Graphite is one of the most difficult materials to classify accurately. Not because it is particularly hard — calcite and quartz are both harder — but because of its shape. Natural graphite and most artificial graphite particles are lamellar: flat, plate-like structures with a high aspect ratio. A classifier that uses aerodynamic drag to separate particles reads aspect ratio as well as size. A flat graphite flake with a projected area equivalent to a 30-micron sphere will behave aerodynamically like a much finer spherical particle. It gets pulled through the classifier into the fine product stream at a size that should be reporting to the reject stream.
The result is a graphite product with a broader D90 and higher D97 than the classifier settings would predict. It’s because that the flat particles are systematically misclassified as fine. For anode graphite production, where a D90 above specification causes uneven lithium intercalation and can trigger electrode cracking during high-rate charging, this is not a tolerable error.
This article covers why graphite is hard to classify, the specific PSD targets for different anode applications. It also introduces how to configure a classifier to compensate for graphite’s lamellar behaviour, and what classification produces in practice.
Why Graphite Is Harder to Classify Than Other Battery Materials
The Lamellar Shape Problem
Air classification separates particles by the ratio of aerodynamic drag force to particle mass. For a sphere, this ratio is well-defined: drag scales with the projected area (proportional to diameter squared) and mass scales with volume (proportional to diameter cubed). The cut point — the particle diameter at which drag and centrifugal force are balanced — is predictable and consistent.
Graphite flakes break this relationship. A graphite particle with a diameter of 25 microns measured by laser diffraction (which measures the equivalent spherical diameter) may have an actual plate geometry of 40 microns across and 5 microns thick. In a classifier, that particle presents a much larger face to the airflow than a 25-micron sphere would. The drag force is higher. The particle reports to the fine product fraction when geometrically it should be in the coarse reject fraction.
The practical consequence: if you set a classifier to produce D90 = 25 microns based on the cut point calculation for a spherical material, anode graphite will give you D90 = 30-35 microns on the laser diffraction analysis of the product. The classifier is performing correctly — it is classifying by aerodynamic behaviour. The product specification, however, is written in terms of equivalent spherical diameter as measured by laser diffraction, not aerodynamic diameter. To hit D90 = 25 microns on a graphite product, you need to set the classifier significantly tighter than you would for an equivalent spherical material.
Electrostatic Agglomeration
Fine graphite powder (D50 below 15 microns) is electrically conductive and builds up static charge during classification, particularly in dry conditions at low humidity. Charged particles attract each other and form soft agglomerates that behave aerodynamically as large particles and report to the coarse reject stream. The result is lower yield and poorer classification efficiency — the fine fraction that should be product is being rejected and recirculated.
Managing electrostatic agglomeration in graphite classification requires either humidity control in the process air (60-70% relative humidity suppresses static accumulation significantly), antistatic grounding of all process equipment, or in some installations, a mild ionising bar at the classifier inlet. None of these are standard on a general-purpose classifier — they are design considerations for graphite-specific equipment.
Low Bulk Density and Dustiness
Natural and artificial graphite have bulk densities of 0.3-0.8 g/cm3 — much lower than mineral fillers (calcium carbonate at 0.8-1.2 g/cm3, quartz at 1.2-1.5 g/cm3). Low bulk density means graphite is easily fluidised and difficult to feed at a controlled, consistent rate. Feed rate inconsistency directly widens the product PSD: when feed rate spikes, particle concentration in the classification zone increases, and the effective cut point shifts coarser through the crowding effect. A graphite-specific classifier installation needs a controlled-rate feeder — vibratory or screw — with a mass flow controller rather than a volumetric feeder.
PSD Specifications for Anode Graphite by Application
Not all anode graphite requires the same particle size. The target PSD depends on the cell format, electrode design, and electrochemical requirements of the application.
| Application | D50 Target | D90 Target | Key Requirement |
| Natural graphite (standard anode) | 14-18 um | 30-38 um | Narrow span; D90 hard limit prevents electrode cracking |
| Artificial graphite (high-rate anode) | 10-14 um | 22-28 um | Tighter PSD for fast charge capability; low fines fraction |
| Spheroidised graphite (premium anode) | 15-20 um | 32-40 um | Very tight span; sphericity combined with controlled D90 |
| Tailings recovery (conductive additive) | 5-10 um | 15-20 um | Fine fraction from spheronisation; acceptable for blending |
| Silicon-graphite composite anode | 6-12 um | 18-25 um | Matched to silicon particle size; uniform composite distribution |
Specifications vary by cell manufacturer and electrode design. Confirm against your buyer’s incoming inspection protocol before setting classification parameters.
The D90 specification is almost always the harder constraint to meet than D50. D50 is set primarily by classifier wheel speed and responds predictably to parameter adjustment. D90 is harder to control because it represents the tail of the distribution — the particles that are just large enough to be rejected by the classifier but sometimes slip through due to shape effects, turbulence, or feed rate variation. For anode graphite, D90 exceedance typically means a small number of oversized lamellar particles that would otherwise be manageable cause electrode defects at a rate that fails the cell manufacturer’s incoming QC.
Configuring a Classifier for Anode Graphite
Rotor Speed: Set Tighter Than You Would for a Spherical Material
Because graphite’s lamellar morphology causes flat particles to misreport to the fine fraction, you need to set the classifier rotor speed higher than the equivalent spherical material calculation would suggest for your D90 target. A practical starting point: for natural graphite targeting D90 35 microns, set the initial rotor speed as if you were targeting D90 25-28 microns on a spherical mineral. Then measure the actual product PSD by laser diffraction and adjust in steps until the measured D90 matches the specification.
Document this morphology correction factor for your specific graphite source and crystallinity — it varies between natural flake graphite, artificial graphite, and spheroidised graphite because the aspect ratios are different. A process recipe that works for one graphite grade will not transfer directly to another.
Airflow: Balance Drag and Centrifugal Force for Flat Particles
Higher airflow increases drag on all particles, which tends to pull more material into the fine product stream. For lamellar graphite where flat particles already have inflated drag, increasing airflow beyond the minimum needed for material transport makes the shape misclassification problem worse. Keep airflow at the minimum level that maintains stable fluidisation in the classifier feed zone — typically 10-20% below what you would use for an equivalent mineral.
The combined effect of tight rotor speed and moderate airflow is a classification zone where centrifugal rejection dominates over aerodynamic drag for the coarser flat particles, improving D90 control without sacrificing throughput.
Feed Rate: Hold Stable with a Mass Flow Controller
Set the feed rate at 60-75% of the classifier’s rated capacity for graphite — lower than for mineral materials of equivalent fineness, because graphite’s low bulk density and easy fluidisability make the crowding effect more pronounced at high feed rates. More importantly, hold the feed rate constant. A mass flow controller on the feed screw, with a tolerance of plus or minus 5% of the set point, is the practical standard for anode graphite classification lines. Feed rate variation above this tolerance will show up as D90 variation in the product.
Humidity Control for Fine Grades
For anode graphite classification below D50 12 microns — the range where electrostatic agglomeration becomes significant — conditioning the process air to 60-70% relative humidity is effective at suppressing static charge. This requires a humidification system on the classifier inlet air, which adds equipment cost but is justified for continuous production of fine anode grades. Alternatively, antistatic additives can be introduced at very low levels (0.05-0.1% by weight) to the graphite feed to suppress agglomeration without humidification, but this must be compatible with the downstream electrode chemistry.
| Graphite Classification Parameter Starting Points Rotor speed correction: Set 15-25% higher than equivalent spherical material calculation for the same D90 target — adjust based on measured PSD Airflow: 10-20% below rated capacity for equivalent mineral; minimum for stable fluidisation Feed rate: 60-75% of classifier rated capacity; control to +/-5% with mass flow controller Humidity (D50 < 12 um): Condition process air to 60-70% RH to suppress electrostatic agglomeration PSD verification interval: Sample every 30 minutes during the first 4 hours of a new grade — graphite takes longer to reach steady state than mineral materials |
Production Results
CASE STUDY 1
Natural Graphite Anode Material: D90 Reduced from 42 to 31 Microns by Classifier Reconfiguration
The situation
A natural graphite anode material producer was consistently failing their cell manufacturer customer’s D90 maximum of 35 microns. Incoming QC at the cell plant was rejecting approximately 20% of batches. Their existing air classifier was configured with parameters derived from the classifier supplier’s standard mineral settings — rotor speed and airflow had not been adjusted for graphite’s lamellar morphology. Laser diffraction analysis of their product showed D50 16.2 microns (within specification) but D90 40-44 microns (above the 35-micron limit) on repeated samples.
What changed
EPIC Powder Machinery’s application engineer conducted a classifier audit and identified the rotor speed as the root cause: it had been set for a D90 target of 35 microns using a spherical particle calculation, which resulted in the actual product D90 being 6-9 microns wider than intended due to the lamellar morphology effect. Rotor speed was increased by 22%; airflow was simultaneously reduced by 12% to prevent excessive fine particle rejection. Feed rate wa
s reduced from 100% to 68% of rated capacity and stabilised with a mass flow controller.
Results
- D50: 15.8 microns — essentially unchanged from before (the median size was already correct)
- D90: 31.2 microns — 26% reduction, within the customer’s 35-micron limit with margin
- Batch rejection rate: reduced from 20% to below 2% at the cell manufacturer’s incoming QC
Throughput: reduced by 14% at the new feed rate setting — accepted as a necessary trade-off for specification compliance
CASE STUDY 2
Artificial Graphite Anode: Eliminating Electrostatic Agglomeration for Fine-Grade Production
The situation
An artificial graphite producer was manufacturing a fine anode grade targeting D50 11 microns, D90 24 microns for a high-rate battery application. The classification yield was only 61%, well below the expected 80-85% for this PSD target. Most of the missing material was being rejected to the coarse return stream despite having measured particle sizes well within specification. Electron microscopy of the coarse return material showed aggregates of fine graphite particles clumped together rather than single oversized particles — classic electrostatic agglomeration.
What changed
A humidification system was installed on the classifier inlet air, conditioning the process air to 65% relative humidity before it entered the classification zone. All metal surfaces in the product contact path were bonded and earthed. The classifier parameters were re-optimised after the humidity control was in place.
Results
• Classification yield: increased from 61% to 83% — 22 percentage points — recovering the fine material that had been lost to agglomeration
• D50: 11.4 microns — within specification
• D90: 23.1 microns — within specification
Production cost per tonne: reduced by approximately 18% through the combination of higher yield and lower recirculation energy
| Classifying Anode Graphite or Another Battery Material? EPIC Powder Machinery’s application engineers configure air classifiers specifically for the challenges of graphite and carbon material classification — lamellar morphology compensation, electrostatic management, and D90 hard-cut performance. We offer free classification trials on your graphite feed material and return full PSD data before you commit to equipment.Send us your feed PSD, target D50 and D90, and throughput requirement and we will recommend the right classifier configuration. Request a Free Classification Trial: www.powder-air-classifier.com/contact Explore Our Anode Graphite Classifier Range: www.powder-air-classifier.com |
Frequently Asked Questions
Why does my anode graphite classifier produce a wider D90 than the classifier settings predict?
This is the single most common problem in graphite classification and it is caused by the lamellar morphology of graphite particles. Air classifiers separate particles based on aerodynamic behaviour — specifically the ratio of drag force to particle mass. Flat, plate-like graphite particles present a much larger face to the airflow than a sphere of equivalent laser diffraction diameter. It means drag force is disproportionately high for their mass.
These flat particles get pulled into the fine product stream when geometrically they should be reporting to the coarse reject fraction. The result is that the actual product D90, measured by laser diffraction, is 5-15 microns wider than the classifier cut point calculation predicts for spherical particles. The solution is to set the classifier rotor speed 15-25% higher. It should be more than the equivalent spherical material calculation would suggest for your D90 target. Then verify with actual PSD measurement and adjust from there.
What D90 specification is typical for natural graphite anode material, and how strictly do cell manufacturers enforce it?
For standard natural graphite anode material used in consumer lithium-ion cells, D90 targets are typically in the 30-38 micron range, with D50 around 14-18 microns. For high-rate and fast-charging applications, specifications tighten: D90 22-28 microns and D50 10-14 microns. Cell manufacturers generally enforce D90 as a hard incoming QC parameter. It’s a batch that exceeds D90 by even 2-3 microns can be rejected. The oversized graphite particles in the electrode can cause localised plating of metallic lithium during fast charging. It is both a capacity and a safety concern. D50 tolerance is typically wider (plus or minus 2 microns) because median size affects electrode energy density. But it is less directly tied to safety failure modes. If your batches are passing D50 but failing D90, the lamellar misclassification issue described above is the most likely cause.
Epic Powder
Epic Powder, 20+ years of experience in the ultrafine powder industry. Actively promote the future development of ultra-fine powder, focusing on crushing, grinding, classifying and modification process of ultra-fine powder. Contact us for a free consultation and customized solutions! Our expert team is dedicated to providing high-quality products and services to maximize the value of your powder processing. Epic Powder—Your Trusted Powder Processing Expert!
“Thanks for reading. I hope my article helps. Please leave a comment down below. You may also contact EPIC Powder online customer representative Zelda for any further inquiries.”
— Emily Chen, Engineer
