Well, I think it may be a good idea to continue this specsmanship game a little longer, because it will help us and others understand the specs better and how they
might relate to a small, practical detector setup. I do like the interchangeable head idea, and already use it here in nearly everything, partly because it also
amounts to an interchangeable back end too, and one sensor might give more than one kind of information, say counts vs pulse heights or total energy. Some detectors
flag the difference between a gamma hit and a neutron hit not by total light, but by pulse width mainly. Recoil protons in plastic from a neutron hit can individually
take on any energy from near zero to about half the energy of the neutron, with a fairly square distribution. The general case is that if the scintillator is
long enough in the direction the neutron came in, and large enough, there will be severl recoil protons per neutron, with some time smear, so instead of
providing a pulse height that tracks neutron energy, you get a wider pulse for neutrons, and the total energy in that wider pulse is the thing that somewhat
correlates with the incoming neutron energy. Gammas will tend to make a skinny pulse, the width of which is determined by the scintillator/phototube
characteristics. Usually the scintillator is the slow part, as most modern phototubes will keep the time spread under 5 ns or so. The plastic scintillators only
add to that a little, while nearly all the others are much slower and kind of negate the ability to use pulse width as a meaningful input.
One extremely important aspect that has been left out so far is a spec called "radiation length". This is the length of scintillator that it will take to completely
absorb radiation of a given type/energy and change it to photons, and it matters a lot in all detector types. To give an example, it might take a 3" deep NaI
to absorb the same energy gamma fully as it does a 1" BGO. The specs Lutz has been giving, if I remember my books correctly is photons per
absorbed energy.
Now, if you are simply counting pulses, you may not care one bit about that as long as for the things of interest, you get enough photons to get a count out of
the phototube noise -- once you "digitize" a pulse into a 1 or a 0 (for pulse or not-pulse) it no longer matters. In that case what
does matter is whether
the gamma or beta or neutron hits the scintillator at all. Obviously a 10 sq cm detector has a better probability of intercept than a 1 sq cm one does. 10 times
better to a first approximation, neglecting edge effects where a marginal strike may not pass through enough of it to make photons over the counter threshold
you've set up. In which case, the bigger guy is better yet, as it has more sq area vs edge area.
If you want pulse height information, you must have a depth of the scintillator that is long enough to completely absorb the radiation or the pulse height
spreads out for a single line source, as some but not all the gammas make a full height pulse. The others, perhaps due to geometry issues scatter out
of the crystal before being fully converted. In general, this means a good spectroscopy head will have some kind of shielding with a hole in it to only
allow photons into the scintillator that are coming straight on to the long axis of the crystal. For example, if one wanted to use one of the cheap and
readily available NaI heads in both modes, one might have a chunk of cast lead to insert it into or spectroscopy, but use it without for general gamma
detection and in the latter case let it be sensitive from the sides (more sq area) for that use. So you can have two uses for one head as well.
Here is a table from one of my books on the topic. There's still not enough here to really do a good design, but it's a step up.
- Scint Table, part 1
- Table, part 2
As usual, click these to get a readable version. Note that even with this much more information, we still don't know all we'd like to, but at
least we know a little more. We see that for example, BGO can absorb radiation better per unit length, so you'd need less in a pulse height analysis
situation, and indeed it is very popular in PET scanners for this reason -- it has "good enough" resolution for that, in a much smaller size. In fact,
other than the tiny ones from PET scanners, I know of no other affordable source of these. Now, this table was put together by the guys designing
detectors for places like CERN, and doesn't have some other info important to us in it other than not covering organics at all (I have another somplace for that).
They don't care much about low energy stuff, so good things for that are left out, as are slow things (from their perspective, a few nanoseconds is slow
as they have an 18 ns cycle time at CERN).
A consideration not mentioned here is how badly they attenuate their own light, which begins to matter when you go to bigger sizes for more chance
of intercepting the radiation you want to be sensing. For example all the Zinc based things are horrible in this regard along with being slow (which we
might not care too terribly much about within limits for our design unless it's really slow).
So the pure sensitivity number compared to NaI is more or less meaningless IF: A given hit gets you out of phototube noise well enough. That takes less
for go-nogo than it does for a good PHA spectrum. And those numbers are all for fully absorbed hits -- which may not be the case if the radiation length
is long, but the crystal not. Another important thing to know -- those are theory numbers, not what you can practically get out of one end of a given crystal,
some is lost, some reflects back from the index of refraction gradient, and hey -- phototubes aren't frequency flat and neither are these crystals.
So as with most design, it takes more than a single number to know what's best for a given application. And as always, price and availability figure in there
someplace.
Unfortunately, I can't find a similar set of tables of all the plastic scintillator combinations. Of course, you can go to Eljen or similar place and get individual spec
sheets on them. I can only state what I know for now. Most plastic or liquid scintillators have a base medium, or solvent, and two fluors in them. The first
fluor is selected for most efficient energy transfer from the base material to it, and the second is selected as a down shifting fluor to get to the range
of visible light so an inexpensive detector can manage it. Percentages of both types are adjusted to suit the needs of individual applications. Most plastic
scintillators are very fast -- in the few nanosecond range, but of course this family includes some that use zinc compounds that are so slow as to
be useless for us unless we want to simply average it all and put it on an analog panel meter. For anyone who has not built one of these, I would point out
that radiation is about the most random thing in the known universe, and even a fairly slow rate of counting will have situations where two events happen
inside any long deadtime. Even microsecond dead times will thus undercount 1k cps sources fairly badly, so speed is much more of an issue than you'd think.
I still like plastic best of all, given the above and other considerations (like I can get it). Here is my reasoning on that.
You can make it big and still afford it. No matter how efficient a tiny crystal is, radiation that misses it entirely will not be seen.
The optical properties are a lot better for getting the light out -- it has low index, and they tend to absorb little of their own light, so you can make them big.
You can machine them to whatever form factor you desire, and this includes making them shaped so they guide light to the optical sensor the best.
The stuff is impervious to atmospheric degradations. It may not live next to an interaction point at CERN, or inside a reactor - but we don't do that.
It sees neutrons with really decent efficiency.
It's fast, down in the single digit nanoseconds.
The main downside is that it's not great for spectrometry (radiation length tends to be long) -- that might want to be a different thread here, as that kind of thing isn't generally hand-held anyway, you need some real display area to even see a good spectrum. Though actually in tests here, it's not really bad either if you've got a big enough piece and
a "good geometry" experiment so that most energy doesn't scatter out before being changed into photons. NaI and BGO rule there until you get into
cryo things like Ge.
I would note that significant improvements to the "henny penny" design are possible -- look at this picture of the guts.
- Henny's guts
Just for starters, unless there is some serious optical trickery going on under that tape -- most of the photons aren't finding their way to the tube.
Next, notice some off the shelf power supply is used that simply drives a normal resistive divider dynode chain -- little thought taken to efficiency there.
Also note the PCB and the more or less obsolete, high power, components used -- again, little thought to battery life, so you also notice:
The big battery pack it needs to run for a fairly short time.
Yet this is far and away the most sensitive detector to ever grace my lab. I would think that with just the improvements possible and pretty easy to do,
we could shrink this puppy quite a bit, gain longer battery life, and make it even more sensitive by fixing the dumb optics design. The two biggest remaining
pieces would be the scint, and the display -- the rest can get tiny and power efficient easily if we just use more modern electrical parts. Which shrinks the
batteries needed and so on. It wouldn't quite fit in a shirt pocket unless you made the scintillator a lot smaller, and the existing cheap LCD displays aren't
really thin enough, but given those limitations, I think we could design something really good - and generally useful.
So, no matter what, I'd suppose the search should be on for a good blue sensitive phototube, small, that we can drive the dynodes independently.
The ones I have access to at a decent price are bigger than this one by a good bit, though we could take the sockets off them to get to the
dynodes and ditch the R string in there. We want a gain level roughly in the million range.
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.