๐Ÿงฎ Brain Teaser

Why ฯ€ Is Less Than 22/7

Evaluate

Iโ€…โ€Š=โ€…โ€Šโˆซ01x4(1โˆ’x)41+x2โ€‰dx.I \;=\; \int_0^1 \frac{x^4(1-x)^4}{1+x^2}\,dx .

Then:

  1. Explain why your answer proves the inequality ฯ€<227\pi < \dfrac{22}{7}.
  2. (Bonus) With one more line of work, show
11260โ€…โ€Š<โ€…โ€Š227โˆ’ฯ€โ€…โ€Š<โ€…โ€Š1630.\frac{1}{1260} \;<\; \frac{22}{7}-\pi \;<\; \frac{1}{630}.

Hint to resist: Try to predict the remainder of x4(1โˆ’x)4x^4(1-x)^4 modulo x2+1x^2+1 without doing long division.


Field: Integration / Real Analysis (with a pinch of algebra)

Why it is interesting / classic: This is Putnam 1968 A1, and the identity goes back to D. P. Dalzell (1944). A positive integrand on (0,1)(0,1) produces exactly 22/7โˆ’ฯ€22/7 - \pi. That turns the schoolbook approximation of ฯ€\pi into a one-line rigorous inequality. The numerator looks arbitrary, but the exponent 44 is the reason everything is rational except for a single ฯ€\pi.

Difficulty: 2 / 5

Tags: polynomial division, evaluation at ii, rational approximations of ฯ€\pi, Beta integral

polynomial divisionevaluation at ipi approximationBeta integralPutnam

Answer: Why ฯ€ Is Less Than 22/7

Key Idea / Intuition

Dividing a polynomial by 1+x21+x^2 leaves a polynomial quotient, which integrates to a rational number, plus a remainder over 1+x21+x^2, which produces arctanโก\arctan and hence ฯ€\pi. The remainder is the polynomial's value at x=ix=i, and (1โˆ’i)4=โˆ’4(1-i)^4 = -4 is a real constant. So the remainder is just โˆ’4-4, and the "ฯ€\pi part" of the integral is exactly โˆ’4โ‹…ฯ€4=โˆ’ฯ€-4\cdot\frac{\pi}{4} = -\pi. The integrand is visibly positive, so (rational)โˆ’ฯ€>0(\text{rational}) - \pi > 0. The rational number turns out to be 22/722/7.


Formal Proof / Solution

Step 1: Find the remainder without long division. Write p(x)=x4(1โˆ’x)4=q(x)(x2+1)+r(x)p(x) = x^4(1-x)^4 = q(x)(x^2+1) + r(x) with degโกrโ‰ค1\deg r \le 1. Substituting x=ix = i gives

r(i)=p(i)=i4(1โˆ’i)4=((1โˆ’i)2)2=(โˆ’2i)2=โˆ’4.r(i) = p(i) = i^4 (1-i)^4 = \big((1-i)^2\big)^2 = (-2i)^2 = -4 .

Since rr has real coefficients, r(i)=ai+b=โˆ’4r(i) = a i + b = -4 forces a=0a = 0 and b=โˆ’4b = -4. So the remainder is the constant โˆ’4-4.

Step 2: Find the quotient. Expand the numerator:

p(x)=x8โˆ’4x7+6x6โˆ’4x5+x4.p(x) = x^8 - 4x^7 + 6x^6 - 4x^5 + x^4 .

Then

x4(1โˆ’x)41+x2=x6โˆ’4x5+5x4โˆ’4x2+4โ€…โ€Šโˆ’โ€…โ€Š41+x2.\frac{x^4(1-x)^4}{1+x^2} = x^6 - 4x^5 + 5x^4 - 4x^2 + 4 \;-\; \frac{4}{1+x^2}.

Check: (x6โˆ’4x5+5x4โˆ’4x2+4)(x2+1)=x8โˆ’4x7+6x6โˆ’4x5+x4+4(x^6 - 4x^5 + 5x^4 - 4x^2 + 4)(x^2+1) = x^8 - 4x^7 + 6x^6 - 4x^5 + x^4 + 4. Subtracting 44 gives p(x)p(x). โœ“

Step 3: Integrate.

I=17โˆ’46+55โˆ’43+4โˆ’4arctanโก(1)=17โˆ’23+1โˆ’43+4โˆ’ฯ€=227โˆ’ฯ€.I = \frac17 - \frac46 + \frac55 - \frac43 + 4 - 4\arctan(1) = \frac17 - \frac23 + 1 - \frac43 + 4 - \pi = \frac{22}{7} - \pi .

Step 4: The inequality. On (0,1)(0,1) the integrand x4(1โˆ’x)41+x2\dfrac{x^4(1-x)^4}{1+x^2} is continuous, non-negative, and not identically zero. Hence I>0I > 0, which gives

ฯ€<227.\pi < \frac{22}{7}.

Step 5 (Bonus): Bracketing the error. On [0,1][0,1] we have 1โ‰ค1+x2โ‰ค21 \le 1+x^2 \le 2, with strict inequalities on the interior. Therefore

12โˆซ01x4(1โˆ’x)4โ€‰dxโ€…โ€Š<โ€…โ€ŠIโ€…โ€Š<โ€…โ€Šโˆซ01x4(1โˆ’x)4โ€‰dx.\frac12 \int_0^1 x^4(1-x)^4\,dx \;<\; I \;<\; \int_0^1 x^4(1-x)^4\,dx .

The remaining integral is a Beta integral:

โˆซ01x4(1โˆ’x)4โ€‰dx=B(5,5)=4!โ€‰4!9!=576362880=1630.\int_0^1 x^4(1-x)^4\,dx = B(5,5) = \frac{4!\,4!}{9!} = \frac{576}{362880} = \frac{1}{630}.

Hence

11260<227โˆ’ฯ€<1630,\frac{1}{1260} < \frac{22}{7} - \pi < \frac{1}{630},

which gives 3.14127โ€ฆ<ฯ€<3.14206โ€ฆ3.14127\ldots < \pi < 3.14206\ldots.


Remarks

  • Why the exponent 4? You need (1โˆ’i)n(1-i)^n to be real. Since 1โˆ’i=2โ€‰eโˆ’iฯ€/41-i = \sqrt2\,e^{-i\pi/4}, this happens exactly when 4โˆฃn4 \mid n. With other exponents the remainder has an xx-term, and integrating it produces a lnโก2\ln 2 alongside ฯ€\pi.
  • Generalization (Lucas, 2005): The integrals โˆซ01x4n(1โˆ’x)4n1+x2โ€‰dx\displaystyle\int_0^1 \frac{x^{4n}(1-x)^{4n}}{1+x^2}\,dx all equal (rational) ยฑโ€‰ฯ€\pm\,\pi, up to a power-of-4 factor. As nn grows they give increasingly sharp rational approximations to ฯ€\pi.
  • Sources: Putnam 1968, Problem A1; D. P. Dalzell, J. London Math. Soc. 19 (1944).

Source: Putnam 1968 A1; D. P. Dalzell (1944)

Type: IntegrationSource: Putnam 1968 A1; D. P. Dalzell (1944)Edit on GitHub โ†—