Waite Solid-state NMR FacilityWaite Solid-state NMR FacilityThe Waite Solid State NMR Facility is jointly operated by the
University
of Adelaide Department of Soil and Water and CSIRO Division
of Land and Water.
The facility specialises in the characterisation of natural organic matter (NOM)
derived from soils, sediments and waters.
|
Inquiries regarding access to this facility, collaborations, consultation and prices for analyses should be directed to: Waite Solid State NMR Facility Department of Soil and Water Waite Campus, University of Adelaide Glen Osmond SA 5064 Australia Ph: + 61 8 8303 7436 Fax: + 61 8 8303 6511 Email: ronald.smernik@adelaide.edu.au |
These pages contain information about the characterisation of NOM using solid state 13C NMR spectroscopy. Free, downloadable spreadsheets are included for the analyses described. Any comments, suggestions or questions regarding these pages would be gratefully received (ronald.smernik@adelaide.edu.au). The intention of these pages is to stimulate discussion and interaction amongst a thinly spread research community and to provide starting point for those interested in research in this area. These pages do not contain information about basic NMR theory, or solid state NMR because this type of information can be accessed easily elsewhere on the web. The Varian NMR webpage http://www.varianinc.com/nmr/apps/varianusers.html has links to many such sites.
Introduction
Role of NOM in the environment
Detection and characterisation of NOM
Advantages and disadvantages of solid state 13C NMR spectroscopy for NOM characterisationNMR quantitation
Choice of polarisation technique
Recycle delay
Contact time
Spinning side-bands
Background rotor, end-cap and stator signal
Paramagnetic species
Remote protonation
High degrees of molecular motion
Spin counting experiments.More sophisticated NMR experiments
Variable contact time experiments for determination of T1r H
Inversion-recovery experiments for determination of T1H
Proton spin relaxation editing (PSRE)
Dipolar dephasingReferences
Downloadable spreadsheets
Dipolar dephasing
Download Excel workbook
Instructions and tips for using spreadsheets
Sample data and output
Quantitation in solid state 13C NMR spectroscopy of NOM
The potential to determine quantitative distributions of functional groups is the major attraction of solid state 13C NMR spectroscopy in the study of NOM. However, realising this potential requires careful control of the experimental set-up. Even with optimal set-up, quantitation may not be possible with some samples. For other samples, quantitation may need to be played off against signal-to-noise ratio and cost. The issue of quantitation in solid state 13C NMR spectroscopy has been debated at length in the literature (see Smernik and Oades, 2000a,b and references therein), and here follows my contribution to this discussion.
Some important factors which influence quantitation in solid state 13C NMR spectroscopy in NOM samples are: the choice of polarisation technique, the recycle delay, the contact time, spinning side-bands, background rotor, end-cap and stator signal, the presence of paramagnetic species, remote protonation and high degrees of molecular motion. The best way to gauge quantitation is the use of spin counting experiments.
Polarisation technique
There are two common methods of generating 13C polarisation (or magnetisation or coherence), direct polarisation (the Bloch decay experiment) or cross polarisation. Direct polarisation (DP) is exactly that; 13C nuclei are irradiated with resonant rf frequency to bring 13C magnetisation into the transverse plane. Nothing much can go wrong with DP, however, relaxation of 13C magnetisation back to equilibrium (T1C relaxation) is notoriously slow, necessitating long recycle delays and resulting in slow repetition rates (pulses per unit time). Cross polarisation (CP) offers improved signal-to-noise (S/N) ratios over DP for two reasons: (i) a signal enhancement of up to four-fold (but usually closer to three-fold) per scan and (ii) recycle delays in CP experiments are dictated by T1H relaxation rates which are usually at least an order of magnitude more rapid than T1C. The combination of these two factors equates to an improvement in sensitivity of often two orders of magnitude for CP over DP. This improved sensitivity can decrease acquisition time (and hence cost), and/or improve S/N ratios. The drawback to the CP experiment is that more can go wrong. Quantitation is more difficult to achieve because CP is more sensitive to paramagnetic impurities and is reliant on the presence of neighbouring 1H nuclei.
Recycle delay
Because NMR is inherently insensitive, an NMR spectrum is generally the compilation of many (hundreds to over a million) individual scans. The strength of the signal is linearly related to the number of scans, whereas random noise increases as the square root of the number of scans. Thus the signal-to-noise (S/N) ratio increases as the square root of the number of scans (e.g. to improve the S/N ratio by a factor of two requires four times as many scans, and thus four times the acquisition time). The recycle delay is the period between consecutive scans. There are a number of considerations in setting the recycle delay, e.g. hardware limitations on switching between transmission and receiving, sample heating, and acquisition time required for the desired resolution. However, for solid state 13C NMR of NOM samples, the overwhelming consideration in nearly all cases is the time required for the nuclei to regain equilibrium magnetisation (T1 relaxation). Irradiation of a system not at equilibrium results in signal loss in a process termed saturation.
The two common methods for determining appropriate recycle delays are: (i) 5-10 times the measured T1 as determined from inversion recovery experiments and (ii) running experiments with an array of recycle delays, then choosing the shortest delay for which there is no apparent saturation.
The inversion-recovery pulse sequence and software for determining T1 from such experiments are standard on most spectrometers. However, it must be kept in mind that there is an implicit assumption of uniform T1 in this analysis. This is a reasonable assumption for CP experiments on most organic compounds where spin diffusion (a process by which rapid relaxation of one nucleus induces rapid relaxation of neighbouring nuclei) ensures a single T1H value throughout. However, it is my experience that in most NOM samples T1H is not uniform. This comes about because spin diffusion is only efficient up to approximately 100 nm and NOM samples are often heterogeneous at scales larger than this. This problem can be overcome by fitting inversion recovery data to a two T1H component model, and statistically comparing the fit to the single component fit. A spreadsheet for the two T1H component fit for inversion-recovery data can be downloaded here and is described in Smernik et al., 2000. For samples for which the two component fit is significantly better (F-ratio test), the recycle delay for CP experiments should be at least five T1H for the slower relaxing (longer T1H) component. Failure to do this can result in preferential loss of signal from the slowly relaxing component through saturation.
For DP (Bloch decay) experiments T1C will not be uniform, even in simple systems, because spin diffusion is inefficient in the isotopically rare 13C population. For DP experiments on NOM samples, direct measurement of T1C is rarely enlightening due to the large range of T1C values and overlapping signals. In this case the recycle delay is more conveniently determined by running experiments with an array of recycle delays, then choosing the shortest delay for which there is no apparent saturation. NOM samples often require recycle delays of 60-90 s.
Contact time
In CP experiments, the contact time, the period during which the magnetisation is transferred from the 1H population to the 13C population, has a significant effect on quantitation. During the contact time there are two competing effects on 13C magnetisation. Magnetisation build-up occurs due to the transfer of from the 1H population. This is often modelled as an exponential build up, with time constant 1/TCH. At the same time there is loss of magnetisation due to 1H relaxation in the rotating frame, characterised by an exponential time constant 1/T1r H. (Magnetisation loss can also occur through 13C relaxation in the rotating frame (T1r C) but is usually minor compared to T1r H relaxation.) Magnetisation (I), as a function of contact time (t) is thus modelled as Equation 1:
which can be rearranged to give the alternate form (Equation 2):
where C is the concentration of 13C nuclei giving rise to the signal, and a = 1-(TCH/T1r H).
So what do these functions look like with T1r H and TCH values typical for NOM samples? In my experience, T1r H values for NOM samples are typically in the range 3-6 ms (see below for determination of T1r H). Although some NOM samples do contain components with different T1r H values, these differences are usually small, and in most cases an assumption of uniform T1r H is justified. TCH values depend on the proximity of nearest proton neighbours. On a 200 MHz spectrometer and with a MAS rate of 5 kHz, typical TCH values are 0.05 ms for directly protonated carbons, and 0.35 ms for non-protonated carbons. However, in materials such as charcoal, where there are carbons many bonds distant from nearest proton neighbours, much longer TCH values of 1 ms or longer are possible. MAS rates higher than 5 kHz will increase these TCH values, whereas lower MAS rates will decrease TCH values. Figures 1, 2 and 3 below show how TCH and T1r H affect relative signal intensity at a range of contact times.
Figure 1 represents the "typical" situation in NOM samples (T1r H = 4 ms). In this example (and the other two), longer TCH values result in slower build-up of magnetisation, as expected, but also greater signal intensity at longer contact times. The curves for TCH = 0.05 ms (typical for protonated carbons) and TCH = 0.35 ms (typical for most non-protonated carbons) intersect at a contact time very close to 1 ms. Therefore using a contact time of 1 ms results in excellent relative quantitation between protonated and most non-protonated carbons in this case. However, at a contact time of 1 ms, the curve for TCH = 1 ms (which may represent some remotely protonated carbons) shows an intensity of only 70% of the other two curves, representing a significant underestimation of such structures. A contact time of 2 ms probably represents the best compromise for the three TCH values, with relative intensities of 61, 66 and 63 for TCH values of 0.05, 0.35 and 1 ms, respectively.
Figure 2 shows the situation for a much shorter T1r H value of 1.5 ms. Such a short T1r H value will only occur in NOM samples where there is an unusually high concentration of paramagnetic centres. In this case relative quantitation between TCH = 0.05 ms and TCH = 0.35 ms is best at 0.8 ms. Figure 3 shows the situation for a much longer T1r H value of 10 ms. Such a long T1r H value is above the usual range for NOM (pure cellulose has a T1r H value of about 10 ms under these conditions). In this case relative quantitation between TCH = 0.05 ms and TCH = 0.35 ms is best at 1.2 ms, but is still good at 1 ms.
Figure 1 Variable contact time signal intensities from Equations 1, 2 for
T1r H = 4 ms.
Figure 2 Variable contact time signal intensities from Equations 1, 2 for
T1r H = 1.5 ms
Figure 3 Variable contact time signal intensities from Equations 1, 2 for
T1r H = 10 ms
So there is no "perfect" contact time which ensures quantitation in CP spectra for all nuclei regardless of TCH and T1r H values. This is what I do about it:
(i) I use a 1 ms contact time which ensures good relative quantitation (within 6%) between typical protonated (TCH = 0.05 ms) and non-protonated (TCH = 0.35 ms) carbons within the range of T1r H (2.5 - 10 ms) values normally encountered in NOM samples.
(ii) I accept that the intensity of carbons remote from nearest proton neighbours will be underestimated. Nothing practical can be done to avoid this in CP spectra, it just has to be kept in mind when interpreting the spectra. So how remote is remote? I’m working on that one, so keep checking here for updates! TCH probably starts becoming a major problem (TCH ³ 1ms) when the nearest proton is about 4 bonds away, and the problem gets worse (TCH gets larger) the further away the protons are. In NOM samples the problem most often arises with condensed aromatic systems such as those found in charcoal. Components with high degrees of molecular motion and high MAS rates exacerbate the problem.
(iii) I measure T1r H using a variable contact time experiment to ensure it is in the range 2.5-10 ms. If T1r H is not in this range I adjust the contact time to ensure relative quantitation between TCH = 0.05 ms and TCH = 0.35 ms carbons.
(iv) I determine what fraction of potential signal is actually observed via spin counting.
(v) I acquire Bloch decay (DP) spectra on a selection of samples in large studies to check quantitation. I include in the selection those samples with low CP observabilities (determined from spin counting). Signal-to-noise can be sacrificed if all you want to do is check that the distribution of resonances are similar.
Spinning side bands
Spinning side bands (SSB’s) occur in solid state NMR spectra when the rate of magic angle spinning (MAS) is lower than the chemical shift anisotropy (CSA). CSA is proportional to field strength; on a 200 MHz spectrometer, CSA for sp2 hybridised carbons (e.g. carbonyl, aromatic, alkene) is ~13 kHz. MAS rates this high are technically difficult to achieve and cause quantitation problems due to long TCH values, as discussed above. So I run at a lower MAS rate of 5 kHz and live with the SSB’s. The TOSS (TOtal Suppression of Side-bands) pulse sequence can be used to get rid of the SSB’s, but since it can cause quantitation problems, I choose not to use it.
How bad are the SSB’s? Not that bad with my set up. For carbonyl and aromatic resonances (first order) SSB’s are usually <10% of the intensity of the central band. Second order SSB’s are sometimes observed but are very small (<1% of the central band intensity) and can be ignored. For sp3 hybridised carbons (e.g. alkyl, O-alkyl), (first order) SSB’s are negligible (<1% of the central band intensity).
SSB’s can be corrected for in integral region data. First order SSB’s of roughly equal intensity appear either side of the central band, at frequencies shifted by the MAS rate. On a 200 MHz spectrometer (13C resonant frequency = 50 MHz) and with a MAS rate of 5 kHz, sidebands appear 5 kHz/50 MHz = 100 ppm from the central bands. The low field (high ppm) sidebands of the aromatic and carbonyl resonances (the only resonances which give rise to significant SSB’s) do not overlap with any resonances in NOM samples. Signal loss from the central band of these resonances can be corrected for by adding twice the intensity of these low field SSB’s. The high field SSB’s of the carbonyl and aromatic resonances overlap with resonances in the O-alkyl and alkyl regions, respectively. To correct the intensity of these resonances for the presence of SSB’s, the intensity of the low field SSB needs to be subtracted. A downloadable spreadsheet for carrying out this analysis can be obtained here.
Background rotor, end-cap and stator signal
The existence of background signal is a consideration in any analytical technique. For solid state 13C NMR spectroscopy, background signal can arise from rotors, rotor end-caps and the stator if these components contain carbon. Rotors are usually made from zirconia or similar materials which contain no carbon and hence give rise to no signal. Rotor end-caps are often made from a plastic (e.g. Kel-F) and can be a significant source of background signal. The stator of a MAS probe does not usually contain carbon, but plastics or glues are sometimes present in the probe and can give rise to significant background signal. It is highly advisable to acquire spectra on an empty rotor, under your usual acquisition conditions to assess the background signal.
On my spectrometer, using Kel-F end-caps, CP signal is small but can be significant for low carbon samples. Spin counting on the empty rotor indicates that background CP signal is equivalent to <1 mg of observable C. Bloch decay (DP) signal is much greater, equivalent to ~100 mg of C and must be corrected for, for all samples. The difference is due to the fluorinated plastic (Kel-F) from which the end-caps are made. This material contains little or no hydrogen, and so does not give rise to much CP signal.
Background signal can be handled in three ways:
(i) It can be ignored. This is reasonable for CP spectra of samples with a high C content.
(ii) It can be corrected for in integral region data. Background CP signal is handled this way in the downloadable spreadsheet available here.
(iii) Background spectra (empty rotor) can be subtracted from the sample spectrum. Background Bloch decay (DP) signal is handled this way in the downloadable spreadsheet available here.
Why do I treat CP and Bloch decay spectra differently? The main reason is that the Bloch decay background signal is very broad; so broad, in fact, that is impossible to define a baseline. This makes it impossible to determine integral region intensities for the background spectrum. The CP background is not so broad and can be integrated in the same way as for other spectra. The second reason for the different treatments is that the background signal for Bloch decay spectra is so large that background correction is necessary for all samples in order to obtain a meaningful spectrum. For CP spectra, background signal does not grossly change the shape of the overall spectrum except for the lowest C samples, and subtle corrections are more easily handled in the integral region analysis. The third reason is that subtraction of background spectrum diminishes the S/N ratio of the spectra and for most CP spectra, this is unnecessary and best avoided. The fourth reason is that the peculiarities of the Varian spectral subtraction software make spectral subtraction non-trivial. I don’t know how spectral subtraction works on non-Varian spectrometers, but on mine it goes something like this:
It is best to subtract the FID’s, rather than the transformed spectra. In this way the corrected spectrum behaves like a real spectrum in that it can be retransformed, phased, zero filled, apodised with different weighting functions etc. Subtracting spectra only gives a "picture of a spectrum" which can’t be manipulated further. When one FID is subtracted from another, the parameter file for the newly created spectrum appears to be basically derived from the first spectrum added to (or subtracted from) the buffer. Thus the commands "clradd jexp1 add jexp2 sub" and "clradd jexp2 sub jexp1 add" will have subtly different results in that the former spectrum will have a parameter file derived from exp1 (the sample spectrum) whilst the latter will have a parameter file derived from exp2 (the background spectrum). This becomes very important when the sample and background spectra have different phasing parameters (lp and rp), because the resultant spectrum subtracts (or adds) the second spectrum phased with the lp and rp values from the first spectrum. To get around this problem, the lp and rp values for both sample and background signal must be the same, say, correct for the background spectrum, and zero order phasing of the sample spectrum then can then be corrected using the phfid parameter (I’ve found that first order phasing has not required changing in 3 years).
On the subject of phasing, it is very difficult to correctly phase the background Bloch decay spectrum because it is so broad. I’ve found the best method is to first acquire a Bloch decay spectrum of a standard (glycine is good because it gives two sharp, well separated resonances) and phase on it. A background spectrum acquired immediately after will now have the correct phasing parameters relative to this spectrum (because phasing drift is negligible on my machine over a day or so). Subtract the background from the glycine spectrum and adjust the fine tuning on this corrected spectrum. Phasing can also be difficult for Bloch decay spectra of samples, especially if they do not contain sharp resonances. Again, acquisition of a glycine spectrum immediately before or after is the best way to get the correct phasing parameters.
Paramagnetic species
The presence of paramagnetic species is probably the number one cause of signal loss in solid state 13C NMR spectra of NOM samples. Paramagnetic species affect NMR spectra via three mechanisms (Smernik and Oades, 2000a; Smernik and Oades, 2000e):- (i) loss of field homogeneity, (ii) increased T1r H relaxation rates and (iii) direct interaction with 13C or coupled 1H nuclei. The consequences of these three mechanisms to NMR quantitation are quite different.
Loss of field homogeneity decreases the signal of all 13C nuclei in a sample (and background signal) equally. Thus it has no effect on relative quantitation. Proportional signal loss is the same for both CP and Bloch decay (DP) spectra via this mechanism. Iron and manganese species appear to be the worst offenders; the presence of 8.1% Mn2+, present as a salt in a physical mixture with cellulose resulted in the loss of 60% of the cellulose 13C signal intensity, whereas the presence of 9.8% Cu2+ did not affect cellulose observability (Smernik and Oades, 2000a).
It is principally via this mechanism (loss of field homogeneity) that paramagnetic (and ferromagnetic) minerals, present as discrete particles, interfere with solid state 13C NMR spectra of whole soils and sediments. These particles can be removed by treatment with dilute HF, using a method such as that described by Skjemstad et al. (1994). The advantages of HF treatment are two-fold. Firstly, the removal of most of the mineral matter concentrates the carbon in the residue, and secondly, the removal of the interfering paramagnetic minerals improves the NMR observability. For example, HF treatment of a whole soil increased the carbon content from 7.20% to 41.75% and its CP observability from 33% to 75% (Smernik and Oades, 2000a). This represents a 13-fold increase in 13C signal per unit mass or about a 6-fold increase in 13C signal for a full rotor, taking into account the lower density of the HF treated material. To obtain a spectra with equivalent S/N ratios would require a 36-fold increase in acquisition time for the whole soil as compared to the HF treated soil. One potential problem with HF treatment is the possibility that organic carbon is lost during the process and that this loss may be selective. Skjemstad et al. (1994) reported recoveries of 85-92% for five surface soils, suggesting that this was not a significant problem for these soils (especially considering that most of the lost carbon was probably due to transfer manipulations and thus non-selective). However, significant carbon loss has been reported on HF treatment for soils with a high proportion of organo-mineral complexes (Dai and Johnson, 1999).
The second mechanism via which paramagnetic species cause loss of NMR signal, decreases in T1r H, only affects CP spectra. Decreases in bulk T1r H can be corrected for in spin counting experiments. However, it would appear that there is some signal loss caused by extremely short T1r H values in the immediate vicinity of paramagnetic centres. These extremely short T1r H values are not fully transferred to the bulk of the sample via spin diffusion. Signal loss via this mechanism is sensitive to lower paramagnetic concentrations, and to a wider range of paramagnetic species. For example, cation exchange of pectin with Cu2+ (11.9% Cu2+) resulted in loss of 81% of the pectin 13C NMR signal (Smernik and Oades, 2000c).
The third mechanism via which paramagnetic species cause loss of NMR signal, direct interaction with 13C or coupled 1H nuclei, affects both CP and Bloch decay spectra. Coupling between 13C nuclear and electronic spins can result in extreme broadening and/or shifting of the 13C NMR resonance. Coupling of 1H nuclear spins to electronic spins can also shift the 1H resonance out of the chemical shift range of efficient 1H decoupling, and hence can indirectly affect the observability of 13C nuclei. These effects are confined to 13C nuclei within a few nm of the paramagnetic centre.
Remote protonation:
Remote protonation, i.e. the lack of a close 1H neighbour is probably the second largest quantitation problem in solid state 13C NMR spectroscopy of NOM. Remote protonation is only a problem for CP spectra; Bloch decay spectra are unaffected. Unfortunately, not much is known on how close neighbouring protons are required to be for efficient cross polarisation, nor how rapidly observability drops with increasing distances to nearest proton neighbours. I am currently working on this problem using model compounds, and will post results on this page as they become available.
High degrees of molecular motion
A third possible cause of NMR signal loss, and one which is not widely recognised, is high molecular mobility. Again, this problem is restricted to the CP experiment. Cross polarisation is very inefficient for liquid or solution samples because rapid tumbling nullifies C-H dipolar coupling; indeed, this is one of the reasons why solution state NMR has so much better resolution than solid state NMR. Therefore, liquid or solution samples give no signal in a CP spectrum. Some restricted molecular motion is apparent in solid state dipolar dephasing experiments and in the build-up curves of VCT experiments. So what happens between these two regimes of molecular motion? How much molecular mobility is required for signal loss to occur through inefficient CP? Comparison of CP and Bloch decay spectra in the alkyl region of soil organic matter samples suggests that signal loss via this mechanism may be important in NOM samples (Smernik and Oades, 2000b). This is another problem which I am working on and will update here as it progresses.
Spin counting
"Spin counting" is the term used to describe calibration of NMR signal intensity against that of a standard material. Spin counting is the best way to gauge how quantitative an NMR spectrum is likely to be. There are many possible variations on spin counting experiments. The methodology described here is the one I use and have found both reliable and practical for use with NOM samples.
The first choice to be made is the reference material. I use an external sample of glycine. The choice of an external reference is based on practicality; external references are much easier to work with. Internal references either have to be mixed with the sample, making both sample and reference unrecoverable, and risking reaction between the sample and the reference, or the reference can be held in a sealed partition which may hinder packing or spinning and allowances must be made for differences in sensitivity throughout the rotor. Another problem with internal references is separating signal derived from the reference and sample.
The main drawback of external references is that the reference and sample spectra are obtained at different times and under different conditions. As far as changes with time go, I have found that the sensitivity of my spectrometer is quite stable over the period of months, with the intensity of the glycine spectrum stable within a range of ±5%. The main factor that will result in internal and external references giving different results is loss of field homogeneity due to the presence of paramagnetic and ferromagnetic materials. This is only one of three effects that paramagnetic materials have NMR spectra of NOM. Loss of field homogeneity causes non-selective signal loss, i.e. it reduces the observability of all 13C nuclei in a rotor. Spin counting with an external reference will indicate both selective and non-selective signal loss, whereas spin counting with an internal reference will indicate only selective signal loss.
Glycine is an ideal standard because it is cheap, stable, readily available and gives rise to a simple NMR spectrum. Glycine is characterised by a long T1r H value (26 ms on my spectrometer, but check this on your own), which ensures minimal signal loss at a 1 ms contact time, and a short T1H (120 ms on my spectrometer, but check this on your own), which allows a relatively short recycle delay and hence a high repetition rate. It is easy to obtain a good CP spectrum of glycine with low S/N ratio in a short time (~1 hour) which can safely assumed to be quantitative. Obtaining a Bloch decay (DP) spectrum for glycine requires a lot longer acquisition time. I use a spectrum obtained with 1000 scans and a recycle delay of 90 s (total acquisition time ~ 25 hours). Remember that the Bloch decay spectrum requires correction for background signal.
Careful attention should be paid to phasing and baseline correction (I use a linear correction between -100 ppm and 300 ppm). The integral limits should be the same as those used for the sample spectra. I integrate between -10 ppm and 300 ppm, to as to include SSB’s for the carbonyl and aromatic resonances. Acquisition and processing parameters which affect integral scaling need to be recorded. With Varian software, the integral scaling is independent of the number of scans, but may be influenced by parameters is, ins and insref, depending on the software version. Check which parameters influence integral scaling on your machine and either ensure they are the same for sample and standard or include them in your calculations.
The sample spectrum should be acquired so as to obtain the best quantitation possible, as described above. Bloch decay spectra should be corrected for background signal. Background correction for CP spectra is also necessary for low carbon samples (say, <10 mg C).
CP signal intensities need to be corrected for differences in T1r H between the sample and the standard. The need for this correction is easily seen from Figures 1-3. As explained above, a 1 ms contact time provides the best relative quantitation between most protonated and non-protonated carbons for most NOM samples. However, the intensity of the spectrum is still strongly dependent on T1r H (compare 1 ms intensities between Figures 1-3). This dependence is dominated by the exp(-t/ T1r H) term of Equations 1 and 2 (the product of the remaining terms is close to unity for TCH and T1r H values observed in most NOM samples, see Smernik and Oades, 2000a). Thus the intensity of each spectrum (sample and reference) needs to be scaled up by a factor of 1/exp(-1/T1r H). This scaling factor ranges from 1.04 for glycine (T1r H = 26 ms), through 1.11 for cellulose (T1r H = 10 ms) and is the range 1.22-1.40 for typical NOM samples (T1r H = 3-5 ms).
The final spin counting correction factor that needs to be considered involves samples which do not completely fill the rotor. On my spectrometer, sensitivity is inhomogeneous through the volume of the rotor. Sensitivity is greatest in the middle third of the rotor which contributes 49% of the signal from a full rotor. The top third and bottom third of the rotor contribute 31% and 20% of the signal, respectively. For samples for which there is not enough material to fill a rotor, ceramic inserts are used. We have three sizes and either use the same size top and bottom, or have the smaller insert at the top so as to maximise signal. Table 1 shows the relative sensitivity for each possible combination of rotors that may be used.
Table 1 Relative sensitivity for different combinations of inserts.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
near future.
Dai, K.H. and Johnson, C.E. (1999). Applicability of solid-state C-13 CP/MAS NMR analysis in Spodosols: chemical removal of magnetic materials. Geoderma. 93, 289-310.
Newman RH, Condron LM (1995) Separating subspectra from cross-polarization magic angle spinning nuclear magnetic resonance spectra by proton spin relaxation editing. Solid State Nuclear Magnetic Resonance 4, 259-266.
Newman RH, Hemmingson JA (1990) Determination of the degree of cellulose crystallinity in wood by carbon-13 nuclear magnetic resonance spectroscopy. Holzforschung 44, 352-355.
Skjemstad JO, Clarke P, Taylor JA, Oades JM, Newman RH (1994) The removal of magnetic materials from surface soils. A solid state 13C CP/MAS nmr study. Australian Journal of Soil Research 32, 1215-1229.
Smernik, R.J. and Oades, J.M. (2000a). The use of spin counting for determining quantitation in 13C NMR spectra of natural organic matter. 1. Model systems and the effects of paramagnetic impurities. Geoderma 96, 101-129.
Smernik, R.J. and Oades, J.M. (2000b). The use of spin counting for determining quantitation in solid state 13C NMR spectra of natural organic matter. 2. HF treated soil fractions. Geoderma 96, 159-171.
Smernik, R.J., Skjemstad, J.O. and Oades, J.M. (2000). Virtual fractionation of charcoal from soil organic matter using solid state 13C NMR spectral editing. Australian Journal of Soil Research 38, 665-684.
Smernik, R.J. and Oades, J.M. (2000c). Effects of paramagnetic cations on solid state 13C nuclear magnetic resonance spectra of natural organic materials. Communications in Soil Science and Plant Analysis, 31, 3011-3026.
Smernik, R.J. and Oades, J.M. (2000d). The use of solid state 13C NMR dipolar dephasing experiments for characterising soil organic matter. European Journal of Soil Science, in press.
Smernik, R.J. and Oades, J.M (2000e). Paramagnetic effects on solid state 13C NMR spectra - quantification and their use in soil organic matter characterisation. Submitted to Journal of Environmental Quality, May 2000.
Download workbook instructions (Word file, copy of instructions and examples sections below)
Instructions and tips for using spreadsheets
These spreadsheets have been provided to aid analysis of solid state NMR data on natural organic matter samples. Each worksheet is designed for analysis of a single NMR experiment and generally does not rely on data from the other sheets, e.g. T1H and PSRE analysis can be performed without the cross polarisation or Bloch decay pages being used.
Input cells are of two types. Orange shaded cells are not required for the analyses. They are for your records only and contain information such as sample and file names and some acquisition details. Green shaded cells require input for the analyses to work.
Beware of modifying these sheets, since removing or adding cells may affect the referencing within or between sheets! Some sample data is provided below, along with the expected output.
(i) Blanks and glycine sheet:
Required input (green shaded cells):
Integral scale (C4)
Raw data (C14-C25) - integrals for the given regions
Partially filled rotor data (F12-I18)
Output:
Background integral areas corrected for integral scale (D14-D25)
Correction factors for partially filled rotors (J12-J18)
Notes and tips:
This worksheet provides data for corrections used in the cross polarisation sheet. The left hand side is for end-cap and rotor background corrections, which become important for samples containing little observable C (say <10 mg). If no data is recorded here, no correction will be made on the cross polarisation worksheet (but the rest of the analysis will still work). The right hand side of this page calculates relative sensitivities for partially filled rotors. Sample data shows that on our set-up the centre of the rotor is considerably more sensitive than the ends. If you use partially filled rotors, you must perform these experiments and fill in the details in this area. This correction factor is required for spin counting when comparing signal from a partially filled rotor of sample with that of a full rotor of glycine. Unless these factors have been calculated, spin counting on partially filled rotors requires running a spectrum of glycine filled to the same extent as for the sample. The cross polarisation and Bloch decay worksheets require an entry exactly matched to an entry in the Top/Bottom column for observabilities to be determined.
(ii) Cross polarisation sheet:
Required input (green shaded cells):
Organic C content (C7, required for spin counting)
Mass of sample in rotor (C8, required for spin counting)
Integral scale (C10)
Inserts (C15, required for spin counting, must match an entry from "Blanks and Glycine" worksheet F12-F18)
Raw data (C21-C32) - integrals for the given regions
T1r H (estimated) (O7) or T1r H (measured) (O8) (required for spin counting)
Mass of glycine (O16, required for spin counting)
Signal intensity for full rotor of glycine (O19, required for spin counting)
Integral scale for glycine spectrum (O20, required for spin counting)
Output:
Relative intensity of integral regions corrected for SSB’s and background signal (G22-G29)
Percentage of sample signal due to background (rotor and endcaps) (A37)
CP observability of sample carbon (O26)
Notes and tips:
The two functions of this worksheet - determining corrected integral regions and determining CP observability (spin counting) - are essentially independent. If spin counting is not required, then a number of the inputs can be ignored (as listed above). If only spin counting is required, simply enter the total intensity in one cell of the raw data column (other than the top one -50 to 0 ppm, which is discarded). The integral region scheme used here (and in the "Bloch decay" and "Blanks & Glycine" sheets) has been developed for a 200 MHz spectrometer and a 5 kHz MAS rate, where spinning side-bands (SSB'’) are therefore 100 ppm from the central band. A different scheme would need to be used for other spectrometer set-ups.
(iii) Bloch decay sheet:
Required input (green shaded cells):
Organic C content (C8, required for spin counting)
Mass of sample in rotor (C9, required for spin counting)
Integral scale (C11)
Inserts (C16, required for spin counting, must match an entry from "Blanks and Glycine" worksheet F12-F18)
Raw data (C22-C33) - integrals for the given regions (for back-ground corrected spectrum)
Mass of glycine (P11, required for spin counting)
Signal intensity for full rotor of glycine (P14, required for spin counting)
Integral scale for glycine spectrum (P15, required for spin counting)
Output:
Relative intensity of integral regions corrected for SSB’s and background signal (F23-F30)
Difference between CP and BD corrected integral regions (H23-H30)
BD observability of sample carbon (P18)
Notes and tips:
This analysis requires input data from Bloch decay spectra (both sample and glycine) which have already been corrected for end-cap signal intensity as described above.
As for the CP worksheet, the two functions of this worksheet - determining corrected integral regions and determining CP observability (spin counting) - are essentially independent. If spin counting is not required, then a number of the inputs can be ignored (as listed above). If only spin counting is required, simply enter the total intensity in one cell of the raw data column (other than the top one -50 to 0 ppm, which is discarded).
Obviously, comparison between CP and BD integrals requires completion of the CP worksheet.
(iv) T1r H sheet:
Required input (green shaded cells):
Contact times (A11-A26) and corresponding total signal intensities (B11-B26) from variable contact time experiments
Subset of data points to be used (prompted when "perform T1r H regression analysis" button is clicked)
Output:
Calculated T1r H (G10), upper and lower bounds of T1r H (G11, G12)
I(0) (G15, extrapolation of linear fit to zero contact time)
r2 value of linear fit (G19), P-value of linear fit (G22)
Plot of signal intensity vs contact time
Notes and tips:
The relationship between signal intensity and contact time is complex, with an early build-up phase followed by a decay phase. The determination of T1r H from variable contact time (VCT) experiments involves finding the slope of the linear part of the ln(intensity) vs contact time plot in the signal decay phase. Choosing the appropriate section of the plot is best done graphically. The best way to achieve this using this spreadsheet is to enter all of the VCT data and perform the analysis on the whole data set. This will produce a ln(intensity) vs contact time plot, from which the linear decay phase can be seen. A second iteration (click the "perform T1r H regression analysis" button again) on this subset of the data allows an accurate determination of T1r H. In choosing this subset, you may need to exclude early data points (short contact times) where signal has not yet built up, and/or late data points (long contact times) for which signal-to-noise is poor and the associated errors are therefore large (and are exacerbated by the log transform).
(v) T1H and PSRE sheet:
Required input (green shaded cells):
recovery delays (A9-A33) and corresponding total signal intensities (B9-B33) from inversion-recovery experiments
"Null time" of experiment i.e. when total signal intensity is closest to zero (prompted when "Perform a One Component T1H Analysis" button is clicked)
Output:
Calculated one component T1H (H17), Iinf (H16, extrapolation of fit to infinite recovery delay), H value (H15, "heterogeneity factor" which describes how efficient the inversion pulse was; this parameter is required for the fit but conveys no information about the properties of the sample)
Calculated two component T1H’s (fast relaxing component AB19, slow relaxing component AA19), Iinf’s (fast relaxing component AB18, slow relaxing component AA18, which convey the relative quantities of the two components), H value (AA17)
Statistical comparison of the one- and two-component fits (F-ratio AA23, P-value AA24)
The delay times which should be used for PSRE analysis (XX and XX) in order to maximise S/N ratios in the PSRE subspectra, and the fractions of the these inversion recovery spectra which are required to generate PSRE spectra (Z33-AC33)
Plots comparing one- and two-component fits to the experimental data
Notes and tips:
Differences between one- and two-component fits are most visible around the null point of the inversion-recovery data. It is therefore important to concentrate data in this region where the biggest changes occur. A data set containing 12-15 recovery delays usually provides the best compromise between number of data points and S/N ratios of each individual inversion-recovery spectrum.
The one component T1H fit is equivalent to that which is standard on most NMR software and involves fitting 3 parameters (T1H, Iinf and H) by minimising least squares differences between experimental and fitted intensities using the excel "solver" routine. The two component fit uses 5 parameters (T1Hslow, T1Hfast, Iinf,slow, Iinf,fast and H) in a similar minimisation of least squares differences between experimental and fitted intensities. The workings of this fitting procedure have been left on show in the spreadsheet, so that anyone can see how it works. The initial estimates of two-component values are derived from the best-fit one-component values; H is initially set to the best-fit one-component H value, T1Hslow and T1Hfast are set to twice and half the best-fit one-component T1H value, respectively, and Iinf,slow and Iinf,fast are both set at half the best-fit one-component Iinf value. The "solver" routine is then used to find the best fit to this model. Usually (but not always) the two-component model will result in a better fit (smaller RMS error). However, since the two-component fit involves two extra parameters, we need to check whether the decrease in RMS error justifies the extra parameters. This is done using an F-ratio test, which basically compares the improvement in fit per extra parameter, to the residual error per degree of freedom in the 2-component fit. A P-value of <0.05 indicates that the two-component fit is statistically better than the one-component fit. If, and only if, the P-value is <0.05, is PSRE analysis justified.
For samples where PSRE analysis is justified, the component inversion-recovery spectra and the proportions of these components required to generate the fast- and slow-relaxing PSRE subspectra are identified in the bottom right hand corner of the worksheet.
(vi) PSRE Slow&Fast sheet:
Required input (green shaded cells):
integral scales for slow (C8) and fast (K8) relaxing subspectra
Raw data for slow (C15-C26) and fast (K15-K26) relaxing subspectra - integrals for the given regions
Output:
Relative intensity of integral regions corrected for SSB’s for slow (G16-C23) and fast (O16-O23) relaxing subspectra
Notes and tips:
This worksheet duplicates the SSB correction used in the "Cross Polarisation" worksheet and allows side-by-side comparison of functional group distributions between slow and fast relaxing PSRE subspectra. No correction is made for background signal because this signal will be partitioned between the two subspectra, depending on the respective T1H values. PSRE analysis is rarely used for low carbon samples where background signal is a problem in CP spectra, simply due to the large amount of spectrometer time that would be required for a low carbon sample to carry out an inversion-recovery experiment with sufficient signal-to-noise for PSRE analysis.
Sample data and output
Sample data is provided for 3 samples, cellulose, Aldrich humic acid and SS82UV, the residue of a de-ashed soil sample after photo-oxidation in Tables 3, 5, 7, 9 and 11. The expected output is presented in Tables 4, 6, 8, 10 and 12, so that you can check that the spreadsheets are working properly. These samples highlight different aspects of the spreadsheet analyses.
The results for cellulose show that only 0.9% of CP signal is derived from the background (end-caps etc.), and that CP observability is 101%. The distribution of signal is dominated by the O-alkyl and di-O-alkyl regions, as expected. Bloch decay observability is 100% and the distribution of signal intensity is very similar to that of the CP spectrum. Input data for T1r H analysis includes the signal build-up region, as well as the signal decay region. The best way to handle such a data set is to first include the whole data set and look at the graph generated. Then re-run the analysis including only the linear decay phase. For this sample, the contact time range of 1-12 ms was used (A18-A26, C18-C26), which gave a T1r H of 9.60 ms, and a confidence range of 8.74-10.66 ms. The rest of the T1r H output data is of little interest. The one-component T1H value for cellulose was determined to be 2.0492 s. The two-component T1H analysis results in near identical slow and fast T1H relaxation rates, and an undefined P-value, because there is no improvement in the RMS error. This indicates that the cellulose sample has a single, homogeneous, T1H value, and therefore the two-component T1H analysis is invalid. PSRE cannot be performed on such a sample.
The results for Aldrich humic acid show that only 2.5% of CP signal is derived from the background. CP observability is low at 31%. The alkyl and aromatic regions dominate the distribution of signal. Bloch decay observability is 89% and the distribution of signal intensity is very different to that of the CP spectrum, with the BD spectrum dominated by the aromatic region, and much greater contributions from the phenolic and carbonyl regions. Again, the input data for T1r H analysis includes the signal build-up region as well as the signal decay region. Spectra for the two longest contact times were very noisy, so the final T1r H analysis was run on only the contact time range of 1-6 ms (A17-A24, C17-C24), which gave a T1r H of 3.66 ms, and a confidence range of 3.55-3.78 ms. The one-component T1H value for Aldrich humic acid was determined to be 12.4 ms. The two-component T1H analysis resulted in T1H values of 103.2 ms and 8.8 ms. Comparison of the two graphs suggests that the two-component fit is superior. This is confirmed by statistical analysis, with the P-value being < 0.05 (0.0023). PSRE is therefore valid on this sample. The PSRE subspectra are best generated from linear combinations of the 1 s and 0.02 s recovery delay spectra, in the indicated proportions. The PSRE subspectra so generated are very different to each other; the slowly relaxing component being predominantly alkyl, whilst the rapidly relaxing component is much more aromatic.
The results for SS82UV show that 2.5% of CP signal is derived from the background and that CP observability is 49%. Bloch decay observability is 90% and the distribution of signal intensity is again quite different to that of the CP spectrum, with the BD spectrum containing more aromatic and alkyl signal, but less O-alkyl, phenolic and carbonyl signal. T1r H analysis was run on only the contact time range of 3-8 ms (A13-A18, C13-C18), which gave a T1r H of 5.49 ms, and a confidence range of 5.02-6.07 ms. The one-component T1H value for SS82UV was determined to be 14.5 ms. The two-component T1H analysis resulted in T1H values of 95.3 ms and 11.7 ms. Comparison of the two graphs suggests that the two-component fit is superior. This is confirmed by statistical analysis, with the P-value being < 0.05 (0.0144). PSRE is therefore valid on this sample. The PSRE subspectra are best generated from linear combinations of the 1 s and 0.02 s recovery delay spectra, in the indicated proportions. The PSRE subspectra for this sample can be seen in Smernik et al., 2000.
Table 3 CP sample input data
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table 4 CP sample output
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table 5 Bloch decay sample input data
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table 6 Bloch decay sample output
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table 7 T1r H sample input data
|
|
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
||
|
|
|
|
|
||
|
|
|
|
|
||
|
|
|
|
|
||
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table 8 T1r H sample output
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table 9 T1H sample input data
|
|
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table 10 T1H sample output
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table 11 PSRE fast & slow sample input data
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table 12 PSRE fast & slow sample output
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|