# PALMA for DOSY Analysis

## How to measure the best possible DOSY experiment...

You know that the DOSY experiment is not an easy one, it may fail in a number of manner. We provide here a short check-list covering the most typical problems that you may face

This involves three stages

• 1/ ACQUIRE
• 2/ PROCESS
• 3/ ANALYSE
• IN CASE OF PROBLEMS

## 1/ Acquire the best possible DOSY experiment...

First Convection is your enemy ! you're always better in D$_2$O and DMSO than in CDCl$_3$ (water is more viscous that chloroform !), you prefer 3mm tubes than 5mm, you prefer room temperature than anything else.

Also check the temperature control !

• eventually increase the gaz-flow to be sure that the temperature in homogeneous in the tube
• display the EDTE panel, and go to monitoring. If you observe oscillation there, you're off. Re-tune the PDI of the temperature control in your current conditions.

Then insure that all the parameters have been set correctly, typically the following ones should be checked before each experiments:

• the P90 should have been carefully calibrated.
• use a reasonable gradient list, typically not less that 20-24 points, and not more 64 points as it is probably useless. Ranging from 1% to 99% is Ok on the modern probes and the modern PFG systems.
• use a correct pulse sequence, stexxx and ledxxx are usually good choices. Use dstebpxxx if you suspect convection.
• calibrate the $\Delta$ (diffusion delai) and $\delta$ (PFG duration) parameters, so that, whe comparing the first and the last PFG increment, the slowest decaying signal is left in the final experiment to a few % of its initial intensity. I usually do this by running a 2 or 3 points DOSY.
• Should should have a descent SNR. PALMA is able to extract information down to about a Signal-to-Noise Ratio of x20; however in the current version, this is limited to x32. So signals lower than that will not be analyzed.
• the gain RG, should not be set too high as non-linearity in the acquisition chain may perturb the measure. The rule of thumb I advise is the following : in the block were the DOSY acquisition is going to take place, run the RGA command, then divide the obtained RG value by 2, just to be sure.
• use a large enough gradient recovery delay (usually D16). Increase it if you observe phase distortion on the right hand of the sharp peaks.

## 2/ Process the best possible DOSY experiment...

Processing is simple

• No need to have a lot of points in the F2 domain. Usually 4k to 8k is enough, you want SNR rather.
• Phase precisely the F2 axis, this will improve the final spectrum
• Baseline should also be corrected. The flatter, the better
• Then PALMA will take care of everything...

Detect possible problems at acquisition.
For this, look at spectrum after the XF2 step and try to detect the possible problems at this stage:

• Signals seem to decay too rapidly, even for the larger objects, some negative spectra can be spotted for the larger gradients.
This is the sign of convection in you sample. Check the advices on temperature above, and redo your acquisition.
• Distorted peaks, typically difficult to phase, with negative wings
This is typical to a problem with the recovery of the field after the PFG, it may have many causes:
• wrong lock phase of lock offset, leading to slow lock recovery
• PFG recovery delay too short (typically D16 on standard PULPROG)
• The position of the NMR peaks is not stable, they oscillate in the spectrum after XF2, and present a snake-like pattern when in close zoom.
Your sample is unstable and evolves during the time of the experiment. This happens also in classical 2D, but DOSY is much more sensitive to this artefact. Typically, this comes from a temperature fluctuation of the sample, (or eventually of the room). You might consider correcting this problem (look at : this work and in particular Fig.2)
• The phase of the spectrum is different for the first (low gradients) spectra than for the last
This is probably due to a non-linearity in your acquisition chain. Try reducing RG, or buying a newer machine...

## 3/ Analyse

Diffusion coefficients depend on many parameters in the sample, for this reason they provide a lot of information on the sample, as well as on the physico-chemical conditions of the solution.

Remember that:

• All the signals coming from the same molecule should present the same diffusion coefficient
Unless there is some chemical exchange taking place (solvent exchangeable protons for instance)
• Inverse Laplace Transform is difficult, analysis of superposed signals from different molecules is always difficult.
Diffusion coefficients are extracted with a better accuracy from isolated peaks
• The width of the signal in the diffusion axis is both an indication of a possible polydispersity and of the uncertainty of the analysis.
Take it with a grain of salt.
• Diffusion depends on the solution conditions (temperature, ionic strength, interactions, etc...).
When comparing different conditions, it is always preferable to use a reference molecule (for instance the solvent), and to correct for possible variations from one tube to another.

## Finally/ In case of problems

The server may exit with an error, in which case, no result are present in the mail but a the subject of the mail details the error, and additional information may be associated to the mail.

Several error messages are possible:

• file should be a zip of the TopSpin expno directory, containg one (and only one) pre-processed
The data file that was deposited for the computation was not recognized. It should be the EXPNO directory from Bruker, zipped in an archive. Also it should only contains one pdata directory with processed data. This may happen if you send to PALMA an experiment which contains already the result from a previous run. This will probably changed in the future.
• preparing data file for processing
An error occured during the initial import of the NMR data-set. One of the file (data-set of Bruker parameter files) may be corrupted or missing. Check the zip file and your data-set.
• processing data-file
An error occured during the PALMA processing itself. A classical case is the difflist file missing. This happens when the experiment was run without using the dosy macro beforehand. This macro generate the spectrometer set-up, and also the difflist file which is required for the processing.
• SNR is too low
PALMA processes only the signals which have an estimated signal to noise over a given value (currently 32). In certain cases, no signal over this threshold was found in the data-set. If this happens, the program does not send any results.
• in parsing arguments from system
The program was called with wrong parameters. This should not happen in normal conditions