Dear Kiu,

Mestrenova will do nice 2D plots of Dosy for you, what helps a lot to identify different substances. If you are interested in checking the quality of your fit however, you better get the data points from MestreNova and plot it yourself (intensities Vs gradients plot, I mean). I have experienced some strange behaviour of topspin distributing data points in incomprehensible manner. Thus if you use topspin you better plot the data in a math program as well. There is no other way to know if you have a reasonable fit or a curve trying to fit crazy points. (obs: VnmrJ does show you the fit graph).

When doing the fit, check if the points are deviating in a random manner from the fitted curve, discard your experiment otherwise. Dosy expts are quite sensitive to conditions and may easily give wrong results.

Maybe here is a suggestion for Mestrenova to plot the fit (Echo Intensities Vs Gradient Intensities) as well. It would be a nice feature to help us find out why different experiments give completely different results :-) And believe me, it happens quite often.

Below I show you how I fitted a Dosy of DMSO using the program Octave (additional function leasqr required):

% comments start with a %
% xydata= gradients in %, echo intensities
xydata=[
10                  6.21843e+06
13.3333          5.53812e+06
16.6667         4.49104e+06
20                 4.0029e+06
23.3333         3.22418e+06
26.6667          2.49896e+06
30                  1.8643e+06
33.3333          1.33617e+06
36.6667          923999
40                  613558
43.3333          393992
46.6667          244480
50                  145317
53.3333          83411.1
56.6667          46831.3
60                  38571.2
];

gcal=42.547; % Gauss/m
p1=11.2e-6;
p2=22.4e-6;
p19=1000e-6;
p30=1000e-6;
d16=0.0003;
d20=0.4;
nuclgamma= 2.675222005e4; % rad / (s Gauss)
k=1.3806503e-23; % m2 kg s-2 K-1
T=298.15;
eta=1.991e-3 ; % viscosity of dmso
legend='dmso';

delta = 2p30 ;
DELTA = d20;
exponent_cnst= (nuclgammadelta)^2 * (DELTA -delta/3 - d16/2);
xydata=[xydata(:,1)gcal xydata(:,2)/max(xydata(:,2))];
p2guess=(log(xydata(1,2)/xydata(end,2))/(xydata(end,1)^2 - xydata(1,1)^2));
p1guess=xydata(1,2)/exp(-p2guessxydata(1,1)^2);
pguess=[p1guess p2guess];
fitfunc=@(x,p) p(1)exp(-p(2)x.^2);

[f,coef,cvg,iter,corp,covp,covr,stdresid,Z,r2]=...
leasqr(xydata(:,1),xydata(:,2),pguess, fitfunc);
D=coef(2)/exponent_cnst;
Rh=kT/(6pietaD);
printf("\n\n=== Your Diff coef is %g, Rh=%g ==== \n \n", D , Rh)

%----------------- ploting -------------------------
clf
plot(xydata(:,1),xydata(:,2), 'o', 'markersize', 16)
hold on
lineover=0.05;
linemin=min(xydata(:,1))-lineoverrange(xydata(:,1));
linemax=max(xydata(:,1))+lineoverrange(xydata(:,1));
x=(linemin:range(xydata(:,1))/1000:linemax)';
plot(x, fitfunc(x, coef),'-', 'linewidth', 3)

border=0.1;
Gmin=min(xydata(:,1))-borderrange(xydata(:,1));
Gmax=max(xydata(:,1))+borderrange(xydata(:,1));
Emin=min(xydata(:,2))-borderrange(xydata(:,2));
Emax=max(xydata(:,2))+borderrange(xydata(:,2));
axis([Gmin Gmax Emin Emax])
xlabel('Gradient / Gauss m^{-1}')
ylabel('Echo Intensity')

Hey,

sorry. Have to post the code again. This wiki is using markdown. It changed a little bit what i had typed ...

% xydata= gradients in %, echo intensities
xydata=[
10                6.21843e+06
13.3333            5.53812e+06
16.6667            4.49104e+06
20           4.0029e+06
23.3333            3.22418e+06
26.6667            2.49896e+06
30              1.8643e+06
33.3333            1.33617e+06
36.6667            923999
40              613558
43.3333            393992
46.6667            244480
50              145317
53.3333            83411.1
56.6667            46831.3
60              38571.2
];

gcal=42.547; % Gauss/m
p1=11.2e-6;
p2=22.4e-6;
p19=1000e-6;
p30=1000e-6;
d16=0.0003;
d20=0.4;
nuclgamma= 2.675222005e4; % rad / (s Gauss)
k=1.3806503e-23; %  m2 kg s-2 K-1
T=298.15;
eta=1.991e-3 ; % viscosity of dmso
delta = 2*p30 ;
DELTA = d20;
exponent_cnst= (nuclgamma*delta)^2 * (DELTA -delta/3 - d16/2);
xydata=[xydata(:,1)*gcal  xydata(:,2)/max(xydata(:,2))];
p2guess=(log(xydata(1,2)/xydata(end,2))/(xydata(end,1)^2 - xydata(1,1)^2));
p1guess=xydata(1,2)/exp(-p2guess*xydata(1,1)^2);
pguess=[p1guess p2guess];
fitfunc=@(x,p)   p(1)*exp(-p(2)*x.^2);

[f,coef,cvg,iter,corp,covp,covr,stdresid,Z,r2]=...
          leasqr(xydata(:,1),xydata(:,2),pguess, fitfunc);
D=coef(2)/exponent_cnst;
Rh=k*T/(6*pi*eta*D);
printf("\n\n=== Your Diff coef is %g, Rh=%g ==== \n \n", D , Rh)

%----------------- ploting -------------------------
clf
plot(xydata(:,1),xydata(:,2), 'o', 'markersize', 16)
hold on
lineover=0.05;
linemin=min(xydata(:,1))-lineover*range(xydata(:,1));
linemax=max(xydata(:,1))+lineover*range(xydata(:,1));
x=(linemin:range(xydata(:,1))/1000:linemax)';
plot(x, fitfunc(x, coef),'-', 'linewidth', 3)

border=0.1;
Gmin=min(xydata(:,1))-border*range(xydata(:,1));
Gmax=max(xydata(:,1))+border*range(xydata(:,1));
Emin=min(xydata(:,2))-border*range(xydata(:,2));
Emax=max(xydata(:,2))+border*range(xydata(:,2));
axis([Gmin Gmax Emin Emax])
xlabel('Gradient / Gauss m^{-1}')
ylabel('Echo Intensity')

Dear Kiu Chor Sham, Could you please take a look at this tutorial?

If you have doubts, please contact us

Regards