Effect of sample darkening along the ... Log Out | Topics | Search
Moderators | Register | Edit Profile

NIR Discussion Forum » Bruce Campbell's List » I need help » Effect of sample darkening along the NIR range « Previous Next »

Author Message
Top of pagePrevious messageNext messageBottom of page Link to this message

Cesar Guerrero (cesar)
Member
Username: cesar

Post Number: 11
Registered: 3-2006
Posted on Tuesday, October 17, 2006 - 5:24 am:   

Thank you David!
Yes, generally I used 1st-derivative coupled with other pre-processing methods for scattering correction, as best options.
The heat causes drastic changes in quantity and quality of soil organic matter. In fact, we used the decrease in some peaks to estimate the temperature of the heating based on the changes of thermo-labile substances (organic and maybe inorganic too) of soils. I was working with only five type of soils, and similar results were obtained, but now, I'm increasing the variability of soil types, mainly different in mineralogical composition.
I agree with your idea that other changes (such the particle size and changes in some minerals) have an additional effect to the presence of carbonaceous compounds for explaining the increase in the baseline.

Thank you very much for the suggestions (always welcomed!)

Best regards.
Top of pagePrevious messageNext messageBottom of page Link to this message

David W. Hopkins (dhopkins)
Senior Member
Username: dhopkins

Post Number: 94
Registered: 10-2002
Posted on Monday, October 16, 2006 - 11:33 am:   

Cesar,

This is an interesting application!

I think your 1Der scans show that this is a complicated situation, and you are not just looking at sample darkening. It appears that the presence of the charred organic material is causing baseline offsets and a decrease in the pathlength of light within the samples. It seems to me that the peak heights of some bands are depressed by the charring, and this may be a very good predictor of the temperature. To minimize the affect of the moisture band, which I agree with Howard could be a big problem, you may want to exclude the moisture region that is seen at 1930 nm, but moisture affects the slope to higher wavelengths, too. If you use 1st or 2nd derivatives, the moisture effects can be more effectively minimized. Did you use the 1Der for your calibrations?

To address your question, the baseline shifts you observe in the original spectra could indeed be caused by the black carbon, but may also be influenced by other effects that look like particle size effects on scattering. For example, any mica particles in the soil could be popped by the heat (just like vermiculite) and be contributing to the scattering. You may want to investigate whether the calibrations depend upon soil type and composition.

I agree with you that the conversion to wavelength may be influencing the accuracy of the calibrations. After all, the wavelength values do not enter into the calibrations. I would recommend staying in cm-1, and you could smooth using a range of points to see what gives you the best results. Derivatives also involve smoothing, and you need to optimize the method too. The derivatives in cm-1 or nm will look similar, and will be affected by the spacing of the points. So there is no real advantage to converting to nm for derivatives either. This is an interesting topic that we could probably discuss at length, but I don't want to get into that just now.

Best wishes,
Dave
Top of pagePrevious messageNext messageBottom of page Link to this message

Cesar Guerrero (cesar)
Junior Member
Username: cesar

Post Number: 10
Registered: 3-2006
Posted on Monday, October 16, 2006 - 11:30 am:   

Hello Howard!
Yes, it is a interesting suggestion. Thank you!
I have been developing calibrations trying to avoid the use water bands, and also it works.
Anyway, always I dry the samples (close to 80�C for two days) to avoid this effect.
I've been rewetting and drying the samples several times with the aim to see if the effect of temperature on the NIR spectra remain related with the temperature of heating (obviuosly in those ranges non-related with water).

Thank you very much for your suggestions!
Best regards.
Top of pagePrevious messageNext messageBottom of page Link to this message

Howard Mark (hlmark)
Senior Member
Username: hlmark

Post Number: 52
Registered: 9-2001
Posted on Monday, October 16, 2006 - 10:21 am:   

Cesar - I think there's a point you'll have to be wary of, when applying your calibration results to actual field samples: the fact that on standing, the field samples are likely to regain moisture after being dried by the forest fire, which may affect the measurement of the soils - especially if there should be rain.

Howard

\o/
/_\
Top of pagePrevious messageNext messageBottom of page Link to this message

Cesar Guerrero (cesar)
Junior Member
Username: cesar

Post Number: 9
Registered: 3-2006
Posted on Monday, October 16, 2006 - 9:16 am:   

Sorry...
The density and the particle size of samples remained similar.

Thanks!
Top of pagePrevious messageNext messageBottom of page Link to this message

Cesar Guerrero (cesar)
Junior Member
Username: cesar

Post Number: 8
Registered: 3-2006
Posted on Monday, October 16, 2006 - 9:11 am:   

Hi!
Thank you very much to David and Gavriel for your help and interest.
These spectra are 3 burned soils. In the new pdf, you can see the original unburned soil too, and the spectra after heating at different temperatures (increasing the darkening with temperature as consequence of the carbonization of soil organic matter).
I think that the first option mentioned by David is the most appropriate, because black carbon is formed during heating .
These spectra were measured with a FT-NIR (on reflectance mode), with a spectral resolution of 8 cm-1. The spectra were converted to nm, because when spectra were in nm, our calibrations were slightly better than in cm-1 (I think that could be an effect of the smoothing made during the conversion from wavenumber [spectrum with >2000 points] to wavelength [spectrum with 1000 points]; but I�m not sure of this).
I have been using NIR to estimate the temperature reached on soils after forest fires. I have a set of burned soil samples (which were heated at controlled conditions, in a furnace-oven, in the lab, acting as references or standards). The predictions of the temperature reached with NIR are very accurate!
As you mentioned, the first derivate results in a good option for observe the changes with heating.
Applying the first-derivative, a decrease in absorbance of some peaks is evident (mainly water and others).
As you commented, this increase in the absorbance baseline don't affect the results.
My questions are about the mean of the increase in the spectra baseline. Could be due only to the increase in sample darkening?

A lot of thanks!!!

Best regards.
 and 1st derivative.pdf,application/pdfnew pdf with spectra
spectra (29.2 k)
Top of pagePrevious messageNext messageBottom of page Link to this message

Gavriel Levin (levin)
Intermediate Member
Username: levin

Post Number: 18
Registered: 1-2006
Posted on Monday, October 16, 2006 - 6:06 am:   

Hi, Is this spectrum a translation from FT in wavenumbers into nm?
What type measurement - reflectance? transmission?

What's the matrix?

How can we answer without knowing?
If in transmission - through what? solution with fixed pathlength? if solution - is it a clear (besides being dark, it can be turbid or clear) one? If in transmission not through cuvette, through what?
If in reflectance - were the sample densities identical? surface appearnce, etc?

All these can cause what appears, to the naked eye as mostly verical shifts that have nothing to do with "darkening" due to absorption. If the shift is not truly vertical, then the 1st derivative will reveal it.

Please provide more info so we can try to help.

Gabi Levin
Brimrose
Top of pagePrevious messageNext messageBottom of page Link to this message

David W. Hopkins (dhopkins)
Senior Member
Username: dhopkins

Post Number: 93
Registered: 10-2002
Posted on Sunday, October 15, 2006 - 4:30 pm:   

Cesar,

I think we can distinguish two kinds of "sampling darkening" that extend from the visible into the NIR region. One is wavelength-independent darkening that is observed throughout the visible and entire NIR, due to a truly black pigment, such as carbon black. This is not what you observe.

The other is observed when an intensely colored pigment (absorber) is present. A frequent observation in natural products is that chlorophyl can be dominant, so that the absorbance that peaks at 670 nm in the visible extends to as far as 1400 nm in the NIR. I don't think I have seen chlorophyl extend as far as we see in your samples, because the color seems to reach as far as 2100 nm, where it appears that your spectra finally bottom out. You really have an extreme case here. The absorbance at 800 nm is extremely affected by the color of the samples, but by 2100 nm, the effect is minimal, and we see baseline and slope effects that may be more an indication of particle size or powder compaction effects. Without knowing more about the samples, I won't try to interpret the scans further. Can you tell us more about the samples and what properties you would like to determine?

This need not affect the utility of the NIR to give you useful information, if you design your calibrations carefully. You may want to restrict your calibrations to the 1600 - 2600 nm range, for example, and use derivatives to minimize the impact of the color of the samples.

Best wishes,
Dave
Top of pagePrevious messageNext messageBottom of page Link to this message

Cesar Guerrero (cesar)
Junior Member
Username: cesar

Post Number: 7
Registered: 3-2006
Posted on Sunday, October 15, 2006 - 12:55 pm:   

Dear all,
Please, if anyone can help me, I have some questions:
- In which ranges of NIR must be expected more effect of sample darkening? In ranges closest to the visible (around 800 nm), or in the overall NIR range?
I have some samples (increasing darkening) and I observed an increase in the overall baseline of the absorbance spectra. These samples are not only different in color.
The samples indreasing darkening from a to c (see figure). Sample c is the most dark.

- Are these increasing due mainly to darkening?

Thank you very much in advance.

Best regards.
application/pdfspectra
spectra.pdf (13.2 k)

Add Your Message Here
Posting is currently disabled in this topic. Contact your discussion moderator for more information.