Many engineering simulations aim to find a single, final answer. Think of the stress in a bridge beam under a constant, unchanging load, or the steady airflow over an airplane wing in cruise flight. These are steady-state problems, where the system has reached equilibrium and its properties no longer change with time. The solution is a single snapshot.

But what about a car crash, the turbulent flow of water from a broken dam, or the heating of a circuit board when you power it on? These are transient problems. The system's state is in constant flux, evolving from one moment to the next. A single snapshot won't do; we need to capture the entire sequence of events. To do this, we need to solve our governing partial differential equations not just in space, but also in time.

Computers can't process the smooth, continuous flow of real-world time. Instead, we must discretize it, chopping the timeline into a series of small, distinct intervals. This interval is called the time step, commonly denoted by the symbol $\Delta t$. This process transforms a transient problem into a sequence of static problems, one for each tick of our discretized clock.

The core of time integration is to develop a strategy for advancing, or “marching,” the solution from one time step to the next. Knowing the complete state of our system at the current time, $t$, we use a numerical scheme to calculate the state at the next time step, $t + \Delta t$. This approach is often called a time-marching scheme because we methodically march the solution forward through time, step by step, from a known initial condition.

By repeatedly applying this process, we build a complete history of the system's behavior. We start with an initial state ($U^0$) and compute $U^1$. Then, using $U^1$, we compute $U^2$, and so on, until we reach our desired final time. The result is not one solution, but a series of solutions that, when viewed together, create a movie of the physical process.

The function $f$ in our simple equation hides a lot of complexity. How we define it determines the nature of the time integration scheme. This is where the crucial distinction between explicit and implicit methods comes into play.

An explicit method calculates the next state, $U^{n+1}$, using only known information from the current state, $U^n$. It's direct and often computationally cheap per time step. An implicit method, on the other hand, defines $U^{n+1}$ in terms of both $U^n$ and $U^{n+1}$ itself. This means we have to solve an equation or a system of equations at each step to find the solution. The choice between these two approaches involves a fundamental trade-off between computational cost and numerical stability, which we will explore next.

Now that we've set the stage, we can dive into the specifics of how these different methods work.