Differential Equations Quiz: Analytic Vs Numeric Solutions

What this quiz covers

This quiz focuses on Analytic Vs Numeric Solutions, giving you a quick way to practice the rules, question types, and explanations that matter most for Differential Equations.

How to use this quiz

Try each quiz question before looking at the correct answer. Use the explanations to review missed ideas, then come back to similar questions until the pattern feels familiar.

Question 1

1. The Euler approximation will overestimate the analytic solution because the solution is decaying.
2. The Euler approximation will underestimate the analytic solution because the solution curve is convex. (correct answer)
3. The Euler approximation will be exact for any step size $h$ because the differential equation is linear.
4. The relationship cannot be determined without a specific value for the step size $h$.

Explanation: The analytic solution is $y_A(t) = e^{-2t}$. To determine the relationship, we check the concavity of the solution. The second derivative is $y'' = \frac{d}{dt}(-2y) = -2y' = -2(-2y) = 4y$. Since the initial condition is $y(0)=1$ and the solution is a decaying exponential, $y(t) > 0$ for all $t$. Therefore, $y''(t) = 4y(t) > 0$, which means the solution curve is convex (concave up). Euler's method approximates the solution using the tangent line at the beginning of each step. For a convex function, the tangent line always lies below the curve. Thus, each step of Euler's method will result in a value that is an underestimate of the true analytic solution.

Question 2

1. The analytic solution can be applied to a much broader class of differential equations than any single numerical method.
2. The analytic solution is generally less computationally expensive to obtain than a high-accuracy numerical solution.
3. The analytic solution provides an explicit formula that allows for symbolic analysis of how system parameters affect long-term behavior. (correct answer)
4. The analytic solution avoids the stability issues of numerical methods, which often fail for linear differential equations.

Explanation: The primary advantage of an analytic solution is that it is a closed-form function, $y_A(t)$. This formula provides the exact value of the solution at any time $t$. More importantly, if the original DE contains parameters, the analytic solution will explicitly show how the solution's behavior (like stability, oscillation frequency, or growth rate) depends on those parameters. This allows for a deep, symbolic understanding of the system. In contrast, a numerical solution only provides a set of discrete data points for a single set of parameter values. While distractors touch on related concepts, they are incorrect: A is false (numerical methods are more broadly applicable), B is false (finding analytic solutions can be infinitely expensive if impossible), and D is false (numerical methods are generally very effective for linear DEs, though stability can be an issue for stiff problems).

Question 3

1. The numerical solution will approximate the non-trivial analytic solution $y(t)=t^3$.
2. The numerical solution will approximate the trivial analytic solution $y(t)=0$. (correct answer)
3. The numerical method will fail to execute due to a division-by-zero error at the first step.
4. The numerical solution will oscillate between approximations of the two analytic solutions.

Explanation: The initial value problem has a non-unique solution because the function $f(t,y) = 3\sqrt[3]{y^2}$ does not satisfy the Lipschitz condition at $y=0$. Let's analyze the first step of Euler's method: $y_1 = y_0 + h \cdot f(t_0, y_0)$. Given $t_0=0$ and $y_0=0$, the slope is $f(0,0) = 3\sqrt[3]{0^2} = 0$. So, $y_1 = 0 + h \cdot 0 = 0$. Since the next point is also $(t_1, 0)$, the slope will again be 0, and $y_2=0$. The method will generate the sequence $y_0, y_1, y_2, \dots = 0, 0, 0, \dots$. Therefore, the numerical solver will follow and approximate the trivial solution $y(t)=0$.

Question 4

1. The numerical solution $y_n$ approaches the unstable equilibrium at $y=0$.
2. The numerical solution $y_n$ remains constant at $y=1$ for all $n>0$.
3. The numerical solution $y_n$ takes on values greater than 5 for some $n>0$. (correct answer)
4. The numerical solution $y_n$ becomes a periodic function oscillating around $y=2.5$.

Explanation: The analytic solution to this logistic equation must remain between the equilibria $y=0$ and $y=5$, and since it starts at $y=1$, it will approach $y=5$ from below. However, Euler's method, $y_{n+1} = y_n + h \cdot y_n(5-y_n)$, takes discrete steps based on the slope at the beginning of the interval. If the step size $h$ is large enough, the update can 'overshoot' the equilibrium. For example, if $y_n$ is close to 5, say $y_n=4$, the slope is $4(5-4)=4$. If $h=0.5$, the next step would be $y_{n+1} = 4 + 0.5(4) = 6$, which is greater than the carrying capacity of 5. This overshooting is a common artifact of numerical methods with large step sizes and is a behavior the true analytic solution cannot exhibit.

Question 5

1. Switching to an implicit method of the same order, as they have better stability properties.
2. Decreasing the step size $h$ by a large factor while using the same numerical method.
3. Using a variable step size algorithm that only refines the step size near critical points.
4. Switching from a lower-order method (e.g., Euler's) to a higher-order method (e.g., RK4). (correct answer)

Explanation: When you encounter questions about improving numerical solution accuracy for differential equations, think about the fundamental sources of error and how different strategies address them. The key is understanding how global error scales with different method parameters. The most effective strategy is switching from a lower-order method to a higher-order method (D). Global error for numerical methods scales as $O(h^p)$ where $h$ is step size and $p$ is the method order. A fourth-order method like RK4 has error scaling as $O(h^4)$, while Euler's method (first-order) scales as $O(h^1)$. This means higher-order methods achieve dramatically better accuracy for the same step size, or equivalently, require much larger step sizes to achieve the same accuracy. Option A is incorrect because switching to an implicit method of the same order doesn't improve accuracy—it only improves stability properties for stiff equations. The global error scaling remains identical. Option B, while it will improve accuracy, is inefficient. To reduce error by a factor of 100 using step size reduction alone with Euler's method requires reducing $h$ by a factor of 100, meaning 100 times more computational steps. Option C addresses efficiency but doesn't fundamentally improve the method's accuracy limitations. Variable step sizing optimizes where computational effort is spent but can't overcome the inherent order limitations of the underlying method. Remember: when accuracy is the primary concern and computational time isn't limiting, method order trumps step size optimization. Higher-order methods provide the most bang for your computational buck.

Question 6

1. The problem is 'stiff', which forces the use of an extremely small step size to maintain numerical stability. (correct answer)
2. The method will be unable to start because the initial condition lies on the solution.
3. The presence of trigonometric functions requires a symbolic solver, as numerical methods cannot handle them.
4. The analytic solution is non-monotonic, which causes compounding errors in all explicit numerical methods.

Explanation: When you encounter a differential equation with a large coefficient like -100, you should immediately think about stiffness – a phenomenon that creates serious numerical challenges for explicit methods. Let's examine what makes this problem stiff. The equation $y' = -100(y - \cos(t)) - \sin(t)$ can be rewritten as $y' = -100y + 100\cos(t) - \sin(t)$. The large negative coefficient -100 means that any deviation from the exact solution $y_A(t) = \cos(t)$ will be amplified rapidly. If your numerical approximation drifts even slightly from the true solution, the term $-100(y - \cos(t))$ becomes large and drives the solution away dramatically. For explicit methods like forward Euler to remain stable with such large coefficients, you need an extremely small step size – roughly $h < \frac{2}{100} = 0.02$ or smaller. This makes computation painfully slow. Answer A correctly identifies this stiffness issue. The other options miss the mark: B is wrong because having the initial condition on the solution curve is actually ideal – numerical methods start just fine this way. C is incorrect since numerical methods handle trigonometric functions without difficulty; computers evaluate $\cos(t)$ and $\sin(t)$ routinely. D misunderstands the real issue – while the cosine solution oscillates, this non-monotonic behavior isn't what causes the computational problems here. Study tip: When you see large coefficients (especially negative ones) in differential equations, immediately consider stiffness. This often signals that implicit methods like backward Euler are preferable to explicit ones for numerical stability.

Question 7

1. $y' + ty = t^3$, $y(0)=2$
2. $y' = t^2 / y^2$, $y(0)=1$
3. $y'' + 4y' + 4y = 0$, $y(0)=1, y'(0)=0$
4. $y' = e^{-t^2} + y^2$, $y(0)=0$ (correct answer)

Explanation: We analyze each option: A) This is a first-order linear differential equation, which can be solved using an integrating factor. B) This is a separable equation ($y^2 dy = t^2 dt$), which can be solved by direct integration. C) This is a second-order linear homogeneous equation with constant coefficients, which can be solved using the characteristic equation. D) This equation, $y' = e^{-t^2} + y^2$, is a type of Riccati equation. It is not linear, not separable, and not exact. Finding a closed-form analytic solution for this equation in terms of elementary functions is generally not possible. The term $e^{-t^2}$ does not have an elementary antiderivative, further complicating matters. Therefore, a numerical method is the most essential and practical approach.

Question 8

1. $|2e^{0.2} - 2.42|$ (correct answer)
2. $|2e^{0.2} - 2.41|$
3. $|2e^{0.1} - 2.20|$
4. $|e^{0.2} - 2.42|$

Explanation: First, we perform two steps of Euler's method to find $y_E(0.2)$. Given $t_0=0, y_0=1, h=0.1$. The DE is $y'=f(t,y)=t+y$. Step 1: $y_1 = y_0 + h \cdot f(t_0, y_0) = 1 + 0.1(0+1) = 1.1$. So $y_E(0.1) = 1.1$. Step 2: $t_1=0.1, y_1=1.1$. $y_2 = y_1 + h \cdot f(t_1, y_1) = 1.1 + 0.1(0.1+1.1) = 1.1 + 0.1(1.2) = 1.1 + 0.12 = 1.22$. So $y_E(0.2) = 1.22$. Next, we find the value of the analytic solution at $t=0.2$: $y_A(0.2) = 2e^{0.2} - 0.2 - 1 = 2e^{0.2} - 1.2$. The global error is the absolute difference: $|y_A(0.2) - y_E(0.2)| = |(2e^{0.2} - 1.2) - 1.22| = |2e^{0.2} - 2.42|$. Distractor B comes from forgetting to update $t$ in the second step. Distractor C is the error after only one step. Distractor D comes from an incorrect analytic solution where the integration constant was found to be 1 instead of 2.

Question 9

1. The Euler approximation underestimates the true value because the nonlinear term $y^2$ causes the solution curve to be concave up, making linear approximations fall below the actual curve (correct answer)
2. The Euler approximation is more reliable than the analytic solution because numerical methods avoid the accumulation of algebraic errors that occur in symbolic integration techniques
3. The difference between methods is negligible since both approaches converge to the same limit as the step size approaches zero in the numerical case
4. The analytic solution overestimates because it assumes perfect continuity while the numerical solution accounts for discrete jumps in the derivative function

Explanation: The correct answer is A. For this IVP, $f(t,y) = t^2 + y^2$ has $\frac{\partial f}{\partial y} = 2y > 0$ for $y > 0$, making solutions grow faster than linear approximations suggest. Euler's method uses linear segments that fall below the actual curve when the solution is concave up. B is wrong because numerical methods introduce discretization errors. C is wrong because we're comparing specific finite values, not limiting behavior. D incorrectly suggests analytic solutions have discretization issues.

Question 10

1. The error should decrease by a factor of approximately 4 because improved Euler has second-order accuracy and the analytic solution provides an exact reference for error measurement (correct answer)
2. The error should decrease by a factor of approximately 2 because the differential equation has a singularity at $y = 0$ that limits the convergence rate
3. The error should remain approximately the same because round-off errors begin to dominate when the step size becomes too small relative to machine precision
4. The error cannot be reliably predicted because the comparison assumes the analytic solution is exact, but separation of variables introduces its own approximation errors

Explanation: The correct answer is A. The improved Euler method has global error $O(h^2)$, so halving the step size should reduce the error by approximately $2^2 = 4$. The analytic solution $y = \sqrt{x^2 + 4}$ is exact for this separable equation, providing a true reference. B is wrong because there's no singularity affecting this solution path. C is wrong because $h = 0.05$ is not small enough for round-off to dominate. D is wrong because separation of variables gives exact solutions when properly applied.

Question 11

1. The analytic approach becomes more complex but still feasible using trigonometric substitutions and integration by parts techniques
2. Both analytic and numeric methods require the same computational effort since the periodic harvesting can be handled by Fourier series expansion
3. The analytic solution becomes intractable requiring numerical integration even to express the 'analytic' solution, making purely numerical methods more practical (correct answer)
4. The numerical method becomes less accurate due to the rapid oscillations while analytic methods can capture the periodic behavior exactly

Explanation: The correct answer is C. With time-dependent harvesting $H(t)$, the differential equation becomes $\frac{dP}{dt} = P(1-P/K) - 200(1+0.1\cos(2\pi t))$. This cannot be solved in closed form - even if we could write an 'analytic' expression, it would involve integrals that must be evaluated numerically. This makes numerical ODE solvers more practical. A is wrong because no elementary solution exists. B is wrong about computational equivalence. D is wrong because numerical methods handle smooth periodic functions well.

Question 12

1. The numerical approach will fail near $x = 0$ due to the singularity while the analytic solution using Bessel functions handles this point correctly
2. The analytic solution provides immediate access to asymptotic behavior and zeros of $J_0(x)$, while numerical solutions require extensive computation to reveal these properties (correct answer)
3. The numerical method gives more accurate values because it avoids the infinite series approximations inherent in computing Bessel functions
4. Both approaches have equivalent computational cost since Bessel functions must be computed numerically anyway for specific values

Explanation: The correct answer is B. Expressing the solution in terms of $J_0(x)$ immediately gives access to known properties: asymptotic behavior ($J_0(x) \sim \sqrt{\frac{2}{\pi x}}\cos(x - \frac{\pi}{4})$ for large $x$), zeros, orthogonality relations, etc. Numerical solutions would require extensive computation to discover these properties. A is wrong because numerical methods can handle the singularity with proper initial conditions. C is wrong because both ultimately use numerical computation. D underestimates the value of the analytic structure.

Question 13

1. The system is not actually stiff because stiff systems cannot be solved analytically with such high precision using eigenvalue methods
2. The eigenvalue calculation was performed correctly since numerical methods are always more reliable than analytic solutions for stiff systems
3. Both methods are equivalent in computational complexity for stiff systems since the analytic approach also requires numerical eigenvalue computation
4. The numerical stiff solver is correctly handling the fast transient decay while maintaining accuracy in the slow dynamics over the integration interval (correct answer)

Explanation: When you encounter a stiff differential equation system, you're dealing with dynamics that operate on vastly different time scales. This system has a coefficient of -1000, indicating extremely fast transient behavior alongside slower dynamics. The key insight is understanding what numerical agreement with analytical solutions actually validates. The correct answer is D because when a numerical stiff solver agrees with the analytical solution to high precision, it demonstrates the solver is successfully managing the challenge that defines stiff systems: accurately capturing both the rapid initial transients and the long-term behavior. Stiff solvers use implicit methods with adaptive step sizing to handle the fast dynamics without numerical instability, while maintaining accuracy in the slower evolution over the full integration interval. Answer A is incorrect because stiff systems can absolutely be solved analytically - stiffness refers to numerical challenges, not analytical impossibility. The eigenvalue method works perfectly for linear systems regardless of stiffness. Answer B reverses the logic - we're not validating the eigenvalue calculation using numerics; rather, the analytical solution serves as the benchmark. Answer C misses the point about computational complexity. While eigenvalue computation has its own costs, the agreement validates accuracy, not efficiency equivalence. Remember that stiffness is fundamentally about numerical stability and step-size restrictions in explicit methods. When you see high-precision agreement between analytical and numerical solutions for stiff systems, focus on what this reveals about the numerical method's ability to handle multiple time scales simultaneously.

Question 14

1. Both solutions will be identical for any $h$ because the total dose $D$ is conserved regardless of how the impulse is distributed over time
2. The solutions will differ by a constant offset because numerical methods cannot represent discontinuous jumps in concentration accurately
3. The numerical method will become unstable as $h \to 0$ because the impulse magnitude $D/h$ approaches infinity
4. The numerical solution should converge to the analytic solution because the discrete impulse approximation approaches the true delta function in the limit (correct answer)

Explanation: When you encounter differential equations with delta functions (impulses), you're dealing with the relationship between continuous mathematical models and their discrete numerical approximations. The key insight is understanding how numerical methods handle singularities. The analytic solution to $\frac{dC}{dt} = -kC + D\delta(t-T)$ produces an instantaneous jump in concentration at time $T$, followed by exponential decay. Numerically, since computers can't handle true instantaneous events, the impulse $D\delta(t-T)$ is approximated by adding $D/h$ over one time step of duration $h$. This preserves the total dose $D$ while spreading it over a finite interval. As $h \to 0$, this discrete approximation becomes increasingly narrow and tall, approaching the mathematical definition of a delta function. Standard numerical methods for ODEs are designed to converge to the true solution under such limiting processes, making option D correct. Option A incorrectly assumes dose conservation alone guarantees identical solutions—it ignores that the timing and shape of dose delivery affects the concentration profile. Option B misunderstands numerical methods' capabilities; well-designed schemes can approximate discontinuous behavior arbitrarily well. Option C confuses mathematical singularities with numerical instability—while $D/h$ grows large, the integration $\int (D/h) \, dt = D$ over the time step remains bounded. Study tip: Remember that delta functions in differential equations represent idealized instantaneous events. Numerical methods approximate these through increasingly concentrated distributions that converge to the ideal as discretization becomes finer.

Question 15

1. Numerical methods can systematically approximate all eigenvalues while analytic methods can only provide exact values for the first few eigenvalues using perturbation theory
2. Analytic solutions become impossible to express in closed form while numerical methods can still provide accurate approximations to eigenvalues and eigenfunctions (correct answer)
3. The numerical approach becomes unstable due to the oscillatory coefficient while analytic methods remain robust through Fourier series techniques
4. Both methods require the same computational complexity since the perturbation term can be handled exactly using integration by parts in analytic approaches

Explanation: The correct answer is B. The addition of $0.1\sin(x)$ makes this a variable-coefficient differential equation that generally cannot be solved in terms of elementary functions. While numerical methods (shooting, finite differences, finite elements) can still approximate eigenvalues effectively, no closed-form analytic solution exists. A is wrong because perturbation theory gives approximations, not exact values. C is wrong because numerical methods handle variable coefficients well, and Fourier series don't directly solve this BVP. D is wrong because analytic methods cannot provide exact solutions here.

Question 16

1. The accumulation of round-off errors in floating-point arithmetic makes the numerical solution unreliable for comparison with exact symbolic expressions
2. The analytic solution requires infinite series representations that cannot be evaluated exactly, introducing comparable approximation errors to the numerical method (correct answer)
3. The fourth-order method introduces systematic bias that always overestimates solutions to linear first-order equations with forcing terms
4. The step size is too large for the Runge-Kutta method to achieve convergence to the true solution value at the specified point

Explanation: The correct answer is B. When analytic solutions involve non-elementary functions (like integrals of $e^{x^2/2}$ that appear in this problem), they must be expressed as infinite series or special functions that themselves require numerical evaluation for specific values. This creates a fundamental limitation in comparing 'exact' vs. 'approximate' methods. A is wrong because round-off error is typically much smaller than truncation error. C is wrong because RK4 doesn't systematically over/underestimate. D is wrong because $h = 0.01$ is quite small for RK4.

Question 17

1. The power series method provides exact local information and error bounds while the numerical method gives global approximate solutions with uncertain accuracy
2. The numerical method is superior because it provides solution values throughout the domain while power series have limited radius of convergence
3. Both methods are equally valid approximations, with power series offering analytic insight near the origin and numerics providing global coverage (correct answer)
4. The power series approach is more reliable because it avoids the accumulation of discretization errors that plague all numerical differential equation solvers

Explanation: The correct answer is C. This recognizes that both approaches have complementary strengths: power series provide analytic insight and exact local behavior near $x = 0$, while numerical methods can cover the entire domain $[0,5]$. Neither is inherently superior. A is wrong because adaptive numerical methods do provide error control. B overstates the case against series methods. D is wrong because power series must be truncated for computation, introducing their own approximation errors.

Question 18

1. Numerical methods can directly simulate the approach to the limit cycle while analytic methods can only determine equilibrium points through linearization techniques (correct answer)
2. The numerical approach provides exact trajectories while analytic methods introduce errors through approximation of the nonlinear terms in the system
3. Numerical integration automatically handles the multiple time scales present in the system while analytic methods require separate analysis for fast and slow dynamics
4. The numerical solution converges faster to the steady state because it avoids the infinite series expansions required in analytic approaches

Explanation: The correct answer is A. This nonlinear system has no closed-form analytic solution, but numerical methods can directly simulate trajectories and reveal the approach to the unit circle limit cycle. Analytic methods are limited to local analysis (linearization near equilibria, perturbation theory). B is wrong because numerical methods approximate, not exact. C is wrong because this system doesn't have obviously separated time scales. D is wrong because numerical methods don't 'converge faster' - they provide different information entirely.