Technology for Solutions - Differential Equations

When using technology to implement the shooting method for the boundary value problem $y'' + y = 0$ with $y(0) = 0$ and $y(\pi) = 0$ , a student finds that the method fails to converge to a unique solution regardless of the initial guess for $y'(0)$ . Analysis of the problem using computer algebra shows that the general solution is $y = A\cos(x) + B\sin(x)$ . What does this computational difficulty reveal about the boundary value problem?
Answer: The problem has infinitely many solutions because the boundary conditions are satisfied by $y = B\sin(x)$ for any constant $B$ Applying the boundary conditions: $y(0) = A\cos(0) + B\sin(0) = A = 0$ , so $y = B\sin(x)$ . Then $y(\pi) = B\sin(\pi) = 0$ , which is satisfied for any value of $B$ . This means there are infinitely many solutions, explaining why the shooting method cannot converge to a unique solution. Option A is wrong because the shooting method works for linear BVPs when solutions are unique. Option C is incorrect because this is a mathematical property, not a numerical precision issue. Option D is wrong because the conditions are consistent, just not sufficient to determine a unique solution.
Using symbolic computation software to find series solutions, a student inputs the differential equation $x^2y'' + xy' + (x^2 - \frac{1}{4})y = 0$ and requests a solution near $x = 0$ . The software returns two linearly independent solutions involving Bessel functions $J_{1/2}(x)$ and $J_{-1/2}(x)$ . When the student asks the software to plot these solutions and their numerical approximations using power series, what should they expect to observe about the behavior near $x = 0$ ?
Answer: Both solutions show polynomial-like behavior near the origin with different limiting values For half-integer order Bessel functions, $J_{1/2}(x) \sim \sqrt{\frac{2}{\pi x}}\sin(x)$ and $J_{-1/2}(x) \sim \sqrt{\frac{2}{\pi x}}\cos(x)$ . Near $x = 0$ , these behave like $\sqrt{x}$ and $x^{-1/2}$ respectively, but when expressed as power series solutions, they appear polynomial-like in their local behavior. Option B describes logarithmic singularities that don't occur here. Option C incorrectly describes the oscillatory behavior. Option D overstates the singular behavior visible in typical plotting ranges.
A computer algebra system is used to generate a numerical solution for the system of differential equations $x'(t) = -y$ , $y'(t) = x$ , with initial conditions $x(0) = 2, y(0) = 0$ . The true solution is a circle in the phase plane given by $x(t) = 2\cos(t)$ , $y(t) = 2\sin(t)$ . If the system is solved using the forward Euler method with a fixed step size $h > 0$ , what will be the qualitative shape of the resulting trajectory in the $xy$ -phase plane over a long time interval?
Answer: A spiral that moves outward, away from the origin. When analyzing numerical methods for differential equations, you need to understand how computational errors accumulate over time, especially for systems that should conserve energy. The given system $x' = -y, y' = x$ represents simple harmonic motion in two dimensions. The true solution traces a perfect circle of radius 2, meaning the system conserves energy: $x^2 + y^2 = 4$ for all time. However, the forward Euler method introduces systematic errors. At each step, Euler's method uses $x_{n+1} = x_n + h(-y_n)$ and $y_{n+1} = y_n + h(x_n)$ . This creates a slight increase in the distance from the origin at each time step. You can verify this by computing $x_{n+1}^2 + y_{n+1}^2 = x_n^2 + y_n^2 + h^2(x_n^2 + y_n^2)$ , showing the "energy" grows by factor $(1 + h^2)$ each step. Over many steps, this creates an outward spiral. Choice A is correct because the forward Euler method's truncation error causes the trajectory to spiral outward from the true circular path. Choice B is wrong because the method doesn't create damping—it actually adds energy to the system. Choice C is incorrect because while the trajectory might look circular initially, the systematic energy increase prevents it from remaining a closed loop. Choice D represents what would happen with backward Euler, which tends to be energy-dissipating rather than energy-increasing. Study tip: Remember that forward Euler tends to be unstable for oscillatory systems, typically adding energy and causing outward spirals, while implicit methods like backward Euler tend to dissipate energy.
Using a computer algebra system to analyze the direction field for $\frac{dy}{dx} = \frac{x + y}{x - y}$ , a student notices that there appear to be curves along which the direction field vectors have undefined slopes. Based on the direction field visualization, which statement best describes the behavior of solutions near these curves?
Answer: Solutions approach these curves asymptotically but cannot cross them, as the differential equation becomes singular along these lines The direction field has undefined slopes where the denominator $x - y = 0$ , or along the line $y = x$ . At these points, the differential equation becomes singular (undefined), and solutions typically approach such lines asymptotically but cannot cross them. Option A is incorrect because solutions cannot simply cross singular lines. Option B is wrong because these are not isoclines (lines of constant slope) but rather lines where the slope is undefined. Option D is incorrect because this is a mathematical property of the differential equation, not a computational error.
When using computer algebra to analyze the system $\frac{dx}{dt} = x(3 - x - 2y)$ , $\frac{dy}{dt} = y(2 - x - y)$ , a student generates a vector field plot that shows four equilibrium points. After using technology to linearize the system around each equilibrium point, the student finds that the Jacobian eigenvalues at one particular equilibrium are $\lambda_1 = 1$ and $\lambda_2 = -0.5$ . Based on this computational analysis, what should the student conclude about the local behavior near this equilibrium point?
Answer: The equilibrium is a saddle point since the eigenvalues have opposite signs, making it unstable with trajectories approaching along one direction and departing along another When analyzing equilibrium points in nonlinear systems, the eigenvalues of the Jacobian matrix at each equilibrium determine the local stability and behavior type. The key insight is that real eigenvalues with opposite signs create a very specific geometric pattern. With eigenvalues $\lambda_1 = 1$ and $\lambda_2 = -0.5$ , you have one positive and one negative real eigenvalue. This configuration creates a saddle point - an inherently unstable equilibrium where trajectories approach along the direction of the negative eigenvalue (the stable manifold) but diverge along the direction of the positive eigenvalue (the unstable manifold). Picture a mountain pass where paths converge from one direction but diverge in the perpendicular direction. Answer A incorrectly suggests spiral behavior, but spirals require complex eigenvalues with nonzero imaginary parts. Real eigenvalues like these produce straight-line motion along eigenvector directions, not rotational motion. Answer B misunderstands the significance of these eigenvalues. Neither value is close to zero, so the linearization gives definitive information about local behavior. Values near zero would indeed make linearization inconclusive. Answer C correctly identifies instability but wrongly classifies this as a node. Nodes occur when eigenvalues have the same sign. Here, opposite signs definitively indicate a saddle point, not a node. Remember: eigenvalue signs determine stability classification. Same signs = node (stable if both negative, unstable if both positive), opposite signs = saddle (always unstable), complex with nonzero imaginary parts = spiral or focus.
A student uses a computational tool to solve an ordinary differential equation. By varying the step size $h$ , the student records the global error at $x=1$ . When the step size is $h = 0.1$ , the error is $E_1 = 2.56 \times 10^{-2}$ . When the step size is reduced to $h = 0.05$ , the error becomes $E_2 = 1.60 \times 10^{-3}$ . Based on this data, what is the apparent order of the numerical method used by the tool?
Answer: Fourth-order ( $O(h^4)$ ) The global error $E$ for a method of order $p$ is proportional to $h^p$ , so $E \approx C h^p$ . We can write the ratio of the errors as $\frac{E_2}{E_1} \approx \frac{C (h_2)^p}{C (h_1)^p} = (\frac{h_2}{h_1})^p$ . Here, $h_1 = 0.1$ and $h_2 = 0.05$ , so $\frac{h_2}{h_1} = \frac{0.05}{0.1} = \frac{1}{2}$ . The ratio of the errors is $\frac{1.60 \times 10^{-3}}{2.56 \times 10^{-2}} = \frac{0.0016}{0.0256} = \frac{1}{16}$ . So, we have $\frac{1}{16} \approx (\frac{1}{2})^p$ . This implies $2^p \approx 16$ , which means $p=4$ . Therefore, the method is fourth-order, such as the classic Runge-Kutta method.
A student uses a numerical solver with a fixed-step Runge-Kutta method to approximate the solution to an initial value problem on the interval $[0, 5]$ . The student then runs the solver again with the step size cut in half. To estimate the global error of the more accurate solution (the one with the smaller step size), the student uses the two approximate values $y_h(5)$ and $y_{h/2}(5)$ . If the method is fourth-order, which of the following is the best estimate for the error in $y_{h/2}(5)$ ?
Answer: $\frac{1}{15}|y_h(5) - y_{h/2}(5)|$ Let $Y(x)$ be the true solution. For a fourth-order method, the global error is approximately $E(x) \approx C h^4$ . So, $Y(5) \approx y_h(5) + C h^4$ and $Y(5) \approx y_{h/2}(5) + C (h/2)^4 = y_{h/2}(5) + C h^4/16$ . Let $E_{h/2} = Y(5) - y_{h/2}(5) \approx C h^4/16$ . We want to find an expression for this error. Subtracting the two approximate equations gives $0 \approx (y_h(5) - y_{h/2}(5)) + C h^4 (1 - 1/16) = (y_h(5) - y_{h/2}(5)) + (15/16) C h^4$ . This means $C h^4 \approx -\frac{16}{15}(y_h(5) - y_{h/2}(5))$ . Substituting this back into the expression for the error $E_{h/2}$ : $E_{h/2} \approx \frac{1}{16} C h^4 \approx \frac{1}{16} [-\frac{16}{15}(y_h(5) - y_{h/2}(5))] = -\frac{1}{15}(y_h(5) - y_{h/2}(5))$ . The magnitude of the error is therefore $\frac{1}{15}|y_h(5) - y_{h/2}(5)|$ . This technique is known as Richardson extrapolation.
Using graphing technology to visualize the phase portrait of the system $\frac{dx}{dt} = y$ , $\frac{dy}{dt} = -x - 0.1y$ , a student observes spiral trajectories around the origin. When the student modifies the system to $\frac{dx}{dt} = y$ , $\frac{dy}{dt} = -x + 0.1y$ and regenerates the visualization, what change should they expect to see in the phase portrait?
Answer: The spirals will maintain the same direction of rotation but will spiral outward away from the origin instead of inward In the original system, the $-0.1y$ term provides damping, causing inward spiraling. When changed to $+0.1y$ , this becomes anti-damping, causing outward spiraling. The direction of rotation remains the same because the primary terms $\frac{dx}{dt} = y$ and $\frac{dy}{dt} = -x$ are unchanged. Option A reverses the inward/outward behavior. Option B incorrectly suggests the rotation direction changes. Option D is incorrect because the system still has oscillatory behavior, just with growing amplitude.
A numerical solver is used to generate the solution to $y' = f(x,y)$ with $y(x_0) = y_0$ . The global truncation error is the difference between the true solution and the numerical approximation at a given point. The local truncation error is the error incurred in a single step. Which statement best describes the relationship between these errors when using a technology solver over many steps?
Answer: The global truncation error is a complex accumulation of local truncation errors, where errors from earlier steps can be amplified or diminished by the differential equation. When analyzing numerical methods for differential equations, understanding how errors propagate is crucial for assessing solution quality. Both local and global truncation errors matter, but their relationship is more complex than simple addition. The correct answer is A because global truncation error involves a sophisticated accumulation process. Each local error from previous steps doesn't just add to the total—it gets carried forward and can be amplified or dampened by subsequent calculations. The differential equation itself influences how these errors evolve. For example, if $f(x,y)$ has a large derivative with respect to $y$ , small errors in earlier $y$ -values can grow exponentially. Conversely, some differential equations naturally damp errors over time. Choice B incorrectly suggests simple linear accumulation. This ignores how the iterative nature of numerical methods means that errors from step $n$ affect the starting point for step $n+1$ , creating a compounding effect rather than mere addition. Choice C is fundamentally wrong because it dismisses all previous computational history. Global error inherently reflects the cumulative effect of the entire solution process. Choice D treats these errors as completely separate entities, which misses the core relationship: global error emerges precisely from how local errors interact through the stepping process. Study tip: Remember that numerical methods are inherently iterative—each step builds on previous results. This means errors have a "memory" that propagates forward, making global error analysis much more nuanced than simply counting individual step errors.
A student uses a command ApproximateSolution = NDSolve[{y'[x] == 1 - 2*x*y[x], y[0] == 0}, y, {x, 0, 2}] in a computer algebra system to find a numerical solution to an initial value problem. The student then plots the result. Which of the following values is the best approximation for ApproximateSolution at x=1.0?
Answer: 0.538 This question requires recognizing the behavior of the solution or performing a quick approximation with a simple method like Euler's. Let's use Euler's method with a reasonably small step size, say $h=0.5$ , to get an estimate. The IVP is $y' = 1 - 2xy$ , $y(0)=0$ . Step 1: $(x_0, y_0) = (0,0)$ . The slope is $y'(0) = 1 - 2(0)(0) = 1$ . The next point is $y(0.5) \approx y_0 + h \cdot y'(0) = 0 + 0.5(1) = 0.5$ . Step 2: $(x_1, y_1) = (0.5, 0.5)$ . The slope is $y'(0.5) = 1 - 2(0.5)(0.5) = 1 - 0.5 = 0.5$ . The next point is $y(1.0) \approx y_1 + h \cdot y'(0.5) = 0.5 + 0.5(0.5) = 0.75$ . This crude approximation (0.75) is positive and closer to 0.538 than the other options. The true solution is related to the error function, and $y(1) \approx 0.538$ . Let's analyze the distractors. 0.000 would imply the function doesn't change from its initial value. 1.000 would happen if the slope remained 1 for the whole interval. -0.538 is a sign error. The initial slope is positive, and for small $x, y$ , the slope remains positive, so the function must increase initially. The value 0.538 is the only plausible positive result.
A student uses technology to solve the initial value problem $\frac{dy}{dt} = t^2 - y^2$ with $y(0) = 0.5$ using Euler's method with step size $h = 0.1$ . After computing the first three steps, the student observes that the numerical approximation appears to be diverging from what they expect the true solution to be. Which of the following is the most likely explanation for this divergence?
Answer: The step size is too large for this particular differential equation, causing numerical instability in regions where the derivative changes rapidly Euler's method can exhibit poor accuracy or apparent divergence when the step size is too large, especially for equations where the derivative changes rapidly. The equation $\frac{dy}{dt} = t^2 - y^2$ is nonlinear and can have regions where the derivative changes quickly, making a smaller step size necessary for accurate approximation. Option B is incorrect because Euler's method works for nonlinear equations. Option C is wrong because initial conditions are typically compatible with differential equations unless there's a specific constraint. Option D is incorrect because the existence and uniqueness theorem guarantees a solution exists for this smooth function.
A student uses adaptive step-size control in their numerical solver for $\frac{dy}{dx} = x^2 + y^2$ with $y(0) = 0$ over the interval $[0, 2]$ . The software's error monitoring shows that the step size automatically decreases dramatically near $x = 1.4$ . Given that this equation can develop finite-time blowup, what is the most likely explanation for this computational behavior?
Answer: The solution is approaching a vertical asymptote, and the algorithm reduces step size to maintain accuracy as $\frac{dy}{dx}$ becomes very large When you encounter questions about adaptive step-size control in numerical methods, think about why algorithms automatically adjust their behavior. These methods monitor local truncation error and modify step sizes to maintain accuracy while managing computational efficiency. The differential equation $\frac{dy}{dx} = x^2 + y^2$ with $y(0) = 0$ is notorious for developing finite-time blowup. As the solution grows, $y^2$ becomes increasingly large, causing $\frac{dy}{dx}$ to grow explosively. Near $x = 1.4$ , the solution is approaching a vertical asymptote where the derivative becomes infinite. The adaptive algorithm detects this rapid change by monitoring the local error estimates, which spike as the solution steepens dramatically. To maintain accuracy in this challenging region, the algorithm automatically reduces step size, trying to track the solution's sharp ascent toward infinity. Option A is backwards—step sizes should decrease, not increase, when solutions become large and change rapidly. Option B mischaracterizes the behavior; the solution is becoming highly nonlinear near blowup, not more linear. A more linear solution would actually allow larger step sizes. Option C incorrectly identifies oscillations as the cause. This equation doesn't produce oscillatory solutions; instead, it creates monotonically increasing solutions that blow up. For differential equations exams, remember that adaptive algorithms are your diagnostic tools. When step sizes decrease dramatically, ask yourself: "What's making the solution hard to approximate?" Usually it's either rapid change (like approaching asymptotes) or high curvature—both signal that the solution is becoming numerically challenging.
A researcher uses the fourth-order Runge-Kutta method to solve $\frac{dy}{dt} = -2y + \sin(t)$ with $y(0) = 1$ over the interval $[0, 5]$ . When comparing step sizes $h = 0.1$ and $h = 0.05$ , the maximum difference between the two approximations is found to be approximately $0.0002$ . Based on this information and the theoretical error behavior of the Runge-Kutta method, what would you expect the maximum difference to be when comparing $h = 0.05$ and $h = 0.025$ ?
Answer: Approximately $0.0000125$ , since the fourth-order method means errors scale as $h^4$ The fourth-order Runge-Kutta method has a global error that scales as $h^4$ . When the step size is halved, the error is reduced by a factor of $2^4 = 16$ . Since the difference between $h = 0.1$ and $h = 0.05$ is $0.0002$ , when going from $h = 0.05$ to $h = 0.025$ , we expect the difference to be approximately $0.0002/16 ≈ 0.0000125$ . Option A uses first-order scaling, Option C uses second-order scaling, and Option D uses an incorrect scaling factor.
Consider the initial value problem $y' = x - y^2$ with $y(2) = 1$ . If Euler's method is used with a small positive step size $h$ to approximate the solution, will the first few approximate values $y_1, y_2, \dots$ likely overestimate or underestimate the true solution values?
Answer: Overestimate, because the solution curve is concave down near the initial point. Euler's method approximates the solution by following the tangent line. Whether it overestimates or underestimates depends on the concavity of the solution curve. We can find the concavity by computing the second derivative. Given $y' = x - y^2$ , we differentiate with respect to $x$ : $y'' = \frac{d}{dx}(x - y^2) = 1 - 2y \cdot y'$ . At the initial point $(x_0, y_0) = (2, 1)$ , we have $y'(2) = 2 - 1^2 = 1$ . Now, we evaluate $y''$ at this point: $y''(2) = 1 - 2(1)(1) = -1$ . Since $y'' < 0$ , the solution curve is concave down near $(2, 1)$ . For a concave down curve, the tangent line lies above the curve, so Euler's method will produce an overestimation.

Start this quiz