Limit & Series (Numerical) – Tutorial

Limit, Sequence Limit, and Series – Numerical Approximation Guide (University Level)

This calculator uses numerical approximation instead of analytic solutions. It finds approximate values rather than exact ones. This approach is especially useful for:

• Quick checks on limits that are hard to compute by hand,
• Observing sequence/series convergence,
• Detecting numerical instabilities that hinder convergence (undefined points, overflow, cancellation errors)

---

1) How is a limit (x → a) computed numerically?

1.1 Core idea

A limit asks: What value does f(x) approach as x approaches a? Numerically, we do:

• Right-side approach: x = a + h
• Left-side approach: x = a − h
• We decrease h progressively: 10⁻¹, 10⁻², 10⁻³, ...

If values from left and right converge to the same number, the limit exists. If they don't converge (large difference, oscillation, diverging to infinity), we say the limit is indeterminate or does not exist.

1.2 Why can numerical limits sometimes be wrong?

• Undefined point: f(a) may be undefined even though the limit exists. (e.g.: sin(x)/x, a=0)
• Cancellation error: Subtracting two very close numbers loses precision. E.g.: in (1 - cos(x)) / x² at very small x, errors grow.
• Floating point rounding: Computers store numbers with limited precision. When h gets very small, values become unstable.
• Oscillation: Functions like sin(1/x) keep changing as x→0; limit does not exist.

1.3 What does this tool do for limits?

• Generates an approximation sequence h = scale·10⁻ᵏ (scale = max(1, |a|)).
• Computes left, right, or both sides depending on selection.
• Marks undefined points as "undefined" without crashing.
• Generates a status report: "converging?" based on the last few values.
• For two-sided: checks whether left and right results are consistent.

---

2) How is a sequence limit (n → ∞) analyzed numerically?

2.1 Core idea

For a sequence limit: as n grows large, what value does a(n) approach? Numerically:

• We select n values (small to very large)
• We compute a(n)
• We assess whether the last values are "stable"

2.2 Why log-spaced sampling?

Going n=1,2,3,4,5... can be misleading because as n grows, changes become smaller. This tool selects n values with logarithmic spacing (e.g.: 1, 2, 5, 10, 20, 50, 100, ...). This gives a better view of convergence.

2.3 If the sequence limit "appears to not exist"?

• If values keep growing → ∞ / −∞ (divergence)
• If values keep oscillating → limit may not exist (e.g. (−1)ⁿ)
• If values converge very slowly → try increasing nMax

---

3) How is a series sum (∑ a(n)) approximated numerically?

3.1 Partial sums

A series sum is understood as:

S = a(1)+a(2)+...+a(n)+...

Numerical methods cannot compute infinite sums directly. Instead:

• A partial sum S_N = ∑(n=n0..N) a(n) is computed
• If S_N settles to a value as N grows, the series converges

3.2 "Term test" (necessary condition)

For a series to converge, it is necessary that:

a(n) → 0

If a(n) does not approach zero, the series definitely diverges. However, a(n)→0 is not sufficient for convergence. (e.g. the harmonic series 1/n → 0 but diverges)

3.3 Why do numerical summation errors occur?

• Rounding error: Adding very small terms to a large sum loses precision.
• Overflow: If terms grow too large, the number system returns "Infinity".
• Slow convergence: Series like 1/n converge very weakly and can be misleading.

3.4 What does this tool do for series?

• Uses Kahan summation for partial sums (reduces rounding errors).
• If the change falls below tolerance several times in a row, it declares "converging" and may stop early.
• Optionally applies Aitken Δ² acceleration (can help with slowly converging series).

---

4) Expression syntax (important!)

• For exponents: use ^ (e.g.: 1/n^2)
• Functions: sin, cos, tan, exp, ln, log, sqrt, abs
• Constants: pi, e
• Variables:
  • In limit mode: x
  • In sequence/series mode: n

Examples

• Limit: sin(x)/x, a=0
• Sequence: (1+1/n)^n
• Series: 1/n^2

This tool gives approximate results. For definitive conclusions about limits/series/sequences, analytical methods (L'Hôpital's rule, comparison tests, integral test, ratio test, etc.) are required.