at room temperature. Silicon PIN diodes specially designed as x-ray detectors can be had for something in the $100 range, and, in some cases, for a bit less if already in a consumer application: PocketGeiger. In the latter case, the device is not meant for XRF -- it's a simple geiger counter, not sensitive to the incoming x-ray photon energy. But it appears to be possible to tap into an internal signal line that MAY permit some form of elemental analysis. I'm hopeful about this and will report results as I get them. I have already found that the boost switching regulator, used to generate the ~26V bias voltage needed to operate the PIN diode, is a significant noise source and needs to be replaced by an external bias voltage source. I didn't find an easy way to eliminate the noise and keep the on-board switching regulator. Moving it off-board might work, but for now I'm using an old linear bench supply to provide the detector's bias voltage and it is working OK.
