Retaining Wall Sliding: The Check That Fails Before Overturning

The Wall Passed, and Then It Walked

A 5.5 metre reinforced block face runs along the low side of a distribution yard, engineered and inspected, with a design file that records 1.6 against sliding and 2.4 against overturning. Both numbers are comfortably above what the reviewing authority asked for. The wall is backfilled in 200 mm lifts, the drainage stone is clean, the grid is the length on the drawing, and for eighteen months nothing happens.

Then a contractor arrives to lay a fibre duct along the toe. The trench is 900 mm deep and follows the wall for sixty metres because that is where the easement runs, and it stays open for nine days over a wet fortnight. Nobody on that crew has seen the design file and nothing on site says one exists. By the time the trench is backfilled with the arisings, the face has moved forward between 30 and 40 mm at mid-height, the vertical joints have opened uniformly rather than in one place, and the coping has cracked at two movement joints. The wall has not rotated. It has translated, on a plane below the visible work, and it will not go back.

Everything in that sequence was foreseeable from the design file, because the 1.6 was a number that counted soil in front of the toe. Sliding is the check with the least margin on a wide wall, it is the check whose resistance can be reduced by somebody who has never heard of the wall, and it is the only structural quantity on a civils job that a third party can remove with a machine in an afternoon.

The Friction Coefficient Belongs to an Interface You Build

Sliding resistance is not a soil property. It belongs to a contact between two materials, and the material on the underside of the wall is chosen by whoever detailed the base. Concrete cast directly against a compacted granular formation gives an interface friction angle approaching the soil's own, because failure has to work through grains keyed into the paste. Concrete cast on a blinding layer, a screeded mud slab or a sheet of polythene gives an interface with a friction angle of its own, and that number is not the soil's.

Named sources exist for this, and they beat a value carried in from another job. NAVFAC DM 7.02, Foundations and Earth Structures, tabulates coefficients of friction by material pair, masonry and concrete against various soils. The AASHTO LRFD Bridge Design Specifications carry friction factors for base sliding within their earth retaining structures provisions. Where a geosynthetic is involved the governing document is a test method rather than a table: ASTM D5321 determines the shear strength of a soil-geosynthetic interface by direct shear, and a manufacturer quoting an interaction coefficient is quoting a D5321 result against a specific fill gradation. That last qualifier does most of the damage on site, because a value developed against a well-graded sand does not transfer to open-graded drainage stone or to a site-won silty fill.

The interfaces that get built without ever appearing on a stability calculation are worth walking the formation to find. A separation geotextile keeping the pad off a soft subgrade is a plane. A drainage geocomposite picking up seepage under a levelling pad is a plane. A polythene slip membrane laid to stop the mix drying into the ground is a plane, and one a groundworker installs as a matter of routine because it makes the pour better, without anyone connecting it to a number in a design file.

Where the friction value came from is therefore a question for the drawing, not only for the calculation. If the design assumed concrete cast against soil, the drawing has to say so, and the specification has to prohibit the blinding a competent groundworker would otherwise put down to get a clean line. This is one of the few structural assumptions on a civils job that a site instruction can quietly invalidate without anybody making a mistake.

Water Pushes Harder and Holds Less Down

Water attacks the sliding check from both ends at once, which is what makes a blocked drain so much worse than an equivalent increase in surcharge. On the driving side, water pressure has no earth pressure coefficient to reduce it: it acts at full hydrostatic intensity. For a granular fill at an ordinary friction angle that hydrostatic term on its own is already around half again to twice the lateral pressure the drained soil was applying over the same metre, and it does not replace the soil's contribution — the buoyant soil still pushes alongside it, so the saturated case lands near double the drained one. On the resisting side, that same water reduces the effective weight of everything below the water line to its buoyant value, which for ordinary fill is roughly half the moist unit weight. Half the effective vertical load is half the friction.

Then there is uplift. Water at base level does not stop at the back of the heel; it acts on the underside of the base as an upward pressure that has to come off the normal force before the friction term is calculated at all. A wall founded below a seasonal water table, or built into a bank that perches water on a clay horizon, can have a meaningful fraction of its vertical load cancelled by pressure that is invisible on a dry inspection.

This is why drainage behind a retaining wall is a structural component rather than a durability one, and why it deserves checking as such: a continuous open-graded chimney behind the face, a collector with a fall and a real outfall, a filter arrangement that keeps fines out of the stone, and an outlet somebody can find and rod in ten years. The sliding check a wall passes is a check on the drained condition. If nothing on site guarantees that condition, the check was run on a wall nobody built.

The Shear Key, and What It Is Honestly Doing

When a cast-in-place wall is short on sliding and the base cannot get any wider, the usual answer is a key cast down from the underside of the base. What the key does is worth being precise about, because it is often described as though it simply pins the wall to the ground.

It does two things. It forces the potential slip surface through soil for part of its length instead of along the concrete-to-soil contact, so the strength mobilised over that stretch is the soil's own rather than a reduced interface value — the more reliable of the two benefits, and the one that survives site variation. And it deepens the plane on which passive resistance acts, developing that resistance below the depth a service trench normally reaches. Real, but it is the same passive resistance relocated: it still needs movement to mobilise, and it is still in front of the wall where somebody may dig.

Whether the key sits under the stem or under the heel has been argued for decades, and the answer turns on which soil the designer trusts and on how the base is reinforced — a key under the stem continues the stem steel, while a key elsewhere is a detail in its own right. What is not in doubt is the cost: a trench cut into the formation, a formation that then has to stay open and clean, and steel that has to be detailed and fixed. On many walls the honest comparison is against extending the heel and accepting a slower gain, or against improving the interface by casting the base directly on a compacted granular formation and deleting the blinding, which costs a note on a drawing.

An Order That Finds the Answer

The sequence matters because most of the wrong answers come from establishing the resisting side before the demand side is settled, and then adjusting inputs until the ratio clears. Fix the demand first, from ground the geotechnical report describes, and let the resistance be whatever the section actually delivers.

1. Take the retained height from the underside of the base, at the section with the highest fill or the deepest scour in front, not the average.
2. Pull friction angle and unit weight for the retained fill and for the foundation soil separately — frequently different materials on the same job.
3. Compute the active thrust with the surcharge that will genuinely be present, including construction plant working behind an unbraced wall.
4. Total the vertical load on the base plane: stem, base slab, the soil column standing on the heel, and anything bearing on that column that is certain to remain.
5. Choose the interface friction value from a named source, for the contact you are going to build, and write that contact onto the drawing so it cannot be substituted on site.
6. Run sliding with no passive resistance whatsoever, and record it. That is the number the wall has to live with on its worst day.
7. Run it again with the passive resistance you believe you have. The difference between the two is the exposure a trench or a regrade converts into movement.
8. On a reinforced wall, repeat at every reinforcement elevation using the manufacturer's tested interaction coefficient rather than a soil value.
9. Check the water case explicitly: fill saturated to the design level, drainage assumed blocked, uplift acting on the underside of the base.
10. Compare against the thresholds in one document, not several, and name that document on the calculation sheet.

What the Number Has to Beat, and in Which Document

Allowable-stress practice has settled on roughly 1.5 against sliding and roughly 2.0 against overturning for permanent walls, with the passive term either neglected outright or heavily discounted, and many designers who do count it raise the sliding requirement to compensate. Those are service-level ratios computed on unfactored loads, and they are the numbers most preliminary checks — including the ones on this site — report.

They are not comparable with a limit-state result. The AASHTO LRFD Bridge Design Specifications replace the ratio entirely with load factors on the driving side and resistance factors on the capacity side, so a wall that satisfies AASHTO carries no directly readable factor of safety. Eurocode 7, BS EN 1997-1, does the same job differently again, applying partial factors to actions, material properties or resistances depending on which Design Approach the National Annex adopts — the same wall on the same ground produces different numbers in two European jurisdictions, and neither is a safety factor. NCMA's manual sets thresholds for segmental walls, and FHWA-NHI-10-024 does so for MSE structures.

Mixing them is the error that gets published. A thrust computed with a partial factor on the soil's friction angle, divided into a resistance computed with an unfactored interface coefficient, produces a ratio that belongs to no document. Pick the one the reviewing authority will read, run everything through it, and name the edition on the calculation sheet. A preliminary ratio is a screening tool for finding which check governs and how much room there is to move: it establishes demand, not adequacy, and adequacy on a permanent retaining structure is a conclusion a qualified geotechnical or structural engineer signs.

Put the Ground in Front on the Record

The failure at the top of this page was not an engineering error. It was a communication failure with an eighteen-month fuse, and it is the single most preventable thing on this page. If the sliding check counted passive resistance, then the depth and quality of soil in front of the wall is a structural assumption, and structural assumptions belong somewhere a stranger with a machine will encounter them: a note on the general arrangement drawing stating the minimum cover to be maintained in front of the toe, the same note in the health and safety file or the operation and maintenance manual, and a line in the asset record if the client keeps one.

Where the wall sits alongside an easement or a service corridor, go further and say what has to happen before anyone excavates within a stated distance of the face: the designer consulted, and no trench left open along the toe. The excavation itself is separately governed as a safety matter, under OSHA's requirements in 29 CFR 1926 Subpart P and the equivalent regime elsewhere, but those rules protect the people in the trench. Nothing in them protects the wall beside it. That has to be written by whoever ran the check, at the time they ran it, because when it is needed they will be long gone and the only thing speaking for the wall will be the drawing.