Water has been the default coolant in engineering for well over a century, and for good reason: it is cheap, safe, abundant, and has a reasonably high specific heat capacity. But water alone is running out of headroom. As electronics get denser, batteries get more energy-hungry, and heat exchangers are asked to shrink while doing more work, the humble thermal conductivity of water; around 0.6 W m⁻¹ K⁻¹ at room temperature, is becoming a bottleneck. Over the past few years, our group has been asking a simple question: what happens if we don't replace water, but upgrade it, by dispersing carefully chosen nanoparticles into it to form a "nanofluid"?
This post pulls together several of our recent, peer-reviewed studies to give a behind-the-scenes look at what we found, why it matters, and where this line of work is heading, particularly for applications like electronics cooling, compact heat exchangers, and high-heat-flux systems where a smaller, more efficient cooling loop can make a real engineering difference.
Why bother with nanofluids?
The idea is not new, it goes back to Choi and Eastman's original proposal in the 1990s, but the engineering payoff has become much clearer as measurement techniques, machine learning, and materials characterisation have matured. Suspending nanometre-scale particles (metal oxides, carbon nanomaterials, magnetic particles) in a base fluid like deionised water can raise thermal conductivity, and in some cases electrical conductivity, while keeping viscosity increases manageable. That last point matters: a fluid that transfers heat brilliantly but is too thick to pump efficiently is not actually useful. Much of our work has focused on finding formulations that strike a workable balance between the two.
From single nanoparticles to ternary hybrids
Our earliest work in this series looked at binary and ternary combinations to see whether mixing nanoparticle types could do better than any one type alone.
In one study, we combined graphene nanoplatelets (GNP), aluminium oxide (Al₂O₃), and ferric oxide (Fe₂O₃), separately, in pairs, and all three together, in deionised water at a fixed 0.10% volume concentration, and tracked electrical conductivity, thermal conductivity, and viscosity from 15–60 °C. The ternary blend outperformed the single- and two-component versions across the board: one formulation showed a 394.72% jump in electrical conductivity and a 37.50% rise in viscosity, while another achieved a 27.46% improvement in thermal conductivity at 60 °C. We also worked out empirical correlations and performance indicators (a thermo-electrical conductivity ratio and a "figure of merit") to help translate these lab numbers into design guidance for heat exchangers, microchannel coolers, and solar thermal systems (full paper: Comparative thermophysical and thermo-electrical performance of single, binary, and ternary (GNP–Al₂O₃–Fe₂O₃) hybrid nanofluids for heat transfer applications).
A related tri-hybrid study went a step further, dispersing Fe₃O₄, Al₂O₃, and multi-walled carbon nanotubes (MWCNT) together across five different mixing ratios and looking not just at heat transfer performance but at stability, pH behaviour, and sedimentation over time, because a nanofluid that settles out within a week is not going to survive in a real cooling loop. We found that the mixing ratio mattered enormously: higher Al₂O₃ content improved dispersion stability, while more Fe₃O₄ and MWCNT boosted electrical conductivity but pushed up viscosity. One balanced formulation (equal parts of all three nanoparticles) came out as the best all-round compromise between thermal conductivity, stability, and flow resistance. We also flagged a less glamorous but practically important finding: these nanofluids trended acidic as temperature and concentration rose, which raises corrosion questions for metallic cooling systems, and we discuss mitigation strategies like surfactants and buffering agents (full paper: Tri-hybrid nanofluids for thermal applications: stability, magneto-hydrodynamics, and machine learning prediction).
Bringing in machine learning
Once you have a reasonably large experimental dataset, the next question is: can you predict how a nanofluid will behave under conditions you haven't tested yet, without running the experiment all over again? This is where machine learning has become a genuinely useful tool in our recent work, rather than just a buzzword.
For an Al₂O₃/MWCNT hybrid nanofluid, we compared classical linear correlations against Artificial Neural Networks (ANN) and an ANFIS model combined with fuzzy c-means clustering (ANFIS-FCM). At the higher end of the concentration range tested (0.25 vol%, 60 °C), electrical conductivity exceeded 2400 µS cm⁻¹ and thermal conductivity improved by roughly 35–50% relative to plain deionised water. The ANFIS-FCM models consistently beat both the linear correlations and the standard ANN models, and even outperformed more elaborate hybrid optimisation approaches (PSO-ANFIS, GA-ANFIS), suggesting that for this kind of nonlinear, multi-variable thermophysical data, a well-tuned fuzzy-clustered model can be both more accurate and more interpretable than throwing more computational firepower at the problem (full paper: Experimental characterization and predictive modelling of Al₂O₃/MWCNT hybrid nanofluid thermophysical properties using ANN, ANFIS, FCM and hybrid techniques).
We used a related approach for magnetic hybrid ferrofluids, Fe₃O₄ paired separately with TiO₂, MgO, and ZnO, at different hybridisation ratios. Across the board, the 80:20 (Fe₃O₄-rich) ratio gave the best combination of stability and thermal conductivity. The standout result was the Fe₃O₄/ZnO combination, which delivered the best balance of thermal conductivity enhancement (31.28% at 50 °C) and low viscosity, along with the strongest thermoelectric conductivity values, a property that matters directly for cooling proton-exchange-membrane (PEM) fuel cells, where the coolant also needs favourable electrical characteristics. Feature-importance analysis from the machine learning models consistently pointed to temperature as the dominant driver of thermal behaviour, ahead of the specific nanoparticle mixing ratio (full paper: Experimental investigation and machine learning modeling of the effects of hybridization mixing ratio, nanoparticle type, and temperature on the thermophysical properties of Fe₃O₄/TiO₂, Fe₃O₄/MgO, and Fe₃O₄/ZnO-DI water hybrid ferrofluids).
Don't forget density
Thermal conductivity and viscosity tend to get most of the attention, but density is just as fundamental, it governs mass flow rate, pressure drop, and pump sizing in any real cooling system. We've also worked on experimental measurement and machine learning modelling of hybrid nanofluid density, applying similar data-driven techniques to this often-overlooked property, as part of a broader effort to build a complete thermophysical picture rather than optimising one property in isolation (conference paper: Experimental Measurement and Machine Learning Modelling for the Density of Hybrid Nanofluids).
Related work from our research group
The five studies above sit within a broader research programme on nanofluid heat transfer. Readers interested in going deeper may find these directly related papers by our research team:
- Nanofluids for heat transfer enhancement: a holistic analysis of research advances, technological progress and regulations for health and safety
- Nanofluids flow boiling and convective heat transfer in microchannels: a systematic review and bibliometrics analysis
- Magnetohydrodynamics of nanofluid internal forced convection: a review and outlook for practical applications
- Pulsating nanofluid-jet impingement cooling and its hydrodynamic effects on heat transfer and Hydrodynamic effects of hybrid nanofluid jet on the heat transfer augmentation
- Experimental and machine learning study on the influence of nanoparticle size and pulsating flow on heat transfer performance in nanofluid-jet impingement cooling
- Numerical Investigation of Heat Transfer Performance of Hybrid Nanofluid in Porous Substrate in Microchannel Heat Sink. Among others listed in the references
Where this leaves us
Taken together, this body of work points toward a fairly consistent set of design principles for engineered nanofluid coolants:
- Hybridization beats single nanoparticles; but the ratio of components matters more than simply adding more nanoparticle types.
- Stability and viscosity cannot be an afterthought; a fluid with excellent thermal conductivity is worthless if it settles out in a week or demands too much pumping power.
- Machine learning, particularly fuzzy-clustered ANFIS models, is proving genuinely useful for capturing the nonlinear coupling between temperature, concentration, and composition, often outperforming both simple correlations and more computationally expensive hybrid algorithms.
- Application context should drive formulation choice; A fluid optimised for PEM fuel cell cooling (where electrical properties matter) looks different from one optimised purely for electronics heat sinks.
The long-term motivation behind all of this is straightforward: if water-based nanofluids can reliably deliver these kinds of thermal conductivity gains without unmanageable viscosity or stability penalties, the payoff is smaller, lighter, more energy-efficient cooling hardware; for data centres, electric vehicle battery packs, solar collectors, and high-heat-flux electronics alike. There is still work to do on longer-term stability, corrosion behaviour, cost-effective large-scale production, and validation in real flow systems, and that is where our ongoing research is headed next.
Author, Co-authors and affiliations
This chain of work has its roots at the University of Pretoria, Nanofluid Research Group, and has been genuinely collaborative, spanning several institutions. Recurring co-authors across the studies featured in this post include Mohsen Sharifpur, Josua P. Meyer, Victor O. Adogbeji, Modaser Momin, Solomon O. Giwa, Saheed Adio, Stephen Oladipo, Devendra Vishwakarma, Christopher Enweremadu, Luke Ajuka, Chris Govinder, Saad F. M. Noraldeen, and Emmanuel O. Atofarati.
Disclosure:
This post was drafted with the assistance of Claude aIl based on the author's own published research, and the image was modified using Dalle; the content has been reviewed and verified by the author prior to publication.
References
[1] V. O. Adogbeji, E. O. Atofarati, M. Sharifpur, and J. P. Meyer, “Experimental investigation and machine learning modeling of the effects of hybridization mixing ratio, nanoparticle type, and temperature on the thermophysical properties of Fe3O4/TiO2, Fe3O4/MgO, and Fe3O4/ZnO-DI water hybrid ferrofluids,” J. Therm. Anal. Calorim., vol. 150, no. 13, pp. 10549–10573, Jul. 2025, doi: 10.1007/S10973-025-14399-Y/FIGURES/16.
[2] M. Momin et al., “Experimental characterization and predictive modelling of Al2O3/ MWCNT hybrid nanofluid thermophysical properties using ANN, ANFIS, FCM and hybrid techniques,” J. Therm. Anal. Calorim., vol. (In Press), pp. 1–31, Jun. 2026, doi: https://doi.org/10.1007/s10973-026-15698-8 1.
[3] V. O. Adogbeji, E. O. Atofarati, K. Govinder, M. Sharifpur, and J. P. Meyer, “Tri-hybrid nanofluids for thermal applications: stability, magneto-hydrodynamics, and machine learning prediction,” Multiscale Multidiscip. Model. Exp. Des., vol. 8, no. 9, pp. 1–29, Sep. 2025, doi: 10.1007/S41939-025-00968-Z/FIGURES/16.
[4] V. O. Adogbeji, E. O. Atofarati, M. Sharifpur, and J. P. Meyer, “Magnetohydrodynamics of nanofluid internal forced convection: A review and outlook for practical applications,” Results Phys., vol. 68, p. 108082, Jan. 2025, doi: 10.1016/J.RINP.2024.108082.
[5] M. Momin, · Saad, F. M. Noraldeen, E. O. Atofarati, · M Sharifpur, and M. Sharifpur, “Comparative thermophysical and thermo-electrical performance of single, binary, and ternary (GNP–Al2O3–Fe2O3) hybrid nanofluids for heat transfer applications,” J. Therm. Anal. Calorim. 2026, pp. 1–18, Aug. 2026, doi: 10.1007/S10973-026-16017-X.
[6] E. O. Atofarati, M. Sharifpur, and J. Meyer, “Hydrodynamic effects of hybrid nanofluid jet on the heat transfer augmentation,” Case Stud. Therm. Eng., vol. 51, p. 103536, Nov. 2023, doi: 10.1016/J.CSITE.2023.103536.
[7] S. A. Adio, E. O. ATOFARATI, O. A. Muritala, Z. Huan, and V. V. Rao, “Nanofluids Flow Boiling and Convective Heat Transfer in Microchannels : A Systematic,” J. Therm. Anal. Calorim., 2025.
[8] E. O. Atofarati, M. Sharifpur, Z. Huan, O. O. Awe, and J. P. Meyer, “Experimental and machine learning study on the influence of nanoparticle size and pulsating flow on heat transfer performance in nanofluid-jet impingement cooling,” Appl. Therm. Eng., vol. 258, p. 124631, Jan. 2025, doi: 10.1016/J.APPLTHERMALENG.2024.124631.
[9] E. O. Atofarati et al., “Life cycle assessment of nanomaterials and laboratory consumables: toward safer and sustainable nanotechnology,” J. Nanoparticle Res. 2026 284, vol. 28, no. 4, pp. 94-, Mar. 2026, doi: 10.1007/S11051-026-06621-1.
[10] V. O. Adogbeji and T. Lagouge, “Synergistic effects of spiral tape inserts and magnetic hybrid nanofluids on heat transfer and entropy generation in transition flow regime,” Multiscale Multidiscip. Model. Exp. Des., vol. 8, no. 10, pp. 1–29, Nov. 2025, doi: 10.1007/S41939-025-01012-W/FIGURES/12.
[11] V. O. Adogbeji and T. Lagouge, “Effect of Fe3O4/ZnO hybridization ratios on heat transfer and transition behavior in the transition flow regime,” Int. J. Therm. Sci., vol. 220, p. 110238, Feb. 2026, doi: 10.1016/J.IJTHERMALSCI.2025.110238.
[12] V. O. Adogbeji, M. Sharifpur, and J. P. Meyer, “Experimental investigation of heat transfer, thermal efficiency, pressure drop, and flow characteristics of Fe3O4-MgO magnetic hybrid nanofluid in transitional flow regimes,” Int. J. Therm. Sci., vol. 209, p. 109515, Mar. 2025, doi: 10.1016/J.IJTHERMALSCI.2024.109515.
[13] S. A. Adio et al., “Thermal and hydraulic performance of pulsating hybrid nanofluid in a straight and helical-twist rectangular microchannel heat sinks,” Therm. Sci. Eng. Prog., vol. 68, p. 104338, Dec. 2025, doi: 10.1016/J.TSEP.2025.104338.
[14] E. O. Atofarati, M. Sharifpur, and J. P. Meyer, “Pulsating nanofluid-jet impingement cooling and its hydrodynamic effects on heat transfer,” Int. J. Therm. Sci., vol. 198, p. 108874, Apr. 2024, doi: 10.1016/J.IJTHERMALSCI.2023.108874.
[15] E. O. Atofarati, “Investigation into heat transfer enhancement using AL2O3-MWCNT hybrid nanofluids in jet cooling,” University of Pretoria, Hatfield, South Africa, 2024.
[16] E. O. Atofarati, S. Mohsen, and J. P. Meyer, “Parametric influences on nanofluid-jet cooling heat transfer,” vol. 1, Elsevier, 2024, ch. 12, pp. 351–398. doi: 10.1016/B978-0-443-13625-2.00014-0.
[17] E. O. Atofarati and C. Enweremadu, “Experimental Measurement and Machine Learning Modelling for the Density of Hybrid Nanofluids,” in Proceedings of the 11th World Congress on New Technologies (NewTech’25), 2025, pp. 142-1-142–27. doi: 10.11159/icert25.142.
[18] V. O. Adogbeji, M. Sharifpur, and J. P. Meyer, “Experimental investigation of heat transfer, thermal efficiency, pressure drop, and flow characteristics of Fe3O4-MgO magnetic hybrid nanofluid in transitional flow regimes,” Int. J. Therm. Sci., vol. 209, p. 109515, Mar. 2025, doi: 10.1016/J.IJTHERMALSCI.2024.109515.
[19] V. O. Adogbeji and T. Lagouge, “Effect of Fe3O4/ZnO hybridization ratios on heat transfer and transition behavior in the transition flow regime,” Int. J. Therm. Sci., vol. 220, p. 110238, Feb. 2026, doi: 10.1016/J.IJTHERMALSCI.2025.110238.
[20] V. O. Adogbeji and T. Lagouge, “Synergistic effects of spiral tape inserts and magnetic hybrid nanofluids on heat transfer and entropy generation in transition flow regime,” Multiscale Multidiscip. Model. Exp. Des., vol. 8, no. 10, pp. 1–29, Nov. 2025, doi: 10.1007/S41939-025-01012-W/FIGURES/12.
[21] I. U. Ibrahim, M. Sharifpur, O. Manca, and J. P. Meyer, “Nanofluid’s convective heat transfer for laminar, transitional, and turbulent flow,” Nanofluid Appl. Adv. Therm. Solut., pp. 151–192, Jan. 2023, doi: 10.1016/B978-0-443-15239-9.00006-0.
[22] I. U. Ibrahim, M. Sharifpur, J. P. Meyer, and S. M. S. Murshed, “Experimental investigations of effects of nanoparticle size on force convective heat transfer characteristics of Al2O3 - MWCNT hybrid nanofluids in transitional flow regime,” Int. J. Heat Mass Transf., vol. 228, p. 125597, Aug. 2024, doi: 10.1016/J.IJHEATMASSTRANSFER.2024.125597.
[23] I. U. Ibrahim, M. Sharifpur, and J. P. Meyer, “Experimental investigations of particle sizes effects on exergy and entropy characteristics of AL2O3 - MWCNT hybrid nanofluid along the transitional flow regime,” Case Stud. Therm. Eng., vol. 51, p. 103575, Nov. 2023, doi: 10.1016/J.CSITE.2023.103575.
[24] I. Umar Ibrahim, M. Sharifpur, and J. P. Meyer, “Nanofluids mixed convection, heat transfer characteristics and effects of volume concentrations in the thermally developing region,” Exp. Therm. Fluid Sci., vol. 172, p. 111626, Jan. 2026, doi: 10.1016/J.EXPTHERMFLUSCI.2025.111626.