📝 Power series in x formula (35 MCQs)

📖 Explanation: The radius of convergence $R$ defines a symmetric interval $(-R, R)$ where absolute convergence occurs. Since the series converges at $x=4$, we know $R \geq 4$. Since it diverges at $x=-7$, we know $R$ cannot be greater than or equal to 7 (otherwise it would converge absolutely at -7). Thus, $4 \leq R < 7$. Option A assumes the endpoints are exactly the convergence boundaries, which is not guaranteed. Option C includes 7, which contradicts the divergence at -7.

📖 Explanation: Differentiation of a power series preserves the radius of convergence $R$, but endpoint behavior can change. For $f(x)$, the series converges at both $x=1$ (p-series p=2) and $x=-1$ (alternating p-series). The derivative is $\sum x^{k-1}/k$. At $x=1$, this becomes the harmonic series which diverges. At $x=-1$, it is the alternating harmonic series which converges. Thus, the interval shrinks from $[-1, 1]$ to $[-1, 1)$. This illustrates that while $R$ is invariant, endpoint inclusion is not.

📖 Explanation: When finding a specific function via integration of a series, \int f&#039;(x) dx = f(x) + C. To determine $C$, one must evaluate at the center of expansion (usually $x=0$ for Maclaurin). Since $\ln(1+0^2) = \ln(1) = 0$ and the derived power series has no constant term (all terms have powers of $x$), substituting $x=0$ yields $0 = 0 + C$, implying $C=0$. This highlights the necessity of checking initial values when reconstructing functions from derivatives via series.

📖 Explanation: Graphical analysis of partial sums provides visual evidence of convergence. Inside the radius of convergence, partial sums settle toward a limit curve. Outside, they diverge. The transition point where behavior shifts from stable approximation to unbounded oscillation indicates the boundary of the interval of convergence. Here, the shift at $|x|=1$ implies $R=1$. The wild oscillation at the boundary suggests divergence rather than conditional convergence, as conditional convergence typically shows damped oscillation settling slowly, whereas 'wildly with increasing amplitude' indicates the terms do not approach zero.

📖 Explanation: The radius of convergence of a sum is at least the minimum of the individual radii, i.e., $R_H \geq \min(R_A, R_B) = 3$. Within $|x|<3$, both converge, so the sum converges. For $3 < |x| < 5$, $B(x)$ converges but $A(x)$ diverges, making the sum diverge. Thus, usually $R_H = 3$. However, if $A(x)$ and $B(x)$ are constructed such that their singularities at $|x|=3$ cancel each other out perfectly (e.g., $A(x) = \frac{1}{3-x}$ and $B(x) = -\frac{1}{3-x} + \frac{1}{5-x}$), the singularity at 3 vanishes, potentially extending $R_H$ to 5. Therefore, stating it 'must' be 3 or 5 without analyzing coefficients is flawed.

📖 Explanation: This addresses the distinction between theoretical convergence and numerical utility. Even if a series converges at a point within its radius, the rate of convergence might be extremely slow, or intermediate terms might grow very large before decaying. In computational modeling, finite precision arithmetic means that if terms grow large (e.g., $10^{16}$) before cancelling out to a small sum, significant digits are lost, rendering the result useless. This 'numerical breakdown' often precedes the theoretical radius of convergence, especially near singularities or in ill-conditioned expansions.

📖 Explanation: While direct differentiation works theoretically, the product rule complexity explodes for $e^{-x^2}$. By the fourth or sixth derivative, the expression involves polynomials multiplied by the exponential, making evaluation at zero prone to algebraic errors. Substitution leverages the known structure of $e^u = \sum u^k/k!$. Replacing $u$ with $-x^2$ instantly yields $\sum (-1)^k x^{2k}/k!$. This question tests the strategic selection of methods. The distractor about substitution validity is false; composition with analytic functions preserves analyticity within the domain.

📖 Explanation: The geometric series $\sum (-x^2)^k$ converges only for $|x|<1$. Integrating term-by-term gives the arctan series, which converges conditionally at $x=1$ (alternating harmonic-like). While the original series diverges at the boundary, the integrated series defines a continuous function on $[0, 1]$. Abel's Theorem states that if a power series converges at an endpoint of its interval of convergence, the function it represents is continuous at that endpoint from within the interval. Since $\arctan(x)$ is also continuous at $x=1$, the equality holds. This connects integration, convergence, and continuity.

📖 Explanation: Using the Root Test concept implicitly related to coefficient growth: $\lim |c_k|^{1/k}$. Applying Stirling’s approximation $k! \approx \sqrt{2\pi k}(k/e)^k$, we get $c_k \approx \sqrt{2\pi k} (k/e)^k / k^k = \sqrt{2\pi k} e^{-k}$. Taking the k-th root: $(\sqrt{2\pi k})^{1/k} (e^{-k})^{1/k} \to 1 \cdot e^{-1} = 1/e$. The radius is the reciprocal of this limit, so $R = e$. This requires connecting coefficient asymptotics to convergence radii without relying solely on mechanical ratio test application, testing deep understanding of factorial growth rates versus exponential bases.

📖 Explanation: A common misconception is treating power series multiplication like vector component-wise multiplication. If $A(x) = \sum a_n x^n$, then $A(x)^2 \neq \sum a_n^2 x^n$. Instead, $A(x)^2 = \sum c_n x^n$ where $c_n = \sum_{j=0}^n a_j a_{n-j}$. For $\frac{1}{1+x}$, term-wise squaring gives $\sum x^k$ (since $(-1)^{2k}=1$), which equals $\frac{1}{1-x}$, clearly wrong. The correct method is either Cauchy product or recognizing $\frac{1}{(1+x)^2} = -\frac{d}{dx}(\frac{1}{1+x})$. This tests understanding of series algebra versus scalar algebra.

📖 Explanation: The original series converges on $(2-3, 2+3) = (-1, 5)$. The function $f(x)$ is analytic on this interval. A Maclaurin series (centered at 0) will converge on the largest symmetric interval $(-R_{new}, R_{new})$ contained within the domain of analyticity. Since $f$ is analytic on $(-1, 5)$, it is certainly analytic on $(-1, 1)$. Thus $R_{new} \geq 1$. However, if the nearest singularity to 0 is actually at $x=-1$, then $R_{new}=1$. If the singularity at -1 was removable or the function extends analytically beyond it, $R_{new}$ could be larger. We cannot assume $R_{new}=R$ because the center shifted relative to the singularities.

📖 Explanation: The Lagrange Error Bound requires finding a maximum $M$ for the $(n+1)$-th derivative on an interval containing $x$ and the center. While $|\sin^{(n+1)}(z)| \leq 1$ is a valid global bound, it can be overly conservative for small $x$ or specific intervals. The Alternating Series Estimation Theorem states the error is bounded by the magnitude of the first neglected term, $|a_{n+1}|$, provided the series alternates and terms decrease. This bound is exact for the worst-case scenario of the remainder's sign and requires no knowledge of intermediate points $z$. It is often sharper and simpler for standard trigonometric/exponential series.

📖 Explanation: The convergence of the binomial series at endpoints depends critically on $m$. At $x=1$, the series behaves asymptotically like $\sum \frac{1}{k^{m+1}}$ (derived via Gamma function properties or Gauss’s test). For absolute convergence, we need the exponent $m+1 > 1$, implying $m > 0$. If $-1 < m \leq 0$, it converges conditionally. If $m \leq -1$, it diverges. This nuanced endpoint behavior distinguishes the binomial series from simpler geometric or exponential series and requires understanding asymptotic analysis of generalized binomial coefficients.

📖 Explanation: Mathematical models often have domains of validity restricted by physical constraints rather than analytical ones. A power series solution to a DE might be analytic everywhere (entire function, $R=\infty$) or have singularities far from the region of interest. However, the physical quantity being modeled (population) might blow up or become negative due to model limitations (e.g., logistic vs exponential assumptions) before any mathematical singularity is reached. The radius of convergence describes where the *series* sums to the *function*, but the *function* itself might cease to represent reality. This distinguishes mathematical existence from physical applicability.

📖 Explanation: Since the integrated series $\sum (-1)^n \frac{x^{2n+1}}{(2n+1)n!}$ evaluated at $x=0.5$ is an alternating series with decreasing terms, the Alternating Series Estimation Theorem applies. The error after $N$ terms is less than the absolute value of the $(N+1)$-th term. To guarantee 4 decimal place accuracy (error $< 0.5 \times 10^{-4}$), one solves the inequality involving the next term. This is a direct application of error bounds for numerical integration via series, superior to trial-and-error or inappropriate tests like the Ratio Test (which determines convergence, not partial sum error).

📖 Explanation: A necessary condition for a function to be represented by a power series (analytic) at a point is that it must be infinitely differentiable at that point. $|x|$ is continuous at 0 but not differentiable there (corner). Thus, no Maclaurin series exists. Note: $e^{-1/x^2}$ IS infinitely differentiable and HAS a Maclaurin series (all zeros), but the series does not converge TO the function (it's not analytic). The question asks what *cannot be represented*, implying the series representation itself doesn't exist or fails to represent the function. $|x|$ fails the basic differentiability prerequisite. $\sqrt[3]{x}$ has infinite derivative at 0, also failing. Between A and D, A is the classic example of non-differentiability preventing series formation.

📖 Explanation: When differentiating $\sum_{k=0}^\infty x^k$, the first derivative is $\sum_{k=1}^\infty k x^{k-1}$. The second derivative is $\sum_{k=2}^\infty k(k-1) x^{k-2}$. At $x=0$, only the $k=2$ term survives: $2(1)(0)^0 = 2$. But $\frac{d^2}{dx^2}(1-x)^{-1} = 2(1-x)^{-3}$. At $x=0$, this is 2. So $\frac{1}{(1-x)^3} = \frac{1}{2} \sum_{k=2}^\infty k(k-1)x^{k-2}$. The student missed the factor of $1/2!$ arising from the chain rule/derivative scaling. Proper re-indexing to $n=k-2$ yields $\sum \frac{(n+2)(n+1)}{2} x^n$, which correctly gives 1 at $x=0$.

📖 Explanation: For small $x$, $\cos(x) \approx 1 - x^2/2$. Direct series converges rapidly. Using $1 - 2\sin^2(x/2)$ involves computing $\sin(0.05)$, which converges even faster due to the smaller argument (error scales with $x^{2n+1}$). Although there are extra arithmetic operations (squaring, multiplying by 2, subtracting from 1), the reduction in required series terms for high precision can outweigh this. However, for very small $x$, $1 - \text{small}$ risks loss of significance (catastrophic cancellation). Thus, while theoretically faster in convergence, numerically it requires care. Option B captures the convergence benefit, which is the primary analytical advantage.

📖 Explanation: Let $y = x^2$. The series becomes $\sum c_k y^k$, which converges for $|y| < 2$. Substituting back, $|x^2| < 2 \implies x^2 < 2 \implies |x| < \sqrt{2}$. Thus the interval is $(-\sqrt{2}, \sqrt{2})$. This tests understanding of variable substitution in power series. Students often mistakenly think replacing $x$ with $x^2$ squares the radius or leaves it unchanged. The correct transformation involves solving the inequality imposed by the original radius on the new variable expression.

📖 Explanation: Pointwise convergence on $\mathbb{R}$ does not imply uniform convergence on $\mathbb{R}$. For $e^x$, the remainder $R_n(x)$ grows without bound as $x \to \infty$ for any fixed $n$. Uniform convergence requires $\sup_{x \in D} |R_n(x)| \to 0$. On $\mathbb{R}$, this supremum is infinite. However, on any bounded interval $[-M, M]$, the Weierstrass M-test confirms uniform convergence. This distinction is crucial in analysis: power series are uniformly convergent on compact subsets of their interval of convergence, but not necessarily on the entire open interval if it is unbounded.

📖 Explanation: Although $\sum \frac{(-1)^k x^k}{k}$ converges only conditionally at $x=1$, term-by-term integration over $[0,1]$ is valid. The resulting series $\sum \frac{(-1)^k}{k(k+1)}$ converges absolutely (terms ~ $1/k^2$). More importantly, since the original series converges at the endpoint $x=1$ and represents a continuous function $-\ln(1+x)$ on $[0,1]$, the integral of the series equals the series of the integrals. This is a consequence of Abel's theorem and properties of uniformly convergent sequences on compact sets where endpoint convergence holds.

📖 Explanation: At $x=1/4$, the general term is $\binom{2n}{n} (1/4)^n$. Using Stirling’s approximation, $\binom{2n}{n} \sim \frac{4^n}{\sqrt{\pi n}}$. Multiplying by $(1/4)^n$ gives terms asymptotic to $\frac{1}{\sqrt{\pi n}}$. This is a p-series with $p=1/2$. Since $p \leq 1$, the series diverges. This problem requires combining radius of convergence calculation with asymptotic analysis of binomial coefficients at the boundary, going beyond standard textbook examples where endpoints are usually simple geometric or alternating harmonic series.

📖 Explanation: Direct evaluation suffers from catastrophic cancellation ($e^x \approx 1$) and division by small numbers. The series $\frac{e^x-1}{x} = \frac{(1+x+x^2/2!+\dots)-1}{x} = 1 + \frac{x}{2!} + \frac{x^2}{3!} + \dots = \sum_{k=0}^\infty \frac{x^k}{(k+1)!}$ removes the singularity analytically. At $x=0$, it evaluates exactly to 1. For small $x$, it avoids subtracting nearly equal numbers. This demonstrates the practical utility of power series in algorithm design for robust numerical computation near singular points.

📖 Explanation: The interval of convergence is centered at 3 with radius 2, spanning $(1, 5)$. Convergence at the left endpoint $x=1$ is given. Behavior at the right endpoint $x=5$ is independent of behavior at the left endpoint for general power series. One endpoint can converge absolutely, conditionally, or diverge regardless of the other. Without specific information about the coefficients $c_k$ (e.g., positivity, monotonicity), no deduction can be made about $x=5$. This tests understanding that endpoint behaviors are distinct and not symmetrically linked by the radius alone.

📖 Explanation: Polynomial (or power series) division is a deterministic algorithm where the coefficient of $x^n$ in the quotient is determined solely by coefficients of degree $\leq n$ in the numerator and denominator. Higher-degree terms in the divisor only affect quotient terms of degree $> n$. Therefore, truncating the calculation after obtaining the desired degree is perfectly valid and exact for those coefficients. This property makes series division a practical tool for finding initial terms without needing infinite precision.

📖 Explanation: A fundamental property of power series is that they define infinitely differentiable (analytic) functions within their open interval of convergence. A function with a jump discontinuity is not even continuous, let alone differentiable. Therefore, no power series can represent such a function on an interval containing the discontinuity. If a power series exists, the function MUST be smooth. This question tests the converse understanding: analyticity imposes strict regularity conditions. Options B and C assume a power series exists for a discontinuous function, which is impossible.

📖 Explanation: Standard series at $x=0.1$: error $< \frac{0.1^{n+1}}{n+1}$. For $10^{-6}$, $n \approx 5$ suffices (actually quite fast). Wait, let's re-evaluate. $0.1^5/5 = 2 \times 10^{-6}$. $0.1^6/6 \approx 1.6 \times 10^{-7}$. So ~6 terms. Transformed series for $\ln(1.1)$: $\frac{1+x}{1-x} = 1.1 \Rightarrow x = 1/21 \approx 0.0476$. Terms involve $(0.0476)^{2k+1}$. First term ($k=0$): $2(0.0476) \approx 0.095$. Second term ($k=1$): $\frac{2}{3}(0.0476)^3 \approx 7 \times 10^{-5}$. Third term ($k=2$): $\frac{2}{5}(0.0476)^5 \approx 10^{-7}$. So only 3 terms needed. The transformed series converges MUCH faster. My initial thought was reversed. Correct answer should reflect transformed superiority. Let me correct: Standard needs ~6, Transformed needs ~3. Transformed is better. But option A says standard requires MORE. That matches. Yes, A is correct.

📖 Explanation: By the Ratio Test for power series, $R = 1/\lim |c_{k+1}/c_k|$. If the limit $L=0$, then $R = 1/0 = \infty$. For $c_k = 1/k!$, $c_{k+1}/c_k = \frac{k!}{(k+1)!} = \frac{1}{k+1} \to 0$. Thus $R=\infty$. This is a foundational concept linking coefficient ratios to domain of convergence. The distractor A confuses limit 0 with radius 0 (which happens when limit is $\infty$).

📖 Explanation: Recurrence links $a_{n+1}$ to $a_{n-1}$, implying odd terms vanish (since $a_1=0$). For even terms: $a_2 = a_0/2 = 1/2$, $a_4 = a_2/4 = 1/(2\cdot4) = 1/8$, $a_6 = a_4/6 = 1/(2\cdot4\cdot6) = 1/48$. General term $a_{2k} = \frac{1}{2^k k!}$. Series is $\sum \frac{(x^2/2)^k}{k!} = e^{x^2/2}$. This connects series solutions of ODEs to known elementary functions, verifying the method's consistency.

📖 Explanation: Real analysis alone cannot explain why a smooth real function has finite radius. Complex analysis reveals that the radius of convergence of a Taylor series centered at $z_0$ is the distance to the nearest singularity in the complex plane. For $1/(1+z^2)$, singularities are at $\pm i$, distance 1 from origin. Thus $R=1$. The geometric series derivation (Option A) reflects this algebraically. Both explanations are valid: A is the real-variable manifestation, B is the fundamental complex-analytic cause. Recognizing both demonstrates comprehensive understanding.

📖 Explanation: The Maclaurin series (centered at 0) fails at $x=1$, but this doesn't preclude approximation by OTHER power series. Expanding about $x=0.5$ gives a series with radius 0.5 (distance to singularity at 1), converging on $(0,1)$. Expanding about $x=0.9$ gives radius 0.1, converging on $(0.8, 1)$. Thus, local polynomial approximation is possible arbitrarily close to 1 from the left, just not AT 1 or using the series centered at 0. The student conflates a specific series' failure with global non-approximability.

📖 Explanation: The product series converges absolutely wherever both factors converge absolutely. Thus, it certainly converges on the intersection of their disks of convergence, implying $R_C \geq \min(R_A, R_B)$. Equality usually holds, but $R_C$ can be larger if singularities cancel (e.g., $A(x)=(1-x)$, $B(x)=1/(1-x)$, product=1, $R_C=\infty > \min( \infty, 1 )$). Therefore, only the inequality is universally guaranteed. This tests precise knowledge of series algebra domains.

📖 Explanation: Asymptotic series diverge for any fixed $\lambda \neq 0$ as $n \to \infty$. However, for small $\lambda$, terms initially decrease, reach a minimum, then increase. Truncating at the smallest term yields an approximation with error proportional to that term (often exponentially small). Adding more terms worsens accuracy. This contrasts sharply with convergent series where more terms always improve accuracy. Understanding asymptotic vs. convergent series is vital in advanced applied mathematics and physics, where formal power series often lack convergence but retain utility.