1.1. Finite Differences

A finite difference is a mathematical expression of the form: Δ𝑓(𝑥)= 𝑓(𝑥+ ℎ)− 𝑓(𝑥) It represents the change in the value of a function 𝑓(𝑥) as 𝑥 is incremented by a small amount ℎ. Finite differences are used to approximate derivatives and are particularly useful in constructing polynomial interpolating functions when data is available at discrete points.

A finite difference is a way to measure how much the value of a function changes when we make a small change to its input (represented by ℎ). The notation Δ𝑓(𝑥) captures this change: it tells us how much the function's output at 𝑥 + ℎ differs from its output at 𝑥. This is very helpful because, in many cases, instead of having a smooth curve or function, we have a set of discrete points, like the results from experiments. Using finite differences allows us to approximate things like derivatives, which tell us how fast a function is changing at a point, even when we only have these discrete data points.

Imagine you are tracking your car's speed over time with a GPS. Every few seconds, it records your position. The GPS cannot tell you your exact speed at every instant but can use the difference between your two recorded positions over time to give an estimate of your speed. Similarly, finite differences use differences in function values to estimate how quickly the function is changing.

There are mainly four types of finite differences:

a. Forward Difference (Δ) Δ𝑓(𝑥)= 𝑓(𝑥+ ℎ)− 𝑓(𝑥)

Higher order forward differences: Δ2𝑓(𝑥) = Δ(Δ𝑓(𝑥))= 𝑓(𝑥 +2ℎ)− 2𝑓(𝑥+ ℎ)+ 𝑓(𝑥) Δ3𝑓(𝑥) = Δ(Δ2𝑓(𝑥)), and so on.

b. Backward Difference (∇) ∇𝑓(𝑥)= 𝑓(𝑥)−𝑓(𝑥−ℎ)

Higher order backward differences: ∇2𝑓(𝑥)= ∇(∇𝑓(𝑥))= 𝑓(𝑥)−2𝑓(𝑥− ℎ)+ 𝑓(𝑥− 2ℎ)

c. Central Difference (𝛿) ℎ ℎ 𝛿𝑓(𝑥) = 𝑓(𝑥+ ) −𝑓(𝑥− ) 2 2 It is a symmetric difference and gives better accuracy in approximating derivatives. d. Shift Operator (𝐸) 𝐸𝑓(𝑥) = 𝑓(𝑥+ℎ) Using 𝐸, we define: Δ = 𝐸− 1, ∇= 1 −𝐸−1 e. Average Operator (𝜇) ℎ ℎ 𝑓(𝑥 + ) +𝑓(𝑥− ) 2 2 𝜇𝑓(𝑥) = 2

There are different types of finite differences that you can use based on how you want to look at changes in the function values. The forward difference takes the difference between the value of the function at a point and the value a little further along, which is good for estimating the rate of change in the forward direction. Backward differences work the other way — they look at a point and the one just before it, helping estimate the rate of change in the backward direction. Central differences average the values around the point, offering a more balanced estimate of change, which can be more accurate. The shift and average operators are tools to simplify these processes and can help compute differences quickly and efficiently.

Think of finite differences like tracking a runner on a track. If you want to find out how fast they are going, the forward difference is like seeing how much further they run after a certain amount of time. The backward difference is like looking at how fast they slowed down at the end of the race. The central difference would be like asking how fast they were going, on average, during a segment that’s symmetrically around a specific point in time.

A difference table is constructed to systematically compute the finite differences. It is especially useful in applying Newton's interpolation formulas.

x f(x) Δf(x) Δ²f(x) Δ³f(x) x₀ f₀ Δf₀ Δ²f₀ Δ³f₀ x₁ f₁ Δf₁ Δ²f₁ x₂ f₂ Δf₂ x₃ f₃

A difference table is a structured way of organizing the function values and their corresponding finite differences, which makes it easier to compute and visualize the differences step-by-step. The first column lists the x values and the second column lists the corresponding function values. The subsequent columns compute the finite differences starting from the first difference, then moving on to the second difference, and so forth. This structured approach also greatly aids in the application of interpolation formulas, making it easier to derive polynomial approximations from the collected data.

Think of a difference table like a recipe that helps you bake a cake, where each step depends on the last. You start with your original ingredients (the x values), each ingredient’s effect on flavor (the function values), and then understand how the flavor changes when you add more or less (the differences). By outlining it all step-by-step in a table, you can see how those flavors build up, just like the different ingredient ratios in a cake can lead to a delicious end product.

1. Linearity: Δ(𝑎𝑓(𝑥)+ 𝑏𝑔(𝑥))= 𝑎Δ𝑓(𝑥)+𝑏Δ𝑔(𝑥)

2. Polynomial Property: If 𝑓(𝑥) is a polynomial of degree 𝑛, then: Δ𝑛+1𝑓(𝑥)= 0

3. Relation with Derivatives (Approximation): 𝑓(𝑥+ ℎ)−𝑓(𝑥) Δ𝑓(𝑥) 𝑓′(𝑥)≈ = ℎ ℎ

The properties of finite differences provide fundamental rules that help in performing calculations effectively. For example, the linearity property indicates that the difference of a combination of functions can be expressed as a combination of their differences. The polynomial property reveals that if you apply this difference operation to a polynomial of a certain degree, the higher differences will yield zero, giving insights into the behavior of polynomial functions. Finally, understanding the relation to derivatives shows how finite differences serve as an approximation tool for calculating rates of change in functions, which can be particularly useful in numerical methods.

Imagine you are trying to understand how a car’s speed changes when you press the gas pedal in different ways. The linearity property is like saying if you press harder or softer (like adding two different driving styles), you can figure out the total speed by understanding each one separately. The polynomial property tells you that if you have a consistent speed, pressing harder won't change your speed type but could make it zero in certain cases. Lastly, the approximating derivative is like your car’s speedometer giving a near-instantaneous speed based on how fast your foot moves.

• Construction of interpolation polynomials (e.g., Newton-Gregory formulas). • Numerical differentiation. • Numerical solutions to ordinary and partial differential equations. • Curve fitting and numerical integration.

Finite differences have several practical applications in fields like math, physics, and engineering. They are essential for building interpolation polynomials, which help estimate unknown values based on known data points. In numerical differentiation, they allow us to approximate derivatives when we cannot calculate them analytically. Finite differences also play a critical role in finding solutions to differential equations — both ordinary and partial — which model real-world systems. Additionally, they are used for curve fitting, enabling us to create models that approximate data trends, and in numerical integration for estimating areas under curves.

Think of finite differences like the tools used for planning a city. Just as city planners use models to predict how many parks, roads, and schools are needed based on existing population data, finite differences help us approximate and understand how different mathematical functions behave based on known points. This is vital for forecasting future population growth, traffic patterns, or resource needs.

--