Viscous Fingering (Saffman-Taylor)

The intuition — a runaway bump

Squeeze a blob of water into a tray of honey and the water doesn't advance as a tidy wall. Within a fraction of a second the boundary sprouts narrow tongues that race ahead, split, and branch — a pattern that looks like frost on a window or the veins of a leaf. That branching front is viscous fingering, and the runaway is the Saffman-Taylor instability.

The whole effect comes down to a feedback loop on the interface. Imagine the front is almost flat, with one tiny bump poking forward into the thick fluid. The pressure that drives the flow has to fall across the resisting (high-viscosity) fluid ahead of the front. A bump that pokes forward sits closer to the low-pressure region, so it feels a steeper local pressure gradient than the flat parts beside it. Steeper gradient → faster local velocity → the bump grows. A bigger bump feels an even steeper gradient, so it grows faster still. That is positive feedback, and positive feedback on a smooth surface is the textbook recipe for an instability.

Reverse the fluids — push the honey into the water — and the same geometry produces negative feedback: a bump now pokes into the easy, low-resistance fluid, so it feels a gentler gradient than its neighbors and falls back into line. The front self-heals and stays flat. The asymmetry is the heart of the phenomenon: fingering only happens when the less viscous fluid is doing the pushing.

The Hele-Shaw cell — fluids as porous media

You can't see fingering inside an oil reservoir, so the physics is studied in a Hele-Shaw cell: two flat plates separated by a thin gap b, often less than a millimetre. In such a thin gap, inertia is negligible and viscosity dominates so completely that, after averaging across the gap, the velocity obeys Darcy's law:

u = −(b² / 12μ) ∇p

Velocity is simply proportional to the pressure gradient, with the gap and viscosity setting the proportionality. This is identical in form to Darcy's law for slow flow through a porous rock, where the permeability k plays the role of b²/12. That equivalence is why a Hele-Shaw cell between two sheets of glass is the laboratory stand-in for an oil-bearing sandstone.

Inside each fluid the pressure obeys Laplace's equation (the velocity field is incompressible and curl-free), so the problem is a moving-boundary Laplacian growth problem — the same mathematical family as electrostatics, heat flow, and dendritic crystal growth. The interesting physics all lives at the moving interface.

The math — when do fingers grow?

Take a flat interface advancing at speed U and add a small sinusoidal ripple of wavenumber q = 2π/λ. Linear stability analysis (perturbation amplitude ∝ e^ωt) gives the growth rate

ω(q) = U·q · (μ₂ − μ₁)/(μ₂ + μ₁)  −  (b²/12)·(σ/(μ₁+μ₂)) · q³

Read the two terms physically:

• The destabilizing term (∝ q) carries the factor (μ₂ − μ₁), where μ₂ is the displaced fluid and μ₁ the injected one. It's positive only when μ₂ > μ₁ — thick fluid being displaced by thin. That's the viscosity contrast feeding the bump.
• The stabilizing term (∝ q³) carries surface tension σ. Curving a short-wavelength ripple costs interfacial energy, and the penalty grows steeply with q, so surface tension always wins at short wavelengths.

Because growth is positive at small q and negative at large q, there is a single fastest-growing wavelength λ* that emerges first and sets the initial finger spacing. Maximizing ω(q) gives a critical wavenumber where the two terms balance; the most-unstable mode scales as

q* ∝ sqrt( U·(μ₂−μ₁) / (σ·b²) )      λ* ∝ b · sqrt( σ / (μ·U) ) = b / sqrt(Ca)

The two dimensionless groups doing all the work:

Mobility ratio   M  = μ_displaced / μ_displacing   (unstable when M > 1)
Capillary number Ca = μ·U / σ                       (sets finger fineness)

Higher Ca — faster push, lower surface tension, or higher viscosity — shrinks λ*, so you get finer and more numerous fingers.

The selected finger — the ½-width law

Saffman and Taylor's 1958 paper did more than predict the onset. For a single finger advancing steadily down a channel of width W, they found an exact family of finger shapes parameterized by the fraction λ_f of the channel the finger fills:

x = (W·(1−λ_f) / 2π) · ln[ ½(1 + cos(2π·y / (λ_f·W))) ]

Their analysis (with zero surface tension) allowed any λ_f between 0 and 1, but every experiment showed the same answer at high speed: the finger fills almost exactly half the channel, λ_f → ½. Resolving that puzzle — why nature selects ½ out of a continuum — took until the 1980s, when McLean, Saffman, Vanden-Broeck, and others showed that surface tension is a singular perturbation that breaks the degeneracy and selects a discrete width. The selected fraction approaches ½ from above as Ca → ∞ and widens as Ca falls. It is one of the cleanest examples in physics of a tiny term (surface tension) controlling a macroscopic outcome.

Regimes and numbers

Quantity / regime	Symbol or value	Effect on the front
Mobility ratio M > 1	thin pushes thick	Unstable — fingers grow
Mobility ratio M < 1	thick pushes thin	Stable — front stays flat
Capillary number Ca	μU / σ	Higher Ca → finer, more fingers
Fastest-growing wavelength	λ* ∝ b / √Ca	Sets initial finger spacing
Steady channel finger	λ_f → ½ W (high Ca)	One finger fills half the channel
Radial injection, σ → 0	fractal dimension ≈ 1.7	Ramified, DLA-like branching
Typical Hele-Shaw gap b	0.1–1 mm	Thinner gap → smaller permeability b²/12
Water vs glycerol contrast	μ ≈ 1 vs ~1000 mPa·s	M ≈ 1000 → strongly unstable

Real-world figures and costs

• Enhanced oil recovery. Water- and gas-flooding push oil with a fluid that's far less viscous (water ~1 mPa·s vs heavy crude 100–10,000 mPa·s), so M is hugely greater than 1. Fingering causes early water breakthrough that strands oil: primary plus secondary recovery typically leaves 50–70% of the oil in place. Polymer flooding adds high-molecular-weight polymers (HPAM, xanthan) to raise the injected water's viscosity toward the oil's, cutting M and recovering an extra 5–15% of original oil in place — billions of barrels at field scale.
• CO₂ geological storage. Supercritical CO₂ injected into saline aquifers is less viscous and less dense than brine (M > 1), so it fingers upward and sideways. Predicting that fingering is central to estimating how much CO₂ stays trapped over the centuries required for climate-relevant storage.
• Chromatography and filtration. Viscous (or density) fingering of the sample band through a packed column smears peaks and degrades separation resolution; column designers fight it with viscosity matching and flow-rate limits.
• Printing and coating. When a coating fluid is split between a roller and a substrate, the receding meniscus fingers — the "ribbing" or "printer's instability" that streaks ink and paint. Coating lines run below a critical capillary number to stay finger-free.
• Geological fluid intrusion. The same Laplacian-growth physics describes how injected fluids, groundwater, and even magma finger their way through fractured crust and sediments.

Where it shows up

• Petroleum engineering. The original and dominant application — sweep efficiency, water breakthrough, polymer and foam flooding all hinge on controlling the mobility ratio.
• Carbon capture and storage. Finger-controlled migration and trapping of supercritical CO₂ in aquifers.
• Hydrology and contaminant transport. Non-aqueous-phase liquids fingering through soil and groundwater control how pollutants spread.
• Microfluidics. Engineers deliberately exploit or suppress fingering to mix or to keep co-flowing streams separate in lab-on-a-chip devices.
• Pattern-formation physics. A benchmark system, alongside dendritic crystal growth and DLA, for studying how a smooth front becomes a complex branched structure.
• Biology-inspired growth. The branching morphology mirrors bacterial colony spread, river deltas, and lung airways, all governed by similar diffusive/Laplacian fields.

Common misconceptions and edge cases

• "Fingering needs turbulence." No — it's a creeping-flow, low-Reynolds-number instability. Inertia is irrelevant; viscosity dominates. It happens at essentially zero Reynolds number.
• "Surface tension causes the fingers." Backwards. The viscosity contrast causes the instability; surface tension stabilizes short wavelengths and selects the finger width. Remove surface tension entirely and the pattern becomes wilder (fractal), not calmer.
• "It only matters in exotic labs." It controls the economics of every waterflooded oil field and the safety case for CO₂ storage — multi-billion-dollar consequences.
• "Density (gravity) and viscosity fingering are the same." Related but distinct. A dense fluid sinking into a light one (Rayleigh-Taylor) is gravity-driven; Saffman-Taylor is viscosity-driven and can occur with the fluids density-matched. Both can act together — "gravity override" in reservoirs combines them.
• "Faster injection always sweeps better." The opposite at the front: higher velocity raises Ca, shrinks λ*, and makes the front more finger-prone. Mobility control, not raw speed, is the lever.
• "A single finger is the whole story." The clean half-width finger is the channel-geometry limit. In radial injection or a porous medium, fingers continually tip-split and branch into a self-similar tree.