unome - 9-4-2010 at 17:34
I've been looking at the schematics of the Michelson (and various other interferometers) and have to ask the question... Do we need a beamsplitter?
Why can't we just use parallel beams of light that are then recombined using a combination of mirrors & lenses?
I cannot 'see' why there should be any real objection to the concept - the same light-source, giving two parallel beams of light, one of which is
totally reflected to the sensor and the other is totally directed toward the moving mirror and from that, back to the sensor...
The two beams would be recombined using the convergence of their beampaths and then either lenses or optical fiber, but the point is the
interferometric response should theoretically be the same, we would have a beam that has been effectively split, one half of which has traveled
further than the other and then has been recombined...
The only reason I ask, is that MIR-glasses (Arsenic Trisulfide for example) would be a bitch to build a two-way mirror on. And from what I have seen a beamsplitter would be problematic.
unome - 10-4-2010 at 03:57
So this is where the post ended up? Ok
See the attached graphic, I cannot see why it would make any difference, given the two beams have but one source (and are therefore effectively split
at the outset anyway) why this wouldn't work...
Arsenic trisulfide glass is available, it is fairly soft and can be cast, so lenses, etc. would be rather simple.. The only really hard bit about
the interferometer itself (apart from the sensors) is the beamsplitter, the rest is front polished mirrors, ie. metal.
A single lens would serve to combine the beams at the top of the attached graphic, which could then be directed wherever it was needed.
Attachment: no.splitter.interferometer.tiff (59kB)
This file has been downloaded 798 times
According to wikipedia it is transparent between 620nm and 11um which means it is perfect for MIR spectroscopy and there are moves afoot to make
optical fiber from it.
[Edited on 10-4-2010 by unome]
bquirky - 10-4-2010 at 05:53
Any peice of glass will act as a %4 beam splitter.
grab a nice smoth microscope slide send the beam onto it at 45Degrees and you will get %4 reflected and 96% transmited (this happens twice becuse
there are two glass-air interfaces) so you actuly get 90sompthing% transmited.
allthough not optimal from a sensitivity point of view. it should be more than enough to detect some fringe paterns.
interferomiter based instruments oftern run with radicly differnet power levels in each arm. an OCT system might have 10mw in the referance arm and
100 Nano watts coming back from the tissue sample.
The beam spliter is important for getting the two beams in the first place but are not so important for combing them to see the interferance patern.
Becuse if the two beams didnt come from the same sorce (or point on the source if not coherant ) they wont have matching wavelength, phase and
polorisation states and the fringes that you expect to see just wont be there. (becuse they will be berried in noise)
The good news is that any old light can be used to generate visible interferance paterns in a generic interferometer
I may be missunderstanding your question. but using lenses to mess with the reletive path lengths of two beams of light sadly dosnt work. Becuse the
light that travles the 'shorter' distance through the middle of the lense. has to spend more time in the glass (due to the fact the lense has to be
thicker in the middle to work) and the glass has a higher refractive index so the light will travel slower over the shorter physical distance where as
the beam of light that has to travle further to get to the edge of the lense it will spend less time travaling slower in the glass due to the thin
edge of the lense.
so in a practical application..
if two paralell beams of light hit a lense (one at the middle of the lens and one closer to the edge.)
and are focused to a point both rays will arive at that point at exactly the same time. dispite the aperance that the beam hiting the outer diamiter
of the lense hase to travle further.
bquirky - 10-4-2010 at 05:56
Allso plain old NaCL crystals are transparant at fairly long IR wavelengths so you might be able to polish two flat surfaces on one to make a cheap
10um beam splitter if you felt keen
(I dont know what ratio that whould give you at that wavelength off the top of my head)
unionised - 10-4-2010 at 06:01
For visible work you can use a piece of plain glass as a beam splitter so I presume that you could use As2S3 glass in the IR.
Unless you have the facilities for working As2S3 and such like properly and safely then I'd not even think about it if I were you. It's true that the
stuff melts at a reasonably accessible temperature but, when it comes down to it, the stuff is full of arsenic.
You might want to look at pelicle beamsplitters too.
bquirky - 10-4-2010 at 06:03
or maby melt some NaCL ontop of a polished glass surface and then pull them appart
unome - 10-4-2010 at 15:06
The As2S3 is also non-toxic, that and the melting temperature are what gave me the (rather than ZnS, which requires much higher temperatures and then further processing) idea... The entire thing would depend upon getting a useful source of it
(that has a low-no content of the toxic oxide), then processing it. The number of lenses needed would be fairly small - collimation of both beams,
then the joining & collimation of the end beam to the sensor(s).
I still have no answer on whether a beam splitter is actually necessary, that is an intriguing question of itself... It suggests that this could be
done with a shitload less mirrors/lenses down the track when they start rolling out MIR Chalgogenide Glass fibers for optical communication... Actually from what I have read, the As2S3/ZnS are both shortlisted for that job... If no
beam-splitter is needed (and I'm only working from logic here, but I see no 'need' for it), then there would only be the one open beam, that from the
input fiber onto the moving mirror and then back to the output fiber.
The shortpath beam would go from the source directly the recombining optics (no mirror) where it would join the long/variable length beam and be
directed into the optics. As these are going to be huge volume components, prices will fall through the floor, including the light/fiber connections
and the dual fiber combiner optics.
The reason I am saying logically, is that the interferometry is based upon the presence of (1) a constant beam (the short one) and (2) a changing beam
(the long, variable one). The amount of light at the end of that is what the FT is worked out from, from a known and a variable, we can determine the
change in intensity (from, for example X x the known value = the intensity of light at a given position). The shortpath beam being a constant, the
difference in the beam path lengths is irrelevant - yes, we are computing the whole thing on light that went out milliseconds earlier than the other,
but one is a constant...
What I am saying is, I don't see why it would be necessary to incorporate a beamsplitter, when we could simply focus the light (using mirrors) into
two separate beam paths at the outset (effectively 'splitting' the beam at the light source, rather than with a beamsplitter - with the benefit being
that we would not need anything other than metal mirrors to affect that, ie. no 50% transmitting mirror would be necessary).
The real reason I am thinking of it is that the fiber optics mentioned are looking more and more like they are going to become a huge-volume reality
(and thus cheap). If we can work out a way to effectively 'sense' the output from the fiber optics, then amateur, homemade FT-IR becomes a
possibility. Especially given the posited existence (and realistically, not that far off) of cheap, OTC components (like chalcogenide glass lenses
& optics, plus the fiber itself).
[Edited on 10-4-2010 by unome]
unionised - 11-4-2010 at 02:01
A2S3 is relatively non toxic because it's very insoluble.
On heating in air it forms the oxide which is volatile, soluble and toxic.
If you could do without the beam splitter then people would do so.
bquirky - 11-4-2010 at 02:03
Im not sure im really understanding what your trying to achieve.
But with regards to the beam splitters ill have a go at starting from first principals.
If you Imagen two identical wave forms in phase meeting up ie
\/\/\/\/\/
\/\/\/\/\/
the peeks will add and you will end up with a wave form with a larger amplitude.
now say that they are out of phase.
/\/\/\/\/\/
\/\/\/\/\/\
the two wave forms will cancel out leaving no signal.
now let say they are ALLMOST the same but out of phase
/\/\/\/\/\/
\/\__/\/\
your output signal will look a bit like this ___/\/\/\___ so your resulting signal is basicly the diffences betwen the two orignal nearly identical
waveforms.
now lets say that they are completly different waveforms
/\!^\|_^-
^?|_/\_$
The diferences between two compleatly diffrent waveforms is going to be random noise (unless you have very specific apriori knowledge of at least one
of the original signals)
So the way you can use this to your advantage is to take two identical wave forms and do something to one of them (like change its length or pass it
through some material or whatever) and then combine them. the interference signal then contains information about what happened to one beam and not
the other which makes interferometers useful as scientific instruments.
but the key hear is that the two waveforms have to be the same to start with !!
Otherwise your not able to atribute the changes in the output signal to anything that happend in the sample arm of the interfaromiter.
Getting two identical beams ( waveforms ) of light can be surprisingly difficult.
If you just take a light globe and grab the light leaving to the left for one beam
and the light leaving to the right for the other. the light comes from different parts of the filament created by different molecules excited at
slightly different temperature.
Evan if the averaged bandwidth wavelength and intensity are exactly the same the specific shape of each wave form at each point is different.
this is true for allmost all light sources evan non thermal radiators.
its like having a serise of random numbers that all add up to 45
123456789 appears the same as
213465798 when looked at on average (measured) but its not.
ok so what if we take a single beam expand it with a lens grab the left side as one beam and the right side as the other ?
well unless the beam is from a spacialy coherent light source with a Gaussian beam shape the beam that you have expanded is really more than one beam
of light ! what you have is essentially a slide projector that is projecting an image of the light source. This becomes quite clear if you grab a
lens and try to focus a beam of light from a incandescent light globe you literally get a picture of the filiment. so if you where to devide this beam
in half you whould be back to having one beam from one part of the source (or filliment) and the other from a different part.
semiconductor lasers will also give you this problem as the physical area of the emmiter (the pn juction) will spit out slightly diffrent waveforms
one one side than the other. Only a nice clean gas laser will have a uniform enough beam to divide in this fasion.
With a beam splitter you dont have these problems because your actuly creating (at the expense of amplitude) a 'copy' of the original beam. each beam
will have exactly the same shape waveform like this
............. /^|>\/\^
/^|>\/\^
............. /^|>\/\^
(the ... are just for formating spaces seem to get munched)
so when you look at the interference pasterns of the two you will be able to take a meaning full mesurement of changes made to one of the beams as a
result of your instrument.
There is however another way to do this. that is by using a pinhole (or slit) when passing through a confined area waves have a tendency to reradiate
this is what happens in the double slit experment http://en.wikipedia.org/wiki/Double_slit_experiment basicly the defraction caused by the sharp edge causes a new wave front to emerge as if
radiated by the slit itself.
But this destroys Smegloads of optical power that a beam spliter whould save you.
Thats my best attempt at describing beam splitters 
unome - 11-4-2010 at 03:33
Hmmm, yes it would be less than trivial to get perfection in the beams - then again I would argue that it would be far from insignificant to achieve
perfection through the mirror too... What is the ratio of the light transmitted & reflected and what are the tolerances?
The reason people have not done without the beamsplitter to date is quite possibly down to the fact that it works and it is used by the major
manufacturers (who can charge whatever they like for parts). MIR Optics were NOT that easy to source, which is rapidly changing, especially given that
optical fiber that transmits in that range is going to revolutionize the instrumentation anyway (FT-IR via FO probes, the papers are already
available).
Why would it be "TOO" complex to align some mirrors around the light source to apportion half the incident light into each of two pathways? I mean,
considering what is done with mirrors within the unit, it would seem that it should work, no?
The short path - the constant is non-variable within a range (which could always be smoothed electronically anyway), and the long & variable path
is 'assumed' (within the bounds of the instruments accuracy) to be the same as the constant at the outset.
I mean, it is difficult to believe that the reflection:transmission would be a constant 50:50 ratio, as you say, the variables like you mentioned
would have a serious effect on what was and was not split down the middle... The spikes etc. in the output, would have a marked effect on the signal
processing for a start, no matter whether you are using a beam splitter or not.
I mean, it wouldn't even necessitate anything as simple as two holes - I'm suggesting using several mirrors around the light source, the output of
which is then collimated using a fairly simple lens... I'm still trying to work out why the light from a single light source would be so different, and this is
something that will be interesting since the MIR fiber is becoming available (so too the optics)... Unfortunately, I can't see high speed
telecommunications/networks needing a lot of half-silvered mirrors that transmit in the MIR, so this will always be the sticking point in terms of
availability and price.
That is why I am asking the question, it is probably more to do with physics than chemistry, but given that the range of light transmitted down each
path would be basically an averaged 50% of the output, why would the waveforms be so out of sync?