Subtract Radiated from Background Spectrum to Analyze Absorption Spectrum

12 replies · 4K views · last reply
Jhon81
Messages
11
Reaction score
0
Hi everyone,
I have an absorption spectrum that I obtained from a sample after radiation and I need to use the background sample as a reference, how can I do that using excel?

what I did is subtract the radiated from the background spectrum then plot it, is that right?

Thanks.
 
I would subtract the background from the signal not the opposite way, but anyway, that is just a sign.

If you don't run into a problem with statistical fluctuations that approach should be fine. More context would be useful, however.
 
To do a correct subtraction you should do it in absorbance mode, not transmittance; but perhaps you already knew that.
 
Thanks, mfb and mjc123 for your comments.
I subtracted the background from the signal but after plotting it seems that the points have spread! why is that?
Line graph of a noisy signal against an unlabeled horizontal axis from 0 to about 3600, with the vertical axis from -300 to 600. The trace stays near zero, rises to a broad peak of about 500 near x = 1500, then dips to about -220 near x = 2000 before settling back near zero.

then I tried to add them and the shape of the plot was looking more nicer, could someone explain what happened to me?
Line plot of a single noisy curve, intensity from about 0 to 25000 against x from 0 to 3500, rising from near zero around 600, peaking near 1800, and falling back to zero by about 3000. Chart title placeholder reads Chart Title.
 
Attachments
  • Line plot of a single noisy curve, intensity from about 0 to 25000 against x from 0 to 3500, rising from near zero around 600, peaking near 1800, and falling back to zero by about 3000. Chart title placeholder reads Chart Title.
    adding.png
    9.5 KB · Views: 905
  • Line graph of a noisy signal against an unlabeled horizontal axis from 0 to about 3600, with the vertical axis from -300 to 600. The trace stays near zero, rises to a broad peak of about 500 near x = 1500, then dips to about -220 near x = 2000 before settling back near zero.
    subtraction.png
    7.4 KB · Views: 842
Why did you add them?
Can you show your sample and background spectra? It looks as if they are very similar, so the subtracted signal is much weaker than either spectrum. That's why the noise-to-signal ratio is much greater on your subtracted spectrum (I assume that's what you mean by the points "spreading").
What are your y-axis units? If they are mAU, then your spectrum looks a bit too strong for accurate quantitative analysis. Can you repeat the experiment at a lower concentration?
 
mjc123 said:
Why did you add them?
Can you show your sample and background spectra? It looks as if they are very similar, so the subtracted signal is much weaker than either spectrum. That's why the noise-to-signal ratio is much greater on your subtracted spectrum (I assume that's what you mean by the points "spreading").
What are your y-axis units? If they are mAU, then your spectrum looks a bit too strong for accurate quantitative analysis. Can you repeat the experiment at a lower concentration?
I added them to see the shape of the spectrum.
This is the spectra of the background and the first irradiated sample, respectively:
Line graph of a single-peaked spectrum with a smooth blue curve: near zero from about 200 to 350, rising to a broad peak near 600 to 650 with values around 12000, then falling to near zero by about 1100. Horizontal axis unlabelled, ticks 0 to 1200; vertical axis ticks -2000 to 14000.
Line graph of a spectrometer intensity curve against wavelength, with x-axis 0 to 4000 and y-axis -2000 to 14000; the curve stays near zero, rises from about 500, peaks near 12000 around 1750, then falls back to zero by about 3600.

As for the unit, I am not sure as it was my first time using that spectrometer, I am going to check and let you know.
If you mean the concentration of irradiated sample it is very law ( I am using FXO with Fe+2 (0.3mM) & XO (0.05mM)) all of them dissolved in 50 mM of H2SO4.
 
Attachments
  • Line graph of a single-peaked spectrum with a smooth blue curve: near zero from about 200 to 350, rising to a broad peak near 600 to 650 with values around 12000, then falling to near zero by about 1100. Horizontal axis unlabelled, ticks 0 to 1200; vertical axis ticks -2000 to 14000.
    background.png
    6.9 KB · Views: 726
  • Line graph of a spectrometer intensity curve against wavelength, with x-axis 0 to 4000 and y-axis -2000 to 14000; the curve stays near zero, rises from about 500, peaks near 12000 around 1750, then falls back to zero by about 3600.
    1st irradiated.png
    7.2 KB · Views: 763
Last edited:
What are your x-axis units here? You have very similar spectra, but one with the maximum at ca. 650 (nm? that would make sense) and the other at ca. 1800 (what? is it simply data point number in excel? I doubt very much if you scanned down to 0 nm!)
If you ran your spectra in a 10 mm cuvette, try using a 1 mm cuvette.
But I think you'll find it difficult getting a good subtraction spectrum. The difference between the spectra is of the same order of magnitude as the noise - as you found!
 
mjc123 said:
What are your x-axis units here? You have very similar spectra, but one with the maximum at ca. 650 (nm? that would make sense) and the other at ca. 1800 (what? is it simply data point number in excel? I doubt very much if you scanned down to 0 nm!)
If you ran your spectra in a 10 mm cuvette, try using a 1 mm cuvette.
But I think you'll find it difficult getting a good subtraction spectrum. The difference between the spectra is of the same order of magnitude as the noise - as you found!

The x-axis unite is nm. I tried to compare the 4th irradiated with the background
background spectra
Line plot of intensity against wavelength in nm, a single smooth peak centred near 600 to 650 nm reaching about 12000, with small shoulders near 550 nm and 630 nm, falling to near zero below about 350 nm and above about 1000 nm, and a small negative tail on the y-axis.

4th irradiated spectra
Line plot of a single smooth curve against x-axis labelled 0 to 1200 with y-axis from -2000 to 12000; the curve is near zero until about 350, rises to a peak of roughly 11000 near 650, and falls back to zero by about 1100.

subtraction of the above spectrums
Line graph of a difference spectrum, x-axis in nm from 0 to 4000, y-axis from -3000 to 500; a broad negative dip reaching about -2400 near 1500 nm, with a small positive bump near 1000 nm and noisy baseline near zero.

Is it good that the spectra are in the minus y-axis or not, but I noticed that the points are not very spread which makes the noise less right?
 
Attachments
  • Line plot of intensity against wavelength in nm, a single smooth peak centred near 600 to 650 nm reaching about 12000, with small shoulders near 550 nm and 630 nm, falling to near zero below about 350 nm and above about 1000 nm, and a small negative tail on the y-axis.
    background.png
    6.8 KB · Views: 710
  • Line plot of a single smooth curve against x-axis labelled 0 to 1200 with y-axis from -2000 to 12000; the curve is near zero until about 350, rises to a peak of roughly 11000 near 650, and falls back to zero by about 1100.
    4th irradiated.png
    10.8 KB · Views: 743
  • Line graph of a difference spectrum, x-axis in nm from 0 to 4000, y-axis from -3000 to 500; a broad negative dip reaching about -2400 near 1500 nm, with a small positive bump near 1000 nm and noisy baseline near zero.
    subtraction2.png
    6.3 KB · Views: 703
This is a bit better. You can see that the peak at 600 nm is reduced in the irradiated spectrum. This presumably corresponds to your negative peak at "1500" in the subtraction spectrum - get your x-axis units right. Yes, you should subtract the background from the irradiated spectrum, so your peak is negative.. Again, what do you mean by the points being "not very spread"?
 
mjc123 said:
This is a bit better. You can see that the peak at 600 nm is reduced in the irradiated spectrum. This presumably corresponds to your negative peak at "1500" in the subtraction spectrum - get your x-axis units right. Yes, you should subtract the background from the irradiated spectrum, so your peak is negative.. Again, what do you mean by the points being "not very spread"?
what do you mean to get my x-axis unit right?
In this picture, you can see that the points are very spread ( the spectra is not smooth)
Line graph of a noisy signal with x-axis ticks from 0 to 3500 and y-axis from -300 to 600; a broad positive peak near x = 1500 reaches about 500, with a negative dip near x = 2000.


but in the second one, it is smoother.

Line graph of a noisy signal against unlabeled x-axis values from 0 to about 3600, with y-axis from -3000 to 500. Small bump near 1000, sharp dip to about -2400 near 1500, then slow recovery to zero by about 3000.
 
Attachments
  • Line graph of a noisy signal with x-axis ticks from 0 to 3500 and y-axis from -300 to 600; a broad positive peak near x = 1500 reaches about 500, with a negative dip near x = 2000.
    subtraction.png
    7.4 KB · Views: 664
  • Line graph of a noisy signal against unlabeled x-axis values from 0 to about 3600, with y-axis from -3000 to 500. Small bump near 1000, sharp dip to about -2400 near 1500, then slow recovery to zero by about 3000.
    subtraction2.png
    6.3 KB · Views: 715
Jhon81 said:
what do you mean to get my x-axis unit right?
Your measured spectra are from 200 to 1100 nm. Why does the difference have x values of 0 to 3500?
Jhon81 said:
In this picture, you can see that the points are very spread ( the spectra is not smooth)
That tells you that there is some noise in the measurement. Not surprising.
Jhon81 said:
but in the second one, it is smoother.
The signal (the true difference) is stronger, so the signal to noise ratio is better.
 
@Jhon81 In one case it appears you are graphing wavelength on the x-axis in nm and in the other case, you are apparently doing wavenumber=## \frac{1}{\lambda} ## (cm^-1). It is difficult to try to analyze the results unless this is consistent. You must at least state that you made the conversion. ## \\ ## Edit: But that doesn't appear to be the case either. It leaves us guessing...
 
It can't be that because the spectra aren't reversed. It seems to be that sometimes you are plotting an x-y chart with wavelength as the x-value and absorbance as the y-value; other times you are just plotting a y chart, with the absorbance as the y value and the data point number (looks like you have about 3700 data points)appears as the x value. You want to get some more practice plotting charts in excel (or whatever you use).