Gas Permeability Measurement: Methods, Formulas, and Ranges
Coretest · technical notes
Gas permeability measurement: methods, formulas, and the range where each one works
Five laboratory methods span ten orders of magnitude in permeability, plus the Boyle’s-law porosity measurement that rides on the same manifold. None covers the whole range, and the physics that limits each is worth knowing before you choose an instrument.
Gas is the working fluid of choice for routine core analysis: it is inert, clean, does not alter clay minerals, and equilibrates quickly. But gas is compressible, it slips along pore walls, and at high rates it departs from Darcy's law. Every measurement method is, at its core, a different strategy for extracting the true rock property from these three complications. Below we walk through the methods in order of the permeability range they serve, with their exact working equations and the practical details that separate a good measurement from a plausible-looking number.
1. Steady-state method
The reference method, and the only one that measures permeability directly from Darcy's law with no transient model in between. A constant gas flow is established through the sample; once inlet and outlet pressures stabilize, permeability follows from the integrated compressible-flow form of Darcy's law:
The p2 form, rather than a simple Δp, is exact for isothermal ideal-gas flow and accounts for gas expansion along the sample. Using the incompressible form with the arithmetic mean pressure gives the same answer only when Δp is small relative to the mean; at large drawdowns the error is real.
The practical details are all in the flow measurement. A thermal mass-flow meter covers roughly 1.5–2 decades and is calibrated per gas: helium, with its high thermal conductivity, is the worst case and needs its own calibration and conversion factors. Spanning the method's full range therefore means a bank of two or three meters, each with its own drift and recalibration schedule. Below ~0.1 mD the required flow drops to fractions of a standard cm³/min and stabilization stretches to hours. A back-pressure regulator on the outlet is worth having: it raises the mean pore pressure, which both suppresses slippage (Section 6) and keeps the flow meter in its comfortable range.
2. Unsteady-state pressure falloff (single reservoir)
The workhorse of automated plug-scale instruments. A reservoir of known volume V is charged with gas and allowed to discharge through the sample; permeability is computed from the recorded pressure decay. No flow meter is required. The reservoir itself is the flow meter. At any instant, isothermal mass balance gives the flow entering the sample:
which is substituted into Eq. (1) and integrated over the recorded p(t). Modern implementations, following Jones (1972) and API RP 40, fit the entire transient rather than a single point, which allows the Klinkenberg slip factor and the inertial coefficient to be extracted from the same decay, since the transient sweeps through a range of mean pressures and rates on its way down.
Dynamic range is set by hardware, not by the model. Switchable reservoir volumes select the decay time constant at the low-permeability end; calibrated outlet restrictors stretch the transient into measurable seconds at the high end. A restrictor with a known pressure–flow characteristic is effectively a passive flow meter: no gas-specific calibration, no zero drift, no recalibration interval. This is how a single instrument covers five to six decades (approximately 0.001 mD to 5 D), where a flow-meter-based design would need a battery of sensors.
The method's honesty conditions: it is model-based, so anything the model does not know about becomes a permeability error. A leak indistinguishable from sample flow biases low-permeability results directly; a temperature drift during the decay masquerades as pressure change. Leak-rate acceptance tests and thermal stability of the manifold are not accessories here. They are the measurement.
3. Pulse decay (two-reservoir, Brace method)
The tight-rock method. The sample sits between upstream and downstream reservoirs (volumes Vu, Vd), both held at elevated pore pressure; a small pulse Δp0 is applied upstream and the differential pressure is tracked as the system relaxes to equilibrium. For a small pulse and negligible sample pore-volume storage, the decay is a single exponential:
Solving for k: k = α μ cg L / [ A ( 1/Vu + 1/Vd ) ].
On a semi-log plot the decay is a straight line, and its slope is the whole measurement, which is precisely what makes the method robust when done properly and treacherous when not. The exact solution (Dicker & Smits, 1988) adds a correction for finite pore-volume storage; for typical plug geometries and reservoir sizing it stays within a few percent of unity.
Because the whole measurement runs at elevated mean pore pressure, gas slippage is suppressed at the source, and the result lands close to the equivalent-liquid permeability without extrapolation. The price is patience and discipline: a single nanodarcy point can take hours, over which a temperature drift of a tenth of a kelvin produces pressure artifacts comparable to the signal itself, and the acceptable leak rate is orders of magnitude below what a routine instrument would tolerate. Above ~0.1 mD the pulse dies faster than the data system can resolve. The method is pointless there.
4. Crushed-rock method (GRI)
For shales and other ultra-tight rocks: the sample is crushed to a controlled particle size and gas uptake into the particles is recorded after a step change in cell pressure. Permeability is obtained by matching the pressure decay to a diffusion-into-spheres model. There is no closed-form working equation, and the result carries a known dependence on the assumed particle geometry and on the crush size itself: two laboratories crushing to different mesh sizes will not, in general, report the same number. The method is fast, works on cuttings, and reaches the nanodarcy range, but it applies no confining stress, destroys the sample, and its results are matrix permeability only, not directly comparable to plug measurements.
5. Probe (profile) permeameter
A pressure probe sealed against the surface of a slabbed core injects gas through a small tip; permeability follows from a hemispherical-flow solution with an empirically calibrated geometric factor. Its role is high-resolution permeability profiling (hundreds of points along a core in an afternoon), not absolute reference values: the measurement is unstressed, near-surface, and sensitive to the seal quality at every point.
6. The porosity companion: grain volume by Boyle’s law
The instruments that measure gas permeability on plugs almost always measure porosity too: the universal permeameter–porosimeter class pairs both on one manifold, one transducer, and one gas. The porosity half is a double-cell helium expansion. A reference volume Vref is charged to p1 and expanded into a matrix cup of calibrated volume Vcell containing the sample at ambient pressure p0; the system settles at p2, and grain volume follows from Boyle’s law:
All pressures absolute; the real-gas correction for helium at these pressures is marginal but belongs in the computation.
Mechanically the matrix cup is the simplest component in the whole instrument; metrologically it is where the discipline lives. Cell volumes are not computed from geometry. They are calibrated by expansion against certified steel billets, and the porosity value is exactly as good as that calibration chain and no better. Three details separate a rigorous implementation from a plausible one: a single absolute transducer reads both p1 and p2, which removes inter-channel error by construction; valve dead volumes must be reproducible cycle to cycle, since every actuation displaces a small volume; and helium warms on expansion (its Joule–Thomson coefficient is negative at room temperature), so the reading is taken against a dp/dt stabilization criterion, not on a timer.
One more distinction hides under the word “porosity”: grain volume measured in the matrix cup at ambient stress, versus pore volume measured directly in the core holder by expanding helium into the pore space under confining pressure. The second is slower but yields porosity as a function of stress, which is what a specification reading “porosity at overburden pressure” is actually asking for. A universal instrument offers both, and reports which one a given number came from.
Which permeability methods share an instrument with the porosimeter? The natural pairing is with the unsteady pressure-falloff method, and it is no accident that the universal permeameter–porosimeter class is built exactly this way: both measurements live on one manifold: the falloff reservoir doubles as the porosimeter’s reference volume, the same absolute transducer and the same helium supply serve both, and the core holder that measures permeability also measures pore volume under stress. A steady-state module integrates almost as easily, sharing the core holder and adding only the flow-metering branch. Pulse-decay tight-rock instruments obtain pore volume in the holder as a by-product of the same transient (it enters through the storage term of the exact solution), though grain volume at ambient conditions still calls for a matrix cup. The crushed-rock cell is itself a Boyle cell. The GRI measurement is, in effect, porosimetry extended in time. Only the probe stands alone: with no closed cell around the sample, there is nothing to expand into.
7. Klinkenberg slippage: the correction that ties it all together
Gas molecules do not come to rest at pore walls; they slip. Measured gas permeability therefore exceeds the true (equivalent-liquid) permeability, increasingly so at low pressure and in fine pores, where the molecular mean free path approaches the pore size:
Measuring kg at three or more mean pressures and extrapolating the straight line to 1/pm = 0 yields k∞. In an automated instrument this pressure sequence must run without operator intervention. That, more than anything, is what "automatic" means in a Klinkenberg-capable permeameter. The wider the sweep of mean pressures, the more stable the extrapolation; an outlet back-pressure capability extends the sweep upward and visibly improves the fit.
How big is the correction? On rocks in the darcy range it amounts to a few percent, inside instrument uncertainty, and the correction degenerates. At single millidarcies it reaches tens of percent. At the microdarcy level and below, the measured gas permeability can exceed k∞ several-fold. The correction is therefore critical exactly where the transient methods operate, and irrelevant where the steady-state method is at home. Empirically, b grows as permeability falls (approximately as a power law in k∞), which is why tight-rock work cannot skip it.
8. Forchheimer inertia: the correction with the opposite sign
At high flow rates the pressure gradient acquires a quadratic term:
Ignored, inertia makes apparent permeability fall with increasing rate, the opposite sign to slippage. On very permeable samples the two effects can superpose so that a naïve Klinkenberg plot acquires an inverted slope, which looks like an instrument fault but is physics. The correct behavior for an automated system is to limit differential pressure on high-permeability samples (or fit β explicitly), and to report measured gas permeability without a slip correction where the correction is degenerate, flagged as such, not silently.
9. Practical nuances that outweigh the choice of method
Confining stress. Permeability is stress-dependent: modestly in clean sandstones, dramatically in microcracked and tight rocks. A number without its net confining stress is incomplete; comparisons between laboratories fail on this point more often than on any other. All plug methods above run in a hydrostatic (Hassler-type) or biaxial core holder for exactly this reason.
Sample preparation. Cleaning and drying come before everything. Oven-drying clay-bearing samples collapses clay structure and overstates permeability; humidity-controlled drying (typically 40 % relative humidity, per API RP 40) preserves clay-bound water. The best instrument cannot recover information destroyed in the drying oven.
Choice of gas. Helium, nitrogen, and dry air are all suitable: inert and non-adsorbing on most reservoir rock. Gases that adsorb (CO2, methane on organic-rich shale) are unsuitable for permeability measurement, because adsorption adds a storage term the flow models do not contain. Real-gas corrections (the z-factor) are negligible for helium and nitrogen at typical laboratory pressures but should be in the computation for completeness.
Temperature. Every transient method converts pressure change into permeability; a temperature change produces pressure change with no flow at all. Thermal stability of manifold and reservoirs sets the noise floor of the entire instrument. The tighter the rock, the more this dominates.
10. What accuracy to expect in practice
Manufacturers' data sheets quote transducer accuracy (0.1–0.15 % of full scale is typical), but the uncertainty of the permeability value is a different and larger number, because it stacks transducer error, volume calibration, temperature stability, leak rate, and the flow model on top of each other. Three figures matter, and they are not interchangeable: repeatability (same sample, same instrument, same day), accuracy (agreement with a reference method), and reproducibility (agreement between laboratories). Vendors advertise the first; users live with the third.
Realistic numbers under good practice: the steady-state method delivers ±2–5 % in its comfortable mid-range and is the benchmark the others are judged against, degrading toward the low-flow limit. Pressure falloff repeats to 1–3 % on the same instrument and agrees with steady state to ±3–5 % mid-range, widening toward ±10 % at both edges of its span. Pulse decay holds ±5–10 % in the microdarcy range; at the nanodarcy level 10–20 % is honest, and the limiting factor is the thermal environment and leak budget rather than the instrument itself. The crushed-rock method repeats within a laboratory to perhaps ±30–50 %, but interlaboratory comparisons on identical shale material have shown spreads up to an order of magnitude, driven by crush size and model assumptions. The probe permeameter is a profiling tool: ±20–50 %, and factor-of-two agreement with plug measurements is considered good. Helium porosity by Boyle expansion, for its part, holds ±0.1–0.2 porosity units under a maintained billet calibration, and quietly degrades when that calibration is neglected.
One deliberately sobering point of perspective: twin plugs cut a few centimetres apart in the same core routinely differ by tens of percent, and a mismatch in net confining stress between two laboratories shifts results by more than any instrument error above. A permeability value is only comparable when it travels with its stress state, its gas, and its correction status: a Klinkenberg-corrected and an uncorrected number on the same tight sample can legitimately differ several-fold with both being “right”.
11. Time per measurement, and what actually breaks
Time. The physics sets the clock: every transient method waits out a decay whose time constant scales inversely with permeability, so measurement time is not a specification the vendor chooses. It is the rock's answer to the reservoir volume and restrictor the instrument selected. Concrete numbers under automation: steady state needs 10–60 minutes per point for flow and temperature to stabilize, stretching to hours below a millidarcy; pressure falloff resolves a point in 2–10 minutes, and a full automated Klinkenberg sequence of three to four mean pressures completes in 15–45 minutes; pulse decay runs from roughly half an hour at 0.1 mD to a full working shift at single nanodarcies; the crushed-rock test takes 15–30 minutes; the probe reads in seconds. Add the overhead every schedule forgets: loading the plug, ramping confining stress, and letting sleeve and sample reach mechanical equilibrium. On tight rock this often takes longer than the measurement itself.
Reliability. A gas permeameter is, mechanically, a simple machine (reservoirs, valves, transducers, a core holder), and a well-maintained one runs for decades: instruments of this class built in the 1990s remain in routine laboratory service today. Failures cluster in predictable places, and nearly all of them are consumables rather than the instrument proper:
| Component | Typical service life | Dominant failure mode |
|---|---|---|
| Rubber sleeve (core holder) | 50–200 loadings / 6–18 months | hardening, extrusion at high confining stress, pinhole leaks |
| O-rings and valve seats | 1–3 years | cycling wear; leaks appearing after sample changes |
| Solenoid / pneumatic valves | 3–5 years (105–106 cycles) | seat leakage, coil failure |
| Pressure transducers | 5–10+ years | slow drift (annual recalibration), overpressure damage |
| Thermal mass-flow meters | 5–10 years | zero drift, contamination; annual per-gas recalibration |
| Confining-pressure pump seals | 1–2 years | gradual pressure loss under load |
Note what the table implies: a transient instrument without flow meters simply has fewer calibration-bearing components to drift. The practical availability figure for a plug-scale gas permeameter under an annual preventive-maintenance routine (a day or two of service plus transducer recalibration) is upwards of 95 %. And most unscheduled downtime is not a component failure at all: it is a fitting leak introduced during a sample change, which the instrument's own leak test finds in minutes, provided the test is actually run. Gas supply hygiene closes the list: particulates and moisture from cylinders are a slow poison for valve seats and flow sensors, and an inlet filter is the cheapest reliability upgrade an instrument can have.
Comparison summary
| Method | Range | Time per point | Flow meter | Confining stress | Typical uncertainty | Key limitation |
|---|---|---|---|---|---|---|
| Steady state | 0.1 mD – 10 D | 10–60 min; hours below 1 mD | yes, per-gas calibration | yes | ±2–5 % | slow and insensitive below 0.1 mD |
| Pressure falloff (1 reservoir) | 0.001 mD – 5 D | 2–10 min; Klinkenberg seq. 15–45 min | no | yes | ±3–5 % (±10 % at edges) | model-based; leak and temperature sensitivity at low k |
| Pulse decay (2 reservoirs) | 1 nD – 0.1 mD | 0.5–8 h | no | yes | ±5–10 %; up to 20 % at nD | slow; unusable at moderate-to-high k |
| Crushed rock (GRI) | 1 nD – 0.01 mD | 15–30 min | no | no | ±30–50 %; ×10 between labs | destructive; crush-size dependent; matrix only |
| Probe permeameter | 1 mD – 10 D | 5–15 s | yes | no | ±20–50 % | semi-quantitative, surface only |
No single method spans the full range of reservoir rocks. A well-designed laboratory instrument therefore either concentrates on one band and does it rigorously, or pairs a transient primary method with a steady-state verification mode for cross-checking in the overlap region. The overlap in Figure 1 is not redundancy, it is quality control.
References. API RP 40, Recommended Practices for Core Analysis, 2nd ed., American Petroleum Institute. · Klinkenberg, L.J. (1941), The permeability of porous media to liquids and gases. · Brace, W.F. et al. (1968), Permeability of granite under high pressure. · Jones, S.C. (1972), A rapid accurate unsteady-state Klinkenberg permeameter. · Dicker, A.I. & Smits, R.M. (1988), A practical approach for determining permeability from laboratory pressure-pulse decay measurements.
