Division rule

(provided $\lim_{x \to a} g(x) \neq 0$)

Deep Breakdown

What it tells me (Situation Awareness): The limit of a quotient equals the quotient of the limits, provided the denominator limit is not zero. This is the division analog of the addition and multiplication limit rules. The condition on the denominator is critical: if $\lim g(x) = 0$, the result may be $\pm\infty$, undefined, or an indeterminate form ($0/0$) requiring L'Hôpital's rule or algebraic techniques. In DA/DS, this rule underpins the convergence analysis of rates: click-through rate $= \text{clicks} / \text{impressions}$, cost-per-conversion, and loss-per-sample all converge as $n \to \infty$ when numerator and denominator converge independently to finite limits with a non-zero denominator.

Flowchart

Real-world use

Where it is used

Convergence of normalized metrics (mean loss = total loss / $n$ as $n \to \infty$), asymptotic analysis of efficiency ratios between two algorithms, computing steady-state rates in Markov chains (stationary distribution is a ratio of counts), and verifying that a ratio of independently converging sequences converges to the ratio of their limits.

Problem it solves

When evaluating $\lim_{n \to \infty} \frac{5n^2 + 3n}{2n^2 - n}$, you could expand or divide top and bottom by $n^2$: $\frac{5 + 3/n}{2 - 1/n}$. Then use the division rule: $\lim(5 + 3/n) / \lim(2 - 1/n) = 5/2$. This converts a complex limit into two simple ones.

How it solves the problem

From the product rule and the limit of $1/g(x)$: if $\lim g(x) = M \neq 0$, then $\lim_{x \to a} 1/g(x) = 1/M$ (this follows from the continuity of $1/x$ at $M \neq 0$). Then $\lim f/g = \lim f \cdot \lim(1/g) = L \cdot (1/M) = L/M$.

Example in a business

Algorithm efficiency ratio: $T_1(n) = 5n^2 + 100$ operations, $T_2(n) = 2n^2 + 50$ operations. As $n \to \infty$: $\lim_{n \to \infty} T_1/T_2 = \lim_{n \to \infty} \frac{5 + 100/n^2}{2 + 50/n^2}$. Both $100/n^2 \to 0$ and $50/n^2 \to 0$ (denominator is not zero). Division rule: $(5 + 0)/(2 + 0) = 5/2$. Algorithm 1 is asymptotically 2.5 times slower than Algorithm 2.

The formula, piece by piece

• $\lim_{x \to a}$: limit as $x$ approaches $a$
• $f(x)$: numerator function with limit $L = \lim_{x \to a} f(x)$
• $g(x)$: denominator function with limit $M = \lim_{x \to a} g(x)$
• Key condition: $M \neq 0$ Full rule: $\lim_{x \to a} \dfrac{f(x)}{g(x)} = \dfrac{\lim_{x \to a} f(x)}{\lim_{x \to a} g(x)} = \dfrac{L}{M}$, provided $M \neq 0$.

When $M = 0$:

• $L \neq 0$: limit is $\pm\infty$ or does not exist (check one-sided limits)
• $L = 0$: indeterminate form $0/0$ — requires L'Hôpital's rule or factoring

The formula in words

"The limit of a ratio equals the ratio of the limits — as long as the denominator limit is not zero." $\lim_{x \to 4} \dfrac{x^2 - 2x}{x + 1} = \dfrac{\lim(x^2 - 2x)}{\lim(x + 1)} = \dfrac{16 - 8}{5} = \dfrac{8}{5} = 1.6$.

Critical Concepts

• The condition $M \neq 0$ is non-negotiable. This is the single most important check before applying the rule. Missing it is a classic error.
• $0/0$ is the most common indeterminate form in calculus. It appears in limits of derivative definitions, convergence rates, and L'Hôpital applications. It requires separate analysis.
• Rational function limits at finite points: for polynomials, if the denominator is non-zero at $a$, just substitute. $\lim_{x \to a} P(x)/Q(x) = P(a)/Q(a)$ when $Q(a) \neq 0$.
• Asymptotic ratios (as $x \to \infty$): divide numerator and denominator by the dominant term ($x^n$), then apply the division rule to the simplified ratio.
• Combined with other rules: the division rule is rarely used alone.

It is combined with addition, constant extraction, and algebraic manipulation to simplify complex limits.

Step-by-step calculation (tiny example)

Step 1: Check the denominator limit is non-zero Evaluate $\lim_{x \to \infty} \dfrac{3x^2 + 2x}{x^2 - 5}$. Divide numerator and denominator by $x^2$ (the dominant term): $\dfrac{3 + 2/x}{1 - 5/x^2}$.

Step 2: Compute each limit separately Numerator: $\lim_{x \to \infty} (3 + 2/x) = 3 + 0 = 3$. Denominator: $\lim_{x \to \infty} (1 - 5/x^2) = 1 - 0 = 1$. Denominator limit $= 1 \neq 0$, so the division rule applies.

Step 3: Apply the division rule $\lim_{x \to \infty} \dfrac{3x^2 + 2x}{x^2 - 5} = \dfrac{3}{1} = 3$.

Code examples

Beginner example (Python)

import numpy as np

# Division rule: lim[f(x)/g(x)] = lim f(x) / lim g(x), provided lim g(x) != 0
# Verify: lim(x->inf) (3x^2 + 2x) / (x^2 - 5) = 3/1 = 3


def f(x: float) -> float:
    return 3 * x**2 + 2 * x


def g(x: float) -> float:
    return x**2 - 5


x_values = [10, 100, 1000, 10000, 100000]
print("lim(x->inf) (3x^2 + 2x) / (x^2 - 5)")
print("After dividing by x^2: lim (3 + 2/x) / (1 - 5/x^2) = 3/1 = 3")
print()
print(f"{'x':>10} | {'f(x)':>15} | {'g(x)':>15} | {'ratio f/g':>12} | {'-> 3?':>8}")
print("-" * 65)
for x in x_values:
    ratio = f(x) / g(x)
    print(f"{x:>10} | {f(x):>15,.0f} | {g(x):>15,.0f} | {ratio:>12.8f} | {abs(ratio - 3) < 0.01}")

# Demonstration: when denominator limit = 0, the rule fails
print("\nFailed case: lim(x->0) x / sin(x)")
print("Both numerator and denominator -> 0 (indeterminate 0/0)")
print("Need L'Hopital: the limit is actually 1, not 0/0")
x_small_values = [0.1, 0.01, 0.001, 0.0001]
for x in x_small_values:
    print(f" x = {x}: x/sin(x) = {x / np.sin(x):.8f}")

Beginner example (R)

# Division rule: lim[f/g] = lim f / lim g, provided lim g != 0
f <- function(x) 3 * x^2 + 2 * x
g <- function(x) x^2 - 5
x_values <- c(10, 100, 1000, 10000, 100000)
cat("lim(x->inf) (3x^2 + 2x)/(x^2 - 5) -> 3\n\n")
cat(sprintf("%10s | %12s\n", "x", "f(x)/g(x)"))
cat(strrep("-", 25), "\n")
for (x in x_values) {
  cat(sprintf("%10.0f | %12.8f\n", x, f(x) / g(x)))
}

Professional example (Python)

import numpy as np
import pandas as pd   # Scenario: computing algorithm efficiency ratio as input size n grows
# T1(n) = 5n^2 + 100 (quadratic algorithm)
# T2(n) = 2n^2 + 50 (also quadratic but smaller constant)
# Efficiency ratio T1/T2 -> 5/2 = 2.5 as n -> infinity (division rule)
# Also analyze: loss-per-sample = total_loss / N -> mean_loss as N -> inf
n_sizes = np.array([10, 50, 100, 500, 1000, 5000, 10000])   # Algorithm efficiency ratio
t1 = 5 * n_sizes**2 + 100
t2 = 2 * n_sizes**2 + 50
efficiency_ratio = t1 / t2   # Loss per sample: lim(N->inf) (constant_total + N*per_sample) / N
FIXED_OVERHEAD = 1000.0   # constant startup cost
PER_SAMPLE_LOSS = 0.35   # constant per-sample contribution
total_loss = FIXED_OVERHEAD + PER_SAMPLE_LOSS * n_sizes
loss_per_sample = total_loss / n_sizes   # -> 0.35 as N -> inf (division rule)
results = pd.DataFrame({ "n": n_sizes, "T1": t1, "T2": t2, "T1_T2_ratio": np.round(efficiency_ratio, 6), "total_loss": np.round(total_loss, 2), "loss_per_sample": np.round(loss_per_sample, 6),
})
print("Limit Division Rule: Asymptotic Ratio Analysis")
print("lim T1/T2 = lim(5 + 100/n^2) / (2 + 50/n^2) = 5/2 = 2.5")
print("lim loss/N = lim(1000/N + 0.35) = 0 + 0.35 = 0.35")
print()
print(results.to_string(index=False))
print(f"\nT1/T2 -> 2.5: last value = {efficiency_ratio[-1]:.6f}")
print(f"loss/N -> 0.35: last value = {loss_per_sample[-1]:.6f}")
print("\nConclusion: division rule confirms the asymptotic ratio for both metrics.")

Professional example (R)

# lim T1/T2 and lim loss/N (division rule)
n <- 10^(1:6)
T1 <- 5 + 100/n^2
T2 <- 2 + 50/n^2
efficiency_ratio <- T1 / T2
loss_per_sample <- 1000/n + 0.35
results <- data.frame(n = n, T1 = T1, T2 = T2, T1_over_T2 = efficiency_ratio, loss_per_N = loss_per_sample)
cat("Limit division: T1/T2 -> 5/2, loss/N -> 0.35\n")
print(tail(results, 3))
cat(sprintf("T1/T2 last: %.6f loss/N last: %.6f\n", tail(efficiency_ratio, 1), tail(loss_per_sample, 1)))

Source: AIO p. 44

#algebra#summation-and-limit-manipulations#limits