M. Vazquez: Just because you can get a good X-ray result on a crystal does not mean that it is 'pure', or that it will provide good microanalytical analyses. You can coat a crystal with Vaseline and it may still X-ray satisfactorily, as X-rays will only observe the repeating crystal pattern. This technique can be used for the X-ray of air-sensitive crystals.

I believe that microanalyses should be required for a new compound, unless you can plead an excellent case for its absence. However, microanalysis will NOT detect impurities if they are stereoisomers, so additional criteria of purity should also be provided. Mass spec and high-res mass spec will also fail to separate stereoisomers, and a correct high resolution peak simply means that your compound might be present, but tells you nothing about the abundance of its presence, nor the purity of the sample. A mass spec with no peaks above your required mass (or M+1) suggests that you possibly have no impurities with higher MW than your compound, but again tells you very little about impurities with lower MW. weight. GC input into a mass spec should separate out impurities, but again the observed mass spec will tell you nothing about non-volatile material in your original sample.

10% error in a micoanalysis suggests your compound may not be what you think it is. Does the observed microanalysis correspond to any other sensible molecular formula? Is it high in hydrogen and oxygen (by difference?) than you wanted - water impurity? Is it lower in carbon, hydrogen and oxygen than you expected - inorganic contaminent, or very volatile compound. Does a repeat analysis give the same result? What is the result after you have recrystallized or re-chromatographed the compound. Is your compound so volatile that mass is lost after weighing? If so, does the analyst know that they must crimp the boat carefully immediately after weighing? Can the analyst burn a similar known pure compound as a standard immediately before they burn your sample?

Some compounds do not burn very well - which might then provide broader g.c. peaks for CO2 and H2O (assuming your microanalytical analysis is by g.c. peak area). The software might need adjusting to correctly integrate the peaks. Tell your microanalyst of your problems and see what they can do. Some compounds behave surprisingly - for instance, steroids with an angular (tertiary) methyl group may analyze low in C and H because the methyl group cracks out as methane and this is so volatile that it can be swept unburnt (!!) through the system. Your analyst should be able to deal with this if you/they suspect this is the problem (won't account for a 10% error though!).

Good luck!

It may be a matter of solvent of crystallization. In which solvent did you crystallize your compound? For example if it is water, then it may come along with signal of residual water present in the deuterated solvent. Do thermogravimetric analysis of your sample as it will help in identifying whether you have anything else apart from your compound.

The accepted deviation is +/- 0.4% so 10% is not acceptable

agreed with Vicente Martí-Centelles and Sagar Sharma

10 % it's a lot, but it doesn't mean you cnnot find a good explaination. It always depends on the molecule, if it's big enough to be microporous, the crystallisation solvent could be kept very tightly. Or if it contains very polar groups which can produce strong hydrogen bonding with the molecule and the solvent of crystallisation, sometimes it's very difficult to remove all the solvent, even with the high temperature required by elemental analysis

If you have crystals, you can then get the monocrystal x-ray diffraction and solve the structure, then hopefully you will be able to locate the solvent (some times the solvent is disordered, and therefore it can not be clearly located).

Another option is to perform a HPLC-MS analysis of your sample

If the HRMS (high resolution mass) is OK then you can think of solvent of crystallization as a culprit or simply try changing place for analysis......it might work better at some other hands!

Timur: you don't state if your error is reproduced. Have you submitted a second independent sample and obtained the same 10% deviation?

If the error is consistent and you suspect solvent co-crystallised with the sample then look for that solvent in the 1H and 13C NMR spectra. It should show up and the integration will give you the ratio. Your MA may then fit for this.

Alternatively, if your compound is a high nitrogen containing compound then you may need to talk to your combustion analysis technician and explain this. There are special techniques for such high nitrogen materials.

Otherwise, your compound is contaminated. Is your C% very much lower or higher? Could you have a salt of your compound? Could there be a hair, glass or silica in your sample? Did you ensure your glassware including the vial was clean and did you perform a hot filtration during your recrystallisation? All these points should be reviewed.

If your compound is pure by HPLC and NMR, check for the following parameters

1. Inorganic salts

2. Water of hydration

3. Residual solvents / % of impurities

4. Nature of solid, if it is moisture sensitive

You can remove the inorganic salts and other fiber materials by dissolving the compound with Acetonitrile and filter through wattmann paper.

If your compound contains a great amount of carbon the combustion may not be complete. We once had this problem, we mixed some amounts of oxidizer in it and got correct results - ask your technician about it!

By the way crystals are not perfect for EA, grind it and dry your EA sample.

Ten percent is much too high. I can only see this as being possible if, as mentionned below, the crystallization solvent is incorporated into your product crystals and this high a percentage would only be possible for a very small product molecule and relatively high molecular weight solvent. Could you have possibly crystallized a salt form? (something like a hydrochloride salt of an amine or Na salt of a carboxylic acid - this would lower your C and H content).

10% thats a lot, in your NMR, you are 100% sure that you do not have any residual peaks (solvents). To look carefully at the purity of your material I would recommand quantitative NMR, and you will know!

HPLC/Xray would be great but it depends I guess if you have access to these gears.

I would like to thank you all for your helpful comments. I am sorry I should have made the question in detail. The material has molecular formula of C(42)H(40)O(4) with a Mw of ~609 g/mol and I have 4 symmetric acetyl group attached to benzenes. I have crystallized it over EtOH/CHCl3 mixture, and there is no peak in NMR for EtOH. But since I took NMR in CHCl3 there still can be CHCl3 residue crystallized with material. I think I should better add CHCl3 to structure to try to meet extra peaks in the mass spec because I have some high mw peaks since CHCl3 low C and H content material compared to my material this might lead to deviation. By the way I have repeated the two independent samples for EA and the results were unfortunately consistent.

Thank you all again.

Firstly, the elemental analysis is usual pain for all chemists, and as most peer reviewed journals only allows ± 0.3 it is really difficult especially with fluctuation in the elemental results from different. To resolve this issue as you have been told before, do HRMS which can replace the elemental analysis. The x-ray diffraction analysis is great, but I think you will find even harder than the elemental issue, specially in obtaining the suitable crystal dimensions.

Timur, OK then take a sample and place it in a vacuum over at 90 oC for 48 h or so. This should drive off any co-crystallised solvent and then rerun the EA. Also consider running your NMR data in CD2Cl2 or acetone-d6 to spot the CHCl3 peak if you suspect that's the contaminant. Also recrystallise from another solvent system.

0.4% deviation maximum for elemental analysis is accepted. You need to observe carrefully your NMR spectra in order to find water or the solvent used for your recristallization. In that case, It is possible to add some solvent molecule. with the solvent , you must find no more than 0.4% deviation.

Christine maybe right but I lean toward Panayiotis' answers. Put the stuff in your vacuum oven and flash the rest of the solvent off. Recrystallize from a different solvent and flash that off in the oven. When I read 10% I thought you might have an amine as an organic acid salt and by itself. That would account for a 10% deviation.

Besides the Solvent problem, avoid any contact with paper. It is better to use craig tube and a centrifuge.

This is a common problem of carbon, hydrogen and nitrogen analysis. Please calibrate the instrument with a standard sample again. If problem persists, this is a problem of sample preparation. Grind the sample to powder and dry it in high vacuum. Add measured amount of oxidant to the sample for pure hydrocarbon sample.

Elemental analysis requires the sample to be pure enough, so it is quite difficult to obtain that pure product. Usually chromatography separation and several times recrystallization procedures are employed. Fewer examples are noticed in recent articles. Why is the elemental analysis needed besides NMR and MS?

To Jianfeng: EA is one of the only techniques that actually determines bulk purity which is a requirement for many journals. Many other spectroscopic techniques require solubilization of the sample and impurities can therefore be removed by filtration prior to analysis. Some impurities also may be invisible to the technique. As such, data collected using these techniques do not really reflect the bulk purity. I could for example run a clean NMR and MS on a sample contaminated with broken glass, but this sample will fail the EA.

Of course EA will not determine whether you have isomers and mixtures of compounds with similar CHN% as these can come back within the required 0.3% limits of the journal, however, in combination with the other techniques EA is very useful and powerful tool.

Its better to dry sample under vacuum as its not easy to remove EtOH from the sample. clean NMR does not guarantee the absence of solvent of recrystallization, I think up to 5% impurity is not easy to detect in NMR. You must have to look the instrumental error (how much error the instrument may have) and then determine the observed error? I guess this may lead to improved results. HRMS is a better substitute.

Although I generally agree with Panayiotis, some other thing should be considered. If your compound is small (I mean low molecular weight) the presence of a tiny amount of solvent will cause a bigger deviation on the results of EA. As Panayiotis said, try to heat in a vacuum oven, but remember that heating organic compounds is always risky, because it may produce decomposition in some extent and this will produce poorer results in the EA. Moreover, it is not always easy to get rid of crystallization solvent by heating even in vacuum. I had the same problem several years ago working with a steroid (perfect NMR, a narrow melting point interval, and poor EA) until X-ray reveled that what I had was a hemi-solvate, this means two host molecules per molecule of ethyl acetate in the asymmetric unit. (see J Chem Cryst 2000, 30(11), 693-695, I can send you the paper if it is not available to you) If the solvent is water, NMR may be unable to "see" it. My advice go to GC-MS, this will give you a hint. If it is solvent, you may follow Panayiotis and heat your compound in a vacuum oven few (aprox 20) degrees below the melting point (provided you have observed that you can melt you compound leave it to solidify and on re-melting it you get the same melting point, but this is only possible with very stable compounds) . Other? Change crystallization solvent, consider your structure and select a solvent that have low affinity for your compound. That means to avoid solvents that can form hydrogen bonds with your compound. At the end my advice is keep trying…crystallization is sometimes tricky and….after all life is not always easy. Good luck!!

In addition to all those considerations, there's always the issue with the technicians and the quality of the analysis service. We've had MANY problems with our facilities using dirty instruments and elemental analysis way off, sometimes even with crystals suitable for X-ray!.

Basically, we do not bother anymore with elemental analysis. We do HRMS and the usual H/C-NMR/MS/chromatography set and we've never had any complaints from the referees.

If you REALLY need to do EA, follow some of the good advice above, and good luck.

One question, what MS did you use? I run many samples on an AGILENT LC-TOF instrument (small molecules) and knowing or suggesting the elements in them (C, O, N, H, S....) the MS spectrum was easy translated into the correct elemental composition. I run known samples (known to me) and samples from another group and they checked what I got as elemental composition and far over 90 % was correct, if not the sample was already degraded. If you have access to an instrument with higher resolution (FT-MS, Orbis) it may be even easier. Then you have two results, from your analysis and the MS-based "suggestion" and you can see how they fit.

If your compound has been purified by colun chromatography, you may silica (very fine particles) in your crystals. Then you must dissolve, centrifuge, take the supernant and evaporate the solvent.

Be sure that your equipment is completely calibrated or run a calibration curve for it. Some of the elemental analyzers need to operate using phosphorous pentoxide to avoid hydration of a given sample. Hope this may help you.

Either your sample is impure or it got impurified or there has been an error duting the analysis.

M. Vazquez: Just because you can get a good X-ray result on a crystal does not mean that it is 'pure', or that it will provide good microanalytical analyses. You can coat a crystal with Vaseline and it may still X-ray satisfactorily, as X-rays will only observe the repeating crystal pattern. This technique can be used for the X-ray of air-sensitive crystals.

I believe that microanalyses should be required for a new compound, unless you can plead an excellent case for its absence. However, microanalysis will NOT detect impurities if they are stereoisomers, so additional criteria of purity should also be provided. Mass spec and high-res mass spec will also fail to separate stereoisomers, and a correct high resolution peak simply means that your compound might be present, but tells you nothing about the abundance of its presence, nor the purity of the sample. A mass spec with no peaks above your required mass (or M+1) suggests that you possibly have no impurities with higher MW than your compound, but again tells you very little about impurities with lower MW. weight. GC input into a mass spec should separate out impurities, but again the observed mass spec will tell you nothing about non-volatile material in your original sample.

10% error in a micoanalysis suggests your compound may not be what you think it is. Does the observed microanalysis correspond to any other sensible molecular formula? Is it high in hydrogen and oxygen (by difference?) than you wanted - water impurity? Is it lower in carbon, hydrogen and oxygen than you expected - inorganic contaminent, or very volatile compound. Does a repeat analysis give the same result? What is the result after you have recrystallized or re-chromatographed the compound. Is your compound so volatile that mass is lost after weighing? If so, does the analyst know that they must crimp the boat carefully immediately after weighing? Can the analyst burn a similar known pure compound as a standard immediately before they burn your sample?

Some compounds do not burn very well - which might then provide broader g.c. peaks for CO2 and H2O (assuming your microanalytical analysis is by g.c. peak area). The software might need adjusting to correctly integrate the peaks. Tell your microanalyst of your problems and see what they can do. Some compounds behave surprisingly - for instance, steroids with an angular (tertiary) methyl group may analyze low in C and H because the methyl group cracks out as methane and this is so volatile that it can be swept unburnt (!!) through the system. Your analyst should be able to deal with this if you/they suspect this is the problem (won't account for a 10% error though!).

Good luck!

No, but you can carry the exact mass determination in this case.

Some further thoughts, although these are only vaguely related to the original question.

It is always worthwhile to consider the limitations of any technique that you use. Elemental analysis can provide two pieces of information. It may provide an empirical formula if the sample is sufficiently pure, and it may provide a measure of purity for a 'known' compound. High resolution mass spectroscopy will possibly provide a molecular formula, but will give no information about the purity of the bulk sample. Both these techniques have trouble dealing with diastereomers. Hopefully you will get some information on the diastereoisomeric problem through nmr, but even here there can be great difficulties. Consider for example 3,17-dimethyleicosane (C22H46). The 3R,17R compound is a diastereomer of the 3R,17S isomer They are different compounds. However C3 is so remote from C17 that it probably cannot 'see' whether C17 is R or S inclined. Thus I predict that these two compounds will give the same elemental analyses, essentially the same mass spec (certainly the same high-res mass spec of the molecular ion), and that they will give superimposable 1H and 13C nmr spectra. They will almost certainly give the same retention on chromatgraphy,, but hopefully they might separate on a chiral g.c. column. But then how do you determine which peak is which? Stereo-selective synthesis of one isomer of known stereochemistry might be required.

Is this an extreme or trivial problem? Certainly not to an insect which might be using one of these two isomers as a sex attractant

So when you use a technique, also consider the times when it will not work because it is stretched beyond its limit.

The column of the EA machine might be saturated. an analysis of a routine commercial compound should let you know if that is the case.

Just because you obtain crystals does not necessarily mean that they are composed solely of your compound. Could it be a hydrate? dihydrate? Consider what recrystallization solvent you used.  

Can anyone have reference document justifying 0.4% variation in CHN content?

I have to respond to query, Please help me.

refer to any publication in any organic or medicinal journals to see this indicated limitation as more than this value could affect the number of hydrogen atoms in molecule.

run solvent through the column to see if it's contaminated

recrystallize the cmpd

is it a silicate mineral? these have average compositions...

There are several reasons to have such deviation, first you must check the accuracy of the elemental analysis equipment, once that you have calibrated it you are certain of that results. Other possibilities are that you have disordered solvent molecules within the crystal lattice and you have not resolved it or it is not possible to solve disorder in that crystal, and this is just because the disorder could be difficult to analyse or to have the correct model of disorder, what are your CFOM, R, Rw and general crystal parameters?? Please let me know these details in order to help you with that results. Best regards Hiram

Thank you for the help.

@ Nicholas Coffey I have to respond to query of Health authority so I think they will not accept reference to journals. Can you provide me reference of any book or any other Regulatory authority guidance document?

It may be water. I suppose that if you have other solvent as solvate you can verify it by running H1NMR of the crystallized material, the signal of the solvent must be there and you can even find the substrate/solvent relation by relative integration

Can anyone have reference document justifying 0.4% variation in CHN content?

I have to respond to query, Please share

You have to check if there is no water left. Watercan be a crystallization water and part of the molecule (CxHyNz, xH2O)

@Darshan: I have looked for such a document before, but it seems it is an upper limit agreed on by the community (i.e. the editorial boards of the journals.)

@pim thanks for the reply. I have responded same to Health authority. Hope they will accept same.

Dear All,

+/- 0.3-0.5% error margin in CHNS results have lot of confusions.

1) like 0.3-0.5% is an absolute error i.e Just difference in calc. and obs. values of the %C, %H, %N and %S

OR

2) 0.3-0.5% is the relative error

calculated wrt, cacl. and obs. % of C, H, N and S,

For example

If calculate %C=20.54 and Observed= 20.14, the difference 0.3 will be the permitted error based upon the difference between two values. if it is ok then the relative error for H, N, and S be much higher.

whether, this 0.3 is based on the difference value just or 0.3% is the relative error with respect to the calculated value, if so, then the resultant difference must be 0.003*20.54= 0.06 i.e. the observed value could only be differ from the calculated ( 20.54) by +/- 0.06.

For years the Australian Journal of Chemistry required