| Author | Message | ||
     venkatarman (venkynir) Senior Member Username: venkynir Post Number: 160 Registered: 3-2004 |
Hi Sathish ; As Sheelagh pointed out a lot publication are there in particle size . We did few experimental studies for PAT project .We are able turn the range than exact size . We have used shift in spectral base data for calibration . | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 517 Registered: 9-2001 |
Sathish - some early work on that was published in: E. Ciurczak, R.P. Torlini, M.P. Demkowitz, "Determination of Particle Size of Pharmaceutical Raw Materials Using Near-Infrared Reflectance Spectroscopy", Spectroscopy 1 (7), 36 (1986). That (and all the other comments made in this thread) said, however, I don't know how well you're going to be able to compare the particle size distribution measurement you get from light scattering, to the average particle-size that an NIR reflectance measurement will give you. \o/ /_\ | ||
| Sheelagh Halsey (shalsey) Senior Member Username: shalsey Post Number: 26 Registered: 12-2006 |
Hi A number of researchers have done work on particle size in pharmaceutical products. I responded in this column back in Sep 2009 with some references that might help. Best regards Sheelagh | ||
| chenchu sathish kumar (sathish) Junior Member Username: sathish Post Number: 7 Registered: 11-2012 |
Dear all, i have been trying to find the way to give a particle size analysis of drug samples by NIR. bt i did not find the way can any one please help me to give any literature or journals in this regard.i want to take the reference values of laser diffraction method Thanks®ards, sathish kumar | ||
| chenchu sathish kumar (sathish) Junior Member Username: sathish Post Number: 6 Registered: 11-2012 |
Dear all, i have been trying to find the way to give a particle size analysis of drug samples by NIR. bt i did not find the way can any one please help me to give any literature or journals in this regard.i want to take the reference values of laser diffraction method Thanks®ards, sathish kumar | ||
| chenchu sathish kumar (sathish) New member Username: sathish Post Number: 5 Registered: 11-2012 |
Dear all, i have been trying to find the way to give a particle size analysis of drug samples by NIR. bt i did not find the way can any one please help me to give any literature or journals in this regard.i want to take the reference values of laser diffraction method Thanks®ards, sathish kumar | ||
| chenchu sathish kumar (sathish) New member Username: sathish Post Number: 4 Registered: 11-2012 |
Dear all, i have been trying to find the way to give a particle size analysis of drug samples by NIR. bt i did not find the way can any one please help me to give any literature or journals in this regard.i want to take the reference values of laser diffraction method | ||
| Francesco Davini (franz) Junior Member Username: franz Post Number: 6 Registered: 2-2009 |
Tony, while thanking again for your and everyone's contribution, I must explain about my signature tag. I am sorry that the joke in it goes lost in translation. "Grappa" is distilled from grapes and is quite alcoholic. It is in my signature since over 10 years and never it has been read by mother tongue readers as a political statement. Best regards, Francesco Davini Energia Nucleare ??? No, Grappa !!! | ||
| Tony Davies (td) Moderator Username: td Post Number: 201 Registered: 1-2001 |
Hello Everyone, You have been busy!! Just to answer a couple of points: Not using PLS (Gabi). My thought was to keep it simple. Scattering is normally a much larger effect than absorption so a PC is likely to catch a large part of it and not the chemistry. Surely the best way of teaching the model to ignore chemistry is to leave it out. Having said that, Don (as usual on these matters) is probaly correct that the best metric would be the Kubelka-Munk function. Those famous photos: I'm pretty sure that Gerry Birth took all the pictures in the chapter in the first edition. Don used them in the second edition so he ought to know. Otherwise we need to ask Phil. If anyone does not know what we are talking about, you really need to see them in either edition of Williams & Norris. The laser photos are also worth thousands of words! Best wishes, Tony PS I'm not keen on nuclear energy but I think we should keep political slogans off this site - please. | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 271 Registered: 9-2001 |
Now that you jogged my memory, though, I think I have a vague memory of someone, who might have been Dave Honigs, made a similar set of pictures later on using CuSO4 as the colored material. Maybe because it would be easier to grind than the green glass that Gerald Birth used. But I don't know where, if ever, Dave published them. \o/ /_\ | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 270 Registered: 9-2001 |
Andrew - They're pretty impressive, aren't they, to make their point? Gerald Birth was the first person (again, as far as I'm aware) to take a set of pictures like that, and those color plates are in the chapter of Karl & Phil's book that is co-authored by Gerald Birth and Harry Hecht, so I think it's safe to assume that those are the same ones. \o/ /_\ | ||
| Andrew McGlone (mcglone) Intermediate Member Username: mcglone Post Number: 16 Registered: 2-2001 |
Howard. I know those colour plates without even looking at them. I eagerly trundle them out if introducing NIR concepts in a presentation or similar. Those pictures, and many of the others in that first chapter, support that old adage that 'a picture paints a thousand words'. Just for interest sake, was it Gerald Birth who actually took them? I've never seen them anywhere else. | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 269 Registered: 9-2001 |
Francesco - While some of the comments made about the effect of light interactions with large particles are, strictly speaking, correct, some of them may have left a wrong impression. It is very possible to find a relationship between the particle size and the transmitance or reflectance of a sample, even for samples with very large particle sizes, due to the interactions between scattering and absorption in the samples. This is shown very nicely in the color plates of both the first and second editions of Williams, P and Norris, K.; "Near Infrared Technology in the Agricultural and Food Industries"; Amer. Assoc of Cereal Chem.; (1987 (1st ed) or 2001 (2nd ed)). While this illustration uses the visible region to make the point, the change in color intensity (reflectivity) with particle size makes a very dramatic illustration that this relationship exists and can be utilized for the purpose. \o/ /_\ | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 268 Registered: 9-2001 |
The earliest paper that I know about, describing particle size measurement by NIR, is Ciurczak, E., Torlini, R.P. and Demkowicz, M.P.; Spectroscopy; 1(7); p.36-39 (1986) \o/ /_\ | ||
| Francesco Davini (franz) New member Username: franz Post Number: 3 Registered: 2-2009 |
Fist of all I wish to thank you all for your contributions! Some explanation is necessary about the kind of samples involved in this project. It is corn grits and sizes are in the range of thousands of microns. Those are analyzed in reflectance. There is little variation in the chemical composition, as the samples come from specifically selected breeds for the same final use. Those chemical parameters are tightly controlled to be in a relatively small range. Though not deep in mathematics and formulas, I know that particles that big do not scatter in the classical meaning of light scattering. I have been working about 15 years ago on a Multi Angle Laser Light Scattering instrument for the analysis of polymers and biopolymers in solution. Almost forgot everything, but for some basic concepts... My thought in this case was that bigger fragments show much more "round" and "even" surface (from the kernel sides), while smaller pieces should have more edges and irregular surfaces. As a consequence they should reflect light in a different way that could be observed on the collected spectra. I guess I will begin to have some spectra of real (unsorted) samples collected, then sieve them and collect the fractions' spectra and see if anything useful comes out. Thank you again Franz Energia Nucleare ??? No, Grappa !!! | ||
| Donald J Dahm (djdahm) Senior Member Username: djdahm Post Number: 30 Registered: 2-2007 |
Replace "uniform" with "broad" in the previous post. | ||
| Donald J Dahm (djdahm) Senior Member Username: djdahm Post Number: 29 Registered: 2-2007 |
Hello, Franz: I would like to add (and perhaps subtract) a bit to what has been said. Of course, I could be wrong. It happens all to frequently. It has often been observed that, in remission spectra, changes in scatter affect a spectrum more than changes in absorption. Indeed, as you have observed, many of our pretreatments are designed to reduce particle size effects. Consequently, as a first response, we often say that measurement of particle size by NIR should be a piece of cake. However, you seem to be interested in particle size distribution, as well as average size. The particle size work with NIR that I am familiar with concentrated on the average size. I would set out to determine these two things (average particle size, and particle size distribution) separately (though you could use the same spectra for analysis). I assume that to obtain the particle size distribution you intend to separate the calibration material into three size groups and then mix them back into various distributions, varying the "% particles below 1750 microns", "% between 1750 and 3500" and "%over 3500". It�s an interesting idea, and I will be interested to see how it works. In the following I will assume that you intend to use remission data. Comments related to average particle size If you are interested in measuring particle size in samples, all of the same chemical composition, the problem is more straightforward than if there are compositional variation. For example, in polymeric materials there is often a change in both molecular weight and particle size, and their effects are confounded. Additionally, there is often a difference in void fraction between samples. For these reasons, multivariate analyses are frequently very useful in the measurement. A few things to keep in mind: Your sieved samples will probably not have as uniform a particle size distribution as your analysis samples. Consequently, we usually expect a bias in the data toward an answer that is �too small�, because small particles are �over-represented� in the spectrum. In measuring particle size, it is the regions of low absorption which contain the most important data. I would avoid all pretreatments which allow one calibration spectrum to affect the baseline of another. Principle component analysis may get rid of the effects of undesirable data portions, but my impulse would be to get rid of the sections of the spectra where there is very large absorption. [The foregoing is said because, you want scatter to dominate the analysis rather than absorption. On the other hand, if you sample only places where there is no absorption, all the intensity comes back to you from an infinitely thick sample, and there is no basis for discrimination based on effective pathlength differences. For samples (of fixed thickness) in diffuse transmission or transflectance, we have a different situation.] Comments related to particle size distribution. As Gabi points out, the effect of particle size on an NIR spectrum is different depending on the relationship between particle size and wavelength. For particle sizes larger that the wavelengths, there is little wavelength dependence of the scatter. However, in my opinion, that will not keep NIR from being an effective tool. (If he�s right, I�ll have to rewrite about 5 pages of my book. There must, of course, be a representative number of particles in the beam.) For particles much smaller than the wavelength, the wavelength dependence is dramatic. You will probably be well advised to use the metric that is most nearly linear with particle size. That would be the Kubelka-Munk function for infinitely thick samples. Your humble (not the usual state) servant, Donald J. Dahm | ||
| Sheelagh Halsey (shalsey) Intermediate Member Username: shalsey Post Number: 20 Registered: 12-2006 |
Hi Franz There have been a number of publications on this subject: Talanta, 2008, 77(2), 647 - Blanco et al Spectroscopy, 1986, 1(7), 36 - Ciurczak et al Analyst, 1998, 123(10), 2043 - Frake et al Analytical Chemistry, 2003, 75(8), 1777 - Higgins et al American Pharmaceutical Review, 2008, 11(6), 62 - Long Analyst, 1999, 124, 33 - O'Neil et al Analyst, 1998, 123(11), 2297 - O'Neil et al Some are looking at median particle sizes and some at distributions. Hopefully these should give you some idea of what can be done. Regards Sheelagh | ||
| Gabi Levin (gabiruth) Intermediate Member Username: gabiruth Post Number: 18 Registered: 5-2009 |
Hi Tony, I am afraid there is a basic problem here - Franz is talking baout particles which are not really particles - if the sizes are correct, i.e., 1750 to 3500 microns we talk about 1.75 to 3.5 millimeter - and I don't think that the scattering laws apply to "particles" which are 1000 times larger than the wavelength used for the measurement. Unless of course if he he means nanometers, which will place the particles in the 1.75 to 3.5 micrometer and these laws will apply. if he is talking about 1.7 to 3.5 micron, then it will be OK to use NIR. This brings me to the second question - why not PLS - you mention that PLS will bring into the picture chemical issues - and actually this may be desirable. If Franz expects absolutley constant and invariant chemistry you are right, PCA should do the job. However, if the substances involved are expected to show some chemistry variability I think he will benefit from using PLS and also to include in the calibration set samples that represent as much of the chemical variability he can lay his hands on - so theat the model will learn how to "ignore" them and predict correctly when such variability occurs in the real process. Thanks a lot, Gabi Levin Brimrose | ||
| Tony Davies (td) Moderator Username: td Post Number: 200 Registered: 1-2001 |
Hello Franz, Welcome to the group! I am sure this is feasible. There are methods for the determination of "hardness" of wheat by NIR. The wheat is given a standard grind and the NIR spectrum obtained. The calibration is essential using the particle size of the wheat. I expect you will find that it will work using the untreated spectra. There are a number of alternatives you could investigate, e.g. Treat the spectrum with a correction pre-treatment (SNV would be my choice), subtract the treated spectrum from the original spectrum and use the difference spectra for your calibration. A more standard chemometric method would be to do Principal Component Analysis (PCA) on you calibration set of spectra, then discover which PC is strongly related to the particle size (expect it to be PC1) and use this PC for your calibration. PLS would probably work almost as well but it will tend to include some of the variation due to chemistry in any PLS factor so PCA is prefered. Good luck! Best wishes, Tony | ||
| Francesco Davini (franz) New member Username: franz Post Number: 2 Registered: 2-2009 |
Hello, I am new to this list. I have some experience at the use of NIR equipment, but little knowledge of chemometrics and spectra pretreatment. Please excuse me for any incorrect terminology. I know that NIR spectra are quite sensitive to the particle size of the sample, due to light scattering effects. This is normally corrected for by means of treatments and choice of calibration models in order to minimize the effect of scattering distortion. I found a lot of literature about this subject. Now, my problem is the opposite! How could the spectra be treated in order to enhance the effect and possibly gather some particle size information? Having reference values is quite easy and only requires a sieve or set of sieves. What I would like to achieve is a calibration for values like "% particles below 1750 microns", "% between 1750 and 3500" and "%over 3500". Anyone ever investigated such an application? Could it be feasible ? Thanks a lot for your suggestions! Franz Energia Nucleare ??? No, Grappa !!! |