Stefan G. Llewellyn Smith
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Stefan G. Llewellyn Smith has a total of 87 co-authors. Here are the top co-authors by number of shared publications:
- Ching Chang
Department of Mechanical and Aerospace Engineering, Jacobs School of Engineering, UCSD, 9500 Gilman Drive, La Jolla, CA 92093-0411, USA
6 shared publications - Sébastien Michelin
LadHyX, CNRS - Ecole Polytechnique, Institut Polytechnique de Paris, 91120 Palaiseau, France
6 shared publications W. R. Young
Scripps Institution of Oceanography, University of California, San Diego, La Jolla CA 92093-0230, USA
6 shared publications- Sarah T. Gille
Scripps Institution of Oceanography University of California San Diego La Jolla CA USA
5 shared publications - William R. Young
University of California San Diego
5 shared publications Anthony M. J. Davis
Department of Mechanical and Aerospace Engineering, Jacobs School of Engineering, UCSD, 9500 Gilman Drive, La Jolla, CA 92093-0411, USA
4 shared publications
Latest Publications
- Open Access
- 15 June 2026
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 1037, A15
Suppression of Rayleigh–Bénard convection and restratification by horizontal convection
We investigate the competition between horizontal convection (HC) and Rayleigh–Bénard convection (RBC) in a fluid layer subject to a uniform destabilising buoyancy flux at the bottom and a horizontally varying buoyancy distribution at the top. The RBC forcing imposes negative horizontal mean vertical buoyancy gradients at the top and bottom of the fluid layer. But if the HC forcing is sufficiently strong then the volume-averaged vertical buoyancy gradient, left angle bracket b Subscript z Baseline right angle bracket $\langle b_z \rangle$ , is positive i.e. opposite in sign to destabilising RBC buoyancy gradients at the boundaries. If left angle bracket b Subscript z Baseline right angle bracket greater than 0 $\langle b_z \rangle \gt 0$ we say that the layer has been ‘restratified’. Using scaling analysis based on power integrals together with two-dimensional direct numerical simulations at Rayleigh numbers up to 10 Superscript 10 $10^{10}$ , we identify two cases: a neutral stratification state, in which HC first offsets RBC so that left angle bracket b Subscript z Baseline right angle bracket equals 0 $\langle b_z \rangle = 0$ , and a strong stratification regime, in which HC dominates and left angle bracket b Subscript z Baseline right angle bracket $\langle b_z \rangle$ is opposite in sign, and greater in magnitude, than the prescribed destabilising vertical buoyancy gradient at the layer boundaries. For the range of parameters explored in this study, we derive scaling laws for the onset of these regimes in terms of the horizontal and vertical flux Rayleigh numbers, upper R a Subscript upper H $\mathit{Ra_H}$ and upper R a Subscript upper V $\mathit{Ra_V}$ , finding upper R a Subscript upper H Superscript upper N Baseline tilde upper R a Subscript upper V Superscript 4 divided by 5 $\mathit{Ra}^{\mathit{N}}_{\mathit{H}} \sim \mathit{Ra_V}^{4/5}$ for the neutral state and upper R a Subscript upper H Superscript upper S Baseline tilde upper R a Subscript upper V $\mathit{Ra}^{\mathit{S}}_{\mathit{H}} \sim \mathit{Ra_V}$ for the onset of strong stratification. The results highlight the controlling role of the top boundary layer in setting the mean stratification and clarify the conditions under which HC suppresses RBC. These findings are relevant to geophysical environments such as subglacial lakes, and the oceans of snowball Earth and icy moons, where bottom heating and horizontal buoyancy variations jointly shape ocean stratification.- 15 June 2026
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 1037, A15
Open AccessSuppression of Rayleigh–Bénard convection and restratification by horizontal convection
We investigate the competition between horizontal convection (HC) and Rayleigh–Bénard convection (RBC) in a fluid layer subject to a uniform destabilising buoyancy flux at the bottom and a horizontally varying buoyancy distribution at the top. The RBC forcing imposes negative horizontal mean vertical buoyancy gradients at the top and bottom of the fluid layer. But if the HC forcing is sufficiently strong then the volume-averaged vertical buoyancy gradient, left angle bracket b Subscript z Baseline right angle bracket $\langle b_z \rangle$ , is positive i.e. opposite in sign to destabilising RBC buoyancy gradients at the boundaries. If left angle bracket b Subscript z Baseline right angle bracket greater than 0 $\langle b_z \rangle \gt 0$ we say that the layer has been ‘restratified’. Using scaling analysis based on power integrals together with two-dimensional direct numerical simulations at Rayleigh numbers up to 10 Superscript 10 $10^{10}$ , we identify two cases: a neutral stratification state, in which HC first offsets RBC so that left angle bracket b Subscript z Baseline right angle bracket equals 0 $\langle b_z \rangle = 0$ , and a strong stratification regime, in which HC dominates and left angle bracket b Subscript z Baseline right angle bracket $\langle b_z \rangle$ is opposite in sign, and greater in magnitude, than the prescribed destabilising vertical buoyancy gradient at the layer boundaries. For the range of parameters explored in this study, we derive scaling laws for the onset of these regimes in terms of the horizontal and vertical flux Rayleigh numbers, upper R a Subscript upper H $\mathit{Ra_H}$ and upper R a Subscript upper V $\mathit{Ra_V}$ , finding upper R a Subscript upper H Superscript upper N Baseline tilde upper R a Subscript upper V Superscript 4 divided by 5 $\mathit{Ra}^{\mathit{N}}_{\mathit{H}} \sim \mathit{Ra_V}^{4/5}$ for the neutral state and upper R a Subscript upper H Superscript upper S Baseline tilde upper R a Subscript upper V $\mathit{Ra}^{\mathit{S}}_{\mathit{H}} \sim \mathit{Ra_V}$ for the onset of strong stratification. The results highlight the controlling role of the top boundary layer in setting the mean stratification and clarify the conditions under which HC suppresses RBC. These findings are relevant to geophysical environments such as subglacial lakes, and the oceans of snowball Earth and icy moons, where bottom heating and horizontal buoyancy variations jointly shape ocean stratification. - 1 December 2025
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 1024
Instability triggered by mixed convection in a thin fluid layer
We investigate the convective stability of a thin, infinite fluid layer with a rectangular cross-section, subject to imposed heat fluxes at the top and bottom and fixed temperature along the vertical sides. The instability threshold depends on the Prandtl number as well as the normalized flux difference ( $f$ ) and decreases with the aspect ratio ( $\epsilon$ ), following a $\epsilon f^{-1}$ power law. Using a three-dimensional (3-D) initial value and two-dimensional eigenvalue calculations, we identify a dominant 3-D mode characterized by two transverse standing waves attached to the domain edges. We characterize the dominant mode’s frequency and transverse wavenumber as functions of the Rayleigh number and aspect ratio. An analytical asymptotic solution for the base state in the bulk is obtained, valid over most of the domain and increasingly accurate for lower aspect ratios. A local stability analysis, based on the analytical base state, reveals oscillatory transverse instabilities consistent with the global instability characteristics. The source term for this most unstable mode appears to be interactions between vertical shear and horizontal temperature gradients.- 1 December 2025
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 1024
Instability triggered by mixed convection in a thin fluid layer
We investigate the convective stability of a thin, infinite fluid layer with a rectangular cross-section, subject to imposed heat fluxes at the top and bottom and fixed temperature along the vertical sides. The instability threshold depends on the Prandtl number as well as the normalized flux difference ( $f$ ) and decreases with the aspect ratio ( $\epsilon$ ), following a $\epsilon f^{-1}$ power law. Using a three-dimensional (3-D) initial value and two-dimensional eigenvalue calculations, we identify a dominant 3-D mode characterized by two transverse standing waves attached to the domain edges. We characterize the dominant mode’s frequency and transverse wavenumber as functions of the Rayleigh number and aspect ratio. An analytical asymptotic solution for the base state in the bulk is obtained, valid over most of the domain and increasingly accurate for lower aspect ratios. A local stability analysis, based on the analytical base state, reveals oscillatory transverse instabilities consistent with the global instability characteristics. The source term for this most unstable mode appears to be interactions between vertical shear and horizontal temperature gradients.- Open Access
Stability of SQG Kolmogorov Flow
Stability analysis is performed on surface quasigeostrophic systems subjected to a Kolmogorov-type "shear force" on the boundaries using linear and nonlinear approaches. For a SQG system of semi-infinite depth forced on the upper boundary, the most linearly unstable mode is 2.74 the energy injection length scale. This is contrary to two-dimensional fluid systems, where the linear instability is greatest for long waves. In the presence of damping, the most linearly unstable mode shifts toward shorter length scales. The nonlinear critical Reynolds number across different damping strengths is found to be qualitatively similar to that of Euler 2D systems. For an SQG system of finite thickness being forced on both boundaries, its behaviour approaches that of a semi-infinite SQG system at the large thickness limit. In the small thickness limit, the behaviour of a symmetrically forced fluid layer approaches that of a 2D system, while an antisymmetrically forced fluid layer is not susceptible to both linear and nonlinear instabilities. With ageostrophic effects, an SQG+ system of semi-infinite depth is much more prone to instabilities than an otherwise identical SQG system in the absence of damping due to the instability of long-wave modes. However, damping significantly suppresses such instabilities. With increasing damping, the most linearly unstable mode moves toward a smaller length scale. Contrary to the zero damping case, when the damping is sufficiently large, ageostrophic effects have a small but measurable stabilising effect.
Open AccessStability of SQG Kolmogorov Flow
Stability analysis is performed on surface quasigeostrophic systems subjected to a Kolmogorov-type "shear force" on the boundaries using linear and nonlinear approaches. For a SQG system of semi-infinite depth forced on the upper boundary, the most linearly unstable mode is 2.74 the energy injection length scale. This is contrary to two-dimensional fluid systems, where the linear instability is greatest for long waves. In the presence of damping, the most linearly unstable mode shifts toward shorter length scales. The nonlinear critical Reynolds number across different damping strengths is found to be qualitatively similar to that of Euler 2D systems. For an SQG system of finite thickness being forced on both boundaries, its behaviour approaches that of a semi-infinite SQG system at the large thickness limit. In the small thickness limit, the behaviour of a symmetrically forced fluid layer approaches that of a 2D system, while an antisymmetrically forced fluid layer is not susceptible to both linear and nonlinear instabilities. With ageostrophic effects, an SQG+ system of semi-infinite depth is much more prone to instabilities than an otherwise identical SQG system in the absence of damping due to the instability of long-wave modes. However, damping significantly suppresses such instabilities. With increasing damping, the most linearly unstable mode moves toward a smaller length scale. Contrary to the zero damping case, when the damping is sufficiently large, ageostrophic effects have a small but measurable stabilising effect.
- Open Access
SQG Point Vortex Dynamics with Order Rossby Corrections
Quasi-geostrophic flow is an asymptotic theory for flows in rotating systems that are in geostrophic balance to leading order. It is characterized by the conservation of (quasi-geostrophic) potential vorticity and weak vertical flows. Surface quasigeostrophy (SQG) is the special case when the flow is driven by temperature anomalies at a horizontal boundary. The next-order correction to QG, QG+, takes into account ageostrophic effects. We investigate point vortx dynamics in SQG+, building on the work of Weiss. The conservation laws for SQG point vortices that parallel the 2D Euler case no longer exist when ageostrophic effects are included. The trajectories of point vortices are obtained explicitly for the general two-vortex case in SQG and SQG+. For the three-vortex case, exact solutions are found for rigidly rotating and stationary equilibria consisting of regular polygons and collinear configurations. As in the 2D case, only certain collinear vortex configurations are rigid equilibria. Trajectories of passive tracers advected by point vortex systems are studied numerically, in particular their vertical excursions, which are non-zero because of ageostrophic effects. Surface trajectories can manifest local divergence even though the underlying fluid equations are incompressible.
Open AccessSQG Point Vortex Dynamics with Order Rossby Corrections
Quasi-geostrophic flow is an asymptotic theory for flows in rotating systems that are in geostrophic balance to leading order. It is characterized by the conservation of (quasi-geostrophic) potential vorticity and weak vertical flows. Surface quasigeostrophy (SQG) is the special case when the flow is driven by temperature anomalies at a horizontal boundary. The next-order correction to QG, QG+, takes into account ageostrophic effects. We investigate point vortx dynamics in SQG+, building on the work of Weiss. The conservation laws for SQG point vortices that parallel the 2D Euler case no longer exist when ageostrophic effects are included. The trajectories of point vortices are obtained explicitly for the general two-vortex case in SQG and SQG+. For the three-vortex case, exact solutions are found for rigidly rotating and stationary equilibria consisting of regular polygons and collinear configurations. As in the 2D case, only certain collinear vortex configurations are rigid equilibria. Trajectories of passive tracers advected by point vortex systems are studied numerically, in particular their vertical excursions, which are non-zero because of ageostrophic effects. Surface trajectories can manifest local divergence even though the underlying fluid equations are incompressible.
- 13 August 2025
- Published by The Royal Society in Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences
- Vol. 481 (2319)
The unified transform method: beyond circular or convex domains
A new transform-based approach is presented that can be used to solve mixed boundary value problems for Laplace’s equation in non-convex and other planar domains, specifically the so-called Lipschitz domains. This work complements Crowdy (2015, CMFT, 15, 655–687), where new transform-based techniques were developed for boundary value problems for Laplace’s equation in circular domains. The key ingredient of the present method is the exploitation of the properties of the Szegö kernel and its connection with the Cauchy kernel to obtain transform pairs for analytic functions in such domains. Several examples are solved in detail and are numerically implemented to illustrate the application of the new transform pairs.
- 13 August 2025
- Published by The Royal Society in Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences
- Vol. 481 (2319)
The unified transform method: beyond circular or convex domains
A new transform-based approach is presented that can be used to solve mixed boundary value problems for Laplace’s equation in non-convex and other planar domains, specifically the so-called Lipschitz domains. This work complements Crowdy (2015, CMFT, 15, 655–687), where new transform-based techniques were developed for boundary value problems for Laplace’s equation in circular domains. The key ingredient of the present method is the exploitation of the properties of the Szegö kernel and its connection with the Cauchy kernel to obtain transform pairs for analytic functions in such domains. Several examples are solved in detail and are numerically implemented to illustrate the application of the new transform pairs.
- 1 December 2024
- Published by IOP Publishing in Fluid Dynamics Research
- Vol. 56 (6) , 061402
The fall of an ellipse in a stratified fluid
The free fall of an ellipse in an infinite linearly-stratified fluid is investigated using a linear two-dimensional, Boussinesq, diffusionless, inviscid model. The oscillations of the ellipse decay because of radiation damping, but unlike the case of a circular cylinder, the ellipse can also rotate and move horizontally. The resulting equations are solved analytically for some simple cases for which there is little or no rotation. Motions with rotation are studied numerically using a spectral method to solve for the wave field in the fluid.
- 1 December 2024
- Published by IOP Publishing in Fluid Dynamics Research
- Vol. 56 (6) , 061402
The fall of an ellipse in a stratified fluid
The free fall of an ellipse in an infinite linearly-stratified fluid is investigated using a linear two-dimensional, Boussinesq, diffusionless, inviscid model. The oscillations of the ellipse decay because of radiation damping, but unlike the case of a circular cylinder, the ellipse can also rotate and move horizontally. The resulting equations are solved analytically for some simple cases for which there is little or no rotation. Motions with rotation are studied numerically using a spectral method to solve for the wave field in the fluid.
- 18 June 2024
- Published by Oxford University Press (OUP) in IMA Journal of Applied Mathematics
- Vol. 89 (3) , 574-597
A transform-based technique for solving boundary value problems on convex planar domains
A new technique is presented that can be used to solve mixed boundary value problems for Laplace’s equation and the complex Helmholtz equation in bounded convex planar domains. This work is an extension of Crowdy (2015, CMFT, 15, 655–687) where new transform-based techniques were developed for boundary value problems for Laplace’s equation in circular domains. The key ingredient of the method is the analysis of the so-called global relation, which provides a coupling of integral transforms of the given boundary data and of the unknown boundary values. Three problems which involve mixed boundary conditions are solved in detail, as well as numerically implemented, to illustrate how to apply the new approach.
- 18 June 2024
- Published by Oxford University Press (OUP) in IMA Journal of Applied Mathematics
- Vol. 89 (3) , 574-597
A transform-based technique for solving boundary value problems on convex planar domains
A new technique is presented that can be used to solve mixed boundary value problems for Laplace’s equation and the complex Helmholtz equation in bounded convex planar domains. This work is an extension of Crowdy (2015, CMFT, 15, 655–687) where new transform-based techniques were developed for boundary value problems for Laplace’s equation in circular domains. The key ingredient of the method is the analysis of the so-called global relation, which provides a coupling of integral transforms of the given boundary data and of the unknown boundary values. Three problems which involve mixed boundary conditions are solved in detail, as well as numerically implemented, to illustrate how to apply the new approach.
- Open Access
- 17 May 2024
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 987, A22
Peristaltic pumping down a porous conduit
A theoretical analysis is presented of peristaltic pumping down a narrow conduit with permeable walls, motivated by the flushing action of lugworms and other marine organisms in sandy burrows. Flow in the conduit is dealt with using lubrication theory; the leakage into the surrounding medium is taken into account by exploiting slender-body theory to solve the associated Darcy problem. By adopting a model for the local force balance on the pumping surface, we bridge between the limits in which the pump operates with either fixed load or displacement. In the latter limit we characterize peristaltic waves with either fixed form or ones that partially collapse the conduit. We construct pump characteristics (the relation between the mean flux and net pressure drop) when the burrow wall is impermeable and pressures are fixed at each end, and compare the results with existing laboratory experiments performed on lugworms. We then consider how the peristaltic dynamics is changed when the wall is made permeable. Last, we consider pumping along an impermeable burrow into a leaky head shaft. The results reveal that the permeability of the conduit wall or end can greatly impact the direction and strength of the recirculating flow.
- 17 May 2024
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 987, A22
Open AccessPeristaltic pumping down a porous conduit
A theoretical analysis is presented of peristaltic pumping down a narrow conduit with permeable walls, motivated by the flushing action of lugworms and other marine organisms in sandy burrows. Flow in the conduit is dealt with using lubrication theory; the leakage into the surrounding medium is taken into account by exploiting slender-body theory to solve the associated Darcy problem. By adopting a model for the local force balance on the pumping surface, we bridge between the limits in which the pump operates with either fixed load or displacement. In the latter limit we characterize peristaltic waves with either fixed form or ones that partially collapse the conduit. We construct pump characteristics (the relation between the mean flux and net pressure drop) when the burrow wall is impermeable and pressures are fixed at each end, and compare the results with existing laboratory experiments performed on lugworms. We then consider how the peristaltic dynamics is changed when the wall is made permeable. Last, we consider pumping along an impermeable burrow into a leaky head shaft. The results reveal that the permeability of the conduit wall or end can greatly impact the direction and strength of the recirculating flow.
- 2 November 2023
- Published by Elsevier in Physica D: Nonlinear Phenomena
- Vol. 457
Steady translating hollow vortex pair in weakly compressible flow
- 2 November 2023
- Published by Elsevier in Physica D: Nonlinear Phenomena
- Vol. 457
Steady translating hollow vortex pair in weakly compressible flow
- Open Access
- 11 April 2023
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 961, A1
Linear and nonlinear stability of Rayleigh–Bénard convection with zero-mean modulated heat flux
Linear and nonlinear stability analyses are performed to determine critical Rayleigh numbers ( ${Ra}_{cr}$ ) for a Rayleigh–Bénard convection configuration with an imposed bottom boundary heat flux that varies harmonically in time with zero mean. The ${Ra}_{cr}$ value depends on the non-dimensional frequency $\omega$ of the boundary heat-flux modulation. Floquet theory is used to find ${Ra}_{cr}$ for linear stability, and the energy method is used to find ${Ra}_{cr}$ for two different types of nonlinear stability: strong and asymptotic. The most unstable linear mode alternates between synchronous and subharmonic frequencies at low $\omega$ , with only the latter at large $\omega$ . For a given frequency, the linear stability ${Ra}_{cr}$ is generally higher than the nonlinear stability ${Ra}_{cr}$ , as expected. For large $\omega$ , ${Ra}_{cr} \omega ^{-2}$ approaches an $O(10)$ constant for linear stability but zero for nonlinear stability. Hence the domain for subcritical instability becomes increasingly large with increasing $\omega$ . The same conclusion is reached for decreasing Prandtl number. Changing temperature and/or velocity boundary conditions at the modulated or non-modulated plate leads to the same conclusions. These stability results are confirmed by selected direct numerical simulations of the initial value problem.- 11 April 2023
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 961, A1
Open AccessLinear and nonlinear stability of Rayleigh–Bénard convection with zero-mean modulated heat flux
Linear and nonlinear stability analyses are performed to determine critical Rayleigh numbers ( ${Ra}_{cr}$ ) for a Rayleigh–Bénard convection configuration with an imposed bottom boundary heat flux that varies harmonically in time with zero mean. The ${Ra}_{cr}$ value depends on the non-dimensional frequency $\omega$ of the boundary heat-flux modulation. Floquet theory is used to find ${Ra}_{cr}$ for linear stability, and the energy method is used to find ${Ra}_{cr}$ for two different types of nonlinear stability: strong and asymptotic. The most unstable linear mode alternates between synchronous and subharmonic frequencies at low $\omega$ , with only the latter at large $\omega$ . For a given frequency, the linear stability ${Ra}_{cr}$ is generally higher than the nonlinear stability ${Ra}_{cr}$ , as expected. For large $\omega$ , ${Ra}_{cr} \omega ^{-2}$ approaches an $O(10)$ constant for linear stability but zero for nonlinear stability. Hence the domain for subcritical instability becomes increasingly large with increasing $\omega$ . The same conclusion is reached for decreasing Prandtl number. Changing temperature and/or velocity boundary conditions at the modulated or non-modulated plate leads to the same conclusions. These stability results are confirmed by selected direct numerical simulations of the initial value problem.