Soil Liquefaction

A sand pile turning to mud

Walk along a wet beach at low tide and stamp your foot on the sand. The sand goes momentarily dark and soft under your boot, then dries again and turns pale. You have just done a tiny demonstration of soil liquefaction. The cyclic shear loading from your foot has rearranged the sand grains, and the pore water between them has been forced upward through the surface. For a fraction of a second the sand had no effective stress at the contact points; the water carried the load.

Now scale that up. A magnitude 7 earthquake delivers tens of cycles of strong shaking at frequencies of 0.5–5 Hz, lasting 20–60 seconds. The ground accelerates back and forth at 0.2–0.5 g. Saturated loose sand experiences exactly the same physics as the wet beach under your boot — but at depths of 5, 10, 15 metres, and over a footprint of square kilometres. The soil layers below the water table try to compact under cyclic shear; the pore water resists by carrying the load; pore water pressure climbs cycle after cycle; and after enough cycles the pressure equals the total overburden stress. Effective stress is zero. Shear strength is zero. The soil flows.

Above the liquefied layer, anything that depended on that layer for support gives way. Reinforced concrete buildings tilt over (Niigata, 1964). Bridges shift on their piers (Loma Prieta, 1989). Buried sewer manholes float up out of the ground (Christchurch, 2011). Sand boils — small mud volcanoes — erupt across kilometres of pavement. Roads buckle and crack in extension where lateral spreading has pulled them apart.

The physics: effective stress and pore pressure

The key concept is Terzaghi's effective stress principle. The total vertical stress in saturated soil at depth z is σ_v = γ_sat · z. The pore water pressure at hydrostatic equilibrium is u = γ_w · z_w, where z_w is depth below the phreatic surface. Effective stress — the stress actually carried by the soil's particle skeleton — is

σ'_v  =  σ_v − u

Shear strength of a cohesionless soil:
τ_f  =  σ'_v · tan(φ')

Shear strength is proportional to effective stress. If u rises until u = σ_v, then σ'_v = 0, and τ_f = 0. The soil has no resistance to shear — by definition, a liquid.

Why does u rise during earthquake shaking? Loose granular soil under shear tends to compact (positive dilatancy is rare; most loose sands are contractive). The grains want to settle into denser packing. But the void space is full of water, which cannot escape on the timescale of an earthquake (the seismic wave period is seconds; consolidation drainage for fine sand is hours to days). Water is essentially incompressible, so to keep the same volume occupied while the grain skeleton tries to contract, the water must take some of the load. Its pressure rises. Each cycle of shear adds a bit more to u. After enough cycles — typically 10–20 for loose clean sand at moderate shaking — u has risen to equal σ_v.

When (and where) does it happen?

Three conditions must all be met:

• Saturation. Soil must be below the water table or within a saturated zone. Dry sand cannot liquefy because there is no pore water to take the load. Partially saturated soil (S < 95%) is largely immune because air compresses to relieve pressure build-up. Saturated zones near rivers, lakes, coasts, and shallow groundwater are the high-risk geometry.
• Loose granular structure. Relative density typically below about 50%. The Standard Penetration Test blow count, normalised to 1 atm effective overburden and 60% hammer energy ratio (N_1,60), serves as the universal proxy. N_1,60 < 15 is "potentially liquefiable" under all but the weakest shaking; N_1,60 > 30 is essentially immune in normal earthquakes.
• Narrow grading. Clean uniform fine sands (D₅₀ ≈ 0.1–0.5 mm) are most vulnerable. Well-graded gravels resist because larger particles transmit load through grain-to-grain contact even as small particles re-shuffle. Plastic clays (plasticity index PI > 7) resist because cohesion provides shear strength independent of effective stress.

The worst geological environment is a Holocene-age (last 10,000 years) loose alluvial sand below a shallow water table. Hydraulic fills — soils placed by dredging or pumped slurry, with no compaction — are particularly susceptible and produce the most spectacular failures. Recent earthquake-resistant building codes in California, Japan, New Zealand, and Chile prohibit construction on such fills without ground improvement.

The Seed-Idriss simplified procedure

The standard engineering procedure, developed by H.B. Seed and I.M. Idriss in the 1970s and refined repeatedly since, compares the demand (cyclic stress ratio) against the capacity (cyclic resistance ratio):

CSR (Cyclic Stress Ratio — earthquake demand)
       = 0.65 · (a_max / g) · (σ_v / σ'_v) · r_d

where  a_max  = peak ground acceleration at the surface
       σ_v    = total vertical stress at depth z
       σ'_v   = effective vertical stress at depth z
       r_d    = depth reduction factor (≈ 1.0 near surface, ≈ 0.85 at 10 m)
       0.65   = factor relating uniform cyclic stress to peak

CRR (Cyclic Resistance Ratio — soil capacity)
       read from CRR-N1_60 chart (or CRR-q_c1N from CPT)
       For clean sand at N1_60 = 15:  CRR_M=7.5 ≈ 0.17
       For clean sand at N1_60 = 25:  CRR_M=7.5 ≈ 0.30
       Magnitude scaling factor MSF adjusts for earthquakes ≠ M 7.5

Factor of safety:
       FS  =  CRR / CSR  ·  MSF

       FS < 1.0          →  liquefaction predicted
       1.0 ≤ FS < 1.25   →  marginal; mitigate
       FS ≥ 1.25         →  acceptable

Inputs are PGA (from a probabilistic seismic hazard analysis), depth profile of total and effective stress (from groundwater and unit weights), and SPT blow counts (or cone penetration q_c, or shear-wave velocity V_s) measured at each candidate layer. Output is a depth profile of FS; layers with FS < 1.25 are flagged for mitigation.

Worked example: a saturated loose sand at 5 m depth

Site:        Holocene fine sand below 1 m of silty crust; water table at 1.5 m.
Soil:        γ_sat = 19.0 kN/m³, γ' (submerged) = 9.2 kN/m³
SPT data:    N_60 = 12 at z = 5 m
Earthquake:  M = 7.0, a_max = 0.30 g (design basis at this site)

Step 1: stresses at z = 5 m
   σ_v   = 1.0 · 18.5 + (5 − 1) · 19.0 = 18.5 + 76.0 = 94.5 kPa
   σ'_v  = σ_v − u = 94.5 − (5 − 1.5) · 9.81 = 94.5 − 34.3 = 60.2 kPa

Step 2: overburden-corrected N1_60
   C_N = √(100 / σ'_v) = √(100 / 60.2) = 1.29
   N1_60 = C_N · N_60 = 1.29 · 12 = 15.5

Step 3: CRR_M=7.5 from chart (clean sand)
   N1_60 = 15.5 → CRR_M=7.5 ≈ 0.18

Step 4: magnitude scaling factor
   MSF = (M_w / 7.5)^(-2.56) = (7.0/7.5)^(-2.56) = 1.20
   CRR_M=7.0 = CRR_M=7.5 · MSF = 0.18 · 1.20 = 0.216

Step 5: CSR
   r_d = 1 − 0.00765·z = 1 − 0.00765·5 = 0.96
   CSR = 0.65 · 0.30 · (94.5 / 60.2) · 0.96 = 0.65 · 0.30 · 1.57 · 0.96 = 0.294

Step 6: factor of safety
   FS = CRR / CSR = 0.216 / 0.294 = 0.73          ✗ LIQUEFIES

Mitigation needed: stone columns to raise N1_60 to ≥ 25 (CRR ≈ 0.36),
or deep foundations bypassing the liquefiable layer.

The factor of safety of 0.73 confirms that this loose Holocene sand will liquefy under the design earthquake. The engineer's choice is between ground improvement (raise CRR) or a deep foundation that bypasses the layer altogether. For a residential subdivision, vibro-compaction with stone columns is common; for a hospital or bridge, deep piles into the underlying non-liquefiable layer plus laterally robust pile caps are standard.

Surface signs of liquefaction

Phenomenon	Mechanism	Field evidence
Sand boils	Excess pore water pressure forces water + entrained sand upward through cracks	Conical sandy deposits 0.1–3 m diameter; sand "volcanoes" on streets and lawns
Lateral spreading	Surficial crust slides on liquefied layer toward free face or down slope	Open ground cracks parallel to riverbanks; offset roads; broken buried utilities
Flow failure	Liquefied mass flows out from beneath embankments, slopes, dams	Embankment slumps; downstream sand fans; reservoir breaches
Bearing failure	Liquefaction beneath a footing — soil cannot carry the building	Building tilts (Niigata Kawagishi-cho); foundation punches into ground; uneven settlement > 100 mm
Buoyancy uplift	Buried tank or pipe with average density < liquefied soil density	Buried sewer manholes pop up 0.5–2 m; underground fuel tanks float to surface
Ground settlement	Loose sand reconsolidates after shaking; pore pressures dissipate	Uniform or differential settlement up to 1% of saturated thickness (10 cm in 10 m of sand)
Pavement extension/compression	Lateral spread pulls and pushes pavement	Tension cracks; compression buckles; offset curbs and rail tracks

Historic case studies

• San Francisco, 1906 (M ≈ 7.9, San Andreas). Extensive liquefaction of hydraulic fills along the bay shore and the Marina, Mission, and South of Market districts. Photographs show open ground cracks, sand boils, and surface waves of soil. The post-quake fire was responsible for most of the property damage, but liquefaction undermined many buildings before they burned.
• Niigata, Japan, 1964 (M 7.5). The defining modern liquefaction event. Kawagishi-cho four-storey apartment buildings tilted up to 80° on their intact footings. Buried sewer manholes floated 1–2 m above ground level. Across vast urban areas, the upper 10–15 m of Holocene river sand lost shear strength. Niigata triggered the modern engineering literature on liquefaction.
• Anchorage, Alaska, 1964 (M 9.2). The same year as Niigata, this Great Alaska earthquake produced massive liquefaction in the Bootlegger Cove Clay overlying loose sand layers in the Anchorage area. The Turnagain Heights landslide moved 86 acres of suburb downhill toward Cook Inlet.
• Loma Prieta, California, 1989 (M 6.9). The Marina District in San Francisco — built on rubble fill from the 1906 quake and the 1915 Panama-Pacific Exposition — liquefied. Several apartment buildings collapsed; the Marina is the textbook example of fill-zone liquefaction in modern American urban planning.
• Christchurch, New Zealand, 2011 (M 6.2). Although smaller in magnitude than Niigata or Loma Prieta, the Christchurch quake was extremely shallow (5 km depth) and produced the highest PGA ever recorded on saturated alluvial soil. Vast areas of eastern Christchurch — the suburbs of Bexley, Avondale, Avonside — liquefied repeatedly during the 2010–2011 sequence. Sand boil deposits covered streets for kilometres. Some 7,000 residential properties were red-zoned and demolished.
• Tohoku, Japan, 2011 (M 9.1). Liquefaction extended over 400 km from the epicentre, particularly in Tokyo Bay reclaimed lands. Urayasu and Maihama suburbs settled 30–80 cm; sand boils erupted across Tokyo Disneyland car parks. Total liquefaction-induced economic damage estimated > $10 billion.
• Palu, Indonesia, 2018 (M 7.5). A "flow liquefaction" event in the Balaroa, Petobo, and Jono Oge districts caused entire neighbourhoods to ride downslope on liquefied silty sand for hundreds of metres, killing more than 2,000 people. A rare and devastating example of liquefaction-driven flow on very gentle terrain.

Mitigation methods

Method	Mechanism	Depth range	Best for
Vibroflotation / vibro-compaction	Vibrating probe inserted, water/air jetted, native soil densified by vibration	0–25 m	Clean sands with N1_60 ≥ 5 (works less well in silts)
Stone columns (vibro-replacement)	Vibroflot creates hole, gravel backfilled and compacted in stages — drainage + densification	0–30 m	Silty sands, low-fines silts; provides drainage path
Dynamic compaction	Heavy weight (10–30 t) dropped repeatedly from 10–30 m onto surface	0–10 m	Shallow loose granular, fills, debris
Compaction grouting	Stiff cement-soil mortar injected at high pressure to displace and densify soil	0–40 m	Restricted-access sites, urban retrofits
Permeation grouting	Low-viscosity grout (sodium silicate, microfine cement) injected to bond grains	0–30 m	Silty/sandy soils, surgical mitigation under existing structures
Jet grouting	High-pressure cement slurry jet erodes and remixes soil into hardened columns	0–25 m	Variable soils, retrofit, basement walls
Wick drains / PVDs	Vertical plastic drains accelerate pore-pressure dissipation	0–30 m	Pre-consolidation of soft clays; sometimes for liquefaction drainage
Excavate and replace	Remove liquefiable layer; backfill with engineered granular fill or LMSL	0–6 m practical	Shallow problems, small footprints
Deep foundations	Piles or drilled shafts pass through liquefiable layer to firm stratum	any	Major structures (buildings, bridges); residual lateral spread loads must be designed for

The most common combinations are stone columns plus deep foundations for major buildings on liquefiable ground, and vibro-compaction plus surface raft for residential subdivisions. Christchurch post-2011 saw widespread use of stone columns and stabilised crust (cement-treated subbase 0.6–1.2 m thick) to protect single-family houses.

Cross-section: a layered profile that liquefies

PRE-EARTHQUAKE                  DURING SHAKING

  ░░░░░  silty crust (1 m)         ░░░░░  cracked, surface waves
  ──────────── water table ────    ──────────── lateral spreads
  ▒▒▒▒▒  loose Holocene sand       ▒▒▒~~~  liquefied — u → σ_v
  ▒▒▒▒▒  N1_60 = 12, saturated     ~~~~~~~  flowing
  ▒▒▒▒▒                            ~~~~~~~  sand boils erupt
  ─────────────                    ─────────────
  ░░░░░  dense Pleistocene sand    ░░░░░  intact (N1_60 = 30)
  ░░░░░  N1_60 = 30, non-liquef.   ░░░░░
  ░░░░░                            ░░░░░
  ───────  bedrock                 ───────  bedrock, seismic input

Above the liquefied layer, a surface crust 0.5–2 m thick of stiffer silty soil typically remains intact. During shaking, this crust slides as a relatively rigid block on the soft liquefied layer beneath — that horizontal motion is the lateral spreading. After shaking ends, pore pressures dissipate over hours; the sand layer re-consolidates and ground-level settlements of 1–5% of the liquefied thickness develop.

Common pitfalls in liquefaction evaluation

• Treating silt as sand. Silts and silty sands behave very differently from clean sands; plastic silts can be non-liquefiable even at low N1_60. Use the Bray-Sancio or Idriss-Boulanger framework for fine-grained soils.
• Ignoring post-liquefaction settlement. Even when the liquefiable layer is "only" 1–2 m thick, reconsolidation settlements of 5–10 cm can damage shallow-foundation buildings even if no tilt occurs.
• Forgetting lateral spread. A deep foundation through a liquefiable layer must resist the lateral push from the spreading crust above — typically a passive-earth-pressure load on the upper several metres of pile. Failing to design for this load broke many piled buildings in Christchurch.
• Assuming dense surface fill is safe. If the deeper layer liquefies and the fill is on top of it, the fill rides the liquefaction. Vibro-densifying a 10 m thick natural deposit is sometimes cheaper than designing the structure to ride a flowing layer.
• One borehole per site. Liquefiable layers can be discontinuous lenses. Investigation must include enough boreholes (typically one per 30 m grid for a building site, plus CPT or shear-wave-velocity profiles) to map the lateral extent.
• Old codes. Pre-1980 building codes did not address liquefaction. Many older structures on liquefiable ground require retrofit; the City of San Francisco runs ongoing programs to identify and upgrade vulnerable apartment buildings in Marina-style fill zones.