Centrifugation is used in published studies to wash or fractionate colloids, but the collected material and retained dispersion depend on the sample and procedure. Washing can remove soluble species, lose particles or change surface chemistry; redispersion is a separate step. This guide compares reported studies, explains a limited sphere model and identifies evidence needed for interpretation. It supplies no universal recipe, size cutoff or demonstrated nanoparticle capability for an ALab™ centrifuge.
Quick answer: Distinguish washing, fractionation and subsequent redispersion. For an isolated sphere under Stokes assumptions, drift depends on diameter squared, density contrast, viscosity and acceleration; sheets and interacting dispersions require additional information. Study conditions describe the authors' samples and equipment, not a transferable recipe. Check the retained phase, medium, concentration, rotor/tube geometry, time, temperature and analytical endpoint. Record missing information rather than reconstructing it. The product section reports specifications separately and establishes no nanoparticle/GO application or paper-to-model pairing.
1. Why Nanoparticle Sedimentation Depends on the Sample
For an isolated rigid sphere in a Newtonian liquid, a steady balance of buoyant centrifugal force and Stokes drag gives the terminal drift relative to the liquid [1][5]. This assumes creeping flow (particle Reynolds number much less than one), no slip and negligible particle interactions. Taking the outward radial direction as positive:
v = d² (ρp − ρf) a ⁄ (18 η)
Here d is sphere diameter, ρp and ρf are particle and fluid densities, a is centrifugal acceleration, and η is dynamic viscosity. With SI units, v is in m/s. Positive density contrast gives outward drift; zero or negative contrast does not. At otherwise unchanged conditions, this sphere model gives v proportional to d² and to density contrast, and inversely proportional to viscosity. These scalings are not a general GO lateral-size law [4][5].
Table T1 makes these scalings concrete with relative, idealized transit times. The figures are an illustration, not a prediction for any instrument, rotor or real sample.
| Idealized sphere in water at 20 °C (bare-core density assumed) | Relative transit time at the same RCF and path length |
|---|---|
| Silica, 100 nm (ρ = 2.0 g/cm³) | 1.00 (reference) |
| Silica, 50 nm | 4.00 |
| Silica, 200 nm | 0.25 |
| Gold, 10 nm (ρ = 19.3 g/cm³) | 5.47 |
| Gold, 20 nm | 1.37 |
| Gold, 50 nm | 0.22 |
| Polystyrene, 200 nm (ρ = 1.05 g/cm³) | 4.83 |
| Illustrative viscosity change only: η_new / η_reference = 1.56; all other inputs fixed | about 1.56 × longer |
T1 is calculated from t_i/t_ref = (100 nm/d_i)²(2.0 − 0.9982)/(ρ_i − 0.9982), with densities in g/cm³. The fluid density 0.9982 and bare-core densities are illustrative inputs, not sample measurements. Assumptions: rigid, smooth, no-slip spheres; Newtonian liquid; creeping-flow terminal drift; dilute, non-interacting suspension; equal constant acceleration and path length; no radial-field variation, ramps, Brownian diffusion, wall effects or convection; no surface shells. The last row changes only the assigned viscosity ratio. An actual temperature change may also change density contrast, stability and other properties; no temperature-specific run time is predicted.
Real dispersions can violate these assumptions in different ways. The direction and magnitude of the departure must be established for the system, not inferred from T1 alone.
Shells affect both buoyant mass and drag. Assume an impermeable, concentric 2 nm radial shell around a 10 nm gold core: outer hydrodynamic diameter 14 nm, core density 19.3, shell density 1.1 and fluid density 0.9982 g/cm³. The same no-slip sphere model gives a coated/bare velocity ratio of about 0.72. This is our composite-sphere calculation, not a prediction for permeable, solvated or interacting ligand layers.
Brownian diffusion also matters. For an assumed 10 nm hydrodynamic sphere at 293.15 K and viscosity 1.002 mPa·s, D = k_B T/(3πηd) gives a one-dimensional free-diffusion RMS displacement √(2Dt) of about 0.23 mm in 600 s. This calculated diffusion scale is not a size cutoff or a sedimentation trajectory. Drift, diffusion, interactions and boundaries must be assessed together [5].
Shape changes the drag. Rods, plates and aggregates do not follow the sphere formula; Sharma and co-workers used this to separate gold nanorods from spherical by-products (Section 4) [5].
Sheets are not spheres. Reviews warn that Stokes-based estimates assume spherical, homogeneous particles, so for GO sheets the equation indicates trends, not numbers [4].
In an actual constant-speed rotor, a = ω²r varies with radial position. A reported RCF therefore needs its reference location or documented conversion basis. T1 deliberately omits that spatial variation; it is not an integrated transit-time model for a real tube.
2. Reading and Reporting a Centrifugation Step
Table T2 preserves study-specific conditions from the identified source passages. Here, “not reported” means the cited source does not give the value. A reported rpm, rotor name or processing temperature does not authorize reconstruction of a missing radius, RCF or spin temperature. See the source-access notes below.
| Study (sample, medium) | Instrument / rotor (as reported) | Speed or force | Radius basis | Time | Temp. | Measured endpoint |
|---|---|---|---|---|---|---|
| Backes et al. 2016 — WS₂ nanosheets, aqueous sodium cholate [2] | Hettich Mikro 220R; rotor 1016 (≤5 krpm, ~10 mL in 28 mL vials) and rotor 1195-A (>5 krpm, 1.5 mL vials), described as fixed-angle by the authors | 1.5–10 krpm in cascade; 240 to 9,740 ×g (15 krpm, 21,915 ×g for monolayer enrichment) | Radius definition not specified in the source; authors give RCF = 106.4·f² and 97.4·f² (f in krpm) | 2 h per step (1 h for the 15 krpm step) | 15 °C | Nanosheet length, thickness and monolayer content from extinction/PL spectroscopy, AFM, TEM |
| Khan et al. 2012 — liquid-exfoliated graphene, NMP [3] | Hettich Mikro 22R; rotor not reported | 4,000, 3,000, 2,000, 1,000, 700, 500 rpm; RCF not reported | Not reported | 45 min each | Not reported | Flake length (TEM, Raman), layer number (TEM) |
| Sharma et al. 2009 — as-made aqueous surfactant-containing gold nanorod dispersion [5] | Jouan MR23i; rotor not reported | 5,600 ×g; rpm not reported | Not reported | 20–40 min (30 min shown) | Not reported | Particle shape fractions by TEM; plasmon peak position |
| Chen T et al. 2009 — GO from expanded graphite, water [11] | Centrifuge model not reported | 4,000 rpm | Rotor radius 14 cm, as reported | 30 min (removal of unexfoliated material) | Not reported | Exfoliated GO by AFM |
| Thomas et al. 2013 — Hummers-route GO, aqueous base-wash sequence [9] | Not reported | 12,500 rpm | Not reported | 30 min for the stated step | Preceding base treatment: 70 °C for 1 h; spin temperature not specified | Solid/debris characterization; 25% debris yield belongs to combined-supernatant work-up, not one spin |
| Ceriotti et al. 2015 — crude LFGO/PGO, conventional comparison wash [10] | Sorvall T1; four 250 mL polypropylene bottles; rotor not specified | 4,000 rpm | Not reported | 90 min for LFGO; 30 min for PGO, in the stated sequence | Not reported | Purification and processing-time comparison; LFGO/PGO identify large-flake-graphite/powder-graphite starting materials |
T2 is not a recipe or an equipment recommendation. Instrument and rotor identifiers, including Backes's geometry wording, belong only to the cited studies. Unknown fields have not been calculated or borrowed from another procedure.
Three lessons follow from this table alone.
rpm alone is insufficient. Conversion needs a radius or applicable reported speed–RCF relation. Backes supplies conversion constants; none is reported for Khan's configuration. Chen T reports a 14 cm radius: our arithmetic, 1.118 × 10⁻⁵ × 14 × 4,000², gives 2,504.32 ×g. The paper's reported inputs are rpm and radius, not that calculated RCF; the paper does not define the reference point.
Force alone does not fix the outcome. Fill height, tube shape, rotor position and run time all decide whether a particle reaches the pellet; Backes and co-workers changed vial size and fill volume between their two rotors [2].
Temperature and ramps need their own record. Viscosity and density contrast can depend on temperature, while the speed history determines the changing field. A set temperature is not measured sample temperature, and a nominal time does not resolve the timer convention. Keep missing study details explicit.
3. Washing Cycles: Pellet, Decant, Redisperse
Washing by sedimentation separates a retained fraction from liquid before a subsequent redispersion step. For an ideal soluble contaminant, assume complete mixing, restoration to the same final volume, a retained liquid fraction q, no adsorption/release and no particle loss. The residual contaminant fraction after n cycles is qⁿ. Choosing q = 0.05 gives 0.05, 0.0025 and 0.000125 after one, two and three cycles. This is dilution arithmetic, not a recommended liquid fraction or wash count. Adsorption, incomplete mixing, changing volume, contamination and particle loss can invalidate it.
The literature adds two cautions that arithmetic does not.
Each wash can change the particles. La Spina et al.'s abstract describes 15 nm citrate-stabilized gold nanoparticles [7]. Three tested centrifugation cycles removed most citrate; further cycles were associated with aggregation. Their tested dialysis sequence left a different residual citrate level and size distribution, and surface analysis detected hydrocarbon contamination. These findings are system-specific, not a three-wash rule. Speed, force, duration and temperature are not given in the abstract; no such conditions are supplied here.
Alternative methods have different trade-offs. Sweeney et al.'s abstract reports more effective purification by diafiltration for their 3 nm thiol-stabilized gold system than the alternatives they tested [8]. Ceriotti et al. reported adequate purification of 1 g of crude LFGO in under 60 min with their laboratory gas-press filtration setup [10]. Their conventional centrifugation comparison used the conditions in T2 and a multisolvent wash sequence. Neither result establishes the preferred method, purity endpoint or equipment for another sample.
For GO there is a further complication, covered in Section 5: what is removed in a wash may be part of what was being called "graphene oxide" in the first place.
4. Size Selection by Differential Centrifugation
Differential centrifugation can enrich populations with different sedimentation behavior through successive collection of sediment or supernatant. Size, density contrast, shape, interactions, time and geometry all contribute, so a fraction need not correspond to a sharp size band. Increasing-force processing of a supernatant and decreasing-speed reprocessing of sediments are distinct workflows. Rate-zonal separation in a gradient is not a synonym for differential centrifugation in a uniform medium [2][3][4].
The following examples distinguish workflow, measured endpoint and limits of transfer.
Graphene, decreasing speeds. Khan et al. studied graphene exfoliated in NMP, not GO [3]. After a 4,000 rpm separation, the sediment was redispersed and processed at decreasing speeds of 3,000, 2,000, 1,000, 700 and 500 rpm, 45 min each, collecting supernatant fractions. TEM mean lengths were 0.94, 1.6 and 3.3 µm for the 3,000, 1,000 and 500 rpm fractions; the reported length trend was approximately rotation rate to the power −0.6. The lowest-speed fraction was also thicker. These are measured populations, not cutoffs; the paper gives no radius or RCF mapping.
WS₂ liquid cascade centrifugation. Backes et al. used increasing-speed processing of supernatants and redispersed the collected sediments in fresh surfactant [2]. T2 records the reported conditions. The two forces bounding a fraction define an operational collection interval, not guaranteed particle-size limits. Their targeted two-step approach should not be confused with the cumulative history of a full cascade. They reported decreasing mean length with increasing RCF and a size-selection/yield trade-off; the later enrichment step was a separate treatment, not an ALab performance result.
Gold nanorod enrichment. Sharma et al. studied a mixed as-made dispersion at 5,600 ×g for 20–40 min, with a representative 30 min result [5]. The side-wall deposit was enriched in thinner, longer rods; the bottom contained mainly spheres/cubes together with shorter, thicker rods. TEM and optical spectra supported enrichment, not complete separation into pure shapes. The authors related the behavior to dimensions, shape-dependent drag and interactions; the result does not assign every rod to a side wall.
Other approaches and secondary GO evidence. Bonaccorso et al. describe two approaches: rate-zonal separation in a high-viscosity gradient and selective sedimentation, demonstrated for 10–30 nm Au/Ag particles [6]. No procedure or ALab equivalence is supplied here. Separately, Tufano's review cites Chen J et al.'s successive-centrifugation work on 0.1 mg/mL GO, with reported mean lateral sizes approximately 260–977 nm [4][12]. The primary Chen J study was not consulted; those secondary mean values are not cutoff bounds and provide no settings for this article.
These studies do not support a universal "X ×g for Y minutes gives particles below Z nm" table. Reported trends belong to particular preparations and analytical endpoints. Fraction widths, recovery, thickness and shape need separate evidence; a speed interval alone does not determine them.
5. Why Graphene Oxide Behaves Differently
GO dispersions can contain individual sheets, multilayer material and aggregates with different oxidation and hydration states. A lateral dimension alone does not describe their buoyant mass, drag or stability [4][13].
Hydration and oxidation can matter. Huang and Yuan's abstract describes GO spanning roughly 400–1,100 nm and 2–17 layers, and reports effects of hydration, interlayer spacing and oxidation on sedimentation [13]. More oxidized sheets in their study sedimented more slowly; lateral size was not the only controlling variable. The implication is a possible source of batch dependence, not a universal ordering for all GO.
A wash can change the retained GO. Thomas et al.'s published methods confirm the treatment and spin reported in T2 [9, p. 3587]. For their Hummers-route sample, the reported 25% oxidation-debris yield came from combined supernatants across the base-wash and subsequent wash/reprotonation sequence, after accounting for salt; it was not the yield from one spin. They distinguished removal of debris from chemical reduction. This supports checking what the separated phases contain, not assigning a fixed debris fraction or prescribed base wash to another GO sample.
The retained phase depends on the step. Chen T et al.'s conference paper reports 4,000 rpm for 30 min at a stated 14 cm rotor radius to remove unexfoliated material [11]. This is a specific preparation step, not a general GO pelleting instruction. Record which fraction was retained; sediment is not automatically the desired product.
For more on how GO itself is characterized, see our benchmark GO characterization summary (FORTH, 2025) and the single-layer GO dispersion study from Imperial College London (2017). These material-characterization articles do not establish an ALab centrifuge pairing.
6. Redispersion: Getting the Pellet Back Into Suspension
When the intended product is a dispersion, recovery includes testing its state after redispersion. Backes reports mild sonication in fresh surfactant [2]; Khan reports 15 min bath sonication of sediment in 16 mL NMP [3]. These are attributed preparation details, not transferable acoustic doses or proof of unchanged particle size in another setup.
NIST SP 1200-2 addresses ultrasonic dispersion primarily from powders and discusses extension to redispersion [14, Sections 2–4]. Its general considerations include:
- Probe and bath arrangements deliver energy differently; method and vessel details matter.
- Sonication can disperse clusters, promote agglomeration or alter material chemistry. A visible suspension does not establish an unchanged product.
- Temperature control and energy assessment belong to the validated setup; a dial setting or nominal elapsed time is not a transferable delivered dose.
- Calorimetric power assessment is one documented approach, not proof that equal power produces equal dispersions across systems.
For sheets, sonication can also cut flakes, so a redispersion step can change the size distribution that a previous spin selected.
Removal of stabilizers may alter colloidal stability; La Spina's repeated-wash result is one system-specific example [7]. Select analytical methods for the required size, shape, thickness, concentration and chemical endpoints. UV-vis, DLS and microscopy answer different questions; a DLS equivalent size is not automatically a GO lateral dimension, and no single signal establishes every aspect of recovery.
7. Specifications of Five ALab™ High-Speed Centrifuges
No first-party nanoparticle/GO centrifugation measurements are reported in this article. The ACS Material product descriptions cited here do not establish these applications. T3 is a separate summary of the specifications and of what is not listed, not a paper-to-model pairing or evidence that any model pellets, washes, size-selects or redisperses a sample. D1524 is the model included from the D1524 / D1524C pair; D1524C is excluded.
| Specification | D1524 | D1524R | D3024 | DG1616 | DG1616R |
|---|---|---|---|---|---|
| Max. speed (setting increment) | 15,000 rpm (100 rpm) | 15,000 rpm (100 rpm) | 15,000 rpm (10 rpm); accuracy ±20 rpm | 16,000 rpm (1 rpm) | 16,000 rpm (1 rpm) |
| Max. RCF (step) | 21,380 ×g (10 ×g) | 21,380 ×g (10 ×g) | 21,380 ×g (10 ×g) | 28,621 ×g (1 ×g) | 28,621 ×g (1 ×g) |
| Tube formats on listed rotors | 0.2/0.5/1.5/2 mL microtubes; 5 mL round-/cone-bottom tube formats; PCR strips; listed adapters where required | Same five rotors | Same five rotors | Rotor table from 2 mL microtubes and PCR strips up to 15/50 mL tubes and 200/400 mL bottles | Same as DG1616 |
| Rotor ratings and radius | All five rotors rated 15,000 rpm / 21,380 ×g; radius not specified | As D1524 | As D1524 | Specified per rotor, e.g. 30×2 mL at 16,000 rpm / 28,620 ×g; 6×15 or 6×50 mL at 12,500 rpm / 17,468 ×g; both at a specified 10 cm radius; reference-point definition not specified | As DG1616 |
| Temperature setting | — | −20 °C to 40 °C; RT to 4 °C in 7 min | — (double cooling air design) | — | −10 °C to 40 °C; RT to 4 °C in as little as 7 min |
| Ramps (specified levels, not times) | 9 acceleration/deceleration levels | 9 levels | Not compared here | 9 acceleration / 9 deceleration levels | 9 acceleration / 9 deceleration levels |
| Imbalance detection | — | Rotor imbalance detection alarm | — | Automatic rotor imbalance warning | Automatic rotor imbalance warning |
Source: ACS Material product specifications (October 2026); products are linked in Keep Exploring. A dash means the specification is not listed. The DG1616 / DG1616R headline is 28,621 ×g, while the AS30-2 rotor is rated to 28,620 ×g. These maxima do not apply to the largest-capacity rotor. Listed pre-cooling times have unspecified test conditions and establish neither sample temperature nor readiness. Ramp levels and alarm/warning names do not establish profiles, detection thresholds or responses. Missing information is not proof of feature absence; none of these specifications validates a materials application.
The following documentation checks do not select an ALab model for any cited study:
- A rotor rating concerns its documented configuration, not every tube/adapter or material.
- Paper rpm needs an applicable radius or documented conversion basis to establish RCF; do not fill missing study fields from ACS Material specifications.
- Even a known RCF does not resolve sample, path, fill, time, ramps or recovery requirements.
- A temperature setting and a named detector do not establish loaded-sample performance or safe containment.
8. A Method-Development Checklist
Use these as evidence and documentation questions under an approved development plan, not an operating sequence or material-specific prescription.
- Identify the source. Record the version and passage you used; distinguish reported values from calculations.
- Define the endpoint. Specify which phase is retained and how purity, recovery and the relevant distributions will be assessed.
- Keep conditions traceable. Record material history, medium, concentration, temperature, rotor, tube/adapter, fill, radius/RCF basis and timing. Do not silently substitute missing conditions.
- Evaluate phase accounting. Where appropriate, characterize representative sediment and supernatant samples; wet pellet mass alone is not nanoparticle recovery.
- Assess washing changes. Use contaminant and material-stability endpoints rather than a fixed cycle count [7].
- Assess redispersion. Record acoustic arrangement, energy assessment, volume, timing and temperature; test the resulting dispersion [14].
- Check the complete assembly. Confirm compatibility, permitted settings, loads and approved mass-balancing patterns; equal fill volumes are not sufficient.
- Address exposure and waste. Include aerosol, splash, solvent and nanomaterial risks in the approved handling and containment plan [14]. No containment or biosealing performance is established for a listed centrifuge here.
- Consider alternatives by endpoint. Filtration, diafiltration and dialysis require their own suitability evidence [8][10].
9. Frequently Asked Questions
What RCF do I need to pellet nanoparticles?
No universal value follows from particle size or the studies here. Shape, buoyant mass, interactions, medium, geometry and time matter; the diameter-squared law belongs to the ideal sphere model. A literature example does not qualify a different sample or instrument. Define and assess the required separation endpoint.
Can I use rpm from a paper directly?
Not as proof of equivalent processing. Establish the paper's rotor, tube/fill configuration and RCF reference radius or reported conversion relation. Then assess the candidate configuration and method separately. An identical rotor name alone does not resolve load, timing, medium or temperature differences.
How many wash cycles should I use?
The literature establishes no universal count. La Spina's results concern one tested citrate-gold preparation, not an allowable cycle threshold for other particles [7]. Assess contaminant removal, particle loss and material change against the defined endpoint.
Can centrifugation give me a specific GO size fraction?
Studies report enriched GO populations, including the secondary mean-size example discussed above [4][12]. Their means are not hard cutoffs. Oxidation, hydration and layer structure can affect sedimentation [13]. Measure the actual lateral-size and thickness distributions; do not infer them from speed and time alone.
Should nanoparticle centrifugation be done cold?
Follow the material's stability and processing requirements. T1's last row changes only an assigned viscosity ratio; real temperature changes may also alter density contrast and colloidal behavior. Record starting conditions and the setting, and assess actual sample-temperature performance rather than assuming a cold run is preferable.
Why won't my pellet redisperse?
Possible causes include loss of stabilizing ions or surfactant and true aggregation. Redisperse in a controlled way and measure the result [14]; if repeated washes are the cause, fewer cycles or another purification method may help [7][8].
10. Keep Exploring Centrifugation
- Laboratory Centrifugation: RCF, Rotors, Sedimentation and How to Choose a Centrifuge: the centrifugation hub covering RCF, rotors, sedimentation, temperature, safety and the full series of guides.
Exfoliated materials: graphene nanoplatelets and dispersions guide.
High-speed models: D1524, D1524R, D3024 and DG1616/DG1616R.
Browse all ALab™ centrifuges.