📝 Taylor series nth Remainder estimation (36 MCQs)

📖 Explanation: This question targets error analysis and conceptual understanding. The Lagrange error bound requires finding an upper bound $M$ for $|f^{(n+1)}(z)|$ for all $z$ between the center $x_0$ and the evaluation point $x$. Evaluating only at the endpoint is insufficient because the derivative might achieve a larger magnitude within the interval. Furthermore, since $e^{-x^2}$ is even, its 5th derivative at 0 is zero, but the bound depends on the interval's maximum, making option C the precise correction to the student's specific procedural error.

📖 Explanation: This tests conceptual understanding of derivative behavior. For $\sin(x)$, every derivative is either $\pm \sin(x)$ or $\pm \cos(x)$, so $|f^{(n+1)}(z)| \leq 1$ for all real $z$. This global bound simplifies error estimation significantly. In contrast, $\frac{d^n}{dx^n}e^x = e^x$, so on an interval $[0, x]$, the maximum is $e^x$. Understanding this distinction is crucial for applying the remainder estimation theorem correctly across different function classes without overestimating error for bounded oscillatory functions.

📖 Explanation: This requires comparing two distinct error estimation methods. For $\ln(1+x)$ at $x=0.2$, the alternating series bound is simply $|a_4| = (0.2)^4/4$. The Lagrange bound involves $\max |f^{(4)}(z)|$ on $[0, 0.2]$. Since $f^{(4)}(z) = -6/(1+z)^4$, the maximum magnitude is at $z=0$ (value 6), giving bound $6(0.2)^4/24 = (0.2)^4/4$. While numerically similar here, conceptually the alternating bound exploits the specific sign-alternating structure and monotonic decrease, often providing a sharper, more direct estimate than the generic derivative-based Lagrange approach which must account for worst-case curvature.

📖 Explanation: This is a graph-based conceptual question. The widening gap visually represents $|R_n(x)|$ increasing. According to the Lagrange formula $|R_n(x)| \leq \frac{M}{(n+1)!}|x-x_0|^{n+1}$, as $x \to R$, both the factor $|x-x_0|^{n+1}$ increases and typically the required bound $M$ for the derivative on the expanding interval also increases. This illustrates why Taylor polynomials provide excellent local approximations near the center but require significantly higher degrees to maintain accuracy near the radius of convergence boundary.

📖 Explanation: This applies error estimation to physical modeling. The series expansion for the pendulum period involves powers of $k^2$. For small $\theta$, $k$ is small, and the series terms decrease rapidly. If the series satisfies alternating series conditions (which modified binomial expansions often do for positive arguments within radius), the Alternating Series Estimation Theorem provides the most straightforward justification: the error is bounded by the first neglected term. This avoids complex derivative calculations required by Lagrange and is more rigorous than informal geometric comparison, directly linking the physical parameter $\theta$ to the mathematical error tolerance.

📖 Explanation: This mixed-concept problem highlights strategic selection of error methods. Integrating $\sum (-1)^k x^{2k}/k!$ term-by-term yields $\sum (-1)^k (0.5)^{2k+1}/((2k+1)k!)$. This is a convergent alternating numerical series. Its truncation error is strictly bounded by the magnitude of the first omitted integrated term. Using Lagrange on $e^{-x^2}$ before integration requires bounding high-order derivatives of a composite function, which is algebraically messy and often yields looser bounds. The post-integration alternating series method leverages the specific structure of the definite integral approximation, demonstrating superior efficiency and precision.

📖 Explanation: This addresses a common point of confusion. The formal Lagrange remainder is $R_n(x) = \frac{f^{(n+1)}(z)}{(n+1)!}x^{n+1}$. Even if $f^{(n)}(0)=0$ making $p_n=p_{n-1}$, the theorem guarantees the bound using $n+1$. However, practically, since $p_3=p_2$ for cosine, one could treat it as $R_2$ using the 3rd derivative (which is $-\sin(z)$) or stick to $R_3$ using the 4th ($\cos(z)$). Option A states the strict definition. Understanding that the formula doesn't 'know' about vanishing terms prevents misapplication, though recognizing when $R_n = R_{n-1}$ allows using whichever bound is tighter is advanced insight.

📖 Explanation: This tests fundamental definition recall and interpretation. By definition, $R_n(x) = f(x) - p_n(x)$. Therefore, if $R_n(x) < 0$, it directly implies $f(x) < p_n(x)$, meaning the polynomial overestimates the function. Students sometimes confuse the *bound* (which is always positive) with the *actual remainder* (which has a sign). The question specifies they calculated the remainder expression (implying the signed quantity), making the sign diagnostic. This reinforces connecting the algebraic sign of the error term to geometric positioning of the curve relative to its tangent polynomial.

📖 Explanation: This compares generic vs. specialized error bounds. For $1/(1-x)$, derivatives grow factorially and blow up near x=1. On [0, 0.5], $f^{(6)}(z) = 6!/(1-z)^7$ is maximized at z=0.5, giving large M. Lagrange must assume worst-case curvature over the whole interval. The geometric series has a closed-form remainder $ar^{n+1}/(1-r)$ derived from exact algebraic manipulation, not derivative bounds. It captures the precise tail behavior. This illustrates that while Lagrange is universally applicable, specialized series forms often yield far sharper estimates by exploiting specific functional structures.

📖 Explanation: This addresses unit consistency in calculus applications. Trigonometric derivatives like $\frac{d}{dx}\sin x = \cos x$ hold ONLY when x is in radians. If x=3 degrees is treated as 3 radians, the derivative chain rule introduces factors of $\pi/180$ that are missing. Thus, $|f^{(n+1)}(z)| \leq 1$ is false for degree-measure inputs plugged directly into radian-based formulas. The error bound would be catastrophically wrong (vastly underestimating error since 3 rad >> 3 deg). This reinforces that mathematical models in calculus have implicit unit assumptions critical for error analysis validity.

📖 Explanation: This tests careful interval analysis for monotonic functions. For $e^x$ on $[-2, 0]$, the function is strictly increasing, so its maximum occurs at the right endpoint $z=0$, not at the evaluation point $x=-2$ or the left endpoint. Many students incorrectly use $e^{|x|} = e^2$ by blindly taking absolute value, vastly overestimating error. Recognizing that $\max_{z \in [-2,0]} e^z = e^0 = 1$ yields a much tighter, accurate bound. This emphasizes analyzing derivative behavior on the specific interval rather than applying rote formulas.

📖 Explanation: This explores scaling relationships in error analysis. The remainder bound scales with $|x|^{n+1}$. Doubling $x$ multiplies this factor by $2^{n+1}$. Additionally, $M$ (max derivative) often grows on larger intervals. To compensate for exponential growth in the error term, $n$ must increase significantly—more than linearly—to bring $\frac{(2a)^{n+1}}{(n+1)!}$ back below tolerance. Factorial denominator eventually wins, but the transient effect requires substantial degree increase. This intuition explains why Taylor series are inherently local and global approximation requires many more terms or different methods.

📖 Explanation: This targets error analysis of theorem prerequisites. The Alternating Series Estimation Theorem requires BOTH $a_k \to 0$ AND $a_{k+1} \leq a_k$ (eventually). For power series like $\sum (-1)^k \frac{k}{k+1} x^k$, the coefficients $\frac{k}{k+1}|x|^k$ may initially increase before decreasing depending on x. Applying the bound before the sequence becomes monotonic yields incorrect error estimates. Students often verify the limit but neglect checking the monotonicity condition at their specific evaluation point, especially when x is close to the radius of convergence where terms may rise before falling.

📖 Explanation: This Olympiad-style question probes deep theoretical limits. For this classic smooth non-analytic function, $f^{(n)}(0)=0$ for all n, so $p_n(x)=0$ and Lagrange gives $R_n(0)=0$. But for any $x \neq 0$, $f(x) > 0$, so the series doesn't converge to f(x). The Lagrange bound fails because while $f^{(n+1)}(z)$ exists for z≠0, as n→∞, the max M on any interval containing 0 grows faster than factorial decay, preventing $R_n \to 0$. This demonstrates that vanishing derivatives at a point don't guarantee series convergence to the function—a subtle distinction between smoothness and analyticity.

📖 Explanation: This mixed-concept application question combines singularity handling with error method selection. While $\frac{\sin x}{x}$ has removable discontinuity at 0, its higher derivatives involve increasingly messy quotient rules. Integrating $\sum (-1)^k x^{2k}/(2k+1)!$ gives $\sum (-1)^k /((2k+1)(2k+1)!)$, a clean alternating numerical series. The AST bound is simply the next term. This avoids singular-point derivative analysis entirely and leverages the integrated series' nice properties. Both practical (complexity) and theoretical (singularity) considerations favor the series-integration approach, illustrating strategic problem-solving in numerical analysis.

📖 Explanation: This challenges fundamental misconceptions about convergence proofs. Two issues exist: (1) M depends on n (it's max of $f^{(n+1)}$), and could grow fast enough to counteract factorial decay (as in non-analytic smooth functions). (2) Even if remainder → 0, this proves series converges TO f(x); but the student's premise assumes the bound → 0 implies convergence, ignoring that M=M(n). For analytic functions like sin/cos/exp, M is bounded or grows slowly, so bound → 0. But this isn't universal. Recognizing M's n-dependence is crucial for rigorous convergence proofs versus mere series convergence.

📖 Explanation: This application question demonstrates strategic series manipulation for error control. At x=0.8, Maclaurin series $\sum (-1)^k (0.8)^{2k+1}/(2k+1)$ converges slowly since 0.8 is near radius 1. The identity transforms to $\arctan(0.111...)$, where argument ≈ 0.11 makes terms decay as $(0.11)^{2k+1}$—dramatically faster. Error bound shrinks exponentially with smaller argument. This illustrates that error estimation isn't just about choosing n, but about reformulating the problem to improve convergence rate. Such transformations are essential in computational mathematics for achieving accuracy efficiently.

📖 Explanation: This analyzes practical limitations of error methods near convergence boundaries. For $\ln(1+x)$ at x=0.9, Lagrange gives $M = \max |f^{(n+1)}(z)| = n!$ at z=0, yielding bound $(0.9)^{n+1}/(n+1)$. This decays slowly since 0.9 is close to 1. The alternating series bound $(0.9)^{n+1}/(n+1)$ is actually identical here! But wait—the key insight is that for x near 1, BOTH bounds are poor, but AST is simpler to compute. However, if we consider x=0.99, Lagrange's M=n! still applies, while AST gives same form. The real issue is that near boundary, many terms are needed regardless. This question reveals that method choice matters less than recognizing inherent slow convergence near radius.

📖 Explanation: This addresses practical implementation concerns. While M=e^{0.1}≈1.105 is tighter, using M=3 is a safe rational upper bound avoiding circular dependency (needing e to approximate e). Solving inequalities numerically can introduce floating-point errors. Adding extra terms beyond theoretical minimum provides robustness against such issues. In scientific computing, slight over-engineering of error bounds is standard practice to ensure reliability. This bridges pure math (exact bounds) and applied numerics (safe, verifiable computation), emphasizing that theoretical minima aren't always practical targets.

📖 Explanation: This graph-based inference question connects visual symmetry to derivative properties. Symmetric error envelope about $x_0$ suggests $R_n(x_0+h) \approx R_n(x_0-h)$. Since $R_n(x) = f(x)-p_n(x)$ and $p_n$ matches f's derivatives, symmetric error implies f's deviation from polynomial is even-symmetric. This occurs when odd-degree terms in Taylor expansion vanish, meaning odd derivatives at $x_0$ are zero (e.g., cos x at 0). Recognizing such patterns helps diagnose function properties from approximation behavior, linking graphical analysis to analytical structure.

📖 Explanation: This tests nuanced understanding of derivative-specific bounds. While M=1 is always safe for sin/cos, it's not always tight. For specific n, we know which derivative appears: for n=0, R_0 uses f'(z)=-sin z, whose max on [π/4,π/3] is sin(π/3)≈0.866<1. For n=1, R_1 uses f''(z)=-cos z, max cos(π/4)≈0.707. Tailoring M to the specific derivative order yields sharper bounds. This advanced insight shows that generic bounds can be improved by exploiting known derivative identities and interval-specific extrema, optimizing error estimates beyond textbook defaults.

📖 Explanation: This error analysis question examines domain restrictions of estimation theorems. At x=0.9, series is $\sum (-1)^k (0.9)^k/k$, alternating with decreasing terms—AST applies. At x=-0.9, it becomes $\sum (0.9)^k/k$, all positive terms. AST requires alternating signs; it cannot bound error for positive series. Students often mechanically apply AST to any series with (-1)^k in formula without checking if the evaluated numeric series actually alternates. This highlights the importance of verifying hypotheses at the specific evaluation point, not just symbolically.

📖 Explanation: This application question showcases structural exploitation. Binomial series $(1+x)^m$ for m=1/2, x=0.1 gives $1 + (1/2)(0.1) - (1/8)(0.1)^2 + ...$. After first term, signs alternate and terms decrease. AST gives error ≤ magnitude of first omitted term, avoiding messy generalized binomial coefficient derivatives needed for Lagrange. This structural insight transforms a potentially complex error analysis into trivial arithmetic. Recognizing when specialized series have alternating tails is a valuable skill for efficient, accurate approximation without heavy calculus machinery.

📖 Explanation: This conceptual question leverages function symmetry. For even f, Taylor polynomials of even degree are even functions, so $R_n(-x) = f(-x)-p_n(-x) = f(x)-p_n(x) = R_n(x)$. Error is symmetric. Verifying bound on [0,a] automatically covers [-a,0]. For odd n, similar parity arguments apply. This halves the domain requiring analysis, reducing computational work. Recognizing and exploiting symmetry is fundamental in efficient mathematical analysis, turning two-sided problems into one-sided ones without loss of rigor.

📖 Explanation: This mixed-concept question connects error estimation to numerical ODEs. In function approximation, error is localized. In ODE solvers using Taylor methods, each step's truncation error feeds into next step's initial condition. Local errors propagate and can grow exponentially (instability). Thus, controlling $|R_n|$ per step is vital for global accuracy. This contextualizes remainder estimation beyond static approximation, showing its role in dynamic system simulation where error dynamics matter as much as instantaneous bounds.

📖 Explanation: This clarifies a pervasive misconception. Coefficients $a_k = f^{(k)}(0)/k!$ relate to derivatives at center, but Lagrange bound uses max of $f^{(n+1)}$ on interval. These differ unless f has special structure (e.g., alternating decreasing series where AST applies). For general series, derivative on interval can exceed center-value implications. Example: $e^x$ at x=1, $a_{n+1}=1/(n+1)!$, but Lagrange M=e^1>e^0. Confusing coefficient size with remainder bound is dangerous. This reinforces distinguishing series representation from error estimation mechanisms.

📖 Explanation: This analyzes domain-limited convergence behavior. Near x=0, $\ln x$ derivatives have poles, so M on [x,1] grows as 1/x^n. Lagrange bound explodes, reflecting series' divergence at x=0 (radius=1, centered at 1). This explains why $\ln x$ Taylor series is impractical near 0 despite formal convergence on (0,2]. Understanding derivative singularities' impact on error bounds guides appropriate series usage and motivates alternative expansions (e.g., about different centers) for different domains.

📖 Explanation: This tests understanding of even function properties in approximation. $e^{-t^2}$ is even, so its Maclaurin series has only even powers. Thus $p_3(t) = p_4(t)$ (the t^3 coefficient is zero). Their integrals are identical. Students might think higher degree always helps, but parity creates plateaus in improvement. Recognizing when consecutive polynomials coincide prevents wasted computation and clarifies that effective degree increases by 2 for even functions, not 1.

📖 Explanation: This addresses practical computational strategy. While Lagrange bound with M=1 is theoretically valid for all x, achieving tolerance for x=100 requires n where 100^{n+1}/(n+1)! < ε, needing enormous n. Using sin(x+2πk)=sin(x) to reduce argument to [-π,π] makes |x| small, drastically reducing required n. This combines mathematical insight (periodicity) with error analysis to make series practically usable. Blind application of bounds without argument reduction is computationally infeasible, highlighting the interplay between theory and algorithm design.

📖 Explanation: This leverages function parity for refined error estimation. For odd f, even-powered Taylor coefficients vanish. So p5=p6, and the first omitted nonzero term is degree 7. While Lagrange with n=5 uses 6th derivative (which is odd, possibly small), recognizing the actual series gap allows using AST-like reasoning: error ≈ |first missing nonzero term|. This gives sharper, more informative bound than generic Lagrange. Exploiting structural properties yields better estimates than brute-force derivative bounds.

📖 Explanation: This reinforces theorem prerequisites. Lagrange remainder formula requires f ∈ C^{n+1} on interval between x0 and x. If f lacks sufficient differentiability (e.g., |x| at 0), the formula doesn't apply. Students sometimes apply bounds mechanically without verifying smoothness. This foundational recall ensures proper application and prevents invalid error claims for non-smooth functions, emphasizing that error theorems have hypotheses as important as their conclusions.

📖 Explanation: This mixed-concept question addresses error analysis in composite functions. Deriving high-order derivatives of $e^{-t}\sin t$ via Leibniz rule is messy, making Lagrange M hard to bound tightly. Multiplying truncated series $(\sum a_k t^k)(\sum b_j t^j)$ lets us estimate error from each factor's truncation and their interaction. This modular approach provides clearer error attribution and often simpler bounds. It illustrates that representation choice affects error analyzability, not just computational ease.

📖 Explanation: This connects radius of convergence to error estimation validity. Though f is C∞ on ℝ, its Taylor series at 0 only converges for |x|<1 (due to complex poles at ±i). For |x|>1, series diverges, so remainder doesn't tend to zero and error bounds are meaningless. This highlights that smoothness ≠ analyticity, and error estimation presupposes convergence. Attempting Lagrange for |x|>1 yields bounds that don't reflect actual approximation quality since series doesn't represent f there.

📖 Explanation: This Olympiad-style question addresses special function challenges. Unlike sin/exp with bounded/self-replicating derivatives, Γ^(n)(1) involves ζ(k) values and polygammas without elementary bounds. Finding rigorous M requires advanced special function theory or numerical verification. This contrasts sharply with textbook examples, exposing real-world complexity in error analysis where derivative bounds aren't readily available, necessitating hybrid analytical-numerical approaches for reliable estimation.

📖 Explanation: This analyzes the gap between worst-case bound and typical error. Lagrange uses supremum M=e^1=e, but actual z in remainder formula is some unknown point in (0,1). On average, z≈0.5, so e^z≈√e≈1.65, roughly half of e. Thus bound is conservatively ~2× actual error. Understanding this systematic conservatism helps interpret bounds realistically—they guarantee safety but often overestimate typical performance. This statistical perspective complements deterministic worst-case analysis.

📖 Explanation: This application question links error estimation to algorithmic adaptation. Adaptive schemes evaluate local |R_n| at each step against tolerance. Exceeding tolerance triggers refinement (smaller h or higher n); significant margin allows coarsening. This feedback loop optimizes computational effort while maintaining accuracy. Understanding remainder as dynamic control signal, not just static bound, is key to modern numerical methods where error estimation drives algorithmic decisions in real-time.