Introduction & Context

Particle-size specification setting is the systematic translation of customer requirements into statistically defensible control limits for a milling, grinding, or classification process. The core metric is the process-capability index \(C_p\), which compares the width of the customer’s specification window to the natural variation of the process. A higher \(C_p\) indicates that the process can routinely deliver material whose particle-size distribution lies entirely within the customer’s Upper Specification Limit (USL) and Lower Specification Limit (LSL). This calculation is routinely embedded in:

Quality-by-Design (QbD) dossiers for pharmaceutical and specialty-chemical registrations.
Process-validation protocols for milling circuits, air-classifiers, and wet-media mills.
Make-or-buy decisions where incoming raw-material variability must be matched to downstream performance.

Methodology & Formulas

Define the specification window
The customer provides an upper and lower size limit, \(USL\) and \(LSL\), expressed in the same units as the measured particle size (e.g. µm).
Quantify process spread
The short-term process variability is assumed normal with standard deviation \(\sigma\). The total spread captured by ±3σ is \(6\sigma\).

Compute the capability index
\[ C_p = \frac{USL - LSL}{6\sigma} \]

Regime	Condition	Interpretation
Not capable	\(C_p < C_{p,\text{req}}\)	Process spread exceeds specification window; action required.
Marginally capable	\(C_{p,\text{req}} \le C_p < 1.67\)	Adequate for routine production with controls.
Highly capable	\(C_p \ge 1.67\)	Robust to normal process shifts; preferred for critical applications.

Back-calculate required standard deviation (if not capable)
Rearranging the \(C_p\) definition gives the maximum allowable variability: \[ \sigma_{\text{req}} = \frac{USL - LSL}{6\,C_{p,\text{req}}} \] Any process whose \(\sigma \le \sigma_{\text{req}}\) will automatically satisfy the customer’s capability requirement.

Start with the failure modes of your process. Map every measurable product or intermediate attribute that can drift because of size variation—e.g., filter-cake resistance, blend segregation, dissolution rate, pneumatic-line saltation velocity, tablet hardness, or catalyst pressure drop. Run a Design-of-Experiments where you deliberately vary D10, D50, and D90 while holding formulation or operating variables constant; record the response that first crosses the acceptable threshold. The size fraction that triggers that failure becomes your critical quality attribute (CQA) and its proven-acceptable range (PAR) becomes your specification. Document the link in the control strategy so regulators see the science behind the numbers.

No—forcing a single median masks the secondary population that may drive segregation, percolation, or dissolution failure. Instead, set a split specification:

D10, D50, D90 for the primary mode (e.g., 5 µm, 25 µm, 55 µm)
Percentage by volume of the secondary mode (e.g., < 10 % above 120 µm)
Optionally add a “fines” limit (e.g., < 15 % below 3 µm) if dusting or poor flow is an issue

Use deconvolution software to quantify each mode and link the limits to functional tests such as powder rheometry or dissolution. When transferring to QC, validate the algorithm settings so the split is reproducible across instruments.

Collect 20–30 representative lots that span the process range. Run both sieve (dry) and laser (wet dispersion) on the same samples. Build a regression or equivalence model:

Plot sieve retained on each mesh vs. laser cumulative volume in the corresponding size band
Use a generalized linear model with particle shape factor as a covariate; elongated particles give poorer correlation
Set equivalence acceptance criteria (e.g., ±5 % absolute on each key fraction)
Validate the model with 3 additional lots; if bias > 3 %, adjust laser optical properties or sieve shaking time

Once verified, you can retire the sieve method or keep it as a PAT surrogate with a correlation factor updated annually.

Worked Example: Setting Particle Size Specification

A pharmaceutical manufacturing line produces a micronized active ingredient. The customer requires the particle size distribution to stay within a lower specification limit (LSL) of 45 µm and an upper specification limit (USL) of 105 µm. The current process operates with a mean particle size of 75 µm and a standard deviation (σ) of 10 µm. Management wants to know whether the process meets the required capability index (C_p) of 1.33 and, if not, what σ must be reduced to.

Knowns

Customer LSL = 45 µm
Customer USL = 105 µm
Process mean (μ) = 75 µm
Current process sigma (σ) = 10 µm
Required C_p = 1.33

Step-by-Step Calculation

Calculate the actual process capability index using the formula: \[ C_p = \frac{USL - LSL}{6\sigma} \] Substituting the known values: \[ C_p = \frac{105 - 45}{6 \times 10} = \frac{60}{60} = 1.0 \]
Compare the actual C_p (1.0) with the required C_p (1.33). Since 1.0 < 1.33, the process does not meet the specification.
Determine the sigma required to achieve the target C_p. Rearrange the capability equation to solve for σ: \[ \sigma_{\text{required}} = \frac{USL - LSL}{6 \times C_{p,\text{required}}} \] Insert the numbers: \[ \sigma_{\text{required}} = \frac{105 - 45}{6 \times 1.33} = \frac{60}{7.98} = 7.5188\ \text{µm} \]
Round the required sigma to three decimal places: \[ \sigma_{\text{required}} = 7.519\ \text{µm} \]
Interpretation: The process standard deviation must be reduced from 10 µm to ≤ 7.519 µm to satisfy the customer’s capability requirement.

Final Answer: The process must achieve a standard deviation of ≤ 7.519 µm to meet the required capability index of 1.33.

"Un projet n'est jamais trop grand s'il est bien conçu." — André Citroën

"La difficulté attire l'homme de caractère, car c'est en l'étreignant qu'il se réalise." — Charles de Gaulle