The Central Idea
Discrete mathematics works with separated steps:
\[ n=0,1,2,3,\ldots \]Continuous mathematics works with values that vary across intervals:
\[ x\in\mathbb R. \]Limits provide the bridge between them.
What Does Discrete Mean?
A discrete model changes in separated increments.
Examples include:
- daily population counts;
- annual interest updates;
- terms of a sequence;
- steps in an algorithm;
- states in a recurrence.
What Does Continuous Mean?
A continuous model allows change at every intermediate value of the independent variable.
Examples include:
- position as a function of time;
- temperature across space;
- continuous population models;
- fluid density;
- solutions of differential equations.
Sequences as Samples of Functions
A sequence
\[ a_n \]can often be viewed as samples of a function \(f\):
\[ a_n=f(n). \]This allows continuous tools to inform discrete behavior.
Example: A Sequence from a Function
If
\[ f(x)=\frac1x, \]then sampling at positive integers gives
\[ a_n=\frac1n. \]Continuous Extension
When a sequence formula \(a_n\) makes sense for real inputs, we may define a continuous extension
\[ f(x) \]such that
\[ f(n)=a_n. \]Derivatives and integrals of \(f\) can then provide information about the sequence.
Monotonicity from a Derivative
If a continuous extension satisfies
\[ f'(x)<0 \]for sufficiently large \(x\), then the sampled sequence
\[ a_n=f(n) \]is eventually decreasing.
Finite Differences
For a sequence \(a_n\), the first finite difference is
\[ \boxed{ \Delta a_n=a_{n+1}-a_n }. \]This is the discrete analogue of a derivative.
Finite Differences Measure Stepwise Change
If
\[ \Delta a_n>0, \]then the sequence rises from step \(n\) to step \(n+1\).
If
\[ \Delta a_n<0, \]it falls.
Derivative as a Limit of Difference Quotients
For a function \(f\),
\[ f'(x) = \lim_{h\to0} \frac{f(x+h)-f(x)}{h}. \]A derivative is therefore a limiting form of discrete change over smaller and smaller increments.
Difference Quotient at Step Size \(h\)
Before taking the limit, define
\[ \frac{f(x+h)-f(x)}{h}. \]This is a finite-step rate of change.
Letting \(h\to0\) turns finite change into instantaneous change.
Unit-Step Differences
For samples at integer spacing,
\[ \Delta a_n=a_{n+1}-a_n \]corresponds to a difference quotient with step size \(1\).
General-Step Differences
On a grid with spacing \(h\),
\[ x_k=x_0+kh, \]define
\[ \Delta_h f(x_k)=f(x_k+h)-f(x_k). \]Then
\[ \frac{\Delta_h f(x_k)}{h} \]approximates \(f'(x_k)\).
Forward Difference Approximation
A basic numerical derivative approximation is
\[ \boxed{ f'(x) \approx \frac{f(x+h)-f(x)}{h} }. \]The approximation improves as \(h\) becomes small under suitable smoothness conditions.
Backward Difference Approximation
\[ \boxed{ f'(x) \approx \frac{f(x)-f(x-h)}{h} }. \]Centered Difference Approximation
Using information on both sides gives
\[ \boxed{ f'(x) \approx \frac{f(x+h)-f(x-h)}{2h} }. \]This is often more accurate than a one-sided difference at comparable step size.
Second Finite Differences
Apply the difference operator twice:
\[ \Delta^2a_n = a_{n+2}-2a_{n+1}+a_n. \]This is a discrete analogue of second-derivative information.
Constant First Differences
A sequence with constant first differences is linear in \(n\):
\[ a_n=mn+b. \]This mirrors the fact that a continuously differentiable function with constant derivative is affine.
Constant Second Differences
Polynomial sequences of degree \(2\) have constant second finite differences.
This parallels the fact that quadratic functions have constant second derivatives.
Higher Differences and Higher Derivatives
Repeated finite differences reduce polynomial degree just as repeated differentiation does.
For a degree-\(k\) polynomial sequence, the \(k\)-th finite difference is constant.
Sums as Discrete Accumulation
A finite sum
\[ \sum_{k=1}^{N}a_k \]accumulates separate contributions one term at a time.
Integration is the continuous analogue of this accumulation.
Riemann Sums
Partition an interval into small subintervals of width \(\Delta x\).
A Riemann sum has the form
\[ \sum_{k=1}^{N}f(x_k^*)\Delta x. \]Integral as a Limit of Sums
As the partition becomes finer,
\[ \Delta x\to0, \]and under suitable conditions,
\[ \boxed{ \int_a^b f(x)\,dx = \lim_{N\to\infty} \sum_{k=1}^{N}f(x_k^*)\Delta x }. \]Discrete Area Becomes Continuous Area
Each term
\[ f(x_k^*)\Delta x \]represents a rectangle area.
The integral arises as the limiting total of increasingly thin rectangles.
Series and Integrals Are Structurally Related
A series accumulates indexed values:
\[ \sum a_n. \]An integral accumulates continuously distributed values:
\[ \int f(x)\,dx. \]The integral test exploits this relationship.
The Integral Test as a Discrete-Continuous Bridge
For positive decreasing \(f\) with \(a_n=f(n)\), the convergence of
\[ \sum_{n=1}^{\infty}a_n \]is linked to the convergence of
\[ \int_1^\infty f(x)\,dx. \]Why the Integral Test Works Geometrically
For decreasing positive \(f\), unit-width rectangles sit above or below the curve in a controlled way.
The discrete sum and continuous area therefore bound one another.
Recurrences as Discrete Dynamics
A recurrence such as
\[ x_{n+1}=F(x_n) \]evolves a system one step at a time.
Differential equations describe analogous evolution continuously in time.
Difference Equations
A recurrence can be written as a difference equation:
\[ x_{n+1}-x_n=G(x_n). \]This describes the change per discrete step.
From Difference Equation to Differential Equation
If each step has duration \(\Delta t\), then
\[ \frac{x_{n+1}-x_n}{\Delta t} \]is a discrete rate of change.
As \(\Delta t\to0\), this suggests a derivative
\[ \frac{dx}{dt}. \]Euler's Method Reverses the Connection
Starting from a differential equation
\[ \frac{dx}{dt}=F(t,x), \]Euler's method creates the discrete update
\[ \boxed{ x_{n+1} = x_n+hF(t_n,x_n) }. \]Continuous Model to Discrete Algorithm
Euler's method shows the bridge in the opposite direction:
a continuous differential law can be approximated by repeated finite steps.
Discrete Exponential Growth
Suppose a quantity grows by a fixed fraction \(r\) each time step:
\[ P_{n+1}=(1+r)P_n. \]Then
\[ P_n=P_0(1+r)^n. \]Continuous Exponential Growth
A continuously growing quantity may satisfy
\[ \frac{dP}{dt}=kP. \]Its solution is
\[ \boxed{ P(t)=P_0e^{kt} }. \]From Repeated Compounding to \(e\)
Divide one unit of time into \(n\) equal periods and compound at rate \(1/n\) per period:
\[ \left(1+\frac1n\right)^n. \]As the compounding frequency increases,
\[ \boxed{ \lim_{n\to\infty} \left(1+\frac1n\right)^n=e }. \]More General Continuous Compounding
For rate \(r\),
\[ \left(1+\frac{r}{n}\right)^n \to e^r. \]Repeated discrete growth approaches continuous exponential growth.
Time-Scaled Compounding
Over time \(t\),
\[ \left(1+\frac{r}{n}\right)^{nt} \to e^{rt}. \]This gives the continuous-compounding formula
\[ P(t)=P_0e^{rt}. \]Discrete Decay to Continuous Decay
The same limit connects repeated percentage loss to
\[ e^{-kt}. \]Continuous decay is the infinitesimal-step limit of repeated discrete decay.
Difference Quotients and Differential Equations
A discrete growth rule
\[ P_{n+1}-P_n = rP_n\Delta t \]gives
\[ \frac{P_{n+1}-P_n}{\Delta t} = rP_n. \]As \(\Delta t\to0\), this becomes
\[ \frac{dP}{dt}=rP. \]Discrete Logistic Growth
A discrete population model may use
\[ P_{n+1}-P_n = rP_n \left( 1-\frac{P_n}{K} \right). \]This updates population one step at a time.
Continuous Logistic Growth
A continuous analogue is
\[ \boxed{ \frac{dP}{dt} = rP \left( 1-\frac{P}{K} \right) }. \]The discrete and continuous models share the same basic growth-limitation mechanism.
Discrete Models Are Not Always Approximations
Some systems are genuinely discrete.
For example, annual censuses, integer populations, and algorithmic steps may be naturally modeled with recurrences rather than differential equations.
Continuous Models Are Idealizations
A continuous model may smooth over small-scale discreteness to provide a simpler large-scale description.
Whether this is appropriate depends on scale and purpose.
Scaling Determines the Bridge
To pass meaningfully from discrete to continuous behavior, the size of each discrete step must shrink while the number of steps increases.
The scaling between those two changes matters.
Fixed Step versus Shrinking Step
Taking more steps of the same fixed size does not automatically produce a continuous limit.
The step size itself must typically tend to zero.
A Standard Scaling Pattern
On an interval of length \(T\), divide time into \(N\) equal pieces:
\[ \Delta t=\frac{T}{N}. \]Then
\[ N\to\infty \quad\Longrightarrow\quad \Delta t\to0. \]Finite Differences Approximate Derivatives Under Scaling
If
\[ t_n=n\Delta t, \]then
\[ \frac{x_{n+1}-x_n}{\Delta t} \]can approximate
\[ \frac{dx}{dt}. \]Finite Sums Approximate Integrals Under Scaling
If
\[ \Delta x=\frac{b-a}{N}, \]then
\[ \sum f(x_k)\Delta x \]approximates
\[ \int_a^bf(x)\,dx. \]Discrete Probability to Continuous Probability
Discrete probability assigns mass to individual outcomes.
Continuous probability uses density over intervals.
Fine discrete grids can approximate continuous distributions.
Probability Mass versus Density
In a discrete model,
\[ P(X=x_k) \]assigns probability directly to a point.
In a continuous model,
\[ P(X=x)=0 \]for individual points, and probability is accumulated over intervals.
Histogram to Density
A histogram uses finite bins.
As bin widths shrink while sample resolution increases, the visual model can approach a smooth probability density.
Binomial to Normal Behavior
Under suitable scaling, discrete binomial distributions can be approximated by a continuous normal distribution.
This is one important manifestation of the central limit theorem.
Random Walks to Diffusion
A random walk changes position by discrete steps.
Under appropriate time and space scaling, large collections of tiny random steps lead toward continuous diffusion and Brownian-motion models.
Discrete Space to Continuous Space
A spatial grid may use points
\[ x_k=kh. \]As
\[ h\to0, \]the grid becomes a finer approximation of a continuous spatial domain.
Finite-Difference Models of Differential Equations
Derivatives in a differential equation can be replaced by finite-difference approximations.
This turns a continuous equation into a large discrete algebraic system.
Example: Second Derivative on a Grid
A standard centered approximation is
\[ \boxed{ f''(x) \approx \frac{f(x+h)-2f(x)+f(x-h)}{h^2} }. \]Why the Second-Difference Formula Looks Familiar
The numerator
\[ f(x+h)-2f(x)+f(x-h) \]is a continuous-grid version of a second finite difference.
Discrete Curvature
A second difference measures how much the first difference itself changes.
This is why it corresponds naturally to curvature and second derivatives.
Summation by Parts
Discrete mathematics has an analogue of integration by parts.
One common form relates sums involving finite differences:
\[ \sum a_n\Delta b_n. \]This illustrates a broader pattern: many calculus identities have discrete counterparts.
Discrete Antiderivatives
If
\[ \Delta A_n=a_n, \]then \(A_n\) plays a role analogous to an antiderivative of the sequence \(a_n\).
Telescoping as a Discrete Fundamental Theorem
Since
\[ \Delta A_n=A_{n+1}-A_n, \]summing gives
\[ \sum_{n=m}^{N}\Delta A_n = A_{N+1}-A_m. \]This is the discrete analogue of recovering net change from a derivative.
Compare with the Fundamental Theorem of Calculus
Continuous accumulation gives
\[ \int_a^bF'(x)\,dx = F(b)-F(a). \]Discrete accumulation gives
\[ \sum \Delta A_n = \text{final value}-\text{initial value}. \]Finite Sums and Integrals Share Boundary Structure
Both summation of finite differences and integration of derivatives collapse internal changes and leave boundary information.
Sequence Limits and Function Limits
A function limit can often be probed through sequences approaching a point.
If every sequence
\[ x_n\to a \]with \(x_n\ne a\) satisfies
\[ f(x_n)\to L, \]then this supports the function-limit statement
\[ f(x)\to L \quad\text{as }x\to a. \]The Sequential Criterion for Limits
In real analysis, function limits can be characterized exactly using convergent sequences.
This creates a deep bridge from discrete indexed objects to continuous local behavior.
Sampling Does Not Capture Everything Automatically
Knowing a function only at integer points may miss behavior between those points.
A continuous extension requires additional assumptions or modeling choices.
Interpolation Bridges Missing Values
Given discrete samples, interpolation constructs a continuous or piecewise continuous function passing through or near those samples.
Linear Interpolation
Between two points
\[ (x_0,y_0) \quad\text{and}\quad (x_1,y_1), \]linear interpolation uses the straight line connecting them.
Polynomial Interpolation
A finite collection of distinct \(x\)-values determines a polynomial of bounded degree passing through the corresponding data points.
This turns discrete data into a continuous algebraic model.
Approximation versus Interpolation
Interpolation typically matches selected data points exactly.
Approximation may sacrifice exact pointwise matching in exchange for simpler or more stable global behavior.
Discrete Fourier Data to Continuous Waves
Finite collections of sampled values can be decomposed into discrete frequency components.
Continuous Fourier analysis extends this idea to functions and integrals over continuous domains.
Sums Become Integrals Under Fine Sampling
A recurring pattern is
\[ \sum_k f(x_k)\Delta x \longrightarrow \int f(x)\,dx \]as the grid spacing tends to zero.
Products Can Also Have Continuous Limits
Repeated multiplicative changes of the form
\[ 1+r\Delta t \]can approach exponential growth when \(\Delta t\to0\).
From Product to Exponential Integral
A variable growth-rate product can resemble
\[ \prod_k \left( 1+r(t_k)\Delta t \right). \]Taking logarithms suggests
\[ \sum_k r(t_k)\Delta t, \]which approaches
\[ \int r(t)\,dt. \]Continuous Accumulated Growth
This leads formally to expressions of the form
\[ \exp\left( \int r(t)\,dt \right). \]Discrete multiplicative accumulation becomes continuous exponential accumulation.
Discrete Conservation Laws
A recurrence may preserve a quantity exactly from one step to the next.
Continuous differential equations may preserve an analogous invariant over time.
Local Rules Produce Global Behavior
Both recurrence relations and differential equations specify local update rules.
Repeated application of those local rules generates global trajectories.
Discrete Stability
A recurrence can amplify or damp small deviations from equilibrium.
Stability can often be studied through the multiplier appearing in the linearized update.
Continuous Stability
Differential equations similarly describe whether nearby trajectories move toward or away from an equilibrium.
The mathematical criteria differ in form but express the same structural idea.
Example: Discrete Linear Stability
Suppose
\[ x_{n+1}=rx_n. \]Then
\[ x_n=r^nx_0. \]If
\[ |r|<1, \]the sequence decays to zero.
Example: Continuous Linear Stability
Suppose
\[ \frac{dx}{dt}=kx. \]Then
\[ x(t)=x_0e^{kt}. \]If \(k<0\), the solution decays to zero.
Matching Discrete and Continuous Rates
If one discrete step multiplies by \(r\), while one unit of continuous time multiplies by \(e^k\), then matching the two gives
\[ r=e^k. \]Therefore,
\[ k=\ln r. \]Discrete Rate and Continuous Rate Are Not Identical
A \(10\%\) discrete growth factor per period is
\[ 1.10. \]The equivalent continuously compounded rate is
\[ \ln(1.10), \]not exactly \(0.10\).
Small-Step Approximation
For small \(r\),
\[ \ln(1+r)\approx r. \]This explains why small discrete percentage changes resemble continuous rates.
Continuum Limits Require Consistency
A discrete model should approximate the intended continuous law as the grid or step size becomes small.
This property is often called consistency.
Convergence of Numerical Schemes
A numerical method is useful when its discrete approximation approaches the true continuous solution as the step size tends to zero.
Stability Matters in Numerical Approximation
Even a locally accurate discrete rule can behave badly if errors grow rapidly from step to step.
Numerical stability therefore matters alongside consistency.
Discrete-to-Continuous Does Not Mean “Exactly the Same”
A continuous model is often a limiting approximation to a discrete process, not an identical description at every finite scale.
Scale Determines Which Description Is Useful
At microscopic scale, individual steps may matter.
At macroscopic scale, a continuous approximation may capture the dominant behavior more efficiently.
Common Error: Treating a Difference as a Derivative
The finite difference
\[ f(x+h)-f(x) \]is not itself a derivative.
A derivative arises from the scaled limit
\[ \frac{f(x+h)-f(x)}{h} \quad\text{as }h\to0. \]Common Error: Treating a Sum as an Integral Without Scaling
A Riemann sum requires the width factor
\[ \Delta x. \]The expression
\[ \sum f(x_k) \]alone does not generally approximate an integral.
Common Error: Sending the Number of Steps to Infinity but Keeping Step Size Fixed
A continuous limit usually requires
\[ \Delta x\to0 \]or
\[ \Delta t\to0. \]More fixed-size steps merely extend the domain.
Common Error: Ignoring Units
Difference quotients divide change in the dependent variable by change in the independent variable.
The units must match the intended rate.
Common Error: Assuming Every Discrete Model Has a Useful Continuous Limit
Some discrete systems retain essential stepwise behavior no matter how large the scale becomes.
A continuous approximation must be justified rather than assumed.
Common Error: Assuming Every Continuous Model Has a Unique Discrete Approximation
Many different numerical schemes can approximate the same differential equation.
They may differ in accuracy, stability, and computational cost.
Common Error: Confusing Sampling with Interpolation
Sampling takes values from a continuous object at discrete points.
Interpolation constructs a continuous object from discrete data.
Common Error: Confusing Approximation with Equality
A discrete scheme may satisfy
\[ \text{discrete model} \approx \text{continuous model} \]for small step size without being exactly equal at finite resolution.
A Reliable Discrete-to-Continuous Procedure
- Identify the discrete quantity and its step size.
- Write the finite difference, sum, or recurrence explicitly.
- Introduce a scaling parameter such as \(h\), \(\Delta x\), or \(\Delta t\).
- Rewrite the expression in a form resembling a derivative or integral.
- Let the step size tend to zero while preserving the relevant total scale.
- Identify the limiting continuous object.
- Check whether the approximation assumptions are valid for the model.
A Reliable Continuous-to-Discrete Procedure
- Start from the continuous derivative or integral relation.
- Choose a grid or step size.
- Replace derivatives with finite differences or integrals with finite sums.
- Derive the resulting recurrence or algebraic system.
- Analyze accuracy and stability.
- Refine the grid and verify convergence.
Worked Example: Difference Quotient
Let
\[ f(x)=x^2. \]The forward difference quotient is
\[ \frac{(x+h)^2-x^2}{h}. \]Expand:
\[ \frac{2xh+h^2}{h} = 2x+h. \]Taking \(h\to0\),
\[ \boxed{f'(x)=2x}. \]Worked Example: Riemann-Sum Limit
Divide \([0,1]\) into \(N\) equal pieces:
\[ \Delta x=\frac1N. \]Using right endpoints \(x_k=k/N\),
\[ \sum_{k=1}^{N} \frac{k}{N}\frac1N = \frac1{N^2} \sum_{k=1}^{N}k. \]Since
\[ \sum_{k=1}^{N}k=\frac{N(N+1)}2, \]the sum becomes
\[ \frac{N+1}{2N}\to\frac12. \]Therefore,
\[ \boxed{ \int_0^1x\,dx=\frac12 }. \]Worked Example: Discrete Growth to Continuous Growth
Suppose one unit of time is divided into \(n\) intervals, and the quantity grows by factor
\[ 1+\frac{r}{n} \]in each interval.
After one unit of time,
\[ P_n=P_0\left(1+\frac{r}{n}\right)^n. \]Letting \(n\to\infty\),
\[ \boxed{ P=P_0e^r }. \]Worked Example: Euler Step
Suppose
\[ \frac{dy}{dt}=2y, \]with step size \(h\).
Euler's method gives
\[ y_{n+1} = y_n+h(2y_n) = \boxed{(1+2h)y_n}. \]Worked Example: Finite Difference for a Quadratic
Let
\[ a_n=n^2. \]Then
\[ \Delta a_n = (n+1)^2-n^2 = 2n+1. \]The second difference is
\[ \Delta^2a_n = 2. \]This mirrors
\[ \frac{d^2}{dx^2}x^2=2. \]Practice 1: First Difference
Find
\[ \Delta a_n \]for
\[ a_n=3n+2. \]Practice 2: Second Difference
Find
\[ \Delta^2a_n \]for
\[ a_n=n^2. \]Practice 3: Difference Quotient
Use
\[ \frac{f(x+h)-f(x)}{h} \]to recover the derivative of
\[ f(x)=x^3. \]Practice 4: Riemann-Sum Structure
On \([0,2]\), divide the interval into \(N\) equal pieces. What is
\[ \Delta x? \]Practice 5: Discrete Exponential Growth
Solve the recurrence
\[ P_{n+1}=1.05P_n, \qquad P_0=100. \]Practice 6: Continuous Growth Model
Solve
\[ \frac{dP}{dt}=0.05P, \qquad P(0)=100. \]Practice 7: Continuous-Compounding Limit
Evaluate
\[ \lim_{n\to\infty} \left(1+\frac{2}{n}\right)^n. \]Practice 8: Euler Update
For
\[ \frac{dy}{dt}=y+t, \]write the Euler update with step size \(h\).
Practice 9: Discrete Fundamental-Theorem Pattern
Simplify
\[ \sum_{n=1}^{N} (A_{n+1}-A_n). \]Practice 10: Explain the Scaling
Explain why taking \(N\to\infty\) in a Riemann-sum construction must also force
\[ \Delta x\to0. \]Practice Answers
Practice 1
\[ \Delta a_n = [3(n+1)+2]-(3n+2) = \boxed{3}. \]Practice 2
First,
\[ \Delta a_n = (n+1)^2-n^2 = 2n+1. \]Then
\[ \Delta^2a_n = [2(n+1)+1]-(2n+1) = \boxed{2}. \]Practice 3
\[ \frac{(x+h)^3-x^3}{h} = \frac{ 3x^2h+3xh^2+h^3 }{h}. \]Therefore,
\[ 3x^2+3xh+h^2. \]Letting \(h\to0\),
\[ \boxed{f'(x)=3x^2}. \]Practice 4
\[ \Delta x = \frac{2-0}{N} = \boxed{\frac2N}. \]Practice 5
Repeated multiplication gives
\[ \boxed{ P_n=100(1.05)^n }. \]Practice 6
\[ \boxed{ P(t)=100e^{0.05t} }. \]Practice 7
\[ \boxed{ \lim_{n\to\infty} \left(1+\frac2n\right)^n =e^2 }. \]Practice 8
Euler's method uses
\[ y_{n+1} = y_n + hF(t_n,y_n). \]Here \(F(t,y)=y+t\), so
\[ \boxed{ y_{n+1} = y_n+h(y_n+t_n) }. \]Practice 9
The sum telescopes:
\[ (A_2-A_1) +(A_3-A_2) +\cdots +(A_{N+1}-A_N). \]Therefore,
\[ \boxed{ A_{N+1}-A_1 }. \]Practice 10
A Riemann integral is obtained by making each subinterval arbitrarily thin. If the interval length remains fixed and it is divided into \(N\) equal parts, then
\[ \Delta x=\frac{b-a}{N}. \]Thus \(N\to\infty\) forces \(\Delta x\to0\), allowing the finite sum of rectangle contributions to approach continuous accumulation.
What to Remember
Discrete and continuous mathematics are connected by limiting processes.
- finite differences approximate derivatives;
- Riemann sums approach integrals;
- difference equations can approach differential equations;
- continuous equations can be discretized into numerical recurrences;
- repeated compounding approaches exponential growth;
- telescoping sums mirror the Fundamental Theorem of Calculus;
- grids become continuous domains as spacing tends to zero;
- discrete models and continuous models must be matched through appropriate scaling.
The deepest unifying idea is that continuous behavior can emerge as the limit of increasingly fine discrete approximations.
Where We Go Next
This lesson completes the core arc of Infinite Processes, Sequences, and Series. The subsection began with sequence convergence, developed infinite sums and convergence tests, moved through power and Taylor series, and then broadened to products, continued processes, approximation, and continuum limits.
These tools now support later work in differential equations, numerical methods, asymptotic analysis, probability, dynamical systems, and mathematical modeling.