| Author | Message | ||
     Michael C Mound (mike) Senior Member Username: mike Post Number: 39 Registered: 7-2007 |
More on LOD... In some recent discussions with EPA-types as well as some of my University colleagues, I have become intrigued with a combination of technologies with NIRS as a centerpiece for characterization of trace metals. One of my pals who specializes in stable isotopes has been testing alcohol and esters in distillation residu�ls. Although he uses Mass Spec, he has had some historical success in tracking trash materials at low levels in rum and vodka derivatives with a combination of techniques, including IR. At this point, though, I don't know what LOD's are possible, although I suspect the MS does the best job. I was thinking that the high attenuations of Fe and Ti, which might be among the baddies in some of my applications, could be used as associative arrays (correlation without causation?)in an indirect model. Any thoughts? Thanks, Mike | ||
| Michael C Mound (mike) Senior Member Username: mike Post Number: 27 Registered: 7-2007 |
Howard, ...A good friend of mine was a (very) distant cousin of Thomas Edison. He told me once (my friend, not Edison, neither he nor I are that old) that almost every day, Edison regularly had dozens of ideas and suggestions for inventions, almost all of them terrible and impractical. The good ones, well, they were pretty good...and, the famous anecdote on this subject was that the great man was asked how many failures he had when he developed the incandescent light bulb. His answer was that he had discovered one right way to do it and 999 ways not to. I don't think I'll ever come up with a light bulb, even a better PDA, but I am not in the least discouraged...I know there is a "light" (couldn't resist) somewhere down that path/road/autobahn... Thanks, again for the advice and good words... And, thanks to all the rest of you who chose to weigh in and participate...all your inputs were valuable. I'll be back with more on this when I have some granularity. Mike | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 151 Registered: 9-2001 |
In other words, you were looking for magic!! Don't feel bad: we're all looking for the magic answers to all our problems. For some people, NIR is it. You're not fortunate enough to be one of them. But hard problems are called "hard" because there are no easy answers. So, yes, I'm not surprised that the road to development of a suitable method is going to be long and difficult. But at least you'll be starting out on the right road to the answer. Imagine how hard it would be if you started off on the wrong road. And in the end, it might not be as bad as it seems now. Oftentimes, once you've gotten past the roadblocks and know how to deal with them, the potholes don't seem like such large craters as they do when viewing them from the beginning of the road (to stretch the analogy a bit, perhaps). \o/ /_\ | ||
| Michael C Mound (mike) Senior Member Username: mike Post Number: 26 Registered: 7-2007 |
Howard, Thanks for the trouble. I had actually scoured the very long and involved EPA listings previously, and knew about the CVAAS and most of the other gas-type CEM's, etc. I was hoping to find some setup that would avoid digestion of analytes, also, but this may not be possible. This reference does show a simplification of sorts, so I guess it is a couple of steps in the direction I was hoping for. Incidentally, on the same EPA site, there is a pretty good February 2007 Armstrong Project Report (Lehigh University, where my cousin is chair of the History Department). It's good background stuff and does a nice job of their experiments and experiences with sampling trains and shows various alternatives for a variety of different instruments and alternatives. Lots of good charts and comparisons. Educational, though a bit of a major time investment to muddle through or even scan...all 313 pages. Fortunately, lots of pictures (my wife tells me that is the only kind of reading I do...along with LARGE PRINT). Scurrilous comments... Thanks, again, Mike | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 150 Registered: 9-2001 |
I dunno, Mike. I haven't yet found how low I can go! But I wasn't recommending a general web search. I think you could go directly to the EPA site (http://www.epa.gov) and look for their standard methods for that analysis. That would tell you exactly which techniques are usable for Hg at such low levels. Concentrate on the science and ignore the politics. Now that I've got your e-address, I can send you the article. It'll take me a little time to get around to scanning it, though. \o/ /_\ | ||
| Michael C Mound (mike) Advanced Member Username: mike Post Number: 25 Registered: 7-2007 |
Howard--- My e-mail address is [email protected] Actually, even a cursory surf on the Web will yield an exhausting (exhaustive) number of references to Hg in coal. Very excited people and equally excitable politics involved. I would say this was an incendiary subject, but I shrink from such low humor (right!). Thanks, again, Mike | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 149 Registered: 9-2001 |
Mike - try http://www.americanlaboratory.com If that doesn't work, then in a pinch, if you send me your e-mail address I can scan the article and send that to you. I don't think I could post it on the discussion group due to copyright issues; I'm sure that would be more than "fair use" would allow. But I think your best bet would be to go directly to the source: look up methods for Hg analysis on the EPA web site. \o/ /_\ | ||
| Michael C Mound (mike) Advanced Member Username: mike Post Number: 24 Registered: 7-2007 |
Howard--- Can you upload or provide a link to the "something you just came across", please? Thanks, Mike | ||
| Michael C Mound (mike) Advanced Member Username: mike Post Number: 23 Registered: 7-2007 |
Thanks. I'll keep you posted... Mike | ||
| Gavriel Levin (levin) Senior Member Username: levin Post Number: 57 Registered: 1-2006 |
Hi, Although not my exact cup of tea, in the past I had to deal with determination of low levels of metallic ions in water. One of the approaches we were looking into and applied reasonably well was to dip in a large volume (this was in rivers, so the dipping was in a side pond) a resin with very high affinity towards the ion in question) and let it tie up the ion in question from a very large volume of the water. Than you can analyze the amount of the ion which is now concentrated in a small volume so in practice the analysis is easier. Then you calculate the concnetration in the large volume from knowledge of the samples volume. Of course, the dissociation constant of the complex that captures the ions needs to be small enough so that the residual concentration after "soaking" will be about 2 orders of magnitude lower than the value we wish to measure. It is not a very accurate method, but if nothing else works, or the alternative is very expensive in equipment cost and implementation, it is still worth a shot. If the mercury in the water is not ionic in nature, then of course the whole thing is different. Gabi Levin | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 148 Registered: 9-2001 |
Mike - here's something I just came across that you would probably be interested in. In the latest issue (Aug. '07) of Amercian Laboratory News edition, there's an article on the approval by EPA of a method of determining Hg in water, that has a LOD of 12 part per trillion. The meothod is based on atomic fluorescence. This new method is replacing a previous method that has an LOD of "only" 0.2 parts per billion. Of course, the EPA is interested in water analysis, but this shows that ultimate sensitivities of the order you need are in existence. It is now your task, should you choose to accept it, to make an analytical system that in fact has the necessary sensitivity work for your non-water samples. Ome approach that I can think of is to ash the sample (which shouldn't be hard, with coal!) and somehow get the Hg into solution; then measure the concentration in the solution. From what I remember of these trace and sub-trace analytical methods, they have two main problems that need to be overcome: 1) to avoid losing the tiny (sub-microscopic, probably, at those levels) amounts of analyte that you have to deal with. This can be trickier than you might think, since van der Waals forces or other normally-weak effects can cause the analyte to stick to the walls of a container, or anything else that they might come into contact with. 2) to avoid contamination, especially if there's any high-concentration (i.e., parts-per-million or higher) samples around. What I would recommend at this point, is that you contact companies that make that type of equipment and ask their advice on how to deal with your samples. They should have applications departments, with scientists who could help you, I think. Good luck and have fun! \o/ /_\ | ||
| Michael C Mound (mike) Advanced Member Username: mike Post Number: 22 Registered: 7-2007 |
Howard, Of course, you are correct in all your comments...and, I guess, I have to concede to the concensus as expressed. So, unless I find a way of "making stones fly upward" (as they say here in Switzerland when the improbable is unlikely to become the probable), I will seek other technologies for my windmill-tilting. Thanks for everyone's gracious and generous good inputs. I'll let those interested know how things turn out (if and when). Mike | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 146 Registered: 9-2001 |
Mike - nobody is downplaying or denigrating the importance of what you are trying to do, so you don't need to defend it. However, the importance of the application has no bearing on the ability of NIR (or any other analytical technique, for that matter) to do it. You're certainly welcome to make the effort to try to generate a successful application for Hg based on NIR, and we would all be interested in knowing the results. But coming to this forum means that you're asking for, and getting, the best advice available on the subject, and that advice is that you're better off spending your time, effort and resources to try something else. No matter how important the analysis, how attractive it is to use NIR because the final use of the method would be easy, or how hard you're willing to try to make it work, the concensus is that the technology simply isn't suitable for what you want to do. Therefore, our recommendation is that you should start with a technology that is known a priori to be suitable for measurement of inorganic materials at low levels, and there are several of those: various forms of atomic spectroscopy, electrochemical methods, automated wet chemistry, possibly even chromatographic methods. Limitations of those methods in terms of sample handling can be overcome with various types of automated accessories, if necessary; there are devices available to automatically perform most, if not all, of the various types of sample preparation steps that might be needed. Yes, it's true that NIR has shown the capability of sidestepping many of the auxiliary manipulations of the sample that "plague" other analytical methods. But you shouldn't put the cart before the horse by focussing on the secondary aspects of the method. In order to make an analysis useful, the primary characteristic to look for is that the methodology first has to WORK, and the concensus is that for your problem, NIR isn't it. \o/ /_\ | ||
| Michael C Mound (mike) Advanced Member Username: mike Post Number: 21 Registered: 7-2007 |
Dear Friends, In my world, what is useful is used (even if not hard science). What was behind my interest in the Hg LOD is the growing concern over identifying and controlling pyritic sources of Hg (inorganic) in coals (organic)that are used in coal-fired power plants. "Inorganic constituents in coal have a significant effect on almost every aspect of coal utilization as well as its impact on the environment. Mercury is one of the elements of special environmental concern. Being highly volatile (it vaporizes at temperature as low as 150�C), mercury poses a special problem for electric utilities. The Clear Air Act of 1990 authorized the U.S. Environmental Protection Agency (EPA) to regulate mercury emissions from electric utilities; on February 24 2004, the EPA proposed a rule supplementing its December 15, 2003 proposal to permanently cap and reduce mercury emissions from power plants. According to this rule, mercury emissions will be reduced by 70 percent when fully implemented. In 2018, the second phase of the mercury program sets a cap of 15 tons. Mercury emissions from power plants are considered the largest anthropogenic source of mercury released to the atmosphere; about 48 tons are emitted annually in the U.S.A. as a result of fossil fuel combustion, mostly from coal-fired power plants. Although the elemental mercury emitted to the atmosphere from coal-fired power plants is not considered harmful, it can chemically transform into a toxic form, methylmercury, that can become concentrated in fish and birds, and from there enter the human body." (quoted from the USGS reports). The latest EPA panic in the USA is Mercury emissions. If a power plant burns 10,000,000 tons/year of coal with 5 ppb of Hg, then the plant emits 50 kgs/year of Hg metal into the atmosphere. The average Hg content of coal is about 5 ppb. So, you see, the theoretical and the practical tend to intersect, and it would be interesting, to say the least, to provide a means, however rough, to help the characterization along. In this instance, how this situation and its solution(s) are resolved are seen by the power plants under the gun as if through the diaphane of the visible, to put it poetically. By that, I mean that how it actually provides some quick answers is less important (the users cannot or are disinclined) to pierce the diaphanous screen of the science, than to be relieved that there is a predictable way to know what broadband threshhold limits can be monitored. What will be the tripwire will be the reproducibility and reliability of a real-time report, rather than a true LOD definition, as has been so elegantly stated. As time goes on, I predict that we will see more anxiety publicly pronounced about this concern. So, why not try for the rough estimate as a first objective? Thanks, Mike | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 145 Registered: 9-2001 |
Bob - David's comment contains part of your answer. A more complete answer would be something like this: There are several contributions to the total error, these include both instrument-induced and sample-induced errors. Instrument errors are more-or-less well-known. I've come to define a term I call "sample error", or "sample noise", which is a sort of catchall term for all the various errors caused by sample-related effects. These include errors induced by models having incomplete or incorrect compensation for interferences in the sample, the effects we variously classify as "repack effect" or "particle size effect" (for particulate solids only, of course), etc. When these effects are random they add as the sum of their squares (or "in quadrature" as engineers call it). These effects do not include systematic effects nor do they include reference lab error, which also adds to the total analytical error, again in quadrature. With proper experimental design it's possible to estimate the magnitude of the various effects. In fact I did that once, for wheat calibrations, and published it: Anal. Chem.; 86, p.2814 (1986). It's not easy, you have to take multiple readings of each sample according to a defined scheme, in order to be able to sort it all out afterward. Generally, what you'll learn is this: the lab error is usually the largest contributor to the total error term (no surprise there). But since the errors add as their squares, it very quickly becomes difficult to determine the effect of the the contribution of the smaller error terms to the total error. It also becomes well-nigh impossible to make any overall difference to the calibration performance by changing any of the lesser effects, no matter how easy it is to measure them unlimited number of times. F'rinstance: suppose the lab error is 0.1, and repack error is 0.02. The combined error is 0.10198. If you decrease the repack error to 0.01 (only half as large as before) the total combined error is 0.100498. Do you know how many samples you'd need to tell that difference of 0.0015 out of a total error of 0.1? LOTS. To reduce the total error to equal that of the lab error, you would have to measure, and average, an "infinite" number of times. It's a losing proposition. The bottom line is that the limiting sensitivity to the changes is the limiting error source, which is the largest error source, and which is almost always the reference lab error. So sure, if your lab error is, let's say 0.1, then since the SD decreases as the square root of the number of readings you average together, then if you average 100 readings you can get one more decimal place. And you can get 2 more decimal places if you measure and average together 10,000 readings. And you could eliminate it entirely be measuring (again) "infinite" number of times - or maybe until you run out of sample. Karl's averages of 64 is close enough to 100 for the difference not to matter a whole lot, although he actually improved his sensitivity by a factor of 8, not ten. But then, you see, he did actually average out the largest error source. It wouldn't have helped much if he was trying to reduce, say, detector noise (by coadding 100 spectra, say). \o/ /_\ | ||
| Bob Rosenthal (rosenthal) New member Username: rosenthal Post Number: 3 Registered: 1-2006 |
Category: General, All others: Limits of Detection Howard, In reading your two comments, it appears that my question was not asked properly. I did not intend for my question to be how to measure samples that are very close to zero concentration of the analyte. Before rephrasing my question, let me provide some background. In the early 1970s Phil Williams provided a set of wheat calibration samples to Karl Norris. If my memory is correct, each of those samples had their protein determinations made by 64 separate Kjeldall determinations. The �true protein� was expressed in two decimal places (e.g., 12.62 percent). However, when Karl performed his NIR analysis (and when Neotec and Technicon performed a similar analysis on those same samples) the NIR determined protein was limited to one decimal point; i.e., in this example 12.6. This was done because the NIR accuracy and precision were limited to a sensitivity of approximately 0.1%. So let me rephrase the question that I previously posed. Was this sensitivity limit of 0.1% due to the inherent electrical noise in the measurement instrument, or even if the NIR instrument had zero noise, you couldn�t resolve closer to the nearest tenth because of limitations in the physical/chemical absorption of the sample. And as I posed it before, if somehow an instrument is built that�s a hundred times more sensitive than any NIR instrument had ever been built, would it be able to resolve 0.01% (or even 0.001%) (i.e.; to the comparable accuracy of the comparison laboratory)? Bob | ||
| David W. Hopkins (dhopkins) Senior Member Username: dhopkins Post Number: 119 Registered: 10-2002 |
Hi Bob, I think the answer to your question is, it depends. It depends on the material you want to measure having a unique spectral band or spectral bands in the region you want to measure. I think that if you can build a more sensitive instrument, you may be able to find that the material you want to measure is in fact different from the other materials. Take one of your favorite organics for example, glucose. It may be that a more sensitive instrument could distinguish glucose from other simple sugars, but present instruments struggle to measure the glucose at the levels necessary to detect the low absorbances of biologically important concentrations. The other problems are the other interfering materials and conditions. Good question. I think the instrument is important, as well as the analyte. Best regards, Dave | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 144 Registered: 9-2001 |
Let me add some more comments: In NIR, we've never been concerned with LOD for two reasons: 1) The vast majority of analyses done using NIR are "very far" away from zero concentation of the analyte, in the sense that the range of interest is usually a fraction of the lowest concentation within the range of interest. Not only does this make the behavior of the analytical method at zero concentration void of interest, it also makes the determination of the value of LOD very uncertain, since the values that are measured create a large uncertainty in the predicted value at zero, due to the large extrapolation needed. 2) Not entirely independently of the first reason, in many situations it is not possible to create samples, even artificial samples, not containing analyte. And as you well know, if we could create such samples, their behavior towrd NIR analysis would not be representative of the behavior of real samples, so that you could not rely on any values calculated using them. 3) Besides the likelihood of exaggerated amounts of variability due to the extrapolation when using "normal" sample sets, there is, additionally, always the possibility of non-linearity affecting values computed from a model based on normal samples, and predicting samples with excessively low concentrations of analyte. \o/ /_\ | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 143 Registered: 9-2001 |
Bob - historically the question of limit of detection was not intended to be diagnostic. It was purely a matter of being able to distinguish a sample containing the analyte with a similar one containing no analyte. To put it into more formal terms, the question to be asnwered is: does the sample at hand contain enough of the analyte to be significantly (in the formally defined statistical sense) different from zero? The lowest value for which the answer is "Yes" is the limit of detection. In practice, it is taken as three times the standard deviation of multiple readings on a sample (or samples) all containing no analyte. This is the classic definition, and the one used by many certification organizations for analytical methods, including USP, for example. If you want to know what happens when you change the conditions, that's a question beyond the definition, or the measurement, of LOD, outside it's scope. Diganosing the situation, then, would be done by making the change and seeing what happens when you actually perform the analysis under the new conditions. \o/ /_\ | ||
| Bob Rosenthal (rosenthal) New member Username: rosenthal Post Number: 2 Registered: 1-2006 |
Catetegory: General, All others: Limits of Detection Hi, I followed the interesting discussion on �Limits of Detection.� Based on that discussion, I do have a question. Let me assume the following: The interest is in measuring an organic constituent. Is the limit of detection a function of instrumentation limitations, or is it a function of the physical/chemical characteristics of a product? Saying this differently, assume I�m able to build an instrument that has 100 times more sensitivity than the best NIR instrument that was ever built, would the expected LOD be improved from the �traditional NIR limit� (0.1 percent) to perhaps either 0.01 percent or even 0.001 percent? Bob | ||
| Michael C Mound (mike) Intermediate Member Username: mike Post Number: 18 Registered: 7-2007 |
Gents, I like your style. Reminiscent of the colloquia I recall from University days, where we mixed it up, so to speak...most stimulating. It is true, as you have surmised, that there has been, was, and still is, sufficient actual interest by the inorganic (stepchildren?) advocates populating this side of the street as to the possibilities of NIRS. This has matured to the point where many users are motivated to develop practical applications including investing time and sweat in this, presently, perhaps, occult work. Whether these efforts should mature into morphs for prospective University or Research Institution long-term projects is a moot point, because regardless of the absence of documenting or providing a solid and irrefutable scientific footing for a proof of concept, so far, ineluctable evidence exists that the darned technology seems to satisfy the need for acceptably good results (at least, for trending quanta). It's not pharma, but it still makes sense to some. I must compliment you once again on your flexibility in thinking and reasoning, as well for sharing your friendly and informed opinions, all of which are both refreshing and encouraging. I'll revert with more interesting and hard information as and when I can do so. Keep tuned. Thanx again, Mike | ||
| Bruce H. Campbell (campclan) Moderator Username: campclan Post Number: 106 Registered: 4-2001 |
I think the reason most of us don't examine non-organics with NIR but instead concentrate on organics has to do with alternative methods. Inorganics are often analyzed with techniques that have exremely low detection limits and are designed to be specific for one or only a few elements, such as atomic absorption spectroscopy. The techniques are well understood. For organics, much of the progress has been in determining proximate values, such as sweetness, degree of ripeness, octane number, etc. These measurements have been rather inexact, possibly crude, take an inordinate amount of time, are not precise and require the user to be highly trained. Biomedics is another area NIR is making good progress, mostly, in my mind, because the present methods are not as useful/exact as wished for. In conclusion, I think NIR is replacing older technigues that are expensive and relatively inexact. So economics is driving much of the development and the area(s)in which NIR is being applied. | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 139 Registered: 9-2001 |
This is an interesting discussion. There's certainly a fair amount of truth to the usual caveats about measuring trace amounts of samples, and the weak absorbances that most materials exhibit, that can be used as justifications for the statements. But I think the biggest single reason that we discount the possibility of measuring trace amounts of anything, especially inorganics, is the history and tradition of NIR as we know and use it. As NIR has been used, it is oriented to measurement of particular analytes on a routine basis, such as for monitoring a proces, and it is very valuable in that application. But because of that, we have become somewhat insular, and somewhat blind to other paradigms for using our favorite spectral region. We forget that "research" can mean something beyond measuring a new analyte not previously reported, or using a new whiz-bang algorithm to calibrate. In fact, I've heard well-respected members of the community disparage work where a scientist did some non-traditional types of measurements in order to elucidate underlying physical or chemical phenonema, dismissing it with a phrase like "why are they bothering with that type of stuff?" because it didn't directly lead to a common calibration model. The military has used NIR for over 50 years to detect exquisitely low "concentrations" of "samples", and in real time: one airplane per million cubic miles of atmosphere, for example. It depends on widening your viewpoint and accepting non-traditional ways of defining and performing "analysis". But even within our standard paradigms, some work on inorganic measurements are known. Ron Rubinovitz has presented talks where he discussed cases where inorganics are measurable. These include, for example, cases where the "inorganic material" includes an organic component (an example would be calcium oxalate, as an example of a general case of an metal salt of an organic acid). Other cases include measurements where the purely inorganic analyte (e.g., sodium chloride) is measurable by its effect on the spectrum of an associated material; an example would be NaCl in water, where the water spectrum is disturbed by the dissolved salt. True enough, these are very specialized situations, and fraught with potential for interferences (and in the case of water, temperature) to affect the spectrum and cause errors. For this reason, such analyses are not usually recommended for routine analytical purposes - so we say "you can't measure the inorganics." But this is our insularity of viewpoint speaking, not the facts of the situation. If someone is able and willing to do the work and make the measurements needed to determine the nature and extent of interference, and the limits of applicability of the method, an "impossible" analysis may well be doable. But this is tantamount to a long-term university research project, not a routine industrial application of NIR, as we are used to seeing it. There seem to be few takers for those types of calibrations. Or research projects. \o/ /_\ | ||
| Michael C Mound (mike) Intermediate Member Username: mike Post Number: 17 Registered: 7-2007 |
For Gabi and Tony, Thanks for being open-minded. To address the comment/question about how geologists got crowded into the picture, I should say that there is ample evidence that both qualitative and quantitative measurements are commonly used in minerals processing for various ore materials. The impetus for this accelerated interest more or less came about as a result of NIRS, among other techniques, being employed in the arcana of space probes (to start with)...where it was considered as being somewhat important to have a modicum of reliable incoming information. These days, there are various flavors (sic!) of minerals-focused analyzers that use NIRS for polymorphic and lattice module characterizations for ore minerals. Quite a bit of impressive investigative effort has gone into this area by a number of specialists, hence my reason to wonder about the negative initial response to using a clearly inorganic substance as a reasonable target analyte. It is true, though, to take this further, when a mineral such as gypsum (CaSO4*2H2O) is measured, the crystalline lattice modules are looked at and the Ca, SO4, and H2O are all "separately" determined to fix on the (theoretically) separated various vibrational characteristics of the "uncombined" molecular translations and rotations (not a scientific explanation, but it works, it would seem). In the gypsum example mentioned, both the hydrated form as well as the anhydrite polymorphs are evaluated because this is important for economic and processing reasons. Thus, thanks again for the opinions and your open-mindedness. Mike | ||
| Gavriel Levin (levin) Senior Member Username: levin Post Number: 56 Registered: 1-2006 |
Hi, The reason we ususally refer to in-organics as non doable by NIR is that in most of the normal applications where they need to be measured they do not provide the signal needed - Of course, in situations where there are crystal water the measurement is possible due to the very specific wavelengths at which the bound O-H absorbs at - usually around 1380 to 1410 nm, but there are problems asscoiated with even this approach - it is all fine and nice if you are absolutley sure that all your mineral is hydrated to the same formula e.e CuSO(4)x6(H2O) - if there is a situation where there could be a mixture of hydrated and non hydrated, or a mixture of hydration complexing configurations then the reference values will start to be ambigous - i.e., you can have a certain value of sulfate (fully hydrated)determined by some reference method associated with a given spectrum, and then if same total sulfate is composed of some hydrated and some not, the spectrum will be different, but the reference value is the same, so how would that work out in quantitative analysis? I admit that I am not familiar with the usage done by geologists, is it just qualitative, ID type, or is it quantitative as well? Regarding Tony's question about the gaseous hydrides such as arsenic hydide - the only reference to that I might make is that when we tried to characterize the gas phase ammonia (NH3) we observed much sharper and clear peaks than when you look at aqueous ammonia. I would imagine that similar hydrides in the gas phase will show similar pattern and be measurable, the only issue will be path length that needs to be sufficiently long. Gabi | ||
| Tony Davies (td) Moderator Username: td Post Number: 168 Registered: 1-2001 |
I think Mike is correct to challenge our conventional thinking. What we are really measuring is the presence of covalently bound hydrogen atoms. These are most commonly found in organic compounds but it is the hydrogen not the carbon that is required for good NIR spectra. (Water is not normally considered as organic!). When it comes to the NIR spectroscopy of rocks then at least three factors come into operation: 1) Some minerals have electronic vibration that absorb NIR. 2) Many minerals contain water of crystallisation. The hydrogen bonding of water is very sensitive to the local environment and give rise to different hydrogen bonding and hence different NIR spectra which may then be characteristic for different minerals. 3) When NIR spectra are measured in the absence of covalently bound hydrogen then we will detect the much weaker absorbers that are normally lost in background noise. A fourth possibility has just occurred to me that some rocks might be characterised but what is growing on them (such as lichen)! In atomic spectroscopy arsenic is released from samples and measured as the hydride. I wonder if anyone has tried to measure the NIR spectra of gaseous hydrides? Tony | ||
| Michael C Mound (mike) Intermediate Member Username: mike Post Number: 16 Registered: 7-2007 |
Dear friends, I appreciate all of your remarks and comments. Thanks to Howard and Gavi. I can accept the fact that the LOD will not meet expectations for trace elements/molecules, which was the point of my inquiry. However, I am somewhat bemused and naturally curious as to why there is a problem with the analyte beiing inorganic. The literaure is chock-full of applications for various minerals (all inorganic), and inorganic solids would seem to have been documented as successfully having been characterized by NIRS. The USGS (United States Geological Survey), for example, has an impressive library of mineral spectra, as do several other institutions. Maybe my thinking is as foggy as my glasses get sometimes... So, in my country (to paraphrase), even the worst vibrator is no worse than a bad vibration. At a modality of any sort, it would seem to have some minimal value as an investigative technique. Many thanks for your thoughts...I guess that is why this is a forum in the true sense of the term. Mike | ||
| Gavriel Levin (levin) Senior Member Username: levin Post Number: 54 Registered: 1-2006 |
Hi guys, I thought we have been through inorganics before. Mercury - the great evil, we would be all rich if we could detect mercury to 5 ppb, or even 10, or even 1000, wouldn't we? (I mean us the equipment guys like me) selling them like crazy. The best NIR can ever do in best situation is for example water dissolved in orgnaic solvent such as dichlorobenzene that has practically zero absorption in the 1940 peak area - and then it can do about 5 to 10 ppm, which is 1000 times more than 5 ppb. There is an old saying in my country that even the best looking lady can only look as good as a good looking one. Gabi | ||
| Kenneth Gallaher (ken_g) Senior Member Username: ken_g Post Number: 28 Registered: 7-2006 |
Never, it's inorganic and its very very trace. There are exceptions both up and down but for most things NIR LOD is 0.1% | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 135 Registered: 9-2001 |
Mike - I thought we'd been over this ground recently. NIR is not recommended as a trace technique. Nor is it recommended for measuring inorganic materials. Bottom line: NOT RECOMMENDED. \o/ /_\ | ||
| Michael C Mound (mike) Member Username: mike Post Number: 14 Registered: 7-2007 |
Recently, there has been some interest in detection of Hg in Soils, etc for environmental and processing reasons. LOD of 5 ppb have been thrown around. Some XRF wags have claimed the ability to measure down to this level, but my experience is that around 10 ppb is a good trick, and probably no better than that. Under what circumstances can NIRS reach this LOD in such a matrix? Thanks, folks! Mike |