| Author | Message | ||
     Bruce H. Campbell (campclan) Moderator Username: campclan Post Number: 104 Registered: 4-2001 |
Howard, The times I had materials that gave me detection limits of much less than 0.1 percent were in non-aqueous media and had concentrations both at the detection level and extending to higher concentrations. I was not working at much higher percentages and exptrapolating. However, I would like to have others indicate if they have found detection limits of much less than 0.1 percent such that the "generalized" limits are based on a number of materials and substances. Bruce | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 115 Registered: 9-2001 |
Gavriel - that's true enough, but every value for LOD is specific to the analyte and the "matrix" (as it's called generically) as well as the analytical method used. \o/ /_\ | ||
| Gavriel Levin (levin) Senior Member Username: levin Post Number: 42 Registered: 1-2006 |
Hi, Let us not forget - NIR is "blessed" with the fact that almost all compounds absorb to a higher or lesser degree in almost every where in the spectrum 1200 to 2500 (less so in liquids, but more so in solids) and thus when you have to look at detection limits it beomes very important to look at what you want detected in what - so if it is a water based solution, and organic solute - the very large absorption of water makes it much more difficult to get low LOD of the solute, while if it is moisture in organic solvent, you can go down to 20 to 30 PPM. So the LOD becomes strongly dependent on the over all make-up of the entire matrix rather than just on general definition of LOD. I hope this contributes something. Gabi Levin Brimrose | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 114 Registered: 9-2001 |
Bruce - one of the problems in determining the detection limit for most NIR analyses is that few of them are for analytes at the lower end of the concentration range, where the reference value is known accurately for samples in the range from below the LOD to, say, 30 times the LOD. Most analyses are done for samples at levels well above the LOD. Then, when you calibrate, the total analytical uncertainty is smallest at the mean value of the sample set, and increases as you move away from that. This has several consequences: 1) The uncertainty at zero concentration (or at the LOD, whatever that turns out to be) will be large compared to the SEE or SEP which is measured around the mean concentration. 2) Measurement at zero/LOD concentration requires a large extrapolation from the calibration data. 3) At the very least, the calibration would require a bias correction to measure samples at around zero/LOD concentration, to even find out what the LOD is, and we can expect that the SD of repeat samples (to get an estimate of the effective noise level, to compare readings with) will indeed be larger than the SEE or SEP. There is no reason, in principle, why a calibration developed at some higher concentration couldn't be applied to samples at zero/LOD concentrations, the bias correction made and the SD to use for estimating LOD computed. But nobody ever does that, and for all the reasons given above, it's (at best) risky to claim an LOD based on a calibration done at higher concentrations. And for those same reasons, I find that I have to be skeptical of a claim for an LOD of much less than 0.1%, unless it can be supported by actual measurements according to the above protocol. As for "upper quantitation limit" I don't think I've ever come across the term. Given the increase in uncertainty as the concentration moves away form the mean value, my guess is that it may simply mean the concentration where the amount of uncertainty in the answer is too large to be useful for the application. \o/ /_\ | ||
| Bruce H. Campbell (campclan) Moderator Username: campclan Post Number: 103 Registered: 4-2001 |
Howard, and others, I can't remember where I encountered the NIR detection limit I mentioned. Nor have I found a reference. However, I did test that in many of the determinations I performed and indeed found that many components of interest did have detection limits of about that number, sometimes a little higher. I should add that in these cases the samples were liquids, perhaps giving better/lower detection limits than for solids. Another aspect of this discussion of detection limits is the upper quantitation limit. I don't remember what that definition is, but I suspect it would echo the lower quantitation limit. It would be of interest to discover what others have found for the upper quantitation limit. For example, if the lower quantitation limit is about 0.3 percent (i.e. three times the detection limit of 0.1 percent) is the upper quantitation thereby 100 +/- 0.3? Or is the variation closer to +/- 0.03? Bruce | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 112 Registered: 9-2001 |
Bruce - there are circumstances where MIR has very excellent sensitivity, and correspondingly low detection and quantitation limits (<PPM, and pushing PPB). Maybe NIR could do as well under similar circumstances, but the key here is that nobody uses it for those types of analyses, and nobody has ever explored the ultimate capabilities of NIR, the way Mid-IR has been explored. I have to ask, though, where you saw the 0.01% figure for the detection limit of NIR; the generally accepted figure, for a long time as far as I know, has been 0.1% for routine operation, essentially the same as mid-IR. \o/ /_\ | ||
| Bruce H. Campbell (campclan) Moderator Username: campclan Post Number: 102 Registered: 4-2001 |
Vjunhan, I don't know if you wanted an explanation of detection level or the reason for the difference in detection levels for MIR compared to NIR, so I will give both. First of all, detection level/limit is defined as the intensity of the signal giving rise to the measurement being three times the level of the noise at that point of the spectrum. Quantitation limit is nine time the noise level. There are other considerations, such as the noise is to be gaussian distributed, which is normally the case in spectroscopy. Another consideration is that when using chemometrics, the noise levels and band intensities for the component of interest are the same throughout the spectrum, or at least in the same range. In my experience, the reason for the lower detection/quantitation limits in NIR compared to MIR is the noise level in NIR is usually much smaller. How much smaller? Again, I have seen about a ten-fold difference. This eventually translates into detection limits in NIR of around 0.0100 percent. For MIR this means detection/quantitation limits of about 0.1 percent. There are a few exceptions for some components but in general, I think most would agree with these values. I was told that one company determines water at one or two ppm in their solvent. I have found a 10 to 15 ppm detection limit for an amine-based material of interest. These though, are two exceptions to the general detection limits I have found. Bruce | ||
| yjunhan (yjunhan) New member Username: yjunhan Post Number: 4 Registered: 7-2006 |
Bruce, could you further explain this point to me? Thanks. | ||
| Bruce H. Campbell (campclan) Moderator Username: campclan Post Number: 101 Registered: 4-2001 |
Another diference between MIR and NIR I have not seen posted in this discussion is that of detection/quantitation limits. In my experience, and that of others I have talked to, NIR has a significantly lower/better limit of detection compared to MIR. Bruce | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 111 Registered: 9-2001 |
No, I didn't forget all that. And I tend to agree with Ken. But: 1) That wasn't the topic of my off-line discussion 2) Donkeys have the same advantages as horses. 3) This is an NIR discussion group, after all. While we don't censor, prevent or object to discussions of other analytical techniques, if you want to know about those other methods in depth, or to compare them, this probably isn't the best place to come to. \o/ /_\ | ||
| Kenneth Gallaher (ken_g) Advanced Member Username: ken_g Post Number: 22 Registered: 7-2006 |
Howard forgot donkeys and oxen and scooters... But there is an important point here that can be lost on folks that are too expert.... That is that there are many many analytical techniques. It it critical that whoever decides an analytical approach have a breadth of knowledge. If you give an expert in method XYZ a problem, and XYZ is all they know...they will try to solve it with XYZ, never mind that ABC would do it better. Or even more likely that the problem will be best solved with some combination of methods and unless the experts can talk with each other intelligently the problem might not be solved. Experts are important but so are the big picture folks. | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 110 Registered: 9-2001 |
This discussion kinda reminds me of a discussion I once had (not on-line) about the advantages of a horse versus a car for transportation. Believe it or not, a horse has a good number of advantages: 1) A horse can find it's way home, even if you're lost, or sleeping, or unconscious, or too drunk to drive. 2) Horses can make more horses, without an enormous infrastructure 3) Horses get energy from renewable sources. There were a couple more, that I forget now. But despite all those, and the fact that horses still have their places, most people use cars for most purposes. So it is, too, with mid- and near-IR. They each have their place. For most routine analysis, and especially process analysis, NIR is usually the better choice. \o/ /_\ | ||
| Kenneth Gallaher (ken_g) Advanced Member Username: ken_g Post Number: 21 Registered: 7-2006 |
I will add one more advantage that was implied but not said explicitly. That is for process NIR transmission spectroscopy of liquids is possible. That is a major advantage because of the optical simplicity - and corresponding simplification of calibration and calibration transfer. Transmission MIR is common in the lab - but not in a flow cell....too many problems with sample filtering as well as obtaining a fringe-free background with the cell in place. | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 109 Registered: 9-2001 |
Don and Gavriel covered the more basic and fundamental differences between mid-IR and NIR, but in practice there is one more type of difference that is at least as important to making measurements in the "real world". Gavriel alluded to this difference when he mentioned fibers, but the difference is more widespread and significant than fibers alone; a very important difference is in the materials that are available and suitable for making optical measurements in the two spectral regions. In the NIR we have available materials such as glass, quartz and sapphire; these materials are hard, strong, inexpensive and chemically inert. That makes them eminently suitable for sample cells, windows, and optics in general. They especially can be used for samples containing water, and even aqueous solutions. In the mid-IR, the main types of material that transmit mid-IR radiation are various salts: NaCl, AgBr, Ki, ZnS, etc., in addition to some ternary and quaternary salts with proprietary names (e.g., KRS-5). Virtually all of these are relatively soft, not particularly strong, easily scratched, etc. In addition some of them are toxic, and many of them are attacked by water (e.g., NaCl) and other common solvents, so that they cannot be used for those types of samples. Even if kept dry, any exposed surfaces can be attacked by humidity in the air, so that they become fogged in fairly short order. \o/ /_\ | ||
| Gavriel Levin (levin) Senior Member Username: levin Post Number: 41 Registered: 1-2006 |
Hi, DJ gave an excellent explanation as to the usefulness of NIR vs. MIR. There are two more aspects where NIR has advantage - one is the fibers. The MIR fibers need to be made from specialty materials such as chalcogenide glasses that do not absorb much in the MIR. These are difficult to make, and even with the best of them the practical lengths do not exceed 10 meters or so. They are very fragile, and tend to change their properties with time, and are sensitive to stresses. The second aspect where NIR is easier to use is the ability to use diffuse reflectance from such distances as 5 cm and more. DJ mentioned the very large absorbances in the MIR - what it does it limits the "depth" of sample that can be "looked into". Unless diluted by non absorbing chemicals such as carbon tetrta chloride for liquids and KBr salt for powders, the maximum path length that can be measured will be on the impractical order of 30 to 80 micron, hich for all practical purposes does not represent the bulk of the material, rather a "boundary layer" that could possibly not represent the over all chemistry of the bulk. I hope this lays a solid foundation for those of us that are involved in NIR. Gabi Levin | ||
| Donald J Dahm (djdahm) Member Username: djdahm Post Number: 12 Registered: 2-2007 |
Well, if there is no advantage to NIR for your problem at hand, go ahead and use MIR. Most of us would prefer that the same instrument would generate the data in any wavelength range, and then we could use which ever we wanted. That is becoming more and more possible for NIR and MIR. Traditionally, there are three main advantages for NIR. The sources are simpler. The detectors are simpler. The sampling arrangements are simpler. The main disadvantage for MIR is that the absorptions tend to be so strong that they do not tend to be as useful for quantitative work. Obviously, there are advantages to MIR, or it would not have been developed first. Qualitiative identitfication is certainly one place where IR shines. NIR would not have developed to the point that it has if it had not been for the development of Chemometrics. The band overlap, and are harder to interpret qualitativley. Sometimes when people refer to NIR, they are not refering to the wavelenght range as much as they are referring to spectroscopy that uses these techniques. | ||
| yjunhan (yjunhan) New member Username: yjunhan Post Number: 3 Registered: 7-2006 |
Thanks so much. I am much clearer now. I had another question in mind, hope you can give me the guide once again. Compared to NIR, MIR gives more information, higher signal intensity, and fingeprint of even certain compoutnds. It looks to me that MIR can do everything NIR can do and even more. So why NIR technique becomes so popular, and as for agricultral samples, NIR are preferable? Thanks | ||
| Tony Davies (td) Moderator Username: td Post Number: 159 Registered: 1-2001 |
Hello Yjunhan, Hydrogen bonding is the key to the differences between your sugars. You are measuring samples that are in the crystalline state and you will get both inter- and intra- hydrogen bonding and this is very dependent on the sterochemistry. I do not know about your sugars in detail. I know a bit about sucrose. There is one -OH in sucrose that is held in the crystal structure so that it cannot get close enough to a hydrogen to form a bond; thus it gives rise to a pure -OH absorption at about 1438nm because it will be the same in every molecule. Other -OHs give rise to hydrogen bonds of varying strengths so all of them will be slightly different. In different sugars the sterochemistry will be different so different hydrogen bonding and different absorptions. When you dissolve sugars in water this structure is lost but each sugar has a slightly different effect on the very complex water spectrum so they can still be distinguished by NIR spectroscopy. All this complexity may sound like "bad news" but it really is the opposite. It makes the potential of NIR spectroscopy much larger. Welcome to the community; there is still much to be discovered! Best wishes, Tony | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 108 Registered: 9-2001 |
Yjunhan - Just as in mid-IR spectroscopy and NMR spectroscopy, chemical bonds, even between the same atoms are not all the same. The reason is that when atoms are near each other they are not independent; they interact with each other. The best examples are the highly electronegative atoms, such as oxygen, chlorine and fluorine. These atoms reduce electron density from bonds in their vicinity, thereby weakening them and causing the associated vibrations to have reduced frequencies. This effect is strongest on the adjacent atoms, but can extend further than that. The presence of a nearby double or triple bond can extend the range of interaction still further, especially if there are conjugated double bonds around. Other atoms, which are electropostive, can donate charge to nearby bonds and raise their vibrational frequencies. Several other effects can influence the absorbance frequencies, including aspects of the intra-molecular environment such as steric hindrance (large molecular clumps such as t-butyl, for example), aromaticity, presence of ionic bonds, etc. There are also effects from the inter-molecular environment, the most well-known being hydrogen bonding. So not all bonds are created equal, even when they involve the same atoms. And the spectrum is not the only aspect that is affected; it's also the reason why different molecules undergo different reactions, after all. I think a good course in structure and mechanism in organic chemistry would clear a lot of this up for you. \o/ /_\ | ||
| yjunhan (yjunhan) New member Username: yjunhan Post Number: 2 Registered: 7-2006 |
I am confused, and really need experts to help me through. As far as I know, NIR signal is related to chemical bonds, so glucose, galactose, and mannose should have identical spectrum since exactly the same chemical bonds existing the their moleculars, just position a little different. But the spectra I acquired from pure glucose, galactose, and mannose look different. |