Comparison of Explicit and

Implicit Finite Element

Methods
Dr. Lonny L. Thompson
March 14, 2025

• Let’s examine the differences, advantages, and disad-

vantages, factors for deciding which method to use for
finite element analysis (FEA), typical applications with
examples, and how to use explicit methods effectively.
• Hybrid Implicit-Explicit methods and mass scaling ap-
proximations are also discussed with examples.
• Deciding between implicit and explicit methods can be
particularly challenging and nuanced in certain cases.
System of Differential Equations to Solve in
Time
For structural dynamics, after finite element discretization
in space and assembly of global matrices, the system of
2nd-order differential equations in time describing the
dynamic equations of motion are:

M ü(t) + C u̇(t) + Ku(t) = F (t)

M : Mass matrix
C : Damping matrix
K : Stiffness matrix
u(t) : Displacement vector as function of time
u̇(t): Velocity vector as time derivative of displacement.
ü(t): Acceleration vector as time derivative of velocity.
F (t) : External time-dependent load vector

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 1 of 30

Implicit and Explicit Time Stepping Methods
1. Implicit Method: Implicit methods solve equilibrium
equations implicitly at the current time step by solving
a system of equations:

M ün+1 + C u̇n+1 + Kun+1 = Fn+1

• Common examples include

the Newmark-β method and
the Backward Euler method.
• In implicit methods, displacements at the next time
step un+1 are solved from equations at the unknown
time step, tn+1.

Keff un+1 = Fn+1 + M ãn + C ṽn

where

Keff = a0M + a1C + K

is an ‘effective’ stiffness matrix, and a0, a1 are param-

eters that depend on sub-step evaluation points and
the scheme used.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 2 of 30

 • The right-hand side terms are known from the previ-
ous step. For Newmark-β, with parameters β, γ, the
update equations are:
M γC un u̇n ün(1 − 2β) 
   


2
+ + K  un+1 = Fn+1 + M 
 
2
+ + 
β∆t β∆t β∆t β∆t 2β
γu u̇ (γ − β) ü ∆t(γ − 2β)
 
n n n
+ C  + + 

β∆t β 2β

• Typically, the implicit solver factorizes the effective

stiffness matrix at each time step with iterations re-
quired for nonlinear problems to update the effec-
tive ‘tangent stiffness matrix,’ resulting in relatively
intensive computation per step but stable for larger
steps.
• Since the solution from the previous step tn is usually
a good approximation to the next step, the number
of iterations required is generally lower than a statics
iterative solution.
• Typically, to dampen unwanted high-frequency nu-
merical artifacts from insufficient spatial finite element
mesh resolution for structural dynamics, implicit meth-
ods such as the HHT-α method are used.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 3 of 30

2. Explicit Method: Explicit methods directly compute
the next step using previously known values without
solving simultaneous equations:

M ün + C u̇n + Kun = Fn

• Typical examples include the Central Difference method

and the Forward Euler method.
• For explicit methods, the equations are solved di-
rectly without iteration.
• Often, the damping is neglected (or assumed lumped),
and the central difference method explicitly calcu-
lates the next-step displacement directly from the
current tn and previous tn−1 steps.

un+1 = ∆t2M −1 (Fn − Kun) + 2un − un−1

No inversion is required for stiffness or damping ma-

trices at each step.
• The mass matrix M is typically diagonalized with, for
example, lumped mass so that the inversion is a triv-
ial scaling.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 4 of 30

Aspect Implicit Method Explicit Method

Matrix Equation (M +C +K)un+1 = Fn+1 un+1 = M −1[. . .]

Form (implicit solution) (explicit direct)
Matrix Inversion Required Not required
(iterative/direct solver) (diagonal mass matrix)
Stability Unconditionally stable Conditionally stable
(CFL condition)
Time Step Size Larger (stable) Very small (CFL
condition)
Computational Higher (requires matrix Low (direct)
Effort per step inversion/iterations)
Computational Moderate to High High for long-duration
cost (overall) (fewer large steps) analyses (small steps)
Solver Type Iterative (nonlinear) Direct calculation
Suitable Stiff or nonlinear, Short-duration
problems steady-state, dynamic, impact,
quasistatic transient
Accuracy Nonlinear, quasistatic, Transient, wave
damped problems propagation

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 5 of 30

Advantages and Disadvantages
Implicit Method:
Explicit Method:
• Advantages:
• Advantages:
Stable for larger time
Simple and direct
steps
computation
Accurate for long-term,
Easily parallelizable
nonlinear analyses
Optimal for dynamic
Handles stiff problems
and impact problems
effectively
Fast per step
• Disadvantages:
• Disadvantages:
Computationally
Stability conditions intensive per step
limit time-step size
Requires robust
Less suitable for iterative solvers
long-duration
More challenging
problems
parallelization

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 6 of 30

Typical Applications
Implicit Method:
Explicit Method:
• Nonlinear structural
analysis (large • Crash tests, automotive
deformation, plasticity) impact analysis

• Quasi-static structural • Wave propagation,

analysis explosions, shock
loading
• Large damping
• High-velocity impact and
• Thermal and coupled ballistic problems
field problems

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 7 of 30

 Decision Criteria

Criteria Explicit Preferred Implicit Preferred

Simulation Short-duration, highly Long-duration,

duration transient quasi-static
Nonlinear Moderate (impact High, static/quasi-static
complexity events) problems
Stability Small steps manageable Stability is priority
concerns
Computational Parallel resources Limited parallelization
resources available

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 8 of 30

• Explicit methods require significantly less memory than
implicit methods.
• The requirement for a small time step size is due to
the stability limit of the numerical integration, not the
physical behavior.
• Implicit FEA time integration requires more memory,
while explicit FEA needs more processor performance
and speed to update with many small time steps.
• Since explicit methods do not require matrix solutions,
they are very easily solved with parallel processing with
little data exchange, allowing these algorithms to scale
speed across many processors much more effectively
than implicit methods.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 9 of 30

Summary
• Due to their stability and permissible larger time steps,
implicit methods are ideal for large-scale, long-duration,
or stiff problems. Explicit methods are suitable for short-
duration, highly dynamic events.
• Implicit methods are preferable for complex, long-term
simulations due to stability and large allowable time
steps. Explicit methods are advantageous for short, dy-
namic simulations requiring rapid computations.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 10 of 30

Explicit Method Example:
Automobile Crash Simulation
• In part, the goal of a frontal crash test involving a car
impacting a rigid barrier at high speed is to predict
the deformation and energy dissipation of the vehicle
structure, accelerations experienced by occupants, and
structural failure locations.
• This type of simulation is well-suited for explicit meth-
ods since it is a short-duration event ( ∼ 100 millisec-
onds) with severe deformation and highly nonlinear be-
havior (material plasticity, large deflections, contact dy-
namics).
• Explicit methods effectively handle rapidly changing con-
tact conditions and crushing due to sudden impact.
• The dynamic explicit time integration must use very
small time steps ( microseconds) to satisfy stability crite-
ria (Courant-Friedrichs-Lewy condition), which is physi-
cally needed anyway to resolve the extremely short im-
pact event accurately.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 11 of 30

Implicit Method Example:
Seismic Response of High-rise Building
• Evaluating the dynamic response of a multi-story steel-
frame building subjected to earthquake ground mo-
tions lasting several seconds or minutes, the primary
goals include determining structural stability, evaluat-
ing drift, and assessing stresses throughout the earth-
quake.
• Implicit methods are best suited for this extended-duration
event, lasting several seconds to minutes, involving low-
frequency vibrations driven by seismic acceleration ap-
plied at the base as boundary conditions with Rayleigh
or modal damping in the dynamic analysis.
• The structural response is dominated by elastic defor-
mation, possibly moderate plastic deformation in local-
ized regions.
• Using implicit time integration, nonlinear equations are
solved at each step iteratively, allowing large, stable
steps ( ∼ milliseconds to seconds). The larger time
steps allowed by implicit methods significantly reduce
computation time for this analysis compared to explicit
methods.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 12 of 30

 Comparative Summary

Aspect Explicit Example Implicit Example

Problem Type Impact, crash Seismic, earthquake

dynamics response
Duration Very short Long (seconds to
(milliseconds) minutes)
Dominant High-frequency wave Low-frequency
Effects propagation vibrations
Time Step Microseconds (tiny Milliseconds to
steps) seconds (larger steps)
Computational Many fast, small steps Fewer larger,
Effort computationally heavy
steps

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 13 of 30

Examples where the choice between Implicit
and Explicit is not as clear

Nonlinear Problems with both Dynamic and

Quasi-Static Characteristics
• Problems like forming processes, incremental loading
of structures, or mechanical interactions with nonlinear-
ities have tradeoffs in the severity of nonlinearity and an
acceptable run-time.
• Implicit methods handle nonlinearity efficiently, with equi-
librium checks within iterations and large time steps.
However, the iterative solver might have complexities
that require many iterations to converge.
• Explicit methods handle severe nonlinearities effectively
without iterations, but small time steps can become
computationally expensive over prolonged load dura-
tions.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 14 of 30

Manufacturing Forming Processes
• An example of the tradeoffs when considering between
explicit and implicit methods is sheet metal stamping
of an automotive panel from a flat sheet; the forming
process involves moderate dynamics (the punch moves
at an intermediate velocity of about 1 m/s), significant
nonlinear deformation frictional contact, and moderate-
duration loading ( ∼ 0.5 seconds).
• The steel is modeled with plastic hardening (nonlinear)
behavior; explicit time-stepping requires an estimated
CFL time step:
v v
E 2.1 × 1011
u u
u u
u
≈ 5189 m/s
u
cmax = u
u = u
t
ρ 7850
t

With a mesh size of around 2 mm:

0.002 m
∆texplicit = ≈ 3.85 × 10−7 s ≈ 0.385µs
5189 m/s
The number of required explicit steps for 0.5-second
duration:
0.5s 6
N= ≈ 1.3 × 10 steps
3.85 × 10−7s
which is time-consuming.
LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 15 of 30
• Using an implicit method allows significantly larger time
steps and efficiently handles moderate dynamics with-
out excessive computational burden. Assuming a con-
servative time step:

∆timplicit ≈ 1 × 10−3 s = 1 ms

0.5 s
N= = 500 steps
1 × 10−3 s
Within each step, iterative nonlinear solutions are solved
using a good initial guess from the previous time step.
• Convergence difficulties may occur during severe and
rapid contact and nonlinear plasticity changes, poten-
tially significantly increasing the computational cost per
step.
• Choosing the best method may involve - preliminary
simulations comparing run times, evaluation of solver
robustness in implicit solutions, and checking explicit
mass scaling accuracy loss effects to speed up the anal-
ysis.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 16 of 30

Complex contact analysis
• For problems involving intermittent or frictional con-
tacts (e.g., bolted joint slippage, multi-body assembly),
implicit methods provide robust solutions by solving
equilibrium-based contact conditions but can have dif-
ficulties converging.
• Explicit methods handle contact easily due to direct
time integration and no iterative equilibrium solving,
yet the stability conditions limit the allowable time step.
• The choice depends on contact complexity and accept-
able simulation run-time versus solver robustness.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 17 of 30

Multi-frequency response analysis
• Problems with mixed-frequency content can be espe-
cially changing for explicit methods, which easily cap-
ture high-frequency behavior but may require prohibitively
small steps to maintain stability through low-frequency
vibrations over longer durations.
• Structures experiencing a mixture of high-frequency and
low-frequency dynamic behaviors, such as structures
with flexible appendages or vibrating equipment, may
be better solved using implicit or hybrid methods (com-
bining implicit and explicit features) to resolve low-frequency
responses efficiently but are also challenged by the solver
frequently having to resolve rapid high-frequency tran-
sient effects.
• Balancing the accurate resolution of both frequency ex-
tremes makes this choice complicated.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 18 of 30

Multi-physics and coupled problems
• Multi-physics or coupled-field problems involving thermal-
mechanical, fluid-structure, or electromagnetic-mechanical
interactions can be challenging.
• Implicit and hybrid methods provide stability and flex-
ibility for solving coupled equations but can become
computationally costly for iteratively solving coupled
nonlinear equations.
• Explicit methods alone are easy to implement coupling
but may limited by severe stability restrictions in one or
both the physical fields interacting.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 19 of 30

Hybrid Implicit-Explicit Methods
• Hybrid implicit-explicit methods, commonly known as
IMEX methods in the CFD community, combine the
strengths of both implicit and explicit integration ap-
proaches within a single simulation.
• These hybrid methods use explicit integration in re-
gions or during phases involving severe nonlinearity,
rapid dynamics, or wave propagation and implicit inte-
gration in regions or during phases with slow dynamics,
stability challenges, or stiff equations.
• IMEX schemes generally partition equations from the
structural dynamic matrix equations

M ü(t) + C u̇(t) + Ku(t) = F (t)

into two subsets:

F explicit : the explicit part is easily evaluated, less stiff,
nonlinear terms, or contact dynamics,
F implicit : the implicit part handles stiff parts, linear
structural response, damping, and slowly varying non-
linearities.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 20 of 30

• The general IMEX scheme is to solve and update iter-
atively.
implicit implicit
M ün+1 + C u̇n+1 + Kun+1 = Fn+1 , ( implicit update )
M ün + C u̇n + Kuexplicit
n = Fn
explicit
, ( explicit update )

• An example is fluid-structure interaction (FSI), where

an explicit fluid dynamics solver handles fast transient
flows, and an implicit solver is used for the structural
deformation for slower structural response.
• An example of a sequential explicit-implicit time inte-
gration is where FEA software can run explicit analysis
first for, say, initial impacts, then transition to implicit
in simulating slower structural relaxation, for example,
metal forming or spring-back simulations.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 21 of 30

Courant Stability Condition on Time-Steps for
Explicit Methods
• The Courant stability condition, also known as the Courant-
Friedrichs-Lewy (CFL) condition, determines the maxi-
mum allowable time step size (∆t) in explicit finite ele-
ment simulations to maintain numerical stability.
• For explicit integration schemes (such as the central dif-
ference method), the CFL condition for structural dy-
namics can be expressed as:

Lmin
∆t ≤
cmax
∆t : Time step size.
Lmin : Smallest characteristic length (element size) in
the mesh.
cmax : Maximum wave propagation speed within the
material.
• The physical meaning of this condition is that the nu-
merical wavefront must not pass through more than
one element in a single time step. Thus, smaller ele-
ment sizes require proportionally smaller time steps.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 22 of 30

• The CFL condition ensures that stress waves do not
propagate further than the smallest dimension of an el-
ement within each time step. If waves travel too far in
a single step, numerical errors grow and accumulate,
causing instability and erroneous oscillations.
• If the CFL condition is violated, numerical instability
occurs with non-physical and divergent solutions pro-
duced in the simulation.
• In structural dynamics, the maximum wave propagation
speed (cmax ) for solids is usually the longitudinal wave
speed.
• For simplified conditions (such as one-dimensional bars),
this reduces to:
v
E
u
u
u
cmax = u
u
ρ
t

E : Young’s modulus
ρ : Material density

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 23 of 30

Practical Example of Courant Stability Condition for
Explicit Analysis
• For a steel component analyzed explicitly: E = 210 GPa
and ρ = 7850 kg/m3, the approximate longitudinal wave
speed is
v
210 × 109 Pa
u
u

3 ≈ 5189m/s
u
cmax = u
u
7850 kg/m
t

• If the smallest mesh element dimension is Lmin = 1 cm,

the CFL condition gives:
0.01 m
∆t ≤ ≈ 1.93 × 10−6 s (≈ 1.93µs)
5189 m/s

• This indicates an extremely small allowable time step is

needed, requiring many steps for longer-duration sim-
ulations.
• Reducing mesh size by half reduces the allowable time
step by half, increasing computational cost further.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 24 of 30

Mass Scaling Approximation for Explicit Meth-
ods
• Strategies for practical implementation of explicit meth-
ods include mass scaling, which artificially increases mass
(carefully controlled) to lower wave speed and allow
larger time steps.
• While this improves efficiency for longer-time integra-
tion, the trade-off is loss of accuracy. When using ex-
plicit methods, the user must carefully avoid excessively
small elements in critical regions unless essential. Oth-
erwise, the simulation time can be excessive.
• Mass scaling is an explicit analysis technique in finite el-
ement simulations where the mass of certain elements
is artificially increased.
• The main purpose is to lower the wave propagation
speed ( cmax ), thereby allowing larger time steps ac-
cording to the Courant stability (CFL) condition.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 25 of 30

• Recall the CFL condition:
v
Lmin E
u
u
u
∆t ≤ , cmax = u
u
cmax ρ
t

• By artificially increasing density (ρ), wave speed de-

creases, increasing the allowable time step.
• However, mass scaling affects the accuracy of the dy-
namic responserof the structure by reducing natural fre-
quencies (ω = k/m), altering the frequency response
of the structure.
• Mass inertial properties are distorted, potentially damp-
ing rapid dynamic responses or inaccurately modeling
structural behavior under dynamic loads.
• Large mass scaling can significantly distort stress wave
propagation, impact results, and energy dissipation, re-
sulting in unrealistic or overly conservative results.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 26 of 30

• To minimize the negative impacts of mass scaling on
explicit solution accuracy:
Only apply mass scaling in local regions that are least
critical to dynamic response accuracy (e.g., distant
from critical impact zones).
Use the smallest mass increase necessary to achieve
a practical increase in time step size, typically limiting
the mass scaling factor to less than 10-20%.
Continuously monitor and check kinetic, internal, and
total energy balances. Excessive mass scaling will
manifest as unrealistic energy distributions or signifi-
cant artificial kinetic energy.
Perform preliminary analyses to quantify the sensitiv-
ity of results to mass scaling. Adjust scaling to ensure
acceptably small deviations from reference solutions.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 27 of 30

Practical Example of Mass Scaling for Explicit Analysis
• An example is a crash simulation where a portion of the
vehicle structure away from the impact zone is overly
refined with the initial time step without scaling: 0.1µs,
which is considered an impractical computational cost.
• By introducing a mass scaling factor of 1.5 ( 50% mass
increase) in these zones and increasing the allowable
time step to 0.15µs, the results can be simulated with
reasonable time and cost.
• The effects of this mass scaling should be checked by
comparing deformation patterns, peak accelerations,
and energy distributions with and without mass scaling.
• Ensuring variations in key response parameters remain
below acceptable engineering uncertainty tolerance (say
<5-10%).
• Best practices are to limit mass scaling to non-critical ar-
eas rather than global scaling and to apply just enough
scaling to achieve computational feasibility without sac-
rificing critical solution accuracy.
• Energy metrics and response parameters should be mon-
itored and tracked throughout the simulation to ensure
solution accuracy.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 28 of 30

Conclusions
• It is not always clear whether implicit or explicit time-
stepping is preferred for structural dynamics applica-
tions.
• Care must be used to correctly apply these methods,
especially explicit methods, to ensure the time-step is
kept below the stability limit.
• If mass scaling is used to increase this step size, this
must be examined carefully so that inaccuracies are within
acceptable limits.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 29 of 30

References:
1. Thomas J.R. Hughes, (1987,2000). The Finite Element
Method: Linear Static and Dynamic Finite Element Anal-
ysis. Dover Publications.
2. Hairer, E., & Wanner, G. (1996). Solving Ordinary Dif-
ferential Equations II: Stiff and Differential-Algebraic
Problems. Springer.

LinkedIn - Dr. Lonny Thompson, Clemson University, Mar 14, 2025. 30 of 30