Soil Compaction: How to Specify and Verify Fill Density

A slab-on-grade poured over fill that was never compacted properly does not fail on the day it is poured. It fails a year later, when the fill has settled unevenly and the slab has cracked along the line of a service trench nobody remembers backfilling. The compaction of that fill was decided by a moisture content and a number of roller passes, both of which were set by someone, and both of which are testable.

This guide covers the two halves of the problem. The first is the specification: what “95 percent compaction” actually means, and why the standard the laboratory result came from matters as much as the percentage. The second is verification: how field density is measured, what the arithmetic looks like, and what to do when the test fails. It closes with a sand replacement test on a fill lift worked through step by step, including the water that has to be added to make it pass.

The boundary with the laboratory sits where it always has. The Proctor test itself, and the moisture-density curve it produces, belong to the soils laboratory. What follows is what the field result means and how it is judged.

Compaction is not consolidation

Both words describe soil getting denser, and they are used interchangeably in the field. They are different mechanisms with different timescales, and confusing them explains most of the bad arguments about fill.

Compaction is densification by mechanical effort, applied to soil at a water content that does not change much during the process. Air leaves the voids. It happens in minutes, under a roller.

Consolidation is densification by the slow expulsion of water under a sustained load, and it takes months or years. It is what a foundation settlement calculation is about.

The quantity both are measured by is the same: the dry density, the mass of solid particles per unit of total volume, with the water excluded.

$\gamma_d = \frac{\gamma}{1 + w} = \frac{M_{wet}}{V\,(1 + w)}$

where $\gamma$ is the wet (bulk) density of the soil mass, $w$ is the water content as a decimal, $M_{wet}$ is the wet mass and $V$ is the total volume. This is the equation the whole business turns on. It is why a fill that looks wet and heavy is not automatically dense: water adds mass and volume together, and contributes nothing to density.

The moisture-density curve

Take a soil, add water in steps, and compact each batch with the same controlled effort. Plot dry density against water content and it rises, peaks, and falls. The peak is the maximum dry density (MDD, $\gamma_{d,max}$), and the water content under it is the optimum moisture content (OMC).

The shape has two mechanical explanations, and they are worth keeping because they predict site behaviour:

• Dry of optimum: the suction in the pore water holds the particles in clusters and the friction is high, so the rammer or roller cannot rearrange them efficiently. Density climbs as water is added because the water lubricates the contacts.
• Wet of optimum: water starts occupying the voids that the soil could otherwise fill with solids. There is not enough air left to expel, and density drops.

The practical consequence is that the curve is usually steeper on the dry side than on the wet side, although not for every soil: some clays give a curve that is close to symmetric. On a soil with a steep dry side, a fill placed 2 percent dry of optimum loses density quickly, while the same fill 2 percent wet of optimum keeps its density but gives up strength and gains pore pressure. Missing low costs density; missing high costs strength. Which of the two hurts depends on what the layer is for.

Standard versus modified Proctor, and why the number is meaningless without it

Two laboratory procedures in common use produce two different curves for the same soil:

Those parameters are the Method A ones. Both standards also offer Method C, which uses the 152.4 mm (6 in) mould, 56 blows per layer and material passing the 19 mm (3/4 in) sieve, for soils with coarse particles. The method designation matters because the moisture-density curve is only valid for the grading it was made from, and a specification that quotes a percentage of “the Proctor” without the method has not named a requirement.

The energy delivered per unit volume is not a table entry to memorise, it can be calculated:

$E = \frac{m\,g\,h\,N_{blows}\,N_{layers}}{V}$

For the standard test:

$E = \frac{2.495 \times 9.807 \times 0.305 \times 25 \times 3}{943.9 \times 10^{-6}} = 593{,}000\ \text{J/m}^3 \approx 593\ \text{kN·m/m}^3$

and for the modified test, with a 4.536 kg rammer, a 457 mm drop and five layers:

$E = \frac{4.536 \times 9.807 \times 0.4572 \times 25 \times 5}{943.9 \times 10^{-6}} = 2{,}693{,}000\ \text{J/m}^3 \approx 2{,}693\ \text{kN·m/m}^3$

The modified test delivers about 4.5 times the energy, so it produces a higher MDD and a lower OMC on the same soil. This is why a specification that says “95 percent compaction” without naming the standard is not a specification. Ninety-five percent of a modified Proctor is a heavier demand than 95 percent of a standard one, and where the two curves differ by more than about 5 percent in density, the modified requirement is heavier than 100 percent of the standard curve. The two numbers cannot be compared to each other, or carried from one project to the next.

Relative compaction and how the specification is written

The acceptance criterion is the ratio of what was achieved to what the laboratory found:

$RC = \frac{\gamma_{d,field}}{\gamma_{d,max}} \times 100$

Two conventions exist in parallel specifications and they are not synonyms:

• Percentage of maximum dry density (relative compaction), the usual formulation in US practice: the field dry density is expressed as a percentage of the laboratory MDD on the same material.
• Minimum dry density in absolute units, used where a design value comes from strength or stiffness (a CBR or plate test) rather than from the laboratory curve.

Typical acceptance values, each relative to the standard named in the row. The number and the testing frequency come from the project specification, and this is the range it normally picks from:

The moisture content is specified alongside the density, and this is the half that gets ignored: most end-result specifications require the material to be placed within a window around the OMC, most often about two percentage points either side of it and one point on more demanding work. A layer can pass on density and still be outside the window, and that is a result the specification can reject.

One more consequence of the definition deserves attention. A field result above 100 percent is not a triumph, it is a signal. It normally means the material being tested is not the material the laboratory curve was made from: the grading has changed, a different borrow source is on site, or the laboratory curve was made on the fraction passing the sieve while oversize particles sit uncorrected in the fill. Heavy rolling on the dry side of optimum, where the laboratory curve says the material cannot reach that density at that water content, is the other way to get there. All of them deserve the same investigation as a failure.

Who carries the risk: method specification or end-result specification

Two ways of writing the same requirement, with opposite risk positions:

• Method specification. The specification prescribes how: the lift thickness, the moisture range, the class of roller and the number of passes. The contractor complies by following the method, and the client accepts the outcome of a correctly followed method. This is the basis of United Kingdom highways earthworks, where the Specification for Highway Works works by method and adds its own testing clauses, and it suits work where the material and the equipment are both well understood.
• End-result specification. The specification prescribes only the outcome: a relative compaction, a moisture window and a testing frequency. The contractor chooses the equipment and the number of passes, and takes the risk of achieving the numbers. This is the usual position in North American earthworks.

In practice most specifications are hybrids: an end-result density with a prescribed lift thickness and a minimum roller class, which keeps the number that matters and controls the two variables that most often break it.

The trap in a method specification is stating a method that does not work for the material actually supplied, because the risk of a failed fill then sits with whoever wrote the method. The trap in an end-result specification is testing too little: a density result describes the spot where the test was taken, and nothing else. Specifications handle that with a stated frequency and a defined consequence for a failing result.

How field density is measured

Four methods cover almost all site work, and they differ in what they suit rather than in accuracy:

• Sand replacement (sand cone), ASTM D1556. A hole is dug in the compacted layer, the excavated soil is weighed and its water content determined, and the volume is found by filling the hole with sand of known density. It is slow, it is the reference method for coarse soils, and it is the one to use on gravelly fills.
• Nuclear gauge, ASTM D6938. Gamma rays measure density and neutrons measure moisture, both within seconds. Fast, and the standard method on large earthworks. The moisture reading is the less reliable of the two, because it infers water from hydrogen and is thrown by organic matter or cementitious material, and the gauge needs its daily standard counts plus a correlation against a direct method on each material. The depth it senses depends on the mode: roughly the top 100 mm in backscatter, up to 300 mm with the source rod inserted in direct transmission.
• Balloon density, ASTM D2167. A rubber balloon measures the hole volume for cohesive soils, faster than the sand cone and less suited to open-graded material.
• Core cutter or drive cylinder. A thin-walled tube of known volume is driven into fine-grained soil and the soil is trimmed and weighed. Cheap and good for cohesive subgrades; useless where gravel is present.

Water content comes from the subsample, by oven drying at 110 ± 5 °C to constant mass (ASTM D2216), or faster by microwave (ASTM D4643) or a calcium carbide gas pressure tester (ASTM D4944). Whatever method is used has to be the one the specification names, because the fast methods carry different tolerances and the carbide tester is limited to the fraction passing the 4.75 mm (No. 4) sieve.

Whichever method is used, the test has three components, and skipping one of them destroys the result: the wet mass of the material in the hole, the volume of the hole, and the water content of the same material. Density and moisture are measured together or they mean nothing.

Worked example: a fill lift that fails, and what it takes to fix it

A 200 mm (8 in) lift of granular fill is placed over 500 m² (5,380 ft²) of a building pad. The specification requires 95 percent of the modified Proctor MDD, with the moisture content within 1 percentage point of OMC. The material is a silty sand; the laboratory curve, made on the fraction passing the 4.75 mm (No. 4) sieve, gives an MDD of 1,960 kg/m³ (122.4 pcf) at an OMC of 10.5 percent.

Input data

Step 1: wet density

$\gamma = \frac{M_{wet}}{V} = \frac{3.990}{2.000 \times 10^{-3}} = 1{,}995\ \text{kg/m}^3\ (124.5\ \text{pcf})$

Step 2: dry density

$\gamma_d = \frac{\gamma}{1 + w} = \frac{1{,}995}{1.080} = 1{,}847\ \text{kg/m}^3\ (115.3\ \text{pcf})$

The 148 kg/m³ difference between the wet and dry density is water. It is the reason the wet density of a saturated fill tells you almost nothing on its own.

Step 3: relative compaction

$RC = \frac{1{,}847}{1{,}960} \times 100 = 94.2\ \%$

The specification asks for 95 percent. The test fails on both counts: the density is 0.8 percentage points short, and the moisture content is 8.0 percent against an OMC of 10.5 percent, which is 2.5 points dry of the window.

Step 4: how much water, and how many passes

The dry density has to rise from 1,847 to at least 0.95 × 1,960 = 1,862 kg/m³, and the material has to be at 9.5 to 11.5 percent moisture before rolling.

The water needed to lift the water content from 8.0 to 10.0 percent, taking the material as it currently lies:

$\Delta m_w = \gamma_d\,(w_{target} - w_{field}) = 1{,}847 \times (0.100 - 0.080) = 36.9\ \text{kg/m}^3$

So about 37 litres of water per cubic metre of fill (7.5 US gal per yd³). Over the lift:

$V_{lift} = 500 \times 0.200 = 100\ \text{m}^3 \quad \Rightarrow \quad 37 \times 100 = 3{,}700\ \text{L} = 3.7\ \text{m}^3\ (977\ \text{US gal})$

Adding that water does not raise the dry density by itself. It only removes the lubrication deficit; the roller delivers the density, and it needs more passes than the layer received the first time. Adding water to the surface and rolling once more is a common mistake: the water has to be mixed into the layer, which in practice means scarifying, watering and re-compacting. The realistic plan is to water in stages, mix, re-compact, and retest, not to spray and hope.

Step 5: retest after reworking

The same material, reworked and re-compacted, gives:

$\gamma = \frac{4.130}{2.000 \times 10^{-3}} = 2{,}065\ \text{kg/m}^3\ (128.9\ \text{pcf})$

$\gamma_d = \frac{2{,}065}{1.102} = 1{,}874\ \text{kg/m}^3\ (117.0\ \text{pcf})$

$RC = \frac{1{,}874}{1{,}960} \times 100 = 95.6\ \%$

The layer is now at 95.6 percent of MDD with a water content of 10.2 percent, inside the window. It passes, and the next lift can go down.

The two numbers that decide the next lift

The same test result sets up the next lift, and it is worth doing the arithmetic before the roller arrives rather than after.

Water already in the lift. 100 m³ of fill at 1,874 kg/m³ dry contains 187,400 kg of dry solids and, at 10.2 percent, about 19,100 kg of water. That is both the quantity placed and the quantity to budget for the next one.

Borrow volume to order. The dry mass of the lift is 187.4 t (206.6 short tons). If the borrow pit material has an in-situ dry density of 1,700 kg/m³ (106.1 pcf), the volume to haul is:

$V_{borrow} = \frac{187{,}400}{1{,}700} = 110\ \text{m}^3\ (\text{about}\ 144\ \text{yd}^3)$

That is 10 percent more bank volume than the 100 m³ of finished fill, which is the shrinkage a quantity surveyor should be allowing for. Ordering by finished volume and hoping is how sites run short in the last week.

When the material has rock in it: the oversize correction

Sand replacement testing on a fill containing cobbles is misleading, because the test measures the fraction that fits in the hole. ASTM D4718 corrects the result for the oversize particles, using a dry density measured on the coarse fraction and the proportion of oversize by dry mass:

$\gamma_{d,corrected} = \frac{100}{\dfrac{P_c}{\gamma_{d,coarse}} + \dfrac{100 - P_c}{\gamma_{d,fine}}}$

where $P_c$ is the percentage of oversize by dry mass, $\gamma_{d,coarse}$ is the dry density of that oversize fraction and $\gamma_{d,fine}$ is the field dry density of the fraction that passed the sieve.

Which sieve defines “oversize” is not a free choice. The correction is based on the fraction retained on the sieve that the laboratory compaction curve was made from, which is the 4.75 mm (No. 4) sieve for a Method A curve, and the 19 mm (3/4 in) sieve only where a Method C curve was used. The reference MDD has to be on the same oversize basis as the field value, which is what AASHTO T224 does for the laboratory curve; comparing a corrected field density against an uncorrected laboratory MDD compares two different quantities.

With $P_c = 20$ percent, $\gamma_{d,fine} = 1{,}800$ kg/m³ (112.4 pcf) and $\gamma_{d,coarse} = 2{,}600$ kg/m³ (162.3 pcf):

$\gamma_{d,corrected} = \frac{100}{\dfrac{20}{2{,}600} + \dfrac{80}{1{,}800}} = \frac{100}{0.0521} = 1{,}918\ \text{kg/m}^3\ (119.7\ \text{pcf})$

The correction moves the result upwards by 118 kg/m³, and it is not a technicality: without it, that same fill reads 91.8 percent relative compaction and gets rejected, while the corrected value against the same MDD is 97.9 percent. Two conditions matter. The correction belongs in the specification, before any test is disputed. And the coarse particle density has to be measured on the actual oversize, not assumed from a rock type.

Why a compaction test fails

The reasons are few and always the same, and the test result tells you which one you have:

• The material is too dry. The most common failure, and the example above. The dry density curve is steep on the dry side, so a small moisture deficit costs a large density result.
• The lift is too thick. Compactive effort decays with depth. A 400 mm lift with the same roller and the same passes leaves the bottom half at a density nobody has measured.
• Not enough passes, or the wrong roller. A smooth drum does little for a silty sand at depth; a heavy vibratory roller on a saturated silt causes pumping instead of densification.
• The wrong laboratory curve. The borrow source changed, or the sample was taken from a different stratum. The result looks like a failure and is really a materials mismatch.
• Oversize particles, uncorrected. Covered above, and it cuts both ways: it can fail a layer that passes, and it can hide a layer that does not.
• Soft subgrade below. The layer is dense and the surface under it is not, so a proof roll or a plate test fails on a layer that passed every density test.
• The test was taken in the wrong place. Density results describe a spot. Specified grids and offsets exist so that a test does not drift, over a project, to the places that always pass.

What a specification for fill compaction should contain

Eleven items, all of them one line, and the absence of any one of them turns into an argument on site:

1. The material: its source, and the grading and classification limits it has to meet.
2. The laboratory standard the maximum dry density comes from (ASTM D698 or D1557), and its method designation.
3. The required relative compaction, as a percentage of that maximum, per application or per layer.
4. The moisture content window, expressed relative to the optimum of the same curve.
5. The maximum lift thickness before compaction, per material type and equipment class.
6. The preparation of the receiving surface: soft spots removed, unsuitable material out, and no frozen ground under the first lift.
7. The testing method (sand replacement, nuclear, balloon) and its frequency, per layer.
8. The acceptance rule, including what a single failing test triggers: retest, rework or rejection.
9. The oversize correction to be applied, including the sieve that defines oversize and who measures the coarse fraction density.
10. Protection during placement: drainage, and a plan for rain on an open fill.
11. The record: test locations, depths relative to the layer, and the results, kept against the layer they describe.

Frequently asked questions

What does 95 percent compaction mean? It means the dry density achieved on site is at least 95 percent of the maximum dry density that the named laboratory test produced on the same material. It is a ratio rather than a fixed density, so the same 95 percent can be a heavy or a light requirement depending on whether the laboratory curve is a standard or a modified Proctor.

Why not specify 100 percent? Because on the curve it is not a target but a peak, and pushing every layer to the maximum dry density costs effort that buys nothing measurable while making the moisture window harder to hit. Ninety-five percent of a modified Proctor is already a dense, well-performing fill. Where a structure needs more than that, the answer is usually a different design, not a denser fill.

Can a compaction test read more than 100 percent? Yes, and it should be investigated rather than celebrated. The usual causes are a change in the material’s grading, a laboratory curve made from a different borrow source, uncorrected oversize particles inflating the field result, or heavy rolling on the dry side of optimum, above what the laboratory curve says the material can reach.

How often should fill be tested? At the frequency the specification names, which is normally tied to the volume or area of each lift, and it has to be tied to the layer being placed rather than to the project. The frequency is a risk allocation: a low frequency is cheaper and passes more bad layers, and the reason to test is to catch them while they can still be reworked.

How long after rolling should the test be taken? After the layer has been completed, not during rolling, and for nuclear gauge work after the moisture has had a chance to even out. Testing a fresh, watered surface with a gauge in backscatter mode reports roughly the top 100 mm, which after watering is the wettest thing in the layer.

Can a failed lift be left in place? Sometimes, and the decision belongs to the designer. A layer below the specified density can be acceptable if the structure above it is tolerant of the resulting settlements, or if the layer is deep enough not to influence them. That is a calculation, not a concession, and it needs the failing density as an input. Where the layer cannot be justified, reworking it while it is still exposed is always cheaper than underpinning the structure that sits on it later.

Where to go next

To work out the compaction curve, the maximum dry density and the optimum moisture content of your own trials, use the free Proctor test calculator.

Compaction is the last step of the earthworks chain and the first step of everything structural, so it connects in two directions. Upstream, the material’s grading and moisture behaviour are what the geotechnical report describes, and the fill specification should follow from it. Downstream, the density that was achieved is one of the inputs to the settlement the structure will tolerate, which is what foundation settlement quantifies and what the shallow foundation tools calculate.

Where the material itself cannot reach the required density at any economic effort, the answer is usually to change the material or to treat it: soil stabilization for the shallow layers, and the ground improvement techniques guide for the deeper alternatives.

References

1. ASTM D698-12(2021), Standard Test Methods for Laboratory Compaction Characteristics of Soil Using Standard Effort (12 400 ft-lbf/ft³ (600 kN-m/m³)), ASTM International, West Conshohocken, PA.
2. ASTM D1557-12(2021), Standard Test Methods for Laboratory Compaction Characteristics of Soil Using Modified Effort (56 000 ft-lbf/ft³ (2 700 kN-m/m³)), ASTM International.
3. ASTM D1556-16, Standard Test Method for Density and Unit Weight of Soil in Place by Sand-Cone Method, ASTM International.
4. ASTM D2167-15(2020), Standard Test Method for Density and Unit Weight of Soil in Place by the Rubber Balloon Method, ASTM International.
5. ASTM D6938-23, Standard Test Methods for In-Place Density and Water Content of Soil and Soil-Aggregate by Nuclear Methods (Shallow Depth), ASTM International.
6. ASTM D2216-19, Standard Test Methods for Laboratory Determination of Water (Moisture) Content of Soil and Rock by Mass, ASTM International.
7. ASTM D4718/D4718M-15(2023), Standard Practice for Correction of Unit Weight and Water Content for Soils Containing Oversize Particles, ASTM International.
8. ASTM D4643-17, Standard Test Method for Determination of Water Content of Soil and Rock by Microwave Oven Heating, ASTM International.
9. ASTM D4944-18, Standard Test Method for Field Determination of Water (Moisture) Content of Soil by the Calcium Carbide Gas Pressure Tester Method, ASTM International.
10. AASHTO T99-22, Standard Method of Test for Moisture-Density Relations of Soils Using a 2.5-kg (5.5-lb) Rammer and a 305-mm (12-in.) Drop, American Association of State Highway and Transportation Officials.
11. AASHTO T180-22, Standard Method of Test for Moisture-Density Relations of Soils Using a 4.54-kg (10-lb) Rammer and a 457-mm (18-in.) Drop, AASHTO.
12. AASHTO T224, Standard Method of Test for Correction for Coarse Particles in the Soil Compaction Test, AASHTO.
13. FHWA, Soils and Foundations Reference Manual, Volume I (FHWA NHI-06-088), Federal Highway Administration, December 2006, Chapter 5 (Laboratory Tests), sections 5.7 to 5.8.
14. National Highways, Manual of Contract Documents for Highway Works, Volume 1: Specification for Highway Works, Series 600: Earthworks (Clauses 608, 612 and 631), 2016, available at standardsforhighways.co.uk.
15. Transportation Research Board, Rock Correction Issues in Compaction Specifications, Transportation Research Record 1462, 1994.