GC Columns and Stationary Phases

A gas chromatography column is a length of tubing. It turns a chemical mixture into a time axis. Inject a microliter of gasoline, and forty peaks emerge over twenty minutes. Each peak elutes at a time set by molecular interaction. That interaction is with a thin film of stationary phase. The film clings to the tube wall. That film — sometimes a hundred times thinner than a human hair — is doing all the real chemistry. The difference between a wasted afternoon and a clean separation is almost always a choice. That choice is made before the sample is ever injected: which column. This matters because separation is the rate-limiting step of nearly every real analysis. A mass spectrometer can identify a compound in milliseconds. But this only works if that compound is not co-eluting with three others of similar mass. The column is where selectivity actually happens. It converts "this mixture has many components" into "component A appears at 4.32 minutes, component B at 4.58 minutes." Everything downstream inherits whatever resolving power the column delivered. This includes quantitation, detector response, and method validation. The big picture is a story of a single major transition. That transition runs from packed columns to open-tubular (capillary) columns. It also involves a single unifying physical idea. Separation power comes from partitioning equilibria. These equilibria occur between a mobile gas phase and a stationary liquid (or bonded) phase. That phase's chemistry can be tuned continuously from nonpolar to strongly polar. Layered on top is a bookkeeping system — the Kováts retention index. It lets a chemist compare retention behavior across laboratories, instruments, and even decades. This works without ever needing identical operating conditions. This article works through packed versus capillary column geometry. It covers the efficiency consequences of each. It also covers the polarity scale used to select stationary phases and predict elution order. It explains the retention index framework as well. That framework converts a raw retention time into a portable, temperature-independent identifier. Worked examples cover van Deemter optimization, retention index calculation, and polarity-based peak reordering. This is the kind of reasoning a chromatographer does before ever touching the injector.

Key takeaways

Packed columns (2–4 mm i.d., particulate support) give high sample capacity but modest efficiency (~2000 plates/m); capillary columns (0.10–0.53 mm i.d.) give 3000–5000 plates/m via lower eddy diffusion and negligible resistance to mass transfer.
The van Deemter equation $H = A + B/u + Cu$ predicts an optimum carrier gas velocity; capillary columns effectively drop the $A$ term to zero (open tube, no packing).
Stationary phase polarity is described qualitatively by "like dissolves like" and quantitatively by McReynolds constants, which sum probe-solute retention shifts relative to squalane.
Nonpolar phases (100% dimethylpolysiloxane) elute compounds roughly by boiling point; polar phases (polyethylene glycol, cyanopropyl phases) elute by boiling point and polarity, reordering peaks.
The Kováts retention index $I$ places any compound on the n-alkane retention scale ( $I = 100n$ for n-alkanes), giving temperature- and instrument-independent identification.
Film thickness and column diameter trade capacity for efficiency; thick films favor volatile analytes, thin films favor high-boiling analytes and fast analysis.
Key idea: retention is governed by a partition equilibrium between gas and stationary liquid phases, and both the column's physical format (packed vs. open-tubular) and its stationary-phase chemistry (polarity) independently control how well that equilibrium separates real mixtures.

Where this lives today

Modern petroleum refining relies on detailed hydrocarbon-type analysis (PIONA: paraffins, isoparaffins, olefins, naphthenes, aromatics) run almost entirely on long, narrow-bore capillary columns coated with nonpolar or moderately polar phases, where retention indices from decades-old Kováts calibrations remain directly usable reference values today. Flavor and fragrance laboratories characterizing essential oils (lavender, citrus, mint) routinely report both a mass spectrum and a retention index for every identified terpene, because many terpenes are structural isomers with near-identical mass spectra but distinguishable retention indices on standard nonpolar and polar phase pairs — a textbook case where GC Detectors alone cannot resolve an identity ambiguity that retention chemistry can. Forensic toxicology screening for drugs and their metabolites in biological fluids uses retention-index-matched libraries as a first-pass filter before confirmatory mass spectral matching, reducing false positive candidate lists dramatically. Environmental analysis of polychlorinated biphenyls (PCBs) and pesticide residues depends on selecting a stationary phase (often a mid-polarity phenyl-methylpolysiloxane blend) chosen specifically to resolve congeners that co-elute on either extreme-polarity phase, directly applying the polarity-selection reasoning developed in Example 3. And in every case, the column sits at the front of an analytical chain that continues into GC Detectors (flame ionization, mass spectrometric, electron capture) and depends on solid grounding in Gas Chromatography Instrumentation for carrier gas control, injector design, and temperature programming — the column's separation is necessary but not sufficient without a detector able to sense what has been separated.

Ask the tutor: How would I redesign a method (phase choice, film thickness, temperature program) to separate a mixture of five structurally similar terpene isomers with nearly identical boiling points?

The physical picture

Inside a GC oven, the column is a coil of tubing — stainless steel or glass for packed columns, fused silica for capillary columns — through which an inert carrier gas (helium, hydrogen, or nitrogen) continuously flows. A packed column is filled with small, inert particles (diatomaceous earth, 100–200 mesh) coated with a thin film of nonvolatile liquid, the stationary phase. A capillary column instead has that liquid film coated or chemically bonded directly onto the inner wall of a narrow, hollow tube, leaving an open channel down the middle for gas flow. This geometric difference — packed with particles versus open-tubular — is the single most consequential design choice in gas chromatography, because it sets the efficiency ceiling of everything downstream. Efficiency is quantified by the height equivalent to a theoretical plate, $H$ , related to the number of theoretical plates $N$ by $H = L/N$ for column length $L$ . Three physical processes contribute to band broadening as an analyte migrates: eddy diffusion (multiple flow paths through packing), longitudinal molecular diffusion (analyte spreading along the flow axis), and resistance to mass transfer (the finite time needed for molecules to equilibrate between gas and liquid phases). In a packed column, carrier gas must snake around thousands of particles per centimeter, creating many parallel flow paths of different length — this is the dominant broadening mechanism and it cannot be reduced by adjusting flow rate. In an open-tubular capillary column there is no packing at all, so this term vanishes identically, which is why capillary columns can achieve plate counts an order of magnitude higher for the same column length. Stationary-phase chemistry is the second axis of control, entirely independent of geometry. A film of 100% poly(dimethylsiloxane) (PDMS, trade names like SE-30, OV-1, DB-1) is essentially nonpolar: retention there is governed almost purely by volatility (boiling point and dispersion forces), so compounds elute in roughly the order of their boiling points. Substituting some methyl groups on the siloxane backbone with phenyl, cyanopropyl, or trifluoropropyl groups introduces dipole and induced-dipole interactions; polyethylene glycol (Carbowax, trade name DB-WAX) is fully polar, capable of hydrogen bonding. On a polar phase, a polar analyte is retained far longer than a nonpolar analyte of identical boiling point, because the stationary phase itself now participates in dipole–dipole and hydrogen-bonding interactions that favor the polar solute staying dissolved in the liquid film rather than escaping into the gas phase.

Figure

Packed column vs. capillary column cross-sections and flow paths

The consequence of these two independent axes — geometry (efficiency, capacity) and chemistry (selectivity, elution order) — is that method development in GC is really a two-variable optimization. You first pick a stationary phase whose polarity matches the separation problem (does the mixture differ mainly in volatility, or mainly in functional-group polarity?), then you pick a column format and dimension (packed for preparative or robust industrial use; capillary — narrow-bore for speed and resolution, wide-bore for capacity) to deliver enough plates in an acceptable analysis time. Retention index calculations, covered below, are the tool that lets a chemist describe where a compound falls on a given phase without re-deriving the whole calibration every time conditions change slightly.

Figure

Elution order reversal: nonpolar vs. polar stationary phase

The core ideas

The retention factor $k$ (capacity factor) links a measured retention time to the underlying partition equilibrium:

k = \frac{t_R - t_M}{t_M} = K \cdot \frac{V_S}{V_M}

where: $t_R$ is the analyte's retention time, $t_M$ is the retention time of an unretained species (the dead time), $K = C_S/C_M$ is the partition coefficient between stationary and mobile phase concentrations, and $V_S$ , $V_M$ are the stationary- and mobile-phase volumes inside the column.

Reading: a compound with a large partition coefficient (strong preference for dissolving in the liquid film) spends most of its time immobilized in the stationary phase and so has a long retention time; $k$ is a dimensionless "how many phase volumes worth of extra time" number, directly measurable from a chromatogram without knowing $K$ itself. Column efficiency is governed by the van Deemter equation, which decomposes plate height into three additive band-broadening mechanisms:

H = A + \frac{B}{u} + C u

where: $H$ is plate height (cm), $u$ is the average linear carrier gas velocity (cm/s), $A$ is the eddy-diffusion term (packing-path irregularity; zero for open-tubular columns), $B$ is the longitudinal-diffusion coefficient (worsens at low velocity because the analyte band has more time to diffuse along the tube), and $C$ is the mass-transfer resistance term (worsens at high velocity because equilibration between phases can't keep up).

Reading: there is an optimum velocity $u_{opt} = \sqrt{B/C}$ that minimizes $H$ (maximizes efficiency); operating faster than optimum trades some resolution for speed, a very common and often correct compromise in routine analysis. The retention index framework converts a raw retention time into a portable number using bracketing n-alkanes:

I = 100 \left[ n + \frac{\log t'_{R(x)} - \log t'_{R(n)}}{\log t'_{R(n+1)} - \log t'_{R(n)}} \right]

where: $n$ and $n+1$ are the carbon numbers of the n-alkanes eluting immediately before and after the analyte $x$ , and $t'_R = t_R - t_M$ is the adjusted retention time (correcting for dead time) for each species, all measured under the same isothermal run.

Reading: by definition, an n-alkane's own retention index is exactly $100 \times$ its carbon number ( $I_{\text{hexane}} = 600$ , $I_{\text{decane}} = 1000$ ); every other compound's index is interpolated logarithmically between its bracketing alkanes, and — crucially — this index is nearly invariant to which instrument, carrier gas flow, or specific column of a given stationary-phase type is used, making it a universal fingerprint for compound identification (tabulated in databases such as NIST alongside mass spectra).

Derivation AOptimum carrier gas velocity from the van Deemter equation

Goal. Starting from $H(u) = A + B/u + Cu$ , find the velocity $u_{opt}$ that minimizes plate height, and the resulting minimum plate height $H_{min}$ .

Step 1. Treat $H$ as a function of the single variable $u$ , holding $A$ , $B$ , $C$ fixed (these depend on column geometry, diffusion coefficients, and phase thickness, not on flow rate). Differentiate:

\frac{dH}{du} = -\frac{B}{u^2} + C

Step 2. Set the derivative to zero to locate the extremum:

-\frac{B}{u^2} + C = 0 \quad\Rightarrow\quad C = \frac{B}{u^2}

Step 3. Solve for $u$ :

u_{opt}^2 = \frac{B}{C} \quad\Rightarrow\quad u_{opt} = \sqrt{\frac{B}{C}}

Step 4. Confirm this is a minimum, not a maximum, by checking the second derivative:

\frac{d^2H}{du^2} = \frac{2B}{u^3} > 0 \text{ for all } u>0, B>0

Since the second derivative is always positive for physically meaningful (positive) values of $B$ and $u$ , every extremum of $H(u)$ found this way is a genuine minimum — confirming $u_{opt}$ minimizes plate height rather than maximizing it.

Step 5. Substitute $u_{opt} = \sqrt{B/C}$ back into the original van Deemter expression to get the minimum achievable plate height:

H_{min} = A + \frac{B}{\sqrt{B/C}} + C\sqrt{\frac{B}{C}} = A + \sqrt{BC} + \sqrt{BC} = A + 2\sqrt{BC}

This is the practical bottom line: even at the ideal flow rate, plate height cannot go below $A + 2\sqrt{BC}$ , and for a well-made capillary column $A \approx 0$ , so $H_{min} \approx 2\sqrt{BC}$ — efficiency is then limited purely by the diffusion (B) and mass-transfer (C) terms, both of which shrink with thinner films and narrower bores.

Go deeper · Derivation B · 2 minRetention index interpolation and its temperature/instrument invariance

Derivation BRetention index interpolation and its temperature/instrument invariance

Goal. Show why the Kováts index formula uses logarithms of adjusted retention time (not raw retention time) and why this choice makes $I$ approximately transferable between instruments running the same stationary phase.

Step 1. For a homologous series (n-alkanes) on a given stationary phase under isothermal conditions, the partition coefficient $K$ increases roughly geometrically with each added CH $_2$ group, because each additional methylene contributes a roughly constant increment to the free energy of solution:

\Delta G_{\text{soln}}(n) \approx \Delta G_0 + n\,\delta g

where $\delta g$ is the free-energy contribution per carbon and $\Delta G_0$ is an intercept term.

Step 2. Relate free energy of solution to the partition coefficient via $\Delta G_{\text{soln}} = -RT\ln K$ , and recall $k \propto K$ and $t'_R \propto k$ (from the retention-factor relation above), so:

\ln t'_R(n) \approx \text{const} - \frac{n\,\delta g}{RT}

This says $\ln t'_R$ (equivalently $\log t'_R$ ) is linear in carbon number $n$ for a homologous series — a straight line on a log-retention-time versus carbon-number plot, which is the empirical basis for Kováts' original observation.

Step 3. Because $\log t'_R$ is linear in $n$ , any analyte eluting between two bracketing alkanes can be located on that line by ordinary linear interpolation in $\log t'_R$ space — which is exactly the formula given above, using $\log t'_R$ rather than $t'_R$ itself. Using raw $t'_R$ for interpolation would introduce systematic curvature error, since retention time itself grows geometrically, not linearly, with carbon number.

Step 4. Consider what happens if oven temperature or carrier flow shifts slightly, uniformly stretching or compressing all retention times by some factor $f$ (a reasonable first approximation for small perturbations): $t'_R \to f\,t'_R$ for every peak, including both bracketing alkanes and the analyte. Then:

\log(f\,t'_{R(x)}) - \log(f\,t'_{R(n)}) = \log t'_{R(x)} - \log t'_{R(n)} + \log f - \log f = \log t'_{R(x)} - \log t'_{R(n)}

Step 5. The multiplicative factor $\log f$ cancels identically in both numerator and denominator of the Kováts formula, since it appears in every log-difference term equally. This is the algebraic reason the retention index is far more robust to day-to-day instrumental drift than a raw retention time would be — small, uniform shifts in flow or temperature largely cancel, leaving $I$ nearly constant for a given compound on a given stationary-phase chemistry (though $I$ still depends on which specific stationary phase and, more weakly, on the absolute temperature, since $\delta g/RT$ is temperature-dependent).

Worked examples

Example 1 — Foundational: locating the van Deemter optimum and comparing packed vs. capillary format. A packed column (2 mm i.d., 2 m length) is characterized by van Deemter constants $A = 0.30$ cm, $B = 0.55$ cm²/s, $C = 0.030$ s. A capillary column (0.25 mm i.d., 30 m) run on the same instrument has $A \approx 0$ (open tube), $B = 0.30$ cm²/s, $C = 0.0020$ s (thinner film, faster mass transfer). Packed column optimum:

u_{opt} = \sqrt{\frac{B}{C}} = \sqrt{\frac{0.55\ \text{cm}^2/\text{s}}{0.030\ \text{s}}} = \sqrt{18.3\ \text{cm}^2/\text{s}^2} = 4.28\ \text{cm/s}

H_{min} = A + 2\sqrt{BC} = 0.30 + 2\sqrt{(0.55)(0.030)} = 0.30 + 2\sqrt{0.0165} = 0.30 + 2(0.1285) = 0.557\ \text{cm}

Plates available over 2 m (200 cm): $N = L/H_{min} = 200/0.557 = 359$ plates. That is a poor number by modern standards but typical of an older packed-column separation — packed columns are chosen for ruggedness and capacity, not raw efficiency. Capillary column optimum:

u_{opt} = \sqrt{\frac{0.30}{0.0020}} = \sqrt{150} = 12.2\ \text{cm/s}

H_{min} = 0 + 2\sqrt{(0.30)(0.0020)} = 2\sqrt{0.00060} = 2(0.02449) = 0.0490\ \text{cm}

Plates over 30 m (3000 cm): $N = 3000/0.0490 = 61{,}200$ plates. The capillary column delivers roughly 170 times more plates than the packed column despite being run under comparable "optimum" conditions — the direct, quantitative payoff of eliminating the eddy-diffusion ( $A$ ) term and shrinking $C$ by reducing film thickness and bore diameter. This is why virtually all modern analytical GC uses capillary columns; packed columns persist mainly for gas analysis (permanent gases, refinery streams) and preparative-scale work where sample capacity outweighs the efficiency deficit.

Example 2 — Applied: computing a Kováts retention index and using it for tentative identification. An unknown compound is run isothermally at 100 °C on a DB-1 (nonpolar, 100% PDMS) capillary column alongside a homologous series of n-alkanes for calibration. Measured retention times (with dead time $t_M = 0.85$ min determined from an unretained marker such as methane):

Peak	$t_R$ (min)	$t'_R = t_R - t_M$ (min)
n-octane (C8)	4.10	3.25
Unknown X	5.62	4.77
n-nonane (C9)	6.40	5.55

The unknown elutes between octane and nonane, so $n = 8$ :

\log t'_{R(X)} = \log(4.77) = 0.6785

\log t'_{R(8)} = \log(3.25) = 0.5119

\log t'_{R(9)} = \log(5.55) = 0.7443

I = 100\left[8 + \frac{0.6785 - 0.5119}{0.7443 - 0.5119}\right] = 100\left[8 + \frac{0.1666}{0.2324}\right] = 100\left[8 + 0.7169\right] = 871.7

So the unknown has a Kováts index of $I \approx 872$ on a nonpolar phase at 100 °C. Cross-referencing this value against a retention-index library (e.g., NIST WebBook nonpolar-phase indices), $I \approx 870$ – $875$ on DB-1-type phases is well matched by 2-methylheptane-type branched alkanes or by simple aromatic compounds such as ethylbenzene ( $I_{lit} \approx 869$ on nonpolar phases) — the retention index narrows the candidate list dramatically before any spectral confirmation (mass spectrum or standard co-injection) is attempted. Note that this index would come out essentially the same on a different DB-1 column of different length, or at a slightly different (but still isothermal, similarly calibrated) run, precisely because of the invariance argument developed above — this portability across labs and instruments is the entire point of using $I$ rather than raw $t_R$ for compound libraries.

Example 3 — Exam/competition level: predicting elution order and resolving a co-elution by switching stationary phase polarity. Two compounds — 1-butanol (bp 118 °C, moderately polar, hydrogen-bond donor) and toluene (bp 111 °C, weakly polar, nonpolar bulk) — must be separated. On a nonpolar DB-1 (100% PDMS) column, retention should track boiling point closely, since dispersion forces dominate and both compounds are similar-sized nonpolar "shapes" from the phase's perspective. Given the close boiling points (118 °C vs. 111 °C, a 7 °C difference), predict whether these compounds are likely to be resolved on DB-1, and propose and justify a phase switch if not. Step 1: Estimate resolution feasibility on the nonpolar phase. On DB-1, retention order should follow boiling point: toluene (111 °C) elutes before 1-butanol (118 °C), but the small $\Delta$ bp (7 °C) combined with PDMS's inability to discriminate polarity means the retention factors $k_{tol}$ and $k_{BuOH}$ will be close — selectivity factor $\alpha = k_2/k_1$ near 1, risking a poorly resolved doublet or a peak-order flip depending on exact operating temperature (the two boiling points are close enough that thermal drift alone could swap elution order run-to-run on a nonpolar phase — an unreliable separation). Step 2: Predict the effect of switching to a polar phase (DB-WAX, polyethylene glycol). PEG stationary phases hydrogen-bond strongly with –OH groups. 1-butanol, as a hydrogen-bond donor, will be dramatically more retained on DB-WAX than its boiling point alone would predict, while toluene — nonpolar, weak dipole only — is retained comparably to (or even less than) its boiling-point-based expectation, since PEG doesn't offer strong dispersion-force stabilization the way PDMS does for large nonpolar aromatics. Step 3: Quantify using McReynolds-style reasoning. McReynolds constants tabulate the retention index shift ( $\Delta I = I_{\text{polar phase}} - I_{\text{squalane}}$ ) for standard polar probes on a given phase; polar phases show large $\Delta I$ for probes like 1-butanol (an alcohol probe) — typically $\Delta I$ on the order of 200–600 index units for strongly polar PEG-type phases — while showing much smaller shifts for aromatic probes like benzene (on the order of 40–120 units). This asymmetric shift is exactly the lever needed: it inflates 1-butanol's retention index far more than toluene's, opening up a wide separation window between the two peaks that did not exist on the nonpolar phase. Step 4: State the resolved conclusion. Switching from DB-1 to DB-WAX is predicted to (a) reverse or widen the peak order — 1-butanol should now elute considerably later than toluene, not merely slightly later — and (b) increase the selectivity factor $\alpha$ well above the near-unity value estimated for the nonpolar phase, because the polar phase's hydrogen-bonding interaction discriminates between these two compounds far more sharply than boiling point (dispersion forces) alone does. This is the general strategy for any co-elution driven by similar volatility but different functional-group polarity: change the stationary phase chemistry, not just the temperature program.

Limiting cases, scope & common traps

Regime / limit	Behavior	Why
$u \to 0$ (very slow flow)	$H \to \infty$	Longitudinal diffusion ( $B/u$ term) dominates; analyte band spreads by simple diffusion while barely migrating
$u \to \infty$ (very fast flow)	$H \to \infty$ (linearly, via $Cu$ )	Mass-transfer resistance dominates; molecules can't equilibrate between phases fast enough, band smears
Open-tubular limit ( $A \to 0$ )	$H_{min} \to 2\sqrt{BC}$	Capillary geometry removes eddy diffusion entirely; efficiency limited only by diffusion and mass transfer
Film thickness $\to 0$	$C \to$ minimal, but $k \to 0$ too	Thin films speed up mass transfer and boost efficiency, but sample capacity and retention both shrink — useful only for very volatile analytes
Homologous series, single compound class	$I = 100n$ exactly	By definition of the n-alkane index scale itself
Nonpolar analyte on nonpolar phase	Elution order ≈ boiling-point order	Only dispersion forces operate; no polarity discrimination
Polar analyte on polar phase	Elution order can invert relative to boiling point	Hydrogen bonding / dipole interactions add a retention term unrelated to volatility

Common trapMisconception: "higher plate count always means a better separation."

Plate count $N$ measures peak narrowness, not how far apart two peaks' centers are. Two poorly-separated compounds can each individually give sharp, high-efficiency peaks (large $N$ ) yet still overlap badly if the stationary phase provides no selectivity ( $\alpha \approx 1$ ) between them. Resolution depends on efficiency, selectivity, and retention factor together (the resolution equation, not covered in depth here, multiplies a $\sqrt{N}$ efficiency term by an $(\alpha-1)/\alpha$ selectivity term). A chemist chasing resolution by lengthening a column when the real problem is wrong stationary-phase chemistry is optimizing the wrong variable — Example 3 above is exactly this situation, where a phase change (selectivity) solves what a longer nonpolar column (efficiency) cannot.

The chemist behind it

The retention index system is named for Ervin Kováts, a Hungarian-born Swiss chemist who worked at the ETH Zürich in the late 1950s alongside pioneers of gas chromatography during the technique's earliest maturation as an analytical method. Kováts recognized that raw retention times were essentially useless for building shared, cross-laboratory compound libraries — a retention time depends on flow rate, column length, temperature, and instrument idiosyncrasies in ways that make numbers from two different labs incomparable at face value. His 1958 proposal to anchor every compound's retention behavior to the logarithmically-spaced retention times of a universal, always-available reference ladder — the n-alkane homologous series — converted an instrument-specific measurement into a portable chemical fingerprint, decades before mass spectral libraries existed at scale to serve the same identification role. The elegance of the approach is that it required no new instrumentation, only a disciplined bookkeeping convention: run your unknowns bracketed by n-alkanes, and the resulting index number becomes comparable to anyone else's index number for the same compound on the same class of stationary phase, anywhere in the world.

Origins · 1 minErvin Kováts and how this was discovered

Ervin Kováts

1927–2018 · Analytical chemistry, ETH Zürich, mid-20th-century GC method development

Kováts worked in a period when gas chromatography was transitioning from a laboratory curiosity into a routine analytical workhorse, and much of the foundational vocabulary of the field — column efficiency, plate theory borrowed from distillation science, and standardized retention metrics — was still being actively invented rather than merely taught. His retention index concept remains, essentially unmodified, the identification standard used today in essential-oil analysis, petroleum characterization, flavor and fragrance chemistry, and forensic toxicology screening.

History

- 1941 — Archer Martin and Richard Synge propose the theoretical plate model for chromatographic separations, later extended to gas–liquid partition chromatography.

- 1952 — Martin and Anthony James publish the first practical gas–liquid partition chromatography separations, launching modern GC.

- 1956 — J. J. van Deemter, F. J. Zuiderweg, and A. Klinkenberg publish the rate-theory (van Deemter) equation describing plate height as a function of flow velocity.

- 1957 — Marcel Golay introduces the open-tubular (capillary) column concept, separating the roles of gas conduit and stationary-phase support.

- 1958 — Ervin Kováts proposes the n-alkane-based retention index system for portable, phase-specific compound identification.

Why do packed columns still exist if capillary columns are so much more efficient?

Packed columns remain useful where capacity, not resolving power, is the priority. Their much larger internal volume tolerates larger sample sizes (milligram quantities) without overloading the stationary phase, which matters for preparative separations or samples with very low analyte concentrations that need a large injection to generate detectable signal. They're also mechanically rugged, cheap to replace, and well suited to simple gas analyses (permanent gases like O $_2$ , N $_2$ , CO $_2$ ) where only a handful of components need separating and high plate counts are unnecessary. In routine trace analysis and complex-mixture work, however, capillary columns dominate because their far higher efficiency is usually the deciding factor.

What actually determines whether a stationary phase is called "polar" or "nonpolar"?

Polarity classification reflects the functional groups on the polymer backbone. Pure poly(dimethylsiloxane), with only methyl groups, interacts with analytes via weak dispersion forces only, so it's classified nonpolar. Replacing methyl groups with phenyl groups adds some polarizability and π-interactions (intermediate polarity); cyanopropyl or trifluoropropyl substituents add strong dipole character; and polyethylene glycol backbones offer hydrogen-bond acceptor oxygen atoms throughout the chain, making them strongly polar. McReynolds constants quantify this rigorously by measuring how much five standard polar/polarizable probe compounds are retained on a given phase relative to squalane (a fully nonpolar reference), giving a numerical polarity fingerprint rather than a qualitative label.

Can two different compounds have the same Kováts retention index on the same column?

Yes, and this is a known limitation, not a flaw in the concept. Retention index depends only on the compound's interaction with the specific stationary-phase chemistry at the given temperature, so structurally different compounds with similar volatility and similar polarity profiles can coincidentally match indices closely or exactly. This is exactly why retention index is used as a filtering tool alongside independent confirmation — typically mass spectrometry — rather than as a stand-alone, unambiguous identification method. A compound whose measured index matches a library value to within a few units is a strong candidate, but confirmation still requires a matching mass spectrum or co-injection with an authentic standard.

How does film thickness interact with column diameter in choosing a capillary column?

Both film thickness and internal diameter set the phase ratio $\beta = r/(2 d_f)$ , where $r$ is the column radius and $d_f$ is film thickness; larger $\beta$ (thin film, wide bore) gives lower retention (analytes spend less time dissolved) and faster analysis, while smaller $\beta$ (thick film, narrow bore) gives stronger retention, useful for very volatile compounds that would otherwise elute too fast to resolve. A very volatile analyte mixture (light hydrocarbons, residual solvents) typically needs a thick film to retain these compounds long enough for separation; a mixture of semi-volatile or high-boiling compounds (fatty acid esters, PAHs) is better served by a thin film, since thick films would retain these compounds far too long and broaden their peaks through excessive mass-transfer resistance.

Does the retention index change if I run the same compound at a different oven temperature?

Yes, modestly. The retention index derivation assumes the free-energy-per-methylene increment $\delta g$ is approximately constant, but $\delta g/RT$ does have residual temperature dependence, so literature retention indices are always reported alongside the isothermal temperature (or temperature program) used. In practice, indices measured on the same phase type across a reasonable temperature range (say, ±20 °C) typically shift by only a few index units for most compounds, which is why retention index libraries remain broadly useful despite this caveat — but for precise identification work, matching the reported temperature conditions as closely as possible improves confidence.

Why does the van Deemter B/u term get worse at low velocity rather than better?

It seems counterintuitive that slower flow — giving the analyte more time to equilibrate between phases — makes the B term worse, but the B term specifically describes longitudinal molecular diffusion, an entirely separate process from phase equilibration. At low velocity, the analyte band spends a long time inside the column, and ordinary diffusion (governed by Fick's laws) has correspondingly more time to spread the band out along the direction of flow, widening the peak. The C term, resistance to mass transfer, actually improves (shrinks) at low velocity because there is more time for the analyte to equilibrate cleanly between gas and stationary phases. The van Deemter optimum exists precisely because these two competing effects — diffusion worsening at low $u$ , mass transfer worsening at high $u$ — pull in opposite directions.

How does stationary phase choice interact with the detector used downstream?

Largely independently, though a few practical couplings exist. Most modern stationary phases (bonded, cross-linked silicones and PEG) are chemically stable and low-bleed enough to be compatible with flame ionization detection, electron capture detection, and mass spectrometric detection alike, so phase selection is driven mainly by the separation problem. One coupling worth noting: phase "bleed" (slow thermal decomposition of the stationary phase releasing volatile fragments) raises the baseline signal in mass spectrometric detection more noticeably than in flame ionization detection, because MS is sensitive to specific low-mass bleed ions (like the characteristic siloxane fragment at m/z 207), so GC-MS methods often specify low-bleed phase formulations. This detector-side consideration is explored further under GC Detectors.