qNMR Purity Recipes Book

 

1. Sample Preparation

 

The technique requires that you co-dissolve your analyte together with an appropriate reference material. The NMR signal for this must be collected under quite careful conditions if the experiment is to be quantitative. Finally, the spectrum must be processed and relevant data extracted to determine the overall analyte purity. I will illustrate the latter task using Mestrenova software.

Disclaimer

This article is intended to help Mestrelab customers get the best results from “purity” determinations using NMR. It is intended solely as a guide to doing this experiment correctly and does not take away your responsibility to show that your procedures are effective, and give the correct result in a robust way.

Sample preparation

Choosing the solvent, and dissolution

It is very important to first determine which NMR (deuterated) solvent fully dissolves your compound. Try very hard to use a single solvent and it should fully dissolves your sample without heating, adding acid, etc. You may need to inspect the liquid with good lighting and a magnifying glass to fully convince yourself that the sample satisfies this criterion, and that it has not, e.g., formed a colloid. It is a good idea to use a vortex mixer to be certain that the (capped!) vial with a solution of your sample has been thoroughly mixed.

Choosing a Reference Standard

The criteria for a suitable internal reference standard have been well documented [1] , and you will need to be satisfied that this compound:

• Is soluble
• Has known purity
• Has one or more signals that do not overlap with a solute signal

Weighing the samples

Depending on the accuracy you need to achieve, you must decide whether or not you need to do replicates, and how many. Three replicates would be common if this is necessary.

How you weigh your samples is probably different for everyone. You can weigh both materials into the same vial, but adjusting the mass of the second compound is then impossible. If you weigh each sample into separate vials then you must be sure that the full contents are quantitatively transferred to the other sample. Because I am more of an NMR spectroscopist than an analyst, I would favour the first method, and carefully add the second compound. Remember: never return material to the stock bottle!

How much should you weigh out?

This question may be answered for you if you only have a little material, but 10-20 mg would be a sensible amount if you have it. You don’t want so little that the mass is uncertain, and not so much that dissolution becomes a concern. Recovering the material is not impossible, but would require a separation.

You should also aim to weigh roughly equimolar amounts of both analyte and reference material. This can be roughly judged by their relative molecular masses. If your analyte has 3X the molecular mass of the reference standard then you should try to weigh about these in a 3:1 ratio, approximately.

The balance, and weighing the samples

It is generally agreed that a 5-figure balance (weighing to 0.01mg) with a sensible weighing capacity is needed for this process. You will always have practical considerations for weighing: draught exclusion and vibration isolation. And, depending on your typical analytes, a safety assessment may be a good idea to determine the safest location of the balance, and necessary safety considerations – fume hoods, Astec cabinets, etc. Decide whether or not you need the masses printed/logged.

After the samples have been weighed they should be kept in a capped glass container until the solvent is added.

Tip

It really is worthwhile setting enough time aside to doing the weighing task well and without disruption.

Sample dissolution

The object is to be certain that all the analyte and reference material has dissolved and forms a single, true solution.

Use a good grade solvent, preferably with the TMS reference material already added. Having this signal in your spectrum is useful as it also allows you to check the overall quality of the NMR spectra your instrument is generating

NMR sample tube preparation

Remember that you want the best quality NMR spectrum you and your spectrometer can consistently deliver. You may need to determine empirically the quality of NMR tube that meets this need. The usual rules apply:

• The tube must be scrupulously cleaned and dried [2]
• The tube must not have chips at the opening

When transferring your solution of analyte and reference material to the NMR tube, please be sure you have a sensible liquid column height. If you use too little the spectrometer cannot shim the sample properly, and the same can be true of very “long” samples. As a guide, use ~0.6mL for a 5mm tube. The NMR tube should be capped to limit atmospheric water uptake and solvent evaporation.

If you choose to weigh the analyte and reference material into separate vessels then you need to ensure a quantitative transfer of one solution to the other.

1. http://www.sigmaaldrich.com/content/dam/sigma-aldrich/docs/Fluka/General_Information/1/analytix_qnmr.pdf
2. Do not dry tubes in too hot an oven as this can distort the tube! An oven temperature of 50-60°C should suffice.

 

2. NMR Acquisition

The NMR Spectrometer

I will have to be a little vague at times in this section, as specifics will depend on your spectrometer vendor and age.

Temperature

Very good temperature regulation of the sample in the probe is required for the best quality spectra. You should also allow ca. 5 minutes for your sample to temperature equilibrate in the probe before starting shimming and signal acquisition.

Acquisition conditions

You will typically use the simplest pulse sequence—repetitive pulse and acquire:

Relaxation delay — Pulse — Acquire

 

Interpulse time

 

The consideration is that all spins should be at their equilibrium position (fully relaxed) before the next pulse is applied. The sum of the acquisition time and relaxation delay is sometimes called the pulse repetition time, and this number needs to be carefully set up.

The consideration is that all nuclei have fully returned to equilibrium between pulses. If you are using 90° pulses, then this should be at least 5 times the T1 of the proton with the longest relaxation time. If you are measuring the same material every time, it would be beneficial to determine this. It’s worthwhile to determine the T1 of your reference material or solvent under your experimental conditions of temperature, etc., at least once, so that you know the minimum pulse repetition time. You may be surprised by this figure—it is often several tens of seconds!

The issue of pulse repetition rate is often overlooked, and yet it is absolutely fundamental to a good experiment.

 

Acquisition time

 

As mentioned, the interpulse delay is the sum of the acquisition time and relaxation time. But you should consider how you divide this time: a longer acquisition time than you normally use for routine 1H spectra will almost certainly be used.

You want to aim for an acquisition time that is long enough to collect the time-domain FID signal until it has decayed to the point where only noise is being digitized. If you use a shorter acquisition time, the signals will be truncated, and the effect in the time-domain spectrum will be that each peak shows ‘wiggles’ (truncation artifacts) at the base, along with baseline artifacts. This will clearly affect the integration of each signal.

 

 

Spectrum digitisation and Peak Digitisation

 

The acquisition time and sweep width ultimately determine the number of Hz per point in the final spectrum. One level of zero filling can be performed, but the digitization in the spectrum is a very important consideration as it strongly influences the fidelity of the peak integration.

As a rule, we suggest aiming for more than 5 data points per peak if (a) GSD is to be successful and (b) conventional integration is to be accurate. Ignore this rule at your peril—it can significantly affect your reported integrals! The number of points you acquire will depend on the sweep width, of course, but aim for a much longer acquisition time than you usually use, perhaps 10 seconds or longer. You won’t be adding to the experiment time, and with today’s large storage capacities, data storage shouldn’t be a concern!

 

Number of scans

 

Time-averaging is often needed to build an adequate signal-to-noise ratio (SNR). So we need to ensure that the SNR is adequate to support good integration. You should target a SNR of 250:1 or better for any peak that will be integrated for qNMR purposes.

 

Receiver gain

 

As with any acquisition, the receiver gain should be set so the NMR signal fills the digitiser (ADC) as best as is possible whilst avoiding early time points being “clipped”. Too small a receiver gain is undesirable, too. Note that automatic setting of the receiver gain must be done using experimental conditions, and may take some time if the interpulse delay is required to be long!

 

Sweep width

 

Choose a sensible sweep width: this would be ca. 20% larger than the separation between the signals at the extremes of the spectrum.

 

Number of data points

 

This will be dictated by the required sweep-width and acquisition time. However, we need to consider the final digitization of each signal in the spectrum: getting this right is crucial to accurate integration. I will come back to this topic in the processing section, but in extreme cases, it may be necessary to collect the spectrum in parts. For typical 1H NMR spectra, you should collect around 64K data points.

 

Shimming

 

I have saved the best for last! For good quantitation, we need good lineshape, and shimming will play a critical role in this. It’s easy to get used to gradient shimming doing the work for us these days, and it works exceptionally well in the majority of cases. But if you only do ‘on-axis’ gradient shimming, remember to check and optimize ‘off-axis’ shim gradients on a regular basis.

A poorly shimmed sample can often result in an unsymmetrical lineshape, and that will cause significant errors for all line fitting methods, including GSD.

 

Broadband 13C decoupling

 

With qNMR you should in principle also integrate the part of the 1H signal away from the central lines. These arise from the 1-bond coupling between the proton and attached 13C. This accounts for 1.1% of the 1H signal area, and the coupling is 135-200 Hz. Collapsing 13C satellites into the central peak has a number of additional benefits because it simplifies the spectrum. Some spectrometers will make this experiment quite easy to perform. But you must be aware that broadband 13C decoupling itself may raise the temperature in the probe quite significantly, and that would affect the lineshape and therefore be undesirable. You can decrease the acquisition time to minimise this effect. Probe and sample heating is often evident if you look at the lock level and see that it decreases during the acquisition time.

 

3. Data Processing

In this section I will consider just the basics of data processing before describing how to determine the purity. If you do this step carefully you will also have a chance to look closely at the data and satisfy yourself whether or not the spectrum quality is adequate for qNMR. Note that you always want to increase the vertical scaling significantly to see the nitty gritty of your data!

 

 

Exponential multiplication

The only Apodisation function you should consider using is Exponential multiplication. The object is to increase the SNR a little, and ensure that all signals decay to ~zero at the end of the FID. Using a ‘matched filter’ is a good place to start. Select ‘Apodization’ under the Processing menu (keyboard shortcut: ‘W’, and you will see a dialogue resembling this:

 

 

Mnova reports the ‘Average T2*’ at the top: 0.300 s in this case. What we need to do is (a) check the box to ensure Exponential multiplication is used, and (b) enter the same 0.30 Average T2* as the Hz value.

When you click on ‘OK’, this Apodisation will be applied and the spectrum automatically updated.

Phasing

Mnova will usually import the phase values saved by the spectrometer. If these are hideously incorrect you can use AutoPhase (Processing > Phase Correction > Automatic). But we need the phase to be correct to 0.1º, and you will almost always have to perform a manual phase correction (Processing > Phase Correction > Manual Phase Correction).

 

When you invoke manual phase correction you will see something like this:

 

 

The blue vertical line shows the tallest peak – which is where the PH0 will be applied. Whilst looking at the base of this peak, place the mouse cursor on the blue pad and drag the mouse whilst holding down the left mouse button. Hold down the ‘Ctrl’ key whilst dragging the mouse to make finer adjustments. Once the biggest line is correctly phased, repeat this process whilst holding down the right mouse button: the other peaks should be phased.

 

 

Baseline correction

It may not be terribly obvious that the baseline needs correction. In the spectrum, below, the lower trace (A) shows one common symptom of poor baseline, here evidenced in the ‘slope’ the integrals are shown with a blue arrow.
The simplest and most robust solution to baseline correction is the ‘Bernstein Polynomial’, accessed using Processing > Baseline > Baseline correction (Keyboard shortcut ‘B’). You will see a dialogue like this:

 

 

Note that there are more sophisticated baseline correction routines available to you, such as the Whittaker Smoother. This must be used with care as it can play havoc with integrals if used incorrectly. Email me if you want some instruction on using this. In most cases the ‘Bernstein Polynomial’ will be the best choice.

 

 

Multiplet determination [1]

The last task before processing task is to determine the multiplets in the spectrum and the number of nuclides for each multiplet.

Automatic Multiplets Analysis

This is invoked by selecting Analysis > Multiplets Analysis > Automatic, and the analysis will be performed automatically.

Automatic Multiplets Analysis with the Molecular Structure

Follow exactly the same process as above, but be sure to include the structures of the analyte and reference compound. This method is more robust in the determination of the number of nuclides.

Checking your procedure

It is very likely that the procedure you adopt will differ a little from what I have described, and you will likely want to perform tests from time to time to ensure that you and your NMR are doing things correctly.
I would therefore suggest that you make a sample of a molecule of high and known purity. Try to use the solvent and internal reference that you use most commonly. The result from this sample will tell you if everything is working well. If you can (and the compound is stable), make perhaps 3 tubes and seal them. This will allow you to check everything from time-to-time. I leave it to you to decide how frequently to do this.

1. Automated purity analysis supported within Mnova NMR, although currently not in the qNMR engine.