📝 Alternating series approximation (40 MCQs)

📖 Explanation: The Alternating Series Test requires that terms eventually decrease monotonically to zero. Strict inequality is not necessary for the test or the error bound to hold; eventual non-increasing behavior suffices. The standard remainder estimate $|R_n| \leq a_{n+1}$ relies on the geometric property that partial sums oscillate around the true sum with diminishing amplitude. As long as $a_k \to 0$ and $a_{k+1} \leq a_k$ for all sufficiently large k, the bounding logic holds. Finite violations at the beginning do not disrupt the asymptotic error estimation.

📖 Explanation: Computational efficiency in alternating series approximation is directly tied to the rate of decay of the term sequence $a_k$. Since the error bound is governed by the magnitude of the first omitted term $a_{n+1}$, a series with terms decaying as $1/k^2$ will reach the threshold $10^{-4}$ much faster than one decaying as $1/\sqrt{k}$. For Series A, $n \approx 100$ suffices, whereas Series B would require $n \approx 10^8$. This highlights that while both converge, practical computability depends heavily on asymptotic decay rates.

📖 Explanation: Students often confuse the inequality $|S - s_n| \leq a_{n+1}$ with equality. The theorem states that the magnitude of the remainder is bounded above by the first neglected term, not equal to it. Furthermore, the sign of the error $S - s_n$ is guaranteed to be the same as the sign of the first omitted term $(-1)^{n+2}a_{n+1}$. While the bound provides a safe worst-case estimate for stopping criteria, treating it as an exact value ignores the cancellation effects inherent in the remaining infinite tail of the series.

📖 Explanation: Graphical analysis of partial sums provides intuitive insight into convergence speed. The vertical distance between $s_n$ and $s_{n+1}$ corresponds exactly to the term magnitude $a_{n+1}$. If this visual gap decreases slowly at $n=50$, it suggests the underlying sequence $a_k$ has a heavy tail. Consequently, at $n=100$, the term $a_{101}$ may still exceed engineering tolerances. Unlike absolute convergence where tails vanish rapidly, slowly decaying alternating series can appear 'settled' on a coarse graph while retaining substantial truncation error, necessitating analytical verification over visual inspection.

📖 Explanation: The Alternating Series Estimation Theorem strictly requires the magnitude sequence $a_k$ to be monotonically decreasing. The function $f(x) = x + \sin(x)$ is increasing overall, but its reciprocal's discrete sampling can exhibit local non-monotonicity because $\sin(k)$ oscillates. Specifically, $k + \sin(k)$ might occasionally decrease or stay flat between integer steps, causing $a_k$ to increase locally. Without establishing eventual monotonicity rigorously or modifying the bound, blindly applying $|R_n| \leq a_{n+1}$ is mathematically unjustified. One must verify $a_{k+1} \leq a_k$ holds for all $k \geq N$ before trusting the error estimate.

📖 Explanation: This problem tests the precise application of the stopping criterion. To guarantee $|S - s_n| < \epsilon$, we need $a_{n+1} < \epsilon$. Here, checking $n=49$ gives bound $b_{50} = 0.0012$, which fails. Checking $n=50$ gives bound $b_{51} = 0.0009$, which satisfies the condition. Therefore, $s_{50}$ is the first partial sum guaranteed to be within tolerance. Note that we do not need to add the term $b_{51}$ itself to achieve the guarantee; the bound relies on the *next* term being small enough. This distinction saves computational effort.

📖 Explanation: The ratio of consecutive terms dictates the geometric-like decay rate of the series envelope. When the ratio is 0.9, terms shrink rapidly, and the error bound $a_{n+1}$ drops below tolerance quickly. As the ratio approaches 1 (e.g., 0.99), the decay becomes extremely sluggish, resembling a harmonic series locally. To maintain the same absolute error $\epsilon$, the index $n$ must increase dramatically because $a_n \approx a_1 r^n$. Solving $r^n < \epsilon$ shows $n \propto \ln(\epsilon)/\ln(r)$; as $r \to 1$, $\ln(r) \to 0$, causing $n$ to explode. This illustrates the sensitivity of numerical approximation to physical parameters.

📖 Explanation: For alternating series with smoothly varying terms, the true sum $S$ lies between $s_n$ and $s_{n+1}$. The errors $S-s_n$ and $S-s_{n+1}$ have opposite signs and similar magnitudes. Averaging them tends to cancel the dominant linear component of the error, leaving a residual related to the curvature of the term sequence. This technique, known as Euler transformation or simple averaging, frequently improves accuracy by an order of magnitude compared to raw partial sums. It leverages the oscillatory structure rather than just the magnitude bound, representing a sophisticated conceptual leap beyond basic theorem application.

📖 Explanation: For series involving factorials, the terms decrease so rapidly that the Alternating Series Estimation Theorem is exceptionally powerful. We solve $\frac{1}{(n+1)!} < 10^{-5}$. Testing values: $1/7! \approx 1.9 \times 10^{-4}$, $1/8! \approx 2.4 \times 10^{-5}$, $1/9! \approx 2.7 \times 10^{-6}$. Thus, we need $n+1 = 9 \Rightarrow n=8$. Wait, re-evaluating: $1/8! < 10^{-5}$ is false. $1/9! < 10^{-5}$ is true. So $n+1=9 \implies n=8$. Correction: Option A says n=7 implying $1/8!$ check? Actually $1/8! \approx 0.000024 > 0.00001$. So n=8 is correct. Let's adjust option A to n=8 for correctness or explain n=8. *Self-correction*: The prompt asks for HOTS. The key is recognizing that for factorials, discrete checking is mandatory; integrals fail. Correct answer should reflect n=8. Adjusted explanation: We seek minimal n where $a_{n+1} < \epsilon$. Since $1/8! > 10^{-5}$ and $1/9! < 10^{-5}$, we need $n+1=9$, so $n=8$.

📖 Explanation: While both methods estimate truncation error, the Alternating Series Estimation Theorem offers immense computational simplicity for series satisfying its hypotheses. The Lagrange form $R_n(x) = \frac{f^{(n+1)}(c)}{(n+1)!}(x-a)^{n+1}$ demands maximizing an unknown derivative over an interval, which can be analytically difficult or impossible for complex functions. In contrast, for an alternating series with decreasing terms, the error is simply bounded by the magnitude of the first omitted term $|a_{n+1}|$. This eliminates calculus optimization steps entirely, making it the pragmatic choice for numerical evaluation of standard transcendental functions.

📖 Explanation: This scenario highlights the critical distinction between mathematical truncation error and computational round-off error. While the theoretical bound $a_{n+1}$ decreases with $n$, floating-point addition introduces rounding errors at each step. For slowly converging alternating series, reaching high precision requires massive $n$. As $n$ grows, the cumulative round-off error can grow proportionally to $\sqrt{n}$ or worse, eventually dominating the diminishing truncation error. The user observes a floor in accuracy not because the math is wrong, but because the numerical representation cannot resolve the tiny corrections against the accumulated noise of previous operations.

📖 Explanation: Defining terms via integrals of positive, decreasing functions (like $1/x^p$) inherently ensures $a_k > 0$ and $a_{k+1} < a_k$ because the integration interval shifts rightward along a decreasing curve. This structural definition bypasses tedious algebraic differentiation or difference checks typically needed to prove monotonicity. Consequently, one can immediately apply the Alternating Series Estimation Theorem with confidence. Moreover, this links the series error bound to integral estimates, providing a bridge between discrete summation and continuous analysis, reinforcing why integral-defined sequences are ideal candidates for alternating series approximation techniques.

📖 Explanation: The standard Alternating Series Estimation Theorem applies specifically to partial sums $s_n$, giving $|S - s_n| \leq a_{n+1}$. It does not formally cover averages of partial sums. While averaging $s_n$ and $s_{n+1}$ often yields superior empirical accuracy by centering the oscillation, deriving a rigorous analytic bound requires additional assumptions about term convexity or using transformed series theorems. Therefore, strictly speaking, only $s_{10}$ carries the immediate textbook guarantee of $1/21$. Recognizing the gap between heuristic improvement and formal proof is a key higher-order skill in numerical analysis.

📖 Explanation: Precision in stating stopping criteria is vital. The theorem guarantees $|R_n| \leq a_{n+1}$. To ensure error $< \epsilon$, we strictly need $a_{n+1} < \epsilon$. If a student stops when the *current* term $a_n = \epsilon$, the *next* term $a_{n+1}$ (which bounds the error of $s_n$) might still be close to $\epsilon$ or, in non-strict cases, equal to it. More importantly, if they interpret 'term equals 0.01' as the stopping signal for $s_n$, they are bounding the error of $s_{n-1}$ by 0.01, not $s_n$. Confusing $a_n$ with $a_{n+1}$ leads to off-by-one errors that violate the specified tolerance.

📖 Explanation: Real-world data rarely satisfies pure mathematical hypotheses like strict monotonicity. When noise violates the condition locally, blindly applying the theorem is risky. However, discarding the model wastes valuable structural information. The robust engineering approach is to separate signal from noise: fit a monotonic decay curve (envelope) to the noisy data. The alternating series framework then applies to this idealized envelope. The error estimate becomes a combination of the theoretical truncation error of the envelope plus the statistical uncertainty from the noise fitting. This adapts the mathematical tool to messy reality without sacrificing rigor.

📖 Explanation: The Alternating Series Estimation Theorem relies fundamentally on the specific ordering of terms. Conditional convergence means rearrangement can change the sum or destroy the oscillatory nesting property $s_{2n} < S < s_{2n+1}$ that underpins the error bound. Grouping terms alters the partial sum sequence structure, potentially making the standard $a_{n+1}$ bound meaningless relative to the new partial sums. Scaling (A) just scales the error; shifting (B) doesn't affect convergence/error of the variable part; modifying the index (D) preserves alternation structure mostly. Rearrangement attacks the foundational hypothesis of ordered partial sums bounding the limit.

📖 Explanation: The convergence rate of the Maclaurin series for $\arctan(x)$ depends critically on $|x|$. The error bound behaves roughly as $|x|^{2n+1}$. For Leibniz ($x=1$), decay is harmonic-slow ($1/n$). For Machin, terms involve $(1/5)^{2n+1}$ and $(1/239)^{2n+1}$, exhibiting rapid geometric decay. To achieve $10^{-d}$ accuracy, Leibniz needs $\sim 10^d$ terms, while Machin needs only $\sim d$ terms. This exponential vs. linear scaling makes Machin's formula computationally feasible for high precision, demonstrating how argument reduction transforms an impractical alternating series into an efficient algorithm.

📖 Explanation: This addresses a subtle misconception. The *upper bound* $a_{n+1}$ is guaranteed to decrease (non-increasing). However, the *actual* error $|S - s_n|$ is not necessarily monotonic. Depending on the specific values of the tail terms, the partial sum $s_n$ might land further from $S$ than $s_{n-1}$ did, even though the safety margin $a_{n+1}$ shrank. The theorem guarantees safety within a shrinking envelope, not monotonic approach to the limit. Distinguishing between the guaranteed bound and the actual trajectory is crucial for understanding numerical behavior versus theoretical guarantees.

📖 Explanation: Decomposing $\frac{1}{k^3+k}$ into $\frac{1}{k} - \frac{k}{k^2+1}$ (or similar) reveals the underlying asymptotic structure. While the combined term clearly decays as $1/k^3$, separation allows analyzing cancellation effects or comparing against known benchmark series more precisely. Sometimes, splitting exposes a dominant alternating component and a faster-decaying correction. This structural insight can refine error estimates beyond the crude $a_{n+1}$ of the combined term, especially if one component can be summed analytically or transforms into a faster-converging form. Algebraic preprocessing is a powerful strategy for optimizing numerical approximation of rational alternating series.

📖 Explanation: The curvature of the envelope formed by odd/even partial sums reflects the second derivative of the underlying term function. If the envelopes are curved (not straight lines connecting points), the midpoint $(s_n + s_{n+1})/2$ does not lie exactly on $S$; it deviates based on the local convexity. Specifically, if the upper envelope is convex down and lower convex up, the average tends to overestimate or underestimate systematically. Recognizing this geometric bias explains why simple averaging improves accuracy but doesn't eliminate error entirely, and motivates higher-order extrapolation methods like Richardson extrapolation for further refinement.

📖 Explanation: Mental estimation requires simplification without losing validity. Since $n^2 + 100 > n^2$, the terms are strictly smaller than $1/n^2$. The alternating series error bound for the target series is $a_{n+1} < \frac{1}{(n+1)^2}$. Finding $n$ such that $\frac{1}{(n+1)^2} < 0.001$ guarantees the stricter condition for the original series. This comparison avoids dealing with the +100 offset mentally. $(n+1)^2 > 1000 \Rightarrow n+1 > 31.6 \Rightarrow n=31$. This provides a safe, quick upper bound on the required terms using a simpler proxy series, leveraging monotonicity of the bounding function.

📖 Explanation: Before applying any error estimation theorem, one must verify convergence prerequisites. Here, $\lim_{n \to \infty} \frac{n}{2n+1} = 1/2 \neq 0$. By the Divergence Test, the series cannot converge. The Alternating Series Test fails at the limit condition. Attempting to use the error bound $a_{n+1}$ would suggest an error approaching 0.5, which is nonsensical for a divergent oscillating sequence. This reinforces that error bounds are conditional on convergence; checking $\lim a_n = 0$ is the mandatory first step in any approximation workflow.

📖 Explanation: If the actual error stabilizes at half the theoretical bound, it indicates the partial sums are oscillating symmetrically and regularly around the limit. This typically happens when $a_k$ follows a smooth, well-behaved asymptotic expansion (like $1/n^p$). The factor of 0.5 arises because $S$ lies roughly midway between $s_n$ and $s_{n+1}$ for large $n$ in regular series. This observation validates using refined estimators like averaging and confirms the series is numerically benign. It transforms the conservative inequality into a precise asymptotic equality, enabling sharper error prediction in practice.

📖 Explanation: Robust numerical code must validate assumptions, not just compute. The Alternating Series Estimation Theorem assumes strict alternation. If upstream data generation introduced sign errors (e.g., two positives in a row), the series might no longer satisfy the theorem's hypotheses. Blindly taking `abs(term)` enforces artificial positivity, hiding the structural violation. The code would stop based on magnitude while the actual series behavior could be divergent or have unpredictable error. Defensive programming requires verifying alternation explicitly or ensuring upstream integrity, rather than assuming mathematical properties hold computationally.

📖 Explanation: While both involve alternating factorial terms, integrating $e^{-x^2} = \sum \frac{(-1)^n x^{2n}}{n!}$ from 0 to 1 yields $\sum \frac{(-1)^n}{n!(2n+1)}$. The extra factor $\frac{1}{2n+1}$ significantly accelerates convergence compared to evaluating $e^{-1} = \sum \frac{(-1)^n}{n!}$. For the same $n$, the integral's error bound is smaller by roughly $2n$. Thus, achieving a given tolerance requires fewer terms for the definite integral than for the function evaluation at $x=1$. Recognizing how operations like integration modify term asymptotics is essential for efficient numerical method selection.

📖 Explanation: If all theoretical conditions are met and the bound calculation is correct, the discrepancy must be computational. The theorem guarantees mathematical proximity of the infinite sum to the partial sum. It cannot protect against floating-point catastrophes. Summing 100+ terms with limited precision can accumulate round-off errors exceeding $10^{-5}$, especially if terms vary widely in magnitude before decaying. This scenario underscores that theoretical error bounds address truncation only; total numerical error = truncation + round-off. Validating high-precision results requires checking both mathematical bounds and numerical stability.

📖 Explanation: The nesting $s_{2n} \leq S \leq s_{2n+1}$ arises directly from the monotonicity of $a_k$. Specifically, $s_{2n+2} = s_{2n} + (a_{2n+1} - a_{2n+2}) \geq s_{2n}$ requires $a_{2n+1} \geq a_{2n+2}$. Similarly, odd partial sums decrease because $s_{2n+1} = s_{2n-1} - (a_{2n} - a_{2n+1}) \leq s_{2n-1}$. Without monotonic decrease, these inequalities break, and partial sums could jump across $S$ erratically, destroying the interval containment that makes $a_{n+1}$ a valid error bound. Positivity and limit-zero are needed for convergence, but monotonicity specifically enables the error estimation mechanism.

📖 Explanation: The Euler Transform exploits the smoothness of the term sequence $a_k$ to construct a new series whose terms involve differences $\Delta^m a_k$. For monotonic sequences like $1/k$, these differences decay much faster than the original terms. The transformed series retains the same sum but achieves the error tolerance with far fewer terms. This is a sophisticated alternative to brute-force summation, converting slow harmonic decay into rapid geometric-like decay through algebraic manipulation. Understanding such transformations represents advanced mastery of alternating series approximation beyond basic theorem application.

📖 Explanation: Mathematical error bounds assume exact terms. When terms carry physical uncertainty $\delta$, summing beyond the point where $a_{k} \approx \delta$ adds noise rather than signal. The theoretical truncation error $a_{N+1}$ becomes irrelevant because it's buried in measurement error. Continuing to sum merely accumulates random fluctuations. The optimal stopping point is dictated by the noise floor, not the mathematical convergence rate. This integrates experimental realism with numerical theory: effective accuracy is $\max(\text{truncation error}, \text{measurement noise})$. Ignoring this leads to false precision and wasted computation.

📖 Explanation: Although both odd and even subsequences decay nicely, the interleaved sequence $1, 1/4, 1/3, 1/16, 1/5, ...$ is not monotonic ($1/4 < 1/3$). The Alternating Series Test and its error bound require the entire sequence $a_n$ to be non-increasing. Local increases violate the nesting property of partial sums. While the series may still converge (by other tests), the simple bound $|R_n| \leq a_{n+1}$ is invalid. One would need to regroup terms or use a different estimation method. This tests understanding that subsequence behavior doesn't guarantee whole-sequence properties.

📖 Explanation: In software, indexing bugs (confusing 0-based vs 1-based, or n vs n+1) are common. Using $a_n$ (the last included term) instead of $a_{n+1}$ (the first excluded term) as the error proxy introduces a deliberate conservatism. Since $a_n \geq a_{n+1}$, this overestimates the error slightly, ensuring the tolerance is met even if an indexing mistake occurs. While mathematically looser, it enhances robustness in production code where correctness trumps minimal efficiency. This reflects practical engineering judgment balancing theoretical optimality with implementation reliability.

📖 Explanation: Absolute convergence guarantees existence but doesn't prescribe the best error estimation method. For an alternating p-series with $p=4$, the alternating series bound $a_{n+1}$ is trivial to compute and extremely tight. Absolute convergence tests (like comparison or integral) establish finiteness but don't yield simple remainder formulas as directly as the alternating series theorem. Dismissing the alternating bound because of absolute convergence misses a computational shortcut. The alternating structure remains exploitable regardless of absolute convergence status. Efficiency comes from leveraging the strongest available structural property, which here is the alternating decay.

📖 Explanation: Visual irregularities in the convergence envelope directly reflect violations of monotonicity in $a_n$. The standard error bound $|R_n| \leq a_{n+1}$ derives from monotonic nesting. If the graph shows the partial sum moving away from S temporarily, it means $a_{n+1} > a_n$ locally, breaking the nesting. At such points, the standard bound is unreliable. Identifying these visual anomalies allows diagnosing hypothesis failures without algebraic derivation. It reinforces that smooth, regular oscillation is the graphical signature of valid alternating series estimation, while irregularity signals caution.

📖 Explanation: The term magnitude $a_n = x^n/n$ depends exponentially on x. For x=0.1, $0.1^n$ drives rapid geometric decay, making $a_{n+1}$ tiny quickly. For x=0.9, the factor $0.9^n$ decays sluggishly, and the harmonic divisor $1/n$ barely helps initially. To achieve $\epsilon = 10^{-6}$, x=0.1 needs ~6 terms; x=0.9 needs ~100+. The error bound's effectiveness is thus highly sensitive to the argument's proximity to the radius of convergence boundary. This illustrates why series are practically useful near the center but computationally expensive near edges, motivating argument reduction techniques.

📖 Explanation: Using the alternating series error bound without verifying monotonicity is a common but dangerous practice. Non-monotonic terms can cause partial sums to swing wider than the next term predicts. For example, if $a_{n+1}$ is anomalously small but $a_{n+2}$ is large, $s_n$ might be far from S despite small $a_{n+1}$. The bound $a_{n+1}$ only works because monotonicity prevents such future surges. Without it, the bound is merely a heuristic with no guarantee. Rigorous approximation demands either proving monotonicity or using alternative error estimation methods like integral bounds or transformation techniques.

📖 Explanation: Leibniz series for $\pi$ converges so slowly that achieving 1 million digits would require $\sim 10^{1000000}$ terms, which is physically impossible. The error bound correctly predicts this impossibility, but the theorem offers no remedy. Modern computing requires acceleration methods (Ramanujan, Chudnovsky) or different representations. The theorem identifies inadequacy but doesn't solve it. In contrast, other options are tractable with reasonable n. This highlights that satisfying theoretical conditions doesn't imply practical computability; the error bound serves as a diagnostic for algorithmic feasibility, not just accuracy verification.

📖 Explanation: While the error bound setup is straightforward, solving $n \ln(n+1) > 100$ analytically for n is impossible with elementary functions due to the transcendental mix. Numerical root-finding (Newton's method, bisection) or iterative testing becomes necessary. This bridges symbolic calculus and numerical methods: setting up the bound is analysis; executing it is computation. Students must recognize when symbolic manipulation hits a wall and switch to algorithmic approaches. The error bound provides the equation; numerical analysis provides the answer.

📖 Explanation: The theorem states $|S - s_n| \leq a_{n+1}$. If $s_n = 0.75$ and bound is 0.02, then S must lie in [0.73, 0.77]. A true value of 0.78 violates this interval, indicating either: (1) the series doesn't satisfy theorem hypotheses (e.g., non-monotonic), (2) the bound was miscalculated, or (3) the partial sum was computed incorrectly. Consistency checking is a vital validation step. If reality contradicts the bound, trust the math: something in the premise or execution is flawed. This reinforces the theorem as a rigorous constraint, not a suggestion.

📖 Explanation: Error propagation principles dictate that uncertainties scale through computations. If downstream operations amplify initial error by K, the upstream tolerance must be tightened proportionally to preserve final accuracy. Setting initial target to $\epsilon_{final}/K$ ensures $K \times (\epsilon_{final}/K) = \epsilon_{final}$. This integrates alternating series approximation into broader numerical workflows. Isolating series error from system error leads to false confidence. Holistic error budgeting requires distributing tolerance across all stages, making alternating series approximation a component of integrated uncertainty management rather than an isolated exercise.

📖 Explanation: Sharpness refers to how closely the bound matches actual error. For geometric $r=1/2$, $R_n = \frac{a_{n+1}}{1-r} = 2a_{n+1}$? Wait, actually for alternating geometric $\sum (-r)^n$, $|R_n| = \frac{r^{n+1}}{1+r}$. Bound is $r^{n+1}$. Ratio is $1/(1+r)$. For r=1, ratio=0.5. For factorial, terms drop so fast that $R_n \approx a_{n+1}$ almost exactly because $a_{n+2} \ll a_{n+1}$. Actually, the bound is sharpest when subsequent terms are negligible. For geometric, subsequent terms form significant fraction. Re-evaluating: The question premise might be inverted or nuanced. Typically, factorial series have error ≈ next term (very sharp), while geometric has error = next_term/(1+r) < next_term. So bound is looser for geometric. Adjusting explanation: For factorial, $a_{n+2}/a_{n+1} \to 0$, so $R_n \approx a_{n+1}$. For geometric, ratio is constant, so $R_n$ is fixed fraction of bound. Thus bound is sharper for factorial. *Correction to option A text in final output to match correct math.*