2.1 Eulerian and Lagrangian descriptions

Eulerian vs Lagrangian Descriptions

Fluid motion can be described from two fundamentally different viewpoints. The Eulerian description watches fluid pass through fixed points in space, like a weather station measuring wind at a specific location. The Lagrangian description follows individual fluid parcels as they move, like tracking a tagged fish through a river. These two frameworks lead to different mathematical formulations, different numerical methods, and different strengths depending on the problem at hand.

Differences in Reference Frames

The core distinction comes down to what you're watching.

• In the Eulerian description, you pick a fixed point in space and record what happens there as fluid flows past. Your coordinate system doesn't move. You're asking: what is the velocity, pressure, or temperature at this location right now?
• In the Lagrangian description, you pick a specific fluid particle and follow it wherever it goes. Your coordinate system moves with the particle. You're asking: what is happening to this particular parcel of fluid over time?

Both descriptions capture the same physical reality. They just organize the information differently.

Advantages of Eulerian Description

• Simplifies treatment of complex geometries and boundary conditions, since the grid stays fixed
• Handles fluid-structure interactions and multiphase flows more naturally on a stationary mesh
• Pairs well with standard numerical methods like finite difference and finite volume schemes
• Provides a natural framework for steady-state flows and flows with large deformations, where tracking individual particles would become unwieldy

Advantages of Lagrangian Description

• Gives a more intuitive picture of fluid motion by following actual particle paths
• Eliminates numerical diffusion that arises from discretizing the advection term in Eulerian methods
• Tracks fluid interfaces and free surfaces precisely, since the particles are the interface
• Simplifies problems involving history-dependent material properties, where you need to know what a specific parcel has experienced over time

Applications of Eulerian Description

• Standard CFD simulations of internal flows in fixed geometries (pipes, channels, ducts, turbine passages)
• Flows with complex boundary physics such as conjugate heat transfer and chemical reactions at walls
• Weather forecasting and climate modeling, where the atmosphere is treated as a continuum on a global grid
• Turbulence research and turbulence model development, where statistical properties at fixed locations are the primary interest

Applications of Lagrangian Description

• Dispersed multiphase flows where you track individual sprays, bubbles, or droplets through a carrier fluid
• Free surface and fluid-structure interaction problems such as ocean waves, sloshing, and breaking waves
• Geophysical flows like lava flows and glacial ice movement, where material deformation histories matter
• Biological flows including blood flow through vessels and cell migration through tissue

Eulerian Description

In the Eulerian framework, you define a fixed region of space and describe how fluid properties evolve at every point within that region. This is the more common approach in most engineering CFD applications.

Fixed Spatial Grid

The fluid domain is divided into a fixed grid or mesh. The grid doesn't move; fluid passes through it. At each grid point or cell center, you store values of velocity, pressure, density, and other quantities. The grid can be structured (regular, like a Cartesian lattice) or unstructured (irregular, using triangles or tetrahedra to fit complex shapes).

Fluid Motion Through the Grid

As the flow evolves, fluid properties at each grid point change over time. The velocity field $\mathbf{u}(\mathbf{x}, t)$ tells you the velocity vector at every point in the domain. You obtain this velocity field by solving the governing equations (continuity, momentum, energy) on the grid.

Field Variables as Functions of Space and Time

All field variables are written as functions of spatial position and time. For instance, the velocity field is $\mathbf{u}(\mathbf{x}, t)$, where $\mathbf{x}$ is the position vector and $t$ is time. The spatial dependence captures how properties vary across the domain at any instant, while the time dependence captures how they evolve at each location.

Partial Time Derivatives

Time derivatives in the Eulerian frame are partial derivatives. The quantity $\frac{\partial \phi}{\partial t}$ gives the rate of change of some field variable $\phi$ at a fixed point in space. This is what a stationary sensor would measure. These partial time derivatives appear directly in the continuity equation, momentum equation, and energy equation.

Advection Term in Equations

Because the observer is stationary but the fluid is moving, you need an extra term to account for the transport of properties by the flow itself. This is the advection term, written as $\mathbf{u} \cdot \nabla \phi$ for a scalar field $\phi$. It captures how much of the observed change at a fixed point is simply due to different fluid being carried past that point. This term is responsible for much of the nonlinearity in the Navier-Stokes equations and is one of the main sources of numerical difficulty in Eulerian methods.

Lagrangian Description

In the Lagrangian framework, you label individual fluid particles and track each one as it moves through space. Instead of asking "what's happening at point $\mathbf{x}$?", you ask "what's happening to particle $\mathbf{X}$?"

Moving Material Points

The fluid is represented as a collection of material points. Each point corresponds to a small fluid element that moves with the local flow velocity. You track the position and properties of each material point over time, building up a picture of the flow from the collective behavior of many particles.

Trajectories of Fluid Particles

The trajectory of a fluid particle is given by $\mathbf{x}(\mathbf{X}, t)$, where $\mathbf{X}$ is the particle's initial position (its "label") and $t$ is time. This mapping tells you where particle $\mathbf{X}$ ends up at time $t$. The velocity of that particle is the time derivative of its position:

$\mathbf{u}(\mathbf{X}, t) = \frac{d\mathbf{x}}{dt}$

Field Variables as Functions of Initial Position and Time

All field variables are expressed as functions of the initial particle position $\mathbf{X}$ and time $t$. For example, the temperature of particle $\mathbf{X}$ at time $t$ is $T(\mathbf{X}, t)$. The initial position $\mathbf{X}$ serves as a permanent label, so you can always identify which parcel you're talking about.

Material Time Derivatives

The natural time derivative in the Lagrangian frame is the material derivative $\frac{D}{Dt}$. It gives the rate of change of a quantity as experienced by a fluid particle moving with the flow. When you follow a specific parcel, the material derivative is just an ordinary time derivative along that parcel's path.

Absence of Advection Term

Since the reference frame moves with the fluid, there's no need for a separate advection term. The transport of properties by the flow is built into the framework automatically. This is a significant advantage: it eliminates the numerical diffusion that plagues Eulerian discretizations of the advection term and simplifies the structure of the governing equations.

Connecting Eulerian and Lagrangian Descriptions

These two descriptions are not independent theories. They describe the same physics from different viewpoints, and several mathematical tools let you translate between them.

Material Derivative in the Eulerian Frame

The material derivative, which arises naturally in the Lagrangian description, can be expressed in Eulerian variables:

$\frac{D\phi}{Dt} = \frac{\partial \phi}{\partial t} + \mathbf{u} \cdot \nabla \phi$

This is the key bridge between the two frames. The left side is the rate of change following a particle. The right side breaks that into two Eulerian contributions: the local rate of change at a fixed point ($\frac{\partial \phi}{\partial t}$) plus the change due to the particle moving into a region with different values of $\phi$ (the advection term $\mathbf{u} \cdot \nabla \phi$).

Velocity Gradient Tensor

The velocity gradient tensor $\nabla \mathbf{u}$ describes how the velocity field varies in space. It captures the local deformation and rotation of fluid elements, connecting the Eulerian velocity field to the Lagrangian deformation of material parcels. It decomposes into:

• The strain rate tensor (symmetric part): describes stretching and compression
• The vorticity tensor (antisymmetric part): describes local rotation

Deformation Gradient Tensor

The deformation gradient tensor $\mathbf{F}$ is a purely Lagrangian quantity that maps the initial configuration of a fluid element to its current, deformed configuration:

$\mathbf{F} = \frac{\partial \mathbf{x}}{\partial \mathbf{X}}$

It encodes all information about how a fluid element has been stretched, rotated, and distorted relative to its original shape.

Jacobian Determinant

The Jacobian determinant $J = \det(\mathbf{F})$ gives the ratio of a fluid element's current volume to its initial volume. If $J = 1$, the element has preserved its volume (as in incompressible flow). The Jacobian is essential for transforming integrals between Eulerian and Lagrangian coordinates and for enforcing mass conservation in numerical schemes.

Numerical Methods for Each Description

The choice of description directly affects which numerical methods you can use. Each class of methods has trade-offs in accuracy, stability, and ease of implementation.

Eulerian Grid-Based Methods

Eulerian methods solve the governing equations on a fixed grid. The most common families are:

• Finite difference methods: approximate derivatives using values at neighboring grid points
• Finite volume methods: enforce conservation laws over discrete control volumes
• Finite element methods: approximate solutions using basis functions on an unstructured mesh

These methods handle complex boundary conditions well and are the backbone of most commercial CFD codes. Their main weakness is numerical diffusion from discretizing the advection term, and they can struggle to sharply resolve moving interfaces.

Lagrangian Particle-Based Methods

Lagrangian methods represent the fluid as discrete particles and integrate their equations of motion forward in time. Two prominent examples:

• Smoothed particle hydrodynamics (SPH): approximates field quantities by kernel-weighted sums over neighboring particles
• Vortex methods: track discrete vortex elements and compute their mutual interactions

These methods excel at free surface flows, large deformations, and fragmentation problems. Their challenges include maintaining uniform particle distributions, imposing solid-wall boundary conditions, and ensuring consistency of the approximation as particles cluster or spread apart.

Arbitrary Lagrangian-Eulerian (ALE) Methods

ALE methods let the computational mesh move, but not necessarily at the fluid velocity. You have freedom to choose how the mesh moves:

• Move it with the fluid (purely Lagrangian) near interfaces or moving boundaries
• Hold it fixed (purely Eulerian) in the interior
• Move it at some intermediate velocity to maintain mesh quality

This flexibility makes ALE methods particularly useful for fluid-structure interaction and problems with moving boundaries, where a purely Eulerian mesh would need constant remeshing and a purely Lagrangian mesh would become too distorted.

Comparison of Numerical Accuracy and Stability

The best choice depends on the specific flow problem, the required accuracy, and available computational resources. Many modern simulations use hybrid approaches that combine elements of all three strategies.