My idea is to select a few infrared wavelengths where the reflectivity of the popular plastics differs the most (e.g. 1000, 1500 nm) and place them around a camera, like those webcams with infrared LEDs (http://4.bp.blogspot.com/_6cM8WmwYUjc/S9ggrx0ceOI/AAAAAAAABHc/yjWREGZwCGY/s1600/Screen+shot+2010-04-28+at+9.47.34+PM.png). By measuring the brightness of the light, it should be able to automatically identify what type of plastic it’s looking at. The problems I’ve seen so far are that, like the thread says, LEDs and detectors only go up to 1+ microns, but FTIR uses much longer wavelengths than that. Also, water vapor is opaque to certain wavelengths, which might affect it.
Interesting - I’ve been chewing this idea over too - my plan was to use an array of different emitters and another array of narrowband sensors - then illuminate the sample by each emitter in turn and build up a profile of the absoption, fluorescence, etc - sort of a budget solid state spectrometer.
Trouble with this approach is that plastics are often tinted or loaded with fillers, plasticisers, etc. Sometimes they’re blended (HDPE and PP for example). I suspect this will introduce enough variables to make false positives a real problem.
The halfbaked alternative idea I had on this was to use a laser to vaporise a small sample, then capture and analyse the gases.
I don’t thin fillers would be much of a problem, if you pay attention to the infrared wavelengths at which the major bonds and pendant groups are visible. After all, there aren’t going to be -CH3 side groups swinging around on silica or carbon or titania or diatomite.
There are lots of cameras that can see in the wavelengths at which resonances identify polymers.
One issue - there are not many sources of infrared available besides thermal emitters, and the long wavelength intensity of an infrared emitter only varies as about the first power of the temperature. Room temperature is 10% of the way to the highest temperatures that solid emitters can handle (e.g. incandescent bulbs with tungsten filaments). So, objects glow somewhat brightly compared to the radiation they can receive from a lamp whose filament is much smaller than the objects are.
I am reminded of the KarTrak system of barcodes installed on all railroad cars in the 1970’s – well fine in theory, in practice railroad cars get so dirty that about 20% of them were misread.
in recycling, many (most?) plastic containers are already somewhat dirty when discarded. They are likely to become more dirty as they are collected & transported. i think this contamination with all kinds of different food, beverage, cleaning, products could easily confuse the key wavelengths enough to cause a high rate of misreads, making this system unreliable.
Ah, thanks for your comments. I think currently plastics are sorted manually, which is very labor intensive. The point about contamination sounds very relevant.
Infrared and near infrared spectrometers are used routinely to analyze plastics. I’m not sure how often they are used in recycling plastics, but I’ve used devices designed for identifying the plastics in recycled carpet.
Its easy to do technically and is done often enough, but I think the bigest problem has already been mentioned: recycling plants don’t deal often with ideal circumstances. There are too many dirty items.
My job allows me to test different applications for such spectrometers, and I’ve personally done some plastics work for recycling applications.
