Why particle size cannot be reduced to “fine sand”
Particle-size distribution describes how grain sizes are distributed within a sample. For an industrial buyer, a simple commercial label is not enough: two sands can have a similar median size while containing very different amounts of fines or coarse grains.
Our industrial silica sand guide notes that grain size and distribution can affect flow, packing, finish and process stability. Useful control therefore links the particle-size curve, test method, tolerances and the actual delivered lot.
1. Reading D10, D50 and D90 correctly
In a cumulative distribution, D10, D50 and D90 are diameters associated with percentiles.
| Indicator | Practical reading |
|---|---|
| D10 | 10% of the distribution lies below this diameter |
| D50 | 50% lies below this diameter; it is the median |
| D90 | 90% lies below this diameter |
This interpretation is complete only when the report states the method and distribution basis. In laser diffraction, Dv(10), Dv(50) and Dv(90), weighted by volume, are commonly reported.
2. D50 is not “the size of the sand”
D50 summarizes the center of a distribution but does not describe its extremes. Two lots with a similar D50 can have different D10 and D90 values and therefore different distribution widths.
For an industrial specification, D50 can be useful, but it should not hide fines, coarse particles or the actual shape of the curve.
3. Why D10 and D90 complement D50
D10 helps track the fine side of the distribution and D90 the coarse side. Together, D10, D50 and D90 provide a richer view than the median alone.
They still do not replace the full curve. The NIST particle-size characterization guide notes that a few percentiles can be misleading when a distribution contains multiple size populations.
4. Span: summarizing distribution width
A common indicator in laser diffraction is the span:
Span = (D90 - D10) / D50
All else equal, a lower value corresponds to a narrower distribution. Span is not, however, a universal silica-sand acceptance limit: its relevance depends on the application, method and specification.
5. The same values can hide different curves
Three percentiles do not describe shoulders, breaks or a possible bimodal distribution. For serious qualification, it is therefore preferable to retain the full cumulative curve and, when available, the differential distribution.
This matters especially when the process is sensitive to a small fine fraction or a few oversized grains.
6. Sieving: a reading directly tied to sieve openings
Sieving separates grains according to their passage through defined openings. ISO 3310-1 covers technical requirements for woven-wire test sieves, while ASTM C136/C136M-25 describes sieve determination of particle-size distribution for fine and coarse aggregates.
Results can be expressed as retained or passing mass percentages by sieve. This is especially useful when a contract defines a grading envelope or limits at specific sieves.
7. Fines may require a complementary method
ASTM C136 states that, within its scope, sieving alone cannot accurately determine material finer than 75 µm and refers to a complementary washing method.
This illustrates a broader rule: the method should match the fraction that actually needs to be controlled.
8. Laser diffraction: another representation of size
ISO 13320:2020 provides the framework for laser diffraction. The technique analyzes light scattering and reports a distribution corresponding to equivalent spheres in its optical model.
ISO 13320 also notes that, for non-spherical particles, the resulting distribution can differ from measurements based on other physical principles such as sieving or sedimentation.
9. Do not mix methods in a lot history
Comparing a laser-diffraction D50 with a D50 calculated from sieving can create a false drift - or hide a real one. To monitor supplier consistency, keep as far as possible:
- the same method;
- the same sample preparation;
- the same reporting convention;
- documented dispersion conditions when relevant;
- a consistent sampling procedure.
10. Lot control starts before the instrument
A precise analyzer cannot make a poor sample representative. A lot may be heterogeneous because of segregation during handling, storage or transport.
Particle-size control should therefore be linked to representative sampling, including lot identification, sampling point, sample reduction and traceability.
11. What should be compared between two lots?
A robust control plan can combine several levels.
| Level | Useful data | Question answered |
|---|---|---|
| Curve | Full distribution | Has the overall shape changed? |
| Percentiles | D10, D50, D90 | Are the fine, median and coarse regions drifting? |
| Width | Span or another agreed indicator | Is the distribution becoming broader? |
| Critical sieves | % passing / retained | Has a contractual limit been exceeded? |
| Fines / oversize | Dedicated method if needed | Is a small problematic fraction increasing? |
The final choice depends on the process; there is no universal set of limits.
12. Define tolerances before delivery
A technical data sheet can show a typical curve without creating a contractual guarantee. To avoid ambiguity, the specification should state the method, preparation, calculation basis, controlled sieves or percentiles, limits or tolerances, frequency, rounding rule and decision process for non-conformity.
13. D10/D50/D90 do not replace chemistry or mineralogy
A compliant particle-size distribution proves neither SiO₂ content, iron content nor mineral phases. These data families answer different questions.
When composition is critical, particle size should be combined with appropriate analyses, such as those covered in our XRF and XRD silica-sand guide.
14. Example of a useful supplier dossier
For repeatable control, the buyer can request product and lot identification, date, sampling method, particle-size method, preparation, full result table, curve, D10/D50/D90 where relevant, critical-sieve results, tolerances and conformity conclusion.
A certificate stating only “particle size compliant” provides less information than a report showing what was measured and how.
15. What to avoid
Avoid turning a typical value into a guarantee, specifying D50 without controlling the tails, directly comparing different methods, or using D10/D50/D90 without stating their basis.
Also avoid assuming that one curve fits every application. Glass, foundry, filtration, mortars, resins or mineral fillers can require different distributions and control methods.
Key takeaways
D10, D50 and D90 are distribution markers, not universal specifications. They become useful when linked to a method, calculation basis, full curve and representative sample.
For lot control, comparability is the priority: same method, same sampling logic, same indicators and clearly defined tolerances.