Nanofluid and Hybrid Nanofluid Models

1. The single-phase (Tiwari–Das) approach

Treat the suspension as one homogeneous fluid whose properties depend on the particle volume fraction φ. The standard correlations, with subscripts f (base fluid) and s (solid particles):

The boundary-layer equations keep their form; only the coefficients change. This is why so much of the literature reduces to the classical ODEs with φ-dependent constant groups multiplying each term.

2. The two-phase (Buongiorno) approach

Buongiorno (2006) argued that two slip mechanisms between particles and fluid actually matter: Brownian diffusion and thermophoresis (particles drifting down temperature gradients). A concentration field C joins the system, and the energy and concentration equations gain the coupled terms

After similarity reduction these appear as the parameters Nb and Nt multiplying θ′φ′ and θ′² — the signature of every Buongiorno-model paper.

3. Hybrid nanofluids

A hybrid nanofluid suspends two particle species (Cu–Al₂O₃/water is the canonical pair, after Devi & Devi 2016). Properties are built sequentially: dope the base fluid with φ₁ of the first particle using the correlations above, then treat that mixture as the new base fluid for φ₂ of the second. The aim is to pair a highly conductive particle with a chemically stable one and beat both mono-nanofluids on heat transfer at equal loading.

Key takeaway

Model choice is the first line of any nanofluid paper: Tiwari–Das when particle loading is the knob being studied; Buongiorno when Brownian and thermophoretic transport are the physics of interest. Both funnel into the same similarity machinery and the same numerical methods.

4. Extended worked examples

Example 1 — full property set for Cu–water at φ = 0.02. Data: water ρ_f = 997, μ_f = 10⁻³, k_f = 0.613; copper ρ_s = 8 933, k_s = 401 (SI). Density: ρ_nf = 0.98×997 + 0.02×8 933 = 1 155.7 kg/m³ (+16%). Viscosity (Brinkman): μ_nf = 10⁻³/(0.98)^{2.5} = 1.052×10⁻³ (+5.2%). Conductivity (Maxwell): numerator 401 + 1.226 − 2(0.02)(0.613 − 401) = 418.24; denominator 402.226 + 0.02(0.613 − 401) = 394.22; k_nf/k_f = 1.061 → k_nf = 0.650 W/m·K (+6.1%). Two percent of metal, six percent more conduction, five percent more friction: the whole engineering trade in one line.

Example 2 — the heat-capacity subtlety. (ρc_p)_nf = (1−φ)(ρc_p)_f + φ(ρc_p)_s = 0.98×(997×4 179) + 0.02×(8 933×385) = 4.083×10⁶ + 0.069×10⁶ = 4.152×10⁶ J/m³K — lower than water's 4.166×10⁶. Metals conduct heat well but store it poorly per volume; nanofluids therefore raise conductivity while slightly reducing volumetric heat capacity, and thermal diffusivity α = k/ρc_p rises on both counts.

Example 3 — hybrid construction, step by step. Cu–Al₂O₃/water with φ₁ = 0.01 alumina then φ₂ = 0.01 copper. Stage 1 (alumina, ρ = 3 970): ρ₁ = 0.99×997 + 0.01×3 970 = 1 026.7; μ₁ = 10⁻³/0.99^{2.5} = 1.0254×10⁻³. Stage 2 treats stage 1 as base: ρ_hnf = 0.99×1 026.7 + 0.01×8 933 = 1 105.8; μ_hnf = μ_f/[(0.99)^{2.5}(0.99)^{2.5}] = 1.052×10⁻³. Conductivity chains Maxwell twice the same way. The bookkeeping is mechanical; the physics claim — that the two species act independently and sequentially — is the model's honest weak point, and papers say so.

Example 4 — what it buys in the boundary layer. In the similarity-reduced momentum equation the viscous term acquires the multiplier μ_nf/μ_f divided by ρ_nf/ρ_f — for Example 1, 1.052/1.159 = 0.908: effective ν drops 9%, thinning the layer and raising wall shear, while the energy equation's k_nf/(ρc_p)_nf multiplier raises thermal diffusion 6.5%. Both push the Nusselt number up — the reason parameter tables in nanofluid papers show Nu increasing in φ, at least until viscosity's pumping-power bill is presented.

5. Common misconceptions

"More particles, more cooling — monotonically." Conductivity gains scale roughly like 3φ (Maxwell, small φ) while Brinkman viscosity grows like 2.5φ and accelerates; beyond a few percent loading, pressure drop and pumping power erase the thermal benefit in most practical loops. Optimal φ is an optimisation, not a maximisation.

"The correlations are laws." Maxwell assumes dilute, well-dispersed spheres with no interfacial resistance; Brinkman assumes the same. Nanoparticle clustering, shapes (tubes, platelets), interfacial (Kapitza) resistance and surfactants push real data both above and below the formulas — the famous scatter of the experimental literature. The models are transparent baselines, which is exactly why theory papers standardise on them.

"Buongiorno replaces Tiwari–Das." Different questions: Tiwari–Das asks "what does loading φ do to a homogeneous coolant?"; Buongiorno asks "how do the particles themselves migrate?" A paper studying thermophoretic deposition needs Buongiorno; one comparing Cu versus Al₂O₃ at fixed φ does not.

"Stability is a lab detail." Sedimentation and aggregation are the primary reason commercial uptake lags the paper count: a suspension that stratifies in a week has no k_nf worth publishing. Serious experimental papers report zeta potentials and settling tests alongside conductivity.

6. Where this shows up

Electronics thermal management is the flagship application — cold plates for power electronics and data-centre liquid cooling, where a few percent conductivity headroom is worth real money. Flat-plate and parabolic-trough solar collectors report measurable efficiency gains with oxide nanofluids; machining coolants (minimum-quantity lubrication) exploit both thermal and tribological effects; and automotive/radiator studies chase smaller frontal areas. On the modelling side, the property correlations of this chapter are the entry ticket to the entire MHD-nanofluid-stretching-sheet literature: every governing-equation section you will ever read in that field opens with precisely these formulas.

7. Practice problems

P1. Al₂O₃–water at φ = 0.05 (ρ_s = 3 970, k_s = 40): find ρ_nf, μ_nf/μ_f, k_nf/k_f. Solution: ρ = 0.95×997 + 0.05×3 970 = 1 145.7; μ ratio = (0.95)^{−2.5} = 1.137; Maxwell: (40+1.226−0.1(0.613−40))/(41.226+0.05(0.613−40)) = 45.16/39.26 = 1.150.

P2. Show Maxwell's formula → 1 + 3φ(k_s−k_f)/(k_s+2k_f) for small φ, and evaluate the slope for Cu–water. Solution: first-order expansion; slope 3(401−0.613)/(401+1.226) ≈ 2.99 — conductivity rises ~3% per 1% loading.

P3. At what φ does Brinkman predict a 50% viscosity rise? Solution: (1−φ)^{−2.5} = 1.5 → 1−φ = 1.5^{−0.4} = 0.850 → φ ≈ 0.15 — far beyond dilute validity; note the model is being stretched, which is the point of the exercise.

P4. A Buongiorno paper reports Nb = 0.3, Nt = 0.1. State physically what raising each does to the concentration boundary layer. Solution: Nb (Brownian) mixes particles toward uniformity, thinning concentration gradients; Nt (thermophoresis) drives particles from hot wall to cool interior, piling concentration away from the wall — the classic opposing pair whose ratio Nt/Nb papers tabulate.

8. Going deeper

The deepest open question is why measured enhancements sometimes exceed Maxwell substantially. Candidate mechanisms — interfacial nanolayers of ordered liquid, percolating aggregates conducting along fractal chains, micro-convection stirred by Brownian motion — each have models (Yu–Choi, Nan, Prasher) that extend Maxwell with extra parameters, and a landmark international benchmark exercise (Buongiorno et al., 2009, ~30 labs) found that careful measurements largely agree with effective-medium theory, attributing outliers to aggregation and measurement artefacts. The field's lesson generalises: when experiment and clean theory disagree, audit the suspension before rewriting the physics.

9. Historical context

The effective-medium mathematics predates the fluids by a century: Maxwell derived the conductivity of a dilute sphere suspension in his 1873 Treatise (for electrical conduction — the thermal analogy is exact), and Hamilton & Crosser (1962) extended it to particle shapes. Einstein's 1906 dissertation gave suspension viscosity its leading term μ(1 + 2.5φ) — the same 2.5 that Brinkman's 1952 formula resums to higher concentrations. The fluids arrived when Choi and Eastman at Argonne National Laboratory coined "nanofluid" in 1995, reporting that nanometre particles (unlike the settling, abrasive micro-slurries of earlier decades) could stay suspended and boost conductivity at tiny loadings. Early reports of enhancements far beyond Maxwell ignited both a research boom and a controversy; the 34-laboratory INPBE benchmark (Buongiorno et al., 2009) largely vindicated classical theory for well-dispersed suspensions. Buongiorno's separate 2006 analysis of slip mechanisms — concluding Brownian diffusion and thermophoresis are the two that matter — supplied the second modelling tradition that now shares the literature with Tiwari–Das.

10. Another way to see it: where Maxwell's formula comes from

The formula is a boundary-value problem, not a fit. Place one sphere (conductivity k_s) in an infinite matrix (k_f) carrying a uniform far-field temperature gradient G. Solving Laplace's equation inside and out with continuity of temperature and flux at the surface gives an exterior field equal to the undisturbed gradient plus a dipole of strength proportional to (k_s − k_f)/(k_s + 2k_f) — the "2" is the sphere's depolarising geometry. For a dilute suspension, superpose the dipoles of number density n: the mixture behaves as a medium whose effective conductivity is shifted by the total dipole moment per volume, and to first order in φ,

Maxwell's full expression is the standard resummation of this result. Every assumption is now visible: spheres (else the shape factor changes — Hamilton–Crosser's n), dilute (dipoles must not interact — fails when aggregates form chains), perfect thermal contact (interfacial Kapitza resistance subtracts from k_s's effect), and steady conduction (Brownian micro-convection is outside the calculation). The experimental scatter of the field maps almost one-to-one onto violations of this list — which is why the derivation is worth owning, not just the formula.

11. Frequently asked questions

Are nanofluids actually used commercially? Niche but real: some machining coolants, heat-pipe working fluids and transformer-oil formulations ship with nanoparticle additives. Mass adoption is held back by cost, long-term stability and erosion questions — engineering issues, not thermal ones.

Which conducts better, one great particle species or a hybrid? At equal total φ, a high-k species alone usually wins on pure Maxwell arithmetic; hybrids are studied for compound benefits (one species for conductivity, one for stability or lubricity) and, frankly, because the model family is publishable. The chapter's formulas let you audit any specific claim in minutes.

Do the property formulas depend on temperature? Through the base fluid, strongly (water's μ halves from 20 to 60 °C); the mixture rules themselves are temperature-blind. Serious simulations evaluate μ_f(T), k_f(T) locally — a "variable properties" tag in the literature.

What does φ = 0.02 mean physically — how crowded is that? Two percent by volume of 30 nm particles is ~10²¹ particles per cubic metre with mean spacing ~4 diameters: close enough to matter thermally, sparse enough that the dilute approximation is defensible — exactly the corridor the models were built for.

12. Further practice

P5. From §10's dilute limit, what particle conductivity achieves 90% of the k_s → ∞ ceiling for water? Solution: the factor (k_s−k_f)/(k_s+2k_f) → 1; at 90%: k_s ≈ 28k_f ≈ 17 W/m·K. Beyond ~20k_f, better particles buy almost nothing — why cheap alumina competes with silver, and why "ultra-high-k particle" claims deserve this two-line audit.

P6. Estimate the Prandtl number of the φ = 0.02 Cu–water fluid of Example 1. Solution: Pr = μc_p/k with mixture values: c_p = (ρc_p)_nf/ρ_nf = 4.152×10⁶/1 155.7 = 3 593; Pr = 1.052×10⁻³×3 593/0.650 = 5.8, down from water's 6.8 — the thermal layer thickens relative to momentum, feeding directly into the θ-equation of the heat-transfer chapter.

P7. A paper claims k_nf/k_f = 1.35 at φ = 0.01 for CuO–water. Audit it. Solution: Maxwell slope ≈ 3 per unit φ → expected ≈ 1.03. A 35% claim at 1% loading exceeds effective-medium theory by an order of magnitude — either extraordinary aggregation physics, or a measurement artefact; INPBE's verdict says bet on the latter.

13. Worked exam problem

Problem. Compute the Tiwari–Das property set for TiO₂–water at φ = 0.03 (TiO₂: ρ_s = 4 250 kg/m³, k_s = 8.95 W/m·K; water as before), and state the percentage changes.

Solution. Density: ρ_nf = 0.97×997 + 0.03×4 250 = 967.1 + 127.5 = 1 094.6 kg/m³ (+9.8%). Viscosity: μ_nf/μ_f = (0.97)^{−2.5} = 1.079 (+7.9%). Conductivity (Maxwell): numerator 8.95 + 1.226 − 2×0.03×(0.613 − 8.95) = 10.176 + 0.500 = 10.676; denominator 10.176 + 0.03×(−8.337) = 9.926; ratio 1.076 (+7.6%). Note the instructive comparison with Cu (Example 1): TiO₂'s conductivity is 45× water's, copper's is 650× — yet the enhancements are 7.6% versus 6.1%-at-lower-φ, nearly saturated in both cases. P5's ceiling effect, live: past k_s ≈ 20k_f, particle quality stops paying, and material choice becomes about density, cost and stability instead.

14. Key takeaways

Tiwari–Das nanofluid modelling is three mixture rules — linear for ρ and ρc_p, Brinkman for μ, Maxwell for k — feeding modified coefficients into otherwise unchanged equations; Buongiorno instead transports the particles with Brownian and thermophoretic fluxes. Both are transparent baselines, not laws: dilute, spherical, well-dispersed assumptions are load-bearing, conductivity buys come with viscosity bills, and every claimed enhancement can be audited against Maxwell's slope of ~3% per 1% loading.

15. Where to go next

These coefficients drop into the stretching-sheet momentum equation and shift the thermal problem's effective Pr; the solutions are computed with bvp4c-style numerics. The kinematic viscosity calculator runs all four mixture rules with steps, for single and hybrid suspensions.