-
- Posts: 1
- Joined: Fri Jul 30, 2010 2:21 pm
The Gaussian distribution is
f(x)= 1/(sigma* sqrt{2\pi) } e^{ -(x-mu)^2/ 2sigma ^2} }
sigma being standard deviation and mu average.
A Gaussian peak in a chromatogram could be described by
f(x)= A/(sigma* sqrt{2\pi) } e^{ -(x-mu)^2/ 2sigma ^2} }
where A is the area of the peak.
The maximum peak height H is readily proven to be found at x =mu, hence the exponential part is equal to 1;
H=f(mu)= A/(sigma* sqrt{2\pi) } e^{ -(mu-mu)^2/ 2sigma ^2} }
=H= A/(sigma* sqrt{2\pi) } = 0.3989A/sigma
A = H*sigma/0.3989
This shows that the area (and hence concentration) is proportional to maximum peak height. But this assumes same /sigma. But I would like to consider that injection and stuff might be different and account for peaks having a bit different sigma.
The full width at half maximum (FWHM) can be shown to be
FWHM=2 sqrt(2 ln(2))*sigma = 2.35*sigma by inserting f(x) = H/2 , find x1 and x2 and then calc the width.
FWHM = 2.35 sigma
sigma= FWHM/2.35
A=H * sigma /0.3989 = H * FWHM / (0.3989 *2.35)
Then the area would easily be calculated from a paper with peaks using only ruler and calculator, by measuring the maximum height H and the FWHM, multiplying them and multiply by a constant.
1.Did I overlook something here, or does it work this way?
It seems very convenient to account for peaks having different sigma this way, although I doubt it has too large practical significance in the high school lab... Also, this still is under assumption that all peaks are perfectly gaussian in shape.
In school we usually have software to calculate peak areas, or just plot peak maximum height.
2. How does a software compute such an area? By using a lot of small integrals?
The integral with respect to x of [tex]
f(x)= A/(sigma* sqrt{2\pi) } e^{ -(x-mu)^2/ 2sigma ^2} }would equal A. Analytical integration yields A=A and does not give the value of the area... or does anyone another way?
3. If I just sit here with a bunch of peaks on a paper, can I use some method applying integrals to find A? Numerical integration seems heavy.
Or any other method at all than A= H * FWHM / (0.3989 *2.35)?
Thanks a lot in advance!