Numbers of Atoms: Finding the smallest traces with gas chromatography
We conclude (for now) our look at chromatography for separations and quantitative analysis by considering instruments that offer even more sensitivity and selectivity than the high-performance liquid chromatography (HPLC) methods described in the last post.
But you said that HPLC can give detection limits in the parts-per-billion range. That seems really low… Can we really do even better than that?
Yes we can, although we won’t get mired in little details of how to optimize HPLC. We will, however, take a bit of time to consider the detection system. Let me remind you that the detector gives a signal that should be proportional to the concentration of the compound in whatever mixture passes through the detector. Simple spectroscopic detectors shine a light of a pre-selected wavelength onto whatever is flowing through it, and different compounds may absorb some of this light at different rates.
Having the detector measure a signal at a single wavelength is fine if we are only interested in analyzing one compound, since the sensitivity can be optimized by selecting the wavelength that is best absorbed by our analyte. The figure to the right shows a spectrum of caffeine – this plot shows how light is absorbed, at different wavelengths, by a sample of caffeine. The wavelength of maximum absorption here is 276 nm. (The absolute absorbance value on the y-axis is irrelevant here, just the location of the highest value.)
If we were only analyzing caffeine by HPLC, we would set the detector to measure absorbance at 276 nm to get maximum sensitivity. Maximum sensitivity means we will be able to detect smaller changes in concentration. But when we analyze several compounds through a single chromatographic separation, we may need to compromise and select a wavelength that works well for all of them, but likely will not be the wavelength of maximum absorption for any of them.
Hmmm… I have an idea. What if we installed multiple detectors, with each one set to measure at a different wavelength?
That’s a great idea, and in fact we can do even better! A modern solution is to use a diode-array detector (DAD). Instead of a single wavelength of light going through the sample and reaching the detector, and light from a lamp that contains many wavelengths is sent through the sample. A prism or a grating then disperses the white light that is not absorbed onto the detector, which measures the leftover light at many different wavelengths at once. This approach allows measurements of hundreds of wavelengths simultaneously. A spectrum is generated for each measurement, just like the caffeine spectrum shown above; with HPLC, a new spectrum can be generated every second or even faster.
A spectrum for each measurement allows the user to independently select the best wavelength to analyze each analyte in the chromatogram. Collecting a spectrum at each point can also assist when several analytes co-elute; with a single signal per measurement it is impossible to quantify both analytes, since we cannot know how much of the signal is caused by each analyte. The spectra allows the user to mathematically distinguish between the analytes – this methodology was part of the Masters work of your humble guide, with a publication to boot.
Speaking of distinctions – doesn’t the detector see both the eluent and the analyte?
Yes, this is a significant handicap of HPLC, as it is certainly possible that the detector will capture both the eluent and the analyte, particularly if both are organic substances – they will tend to absorb light well in the ultraviolet (UV) region, between 200 and 400 nm. This can cause a significant background, and detecting the analyte becomes a bit like reading grey text on a white background. That said, the detector accounts for the background much like a weighing scale is reset to zero after an empty container is placed on it.
I understand that you can account for the mobile phase when measuring the signal. If only we could remove the eluent and only keep the compounds that we want to analyze.
But there is a different approach that minimizes the impact of the eluent… Instead of using a liquid as the mobile phase, we can use a gas as the mobile phase.
A gas? How does that work?
The fundamentals of the technique are similar to HPLC, but instead of the liquid mobile phase, a carrier gas is used to drive the sample along the column. The column is kept in a small oven, and the column is heated after sample injection. Higher temperatures will cause compounds to pass through the column more quickly, as some compounds will exceed their boiling point and evaporate – this is great for eliminating the solvent, but a problem when trying to separate several analytes with similar boiling points. This is where operating a GC can become a bit of an art – finding a compromise, in terms of temperature and carrier gas flow rate, so that the analytes run through the column as quickly as possible, while making use of the stationary phase to ensure separation. Just as with other chromatographic methods, some compounds adsorb onto the stationary phase better than others. GC columns are usually packed with polymers as a stationary phase, and they are typically tens to hundreds of meters long.
Wow, you’d need several football fields to house that instrument…
Of course not! Think of your 50-foot garden hose… Is it a stiff 50-foot-long piece of material?
No, I keep it wound up in the shed.
Exactly. GC columns are wound as well, so that they fit into instruments that are similar in size to HPLC instruments. The gas can still make its way around the column.
Carrier gases are small molecules such as acetylene, etc. As you can imagine, the pressure of the gas (and therefore the flow rate) can be programmed, just like the temperature of the oven in which the column is housed, to optimize separation.
Are similar detection methods used in GC?
UV spectroscopy is not sensitive enough for use in GC, since such minuscule quantities of sample are used. Since the analyte is mixed with inert gases, there are other avenues for detection. A common detector uses flame ionization (FID). As the sample is brought into the flame, it is heated to extremely high temperatures, which causes the organic species to ionize. The signal that is measured by FID is proportional to the concentration of the ionized species, which will be proportional to the concentration of the organic species itself.
Is this a sensitive detection method?
FIDs tend to offer detection limits that are in the parts-per-billion (ppb) – comparable to HPLC, but with fewer complications than UV spectroscopy. It does suffer from the same handicap that plagues single-wavelength detection in HPLC, in that it cannot distinguish the compounds that reach the detector. But it is fairly inexpensive, and works well in a wide range of concentrations.
Another approach, the thermal conductivity detector (TCD), gives a signal whenever there is a change in the gas that passes through the detector – this allows us to note the presence of any analyte with the carrier gas. Again, it is incapable of distinguishing which compounds are in the gas.
Well then, it doesn’t sound like GC offers much advantage over HPLC, and the equipment seems more complicated.
Ah, but there is a popular and highly sensitive, but finicky, detection system that is often combined with GC. It is mass spectrometry, or MS; the combined acronym GC-MS is often used to describe the entire system. In MS, the molecules are bombarded with a stream of high-speed electrons. When an electron strikes the molecule with the right energy, the molecule is ionized (becoming a cation, or a positively-charged species). This molecular ion breaks apart – we say “fragmented” – and some fragments will maintain that positive charge. A magnet in the MS detector causes the charged particles to bend, with the degree of bending related to the fragment’s “mass-to-charge” (m/z) ratio.
A mass spectrum shows the relative abundance (or percentage) of fragments with each m/z ratio. The figure on the left shows the mass spectrum of caffeine. Note the largest peak is at m/z 194; this is the molecular mass of a caffeine molecule (C8H10N4O2). From this spectrum, it is possible to figure out the fragmentation steps followed by the molecule. For caffeine, the steps that lead to the major peaks in the spectrum are shown to the right.
How can someone know that this is the way the molecule broke apart?
Admittedly, figuring these pathways requires practice, but there are general rules in terms of how molecules fragment – and that is well beyond the scope of this post. An experienced user, when looking at the mass spectrum of an unknown molecule, can compare the pattern of the fragmentation to databases of mass spectra to deduce the structure of that molecule. While the analytical chemists are out having fun, organic chemists entertain themselves on Friday nights looking at mass spectra to figure out whether is that white substance at the bottom of their flask.
Oh yeah, orgchem snap!
Oh yeah, I went there.
As it turns out, one of the requirements for using MS is that the sample is vaporized, and since the outflow from the GC is already in the gas phase, it can be streamed directly into the MS instrument.
Quantitative analysis is done in GC-MS similarly to HPLC; standard solutions are used to develop a calibration curve, relating the concentration of the analyte to the height of its peak in the chromatogram. (See our last post for a reminder of calibration curves.) When a sample is analyzed, the height of the peak for the analyte is measured, and the calibration curve is used to calculate its concentration. The mass spectra is very convenient when several compounds are eluting at similar times, allowing unambiguous determination of which peak belongs to an analyte of interest.
So combining gas chromatography and mass spectrometry allows us to lower detection limits?
Yes. A few examples include detection cocaine and heroin at 5 ppb and 10 ppb, respectively, and hydrofluorocarbons and hydrochlorocarbons are detected to less than 200 ppq (parts-per-quadrillion, or pg/L, picograms per litre) and some contaminants in drinking water to low ppq values, through careful application of GC-MS. There are even examples of 1 fg of pesticide detected by GC-MS. (1 femtogram is 10-15 g, or a billionth of a milligram.) GC-MS is widely considered a gold standard in forensic substance identification.
This is great, and I look forward to reading those links. So what have we accomplished in these last few posts on Numbers of Atoms?
We have looked at various ways to separate analytes in compounds – from picking raisins from cereal flakes by hand, to using state-of-the-art equipment to distinguish between minute quantities of substances. As this series is a dynamic product, there may be more discussion on chromatography and other separation methods at a later time, but we will be changing topics in the next portion of Numbers of Atoms.