Gas-solid chromatography is relatively rare, but it is used to separate atmospheric gases. Common solids are charcoal, a synthetic zeolite called "molecular sieve", or a combination of the two. Solids typically adsorb so strongly that some adsorbed components do not pass through the column (in a reasonable time) and are removed by reversing the flow of mobile phase through the column ("backflushing"), usually at a high temperature. (The term "molecular sieve" is often used generically, but was originally a tradename for certain specific types with specific pore sizes.)

Gas-liquid chromatography begins with a more or less inert "support" with a high surface area. That is mixed with a solution of the liquid phase in a volatile solvent. Then the solvent is evaporated in a rotary evaporator, leaving the support with a "coating" of the liquid phase. The coated support, now called a "packing", is "packed" into a column, such as a 6 mm inside diameter stainless steel tube about 200 mm long. No firm packing is done, but the column is often vibrated to be sure that the particles settle without leaving voids. The column is usually coiled, before or after packing. The ends of the column are plugged, often with glass wool, to hold the packing but allow gas flow. The column ends have fittings (often "Swagelok" brand) so that they can be connected to matching fittings in the "oven" of the gas chromatograph.

1. It keeps the column temperature constant.
   
   
2. It allows operation at elevated temperature (faster; perhaps necessary to vaporize the sample).
   
   
3. It allows "temperature programming", a controlled increase in column temperature during an analysis to make the slow-moving components move faster, reducing the time needed for the analysis, and reducing the diffusion which makes peaks broader.

Because the mobile phase is gas, solubility in the mobile phase is not really a factor, but volatility (vapor pressure) of the components being analyzed is a close equivalent. Solubility in the stationary (liquid) phase is also important, and some stationary phases interact chemically with certain types of sample components. The components to be separated must have a reasonable vapor pressure at the column temperature, or they will not move at all. Much work has been done on "derivatizing" components to increase volatility.

Much of the early work in gas chromatography involved separating hydrocarbon mixtures such as gasoline on columns in which the liquid phase was a silicone oil. The process acted somewhat like distillation with the 'pot' temperature gradually increasing, and some of the theoretical treatments were derived from distillation theory. An industrial distillation is done in a vertical column with a series of perforated plates inside. When it is assumed that there is equilibrium between vapor and liquid at each plate, the plates are "theoretical plates". Increasing the number of plates gives better separation. Application of the theory to gas chromatography introduced the term "height of an equivalent theoretical plate", abbreviated HETP. This was (roughly) the distance along the column that gave the same separation as a (theoretical) plate in a distillation column. Great advances have been made in reducing that distance, increasing the number of "theoretical plates" in a length of column, and improving separation, or "resolution". Separation could also be improved by increasing the column length, but that requires longer analysis times, during which the components can diffuse, perhaps actually reducing resolution and reducing the ability to detect them. One of the major theoretical advances was the "van Deemter equation" which related HETP to carrier gas flow rate. The number of "plates" is still important for chromatograph columns.

Molecules equilibrate more rapidly between liquid and gas phases if both phases are very thin. That led to the "capillary" column, which is typically a fused silica capillary of about 0.25 - 0.5 mm inside diameter (ID), with a thin coating (about 0.2 mm) of liquid phase on the wall, and from 10 meters to 100 meters long. This is a "wall coated open tubular" (WCOT) column. The coating on the wall can also be a combination of support and coating, giving a "support coated open tubular" (SCOT) column. The sample must be very small, and the detector must be very sensitive, but the number of theoretical plates is immense (e. g., 100,000) and very complex mixtures can be separated. The capillary usually has a coating of polymer or aluminum on the outside for protection.

Many real samples have components with a very wide range of volatility. If the column temperature is too high, the most volatile components will move practically as fast as the carrier gas, and will not be separated. But with a lower column temperature, the less volatile components will hardly move at all. That situation can be handled by "temperature programming": increasing the column temperature at a controlled rate.
Another problem with high temperature is that the "liquid" may vaporize into the carrier gas, resulting in column "bleed". Significant bleeding interferes with analysis, and can destroy the column by removing the liquid phase. The liquid (coating) is now very often a polymer that resists bleeding and decomposition, and which may improve selectivity for certain types of molecules. More recently, it has been possible to bond molecules of the liquid phase chemically (covalently) to silica or zirconia particles, or to the inner surface of a capillary column, to make "bonded" phases.

Very few analysts pack their own columns or make their own TLC plates now. They buy prepared columns or plates from suppliers. Wide varieties of both supports and coatings are available. For instance, Supelco is a major US supplier of ready-to-use columns and materials. (See http:www.sigmaaldrich.com, and select Supelco. A catalog may be even more helpful.)

So far, I have ignored two major operations in gas chromatography: getting the sample onto the column, and recognizing the components as they elute with the carrier gas.

Samples are applied to the column with an "injector". The early injectors were just that: they provided access to the end of the column, which was sealed by a rubber "septum".
A liquid sample was drawn into a small syringe calibrated in microliters ("Hamilton" is a major syringe tradename). The syringe needle was pushed through the septum, and a quick slap of the syringe plunger injected a narrow pulse of sample onto the column. Similar but larger syringes could be used for gas samples. The rubber septa have been replaced by polymers with more heat resistance, and injectors now have short lengths of tubing ("sample loops") into which the sample is placed before the injector switches the loop into the gas flow. Automated injection systems can be loaded with numerous samples to be injected automatically (overnight, for instance).

As the carrier gas leaves the column, it flows immediately into a "detector". The output of the detector is recorded, giving a "peak" on the recorder chart as each component is detected. Each peak appears at a characteristic "retention time". The degree of separation of the peaks is the "resolution". When separation is so complete that the recorder pen returns to its zero position between peaks, one has "baseline resolution". The peak height was used first as an easy measure of the quantity of that component in the sample. Peak area proved to be a better measure, so the peaks were "integrated". Integration was originally done by cutting a piece of recorder paper below the peak and weighing it; by counting the squares on the recorder chart below the peak; or by measuring its area with a planimeter. Mechanical ("ball and disc") integrators followed, and electronic integrators appeared quickly. Because the amounts of the sample components vary greatly, it may be necessary to "attenuate" strong signals so that the recorder does not overshoot the chart paper. Detectors do not have the same sensitivity for all molecules, so quantitative measurements require use of standards. Sometimes a fixed amount of an "internal standard" is applied to all samples and standards.

The principal detectors are:

Thermal conductivity (TC): A wire in the gas flow is heated by a constant electrical current. The electrical resistance of a hot wire depends on its temperature, which is (approximately) constant in a steady flow of carrier gas. When the carrier gas contains molecules larger than those of the carrier gas, less heat is removed from the wire; its resistance increases; and the voltage across the wire increases. Sensitive to all components, but not very sensitive to any.

Flame ionization detector (FID): The detector contains a small hydrogen-oxygen (or hydrogen-air) flame. Some substances, when they burn in the flame, produce ions which carry current, and the current is measured. Very sensitive, particularly for molecules having C-H bonds. Not at all sensitive to some other molecules (such as CCl4).

Electron capture detector (ECD): A radioactive source produces ions, and an ion current is measured between positive and negative electrodes. Some components (especially those containing C-Cl bonds) capture ions, reducing the current. Very sensitive; complements FID. The preferred carrier gas is argon.

Mass spectrometer (GC-MS): The gas leaving the column (usually a capillary) first goes through a separator which passes the sample molecules to the mass spectrometer while removing most of the carrier gas molecules. (Remember that mass spectrometry is done in vacuum.) In the time-of flight (TOF) mass spectrometer, the sample is broken into ionized fragments, and the ions are accelerated into a "drift tube" by an electrical pulse. The light ions are accelerated more than the heavier ones, and arrive sooner at the other end of the tube, where the ion current is measured versus time to give a mass spectrum. The heaviest ion is quite often the "molecular ion", the whole molecule plus or minus a hydrogen ion. The whole process is repeated at about ¼ second intervals. A computer displays and records the total ion current vs. time, the current of a specific ion (a specific mass/charge ratio) versus time, or the mass spectrum of a component. The computer also has a "library" of mass spectra. It compares the mass spectrum of each peak with the spectra in its library to identify the individual components. This does assume that the component is in the library, but such libraries include 50,000 or more components. An "ion-trap" detector is one type of mass spectrometric detector.

Some of the less broadly useful detectors are the thermionic emission detector (TED), sensitive to nitrogen and phosphorus compounds (also called an NPD); flame photometric detector (FPD), sensitive to sulfur and phosphorus compounds; and the photoionization detector (PID).

Except for the MS detector, chromatography only separates the sample components, but does not identify them. In routine analyses, the analyst knows what components are expected, and can separately run chromatograms of individual known standard compounds or mixtures, so that the substance which elutes at a certain "retention time" is assumed to be a certain compound.

A report or procedure will probably state the instrument manufacturer and model, the carrier gas and its flow rate; perhaps the inlet pressure; the column dimensions; the support and coating (together, the packing), the injection method and injection volume, the detector (perhaps with applied voltages), any integrator used, and attenuator settings. Retention times may appear in the procedure or be shown with a reproduction of the chromatogram.