6.2 Means, AM–GM and Special Sums
The arithmetic mean of positive numbers is never smaller than their geometric mean — a one-line inequality that finds least and greatest values without calculus. With the standard sums $\sum k$, $\sum k^2$, $\sum k^3$ and the telescoping trick, it completes the toolkit for series in JEE Main.
A share price rises by $50\%$ in one year and falls by $50\%$ the next. The average of $+50\%$ and $-50\%$ is $0\%$, so has the investor broken even? Start with $\text{₹}\,100$: after the rise it is $\text{₹}\,150$, after the fall $\text{₹}\,75$ — a loss of a quarter. The growth factors were $1.5$ and $0.5$; their arithmetic mean is $1$, but the factor that, applied twice, gives the same result is their geometric mean $\sqrt{1.5\times0.5}\approx0.866$. The geometric mean came out smaller, and that is no accident: for positive numbers it is always so. This one fact, the A.M.–G.M. inequality, answers a whole family of questions — the largest product of numbers with a fixed sum, the smallest sum of numbers with a fixed product, the cheapest box of a given volume — without any calculus.
The second half of the section is about adding. Some sums turn up again and again: $1+2+\dots+n$, $1^2+2^2+\dots+n^2$ and $1^3+2^3+\dots+n^3$. Once you know these three, any series whose $k$-th term is a polynomial in $k$ can be summed in a few lines. Other series collapse when each term is written as a difference — the telescoping trick. In JEE Main these ideas appear as questions on least values, on two numbers with a given A.M. and G.M., on sums such as $\sum k(k+1)$, and on series of fractions that telescope.
The A.M.–G.M. Inequality
For positive numbers $a_1, a_2,\dots,a_n$ the arithmetic mean and the geometric mean are
$$A=\frac{a_1+a_2+\dots+a_n}{n},\qquad G=\sqrt[n]{a_1a_2\cdots a_n}.$$For $n=2$ these are the $A=\tfrac{a+b}2$ and $G=\sqrt{ab}$ of Section 6.1.
For positive numbers $a$ and $b$ with A.M. $A$ and G.M. $G$:
- $A-G=\dfrac{\left(\sqrt a-\sqrt b\right)^2}{2}\ge0$, so $A\ge G$, with equality if and only if $a=b$.
- $a$ and $b$ are the roots of $x^2-2Ax+G^2=0$, that is, $a,b=A\pm\sqrt{A^2-G^2}$.
Proof. (1) $\dfrac{a+b}{2}-\sqrt{ab}=\dfrac{a-2\sqrt a\sqrt b+b}{2}=\dfrac{\left(\sqrt a-\sqrt b\right)^2}{2}$, a square divided by $2$, so it is $\ge0$, and it is $0$ exactly when $\sqrt a=\sqrt b$. (2) A quadratic with roots $a$ and $b$ is $x^2-(a+b)x+ab=x^2-2Ax+G^2$. $\blacksquare$ Part (2) turns "find two numbers with A.M. $A$ and G.M. $G$" into a quadratic equation — and the discriminant $4(A^2-G^2)\ge0$ is the inequality again.
There is also a picture. On a diameter of length $a+b$, mark the point $P$ that splits it into pieces $a$ and $b$, and draw the perpendicular from $P$ to the semicircle. The radius is $\tfrac{a+b}2=A$. The two triangles cut off by that perpendicular are similar (the angle in a semicircle is a right angle), and give $\tfrac ah=\tfrac hb$, so the perpendicular has length $h=\sqrt{ab}=G$. A half-chord can never be longer than the radius: $G\le A$, with equality only when $P$ is the centre. Interactive 1 lets you slide $P$.
For positive numbers $a_1,\dots,a_n$,
$$\frac{a_1+a_2+\dots+a_n}{n}\ \ge\ \sqrt[n]{a_1a_2\cdots a_n},$$with equality if and only if $a_1=a_2=\dots=a_n$.
Proof. We use one fact about logarithms: $\ln x\le x-1$ for every $x\gt0$, with equality only at $x=1$ (the graph of $\ln x$ lies below its tangent line $y=x-1$ at $x=1$; you will prove this with derivatives in Chapter 7). Let $A$ be the A.M. and put $x=\tfrac{a_i}{A}$ for each $i$:
$$\ln\frac{a_i}{A}\le\frac{a_i}{A}-1\qquad(i=1,\dots,n).$$Add the $n$ inequalities. The right-hand sides add to $\tfrac{a_1+\dots+a_n}{A}-n=n-n=0$, so $\ln\dfrac{a_1a_2\cdots a_n}{A^n}\le0$, that is, $a_1\cdots a_n\le A^n$. Taking $n$-th roots, $G\le A$. Equality needs equality in every line, i.e. $a_i=A$ for every $i$. $\blacksquare$
Two positive numbers have sum $20$. The greatest possible value of their product is
Least and Greatest Values with A.M.–G.M.
Read Theorem 6.9 in two directions. If the product $a_1\cdots a_n=P$ is fixed, then $a_1+\dots+a_n\ge n\sqrt[n]{P}$: the sum is least when all the numbers are equal. If the sum $a_1+\dots+a_n=S$ is fixed, then $a_1\cdots a_n\le\left(\tfrac Sn\right)^n$: the product is greatest when all the numbers are equal. The skill is to arrange the expression so that one of the two is constant.
- Check that every quantity you will average is positive.
- To find a least sum, split it into terms whose product is constant: $x^2+\tfrac{16}{x}=x^2+\tfrac8x+\tfrac8x$ (product $64$). To find a greatest product, adjust factors so that their sum is constant: $x^2(6-2x)=x\cdot x\cdot(6-2x)$ (sum $6$).
- Apply A.M. $\ge$ G.M. to get the bound.
- Find where all the terms are equal, and check that this point is allowed. If it is not, the bound is not the answer.
Keep $a+b=10$ fixed and move the split point. Which split makes the G.M. $\sqrt{ab}$ as large as possible, and how large?
For $x\gt0$, the least value of $x^2+\dfrac2x$ is
Sigma Notation and the Standard Sums
$\displaystyle\sum_{k=m}^{n}a_k$ means $a_m+a_{m+1}+\dots+a_n$: the letter $k$ (the index) runs through the integers from $m$ to $n$, giving $n-m+1$ terms. The numbers $T_n=1+2+\dots+n$ are called triangular numbers ($1, 3, 6, 10, 15,\dots$) — the number of dots in a triangle with $n$ rows.
For constants $\alpha,\beta,c$:
$$\sum_{k=1}^{n}(\alpha a_k+\beta b_k)=\alpha\sum_{k=1}^{n}a_k+\beta\sum_{k=1}^{n}b_k,\qquad\sum_{k=1}^{n}c=nc,$$ $$\sum_{k=m}^{n}a_k=\sum_{k=1}^{n}a_k-\sum_{k=1}^{m-1}a_k.$$These are just regrouping a finite sum. Warning: there is no such rule for products — $\sum a_kb_k$ is not $\left(\sum a_k\right)\left(\sum b_k\right)$, and $\sum k^2$ is not $\left(\sum k\right)^2$.
Proof. The first is the sum of an A.P. (Theorem 6.3). For the second, use the identity $(k+1)^3-k^3=3k^2+3k+1$ and add it for $k=1,\dots,n$. On the left everything cancels except $(n+1)^3-1^3$. So
$$(n+1)^3-1=3\sum k^2+3\cdot\frac{n(n+1)}2+n,$$and solving, $3\sum k^2=(n+1)^3-1-n-\tfrac32n(n+1)$ $=\tfrac{(n+1)\left(2n^2+n\right)}{2}$, which gives $\sum k^2=\tfrac{n(n+1)(2n+1)}6$. For the cubes, notice that $k^2(k+1)^2-(k-1)^2k^2$ $=k^2\big[(k+1)^2-(k-1)^2\big]=4k^3$. Adding for $k=1,\dots,n$, the left side collapses to $n^2(n+1)^2-0$, so $4\sum k^3=n^2(n+1)^2$. $\blacksquare$ (Each formula can also be checked by mathematical induction.) With Theorem 6.10, any sum whose $k$-th term is a polynomial in $k$ of degree at most $3$ can now be evaluated: expand, split, and use the three formulas.
What is $1^3+2^3+3^3+4^3+5^3+6^3$?
If $1+2+\dots+n=36$, then $1^3+2^3+\dots+n^3$ equals
Telescoping Sums and the Method of Differences
The proof of $\sum k^3$ worked because each term was written as a difference $f(k)-f(k-1)$, and in the sum all the middle values cancelled. That is a general method.
If the terms can be written as $a_k=f(k+1)-f(k)$ for some function $f$, then
$$\sum_{k=1}^{n}a_k=f(n+1)-f(1).$$Standard splits (for $k\ge1$):
$$\frac{1}{k(k+1)}=\frac1k-\frac1{k+1},\qquad\frac1{(2k-1)(2k+1)}=\frac12\left(\frac1{2k-1}-\frac1{2k+1}\right),$$ $$\frac{1}{k(k+1)(k+2)}=\frac12\left(\frac1{k(k+1)}-\frac1{(k+1)(k+2)}\right),\qquad\frac1{\sqrt k+\sqrt{k+1}}=\sqrt{k+1}-\sqrt k.$$Proof. In $\big(f(2)-f(1)\big)+\big(f(3)-f(2)\big)+\cdots$ $+\big(f(n+1)-f(n)\big)$ each of $f(2),\dots,f(n)$ appears once with $+$ and once with $-$. Each split is checked by putting the right side over a common denominator; the last one by rationalising, $\tfrac{\sqrt{k+1}-\sqrt k}{(k+1)-k}$. $\blacksquare$ In general,
$$\frac1{(a+kd)(a+(k+1)d)}=\frac1d\left(\frac1{a+kd}-\frac1{a+(k+1)d}\right)$$— the factor in front is $\tfrac1{\text{difference of the two factors}}$.
The method of differences. If a series is not an A.P. or a G.P., write down the differences of consecutive terms. If they form an A.P., the $n$-th term is a quadratic $pn^2+qn+r$ (three unknowns, found from the first three terms); if the differences form a G.P. with ratio $\rho$, the $n$-th term is $p\rho^n+q$. Then sum with Theorems 6.5 and 6.11. For example $2, 5, 10, 17, 26,\dots$ has differences $3, 5, 7, 9,\dots$ (an A.P.), and indeed $a_n=n^2+1$.
1. A.M. $\ge$ G.M. needs positive numbers. For $x\lt0$, $x+\tfrac1x\le-2$, not $\ge2$. State positivity before you use the inequality.
2. The equality case must be attainable. On $x\ge3$, the least value of $x+\tfrac1x$ is not $2$ (that needs $x=1$); the expression increases for $x\ge1$ and the answer is $3+\tfrac13=\tfrac{10}3$.
3. Make the product (or the sum) constant first. $x^2+\tfrac{16}x\ge2\sqrt{16x}$ is true but useless; split into $x^2+\tfrac8x+\tfrac8x$.
4. Limits of $\Sigma$. $\sum_{k=11}^{20}=\sum_1^{20}-\sum_1^{10}$, not $-\sum_1^{11}$; and from $k=m$ to $n$ there are $n-m+1$ terms. $\sum c=nc$, not $c$.
5. Squares and cubes. $\sum k^2\ne\left(\sum k\right)^2$, but $\sum k^3=\left(\sum k\right)^2$. In a telescope, write the first two and last two terms to see which survive: in $\sum\left(\tfrac1k-\tfrac1{k+2}\right)$ two terms survive at each end.
The sum to infinity $\displaystyle\sum_{k=1}^{\infty}\frac{1}{k(k+2)}$ equals
Harmonic progressions and the harmonic mean are listed in the JEE Advanced syllabus but not in the JEE Main 2026 list. The arithmetico-geometric sum below is a standard JEE Advanced technique built on the G.P. method of Theorem 6.5.
Harmonic progression and harmonic mean. Non-zero numbers are in harmonic progression (H.P.) if their reciprocals are in A.P.: $1, \tfrac12, \tfrac13,\dots$ is an H.P. The harmonic mean of positive $a$ and $b$ is $H=\dfrac{2ab}{a+b}$ (the reciprocal of the A.M. of $\tfrac1a$ and $\tfrac1b$). Since $AH=\tfrac{a+b}2\cdot\tfrac{2ab}{a+b}=ab=G^2$, the three means satisfy
$$A\ge G\ge H\qquad\text{and}\qquad G^2=AH,$$with equality throughout only when $a=b$. Average speed over two equal distances travelled at speeds $u$ and $v$ is the harmonic mean $\tfrac{2uv}{u+v}$, not $\tfrac{u+v}2$.
Arithmetico-geometric series. For $S=1+2x+3x^2+\dots+nx^{n-1}$ ($x\ne1$), multiply by $x$ and subtract, exactly as for a G.P.: $(1-x)S=1+x+\dots+x^{n-1}-nx^n$, so $S=\dfrac{1-x^n}{(1-x)^2}-\dfrac{nx^n}{1-x}$. For $|x|\lt1$, letting $n\to\infty$ gives $1+2x+3x^2+\cdots=\dfrac{1}{(1-x)^2}$; for example $\sum_{k\ge1}\tfrac{k}{2^{k-1}}=4$.
Exercises
All letters stand for positive real numbers unless a part says otherwise. "Least" and "greatest" values must be attained — say where.
- 1 Find the A.M. and the G.M. of $4$ and $16$.
- 2 Find the A.M. and the G.M. of $2$, $8$ and $32$.
- 3 The A.M. of two positive numbers is $15$ and their G.M. is $9$. Find the numbers.
- 4 The A.M. and G.M. of two positive numbers are in the ratio $13:12$. Find the ratio of the larger number to the smaller.
- 5 Prove that $\dfrac ab+\dfrac ba\ge2$, and say when equality holds.
- 6 Prove that $a^2+b^2+c^2\ge ab+bc+ca$ for all real $a, b, c$.
- 7 If $a+b+c=12$, show that $abc\le64$.
- 8 The G.M. of two positive numbers is $6$ and their A.M. exceeds the G.M. by $1.5$. Find the numbers.
- 1 Find the least value of $4x+\dfrac{25}{x}$ for $x\gt0$.
- 2 Find the least value of $9^x+9^{1-x}$, $x\in\mathbb{R}$.
- 3 Find the least value of $\dfrac{x^2+2x+16}{x}$ for $x\gt0$.
- 4 Find the greatest value of $x(12-2x)$ for $0\lt x\lt6$.
- 5 If $x+y=6$, find the greatest value of $xy^2$.
- 6 If $xy=24$, find the least value of $2x+3y$.
- 7 Find the least value of $\left(x+\dfrac1x\right)\left(y+\dfrac1y\right)$.
- 8 A rectangular plot of area $400\ \text{m}^2$ is to be fenced on all four sides. Find the least length of fencing needed.
- 9 An open-top box with a square base must hold $32\ \text{m}^3$. Find the least total area of sheet metal (base plus four sides) and the dimensions of the box.
- 1 Evaluate $\displaystyle\sum_{k=1}^{25}k$.
- 2 Evaluate $\displaystyle\sum_{k=1}^{12}k^2$.
- 3 Evaluate $\displaystyle\sum_{k=1}^{9}k^3$.
- 4 Find $\displaystyle\sum_{k=1}^{n}\left(3k^2-2k+1\right)$ in closed form.
- 5 Evaluate $2^2+4^2+6^2+\dots+40^2$.
- 6 Show that $\displaystyle\sum_{k=1}^{n}k(k+1)(k+2)=\frac{n(n+1)(n+2)(n+3)}{4}$, and evaluate it for $n=10$.
- 7 Find $n$ if $1^3+2^3+\dots+n^3=14400$.
- 8 Show that $\displaystyle\sum_{k=1}^{n}k(n-k)=\frac{(n-1)n(n+1)}{6}$, and evaluate it for $n=10$.
- 9 A fruit seller stacks oranges in a square pyramid: $1$ on top, $4$ in the layer below, $9$ below that, and so on, with $15$ layers. How many oranges are there?
- 1 Show that $\displaystyle\sum_{k=1}^{n}\frac1{k(k+1)}=\frac{n}{n+1}$, and evaluate it for $n=49$.
- 2 Find $\displaystyle\sum_{k=1}^{n}\frac1{(3k-2)(3k+1)}$, and its value for $n=10$.
- 3 Evaluate $\displaystyle\sum_{k=1}^{99}\frac1{\sqrt k+\sqrt{k+1}}$.
- 4 Find the sum to infinity $\dfrac1{1\cdot3}+\dfrac1{3\cdot5}+\dfrac1{5\cdot7}+\cdots$
- 5 Show that $\displaystyle\sum_{k=1}^{n}k\cdot k!=(n+1)!-1$, and evaluate it for $n=5$.
- 6 Find the $n$-th term and the sum of the first $n$ terms of $2+5+10+17+26+\cdots$; evaluate $S_{10}$.
- 7 The differences of consecutive terms of $4, 6, 10, 18, 34,\dots$ form a G.P. Find the $n$-th term and $S_8$.
- 8 Show that $\displaystyle\sum_{k=1}^{n}\frac{2k+1}{k^2(k+1)^2}=1-\frac{1}{(n+1)^2}$, and evaluate it for $n=9$.
E1. JEE Main · MCQ For $x\gt0$, the least value of $\dfrac{x^2+3x+9}{x}$ is (A) $6$ (B) $3+2\sqrt3$ (C) $9$ (D) $12$
E2. JEE Main · Numerical Find the value of $1\cdot3+2\cdot5+3\cdot7+\dots+20\cdot41$. Give the answer to the nearest integer.
E3. JEE Main · MCQ $\displaystyle\sum_{k=1}^{20}\frac{1}{(2k-1)(2k+1)}$ equals (A) $\tfrac{40}{41}$ (B) $\tfrac{21}{41}$ (C) $\tfrac{19}{39}$ (D) $\tfrac{20}{41}$
E4. JEE Main · Numerical The arithmetic mean of two positive numbers is $13$ and their geometric mean is $12$. Find the larger number. Give the answer to the nearest integer.
E5. JEE Main · MCQ If $1^3+2^3+\dots+n^3=3025$, then $1^2+2^2+\dots+n^2$ equals (A) $506$ (B) $385$ (C) $285$ (D) $650$
E6. JEE Main · Numerical Positive numbers $x$ and $y$ satisfy $x+2y=12$. Find the greatest value of $xy^2$. Give the answer to the nearest integer.
Research question. A fund advertises its "average annual return". Is that the arithmetic mean of its yearly returns or the geometric mean of its growth factors — and how much does the choice change what an investor actually earns?
Why it matters. Pension funds, mutual funds and national statistics offices all report average growth. If the yearly growth factors are $g_1,\dots,g_n$, the money actually grows by $g_1g_2\cdots g_n=G^n$, where $G$ is their geometric mean. The arithmetic mean is always at least as large (Theorem 6.9), so quoting it flatters a volatile investment. The gap is known in finance as "volatility drag".
What to do.
- Collect ten consecutive yearly returns of a broad stock-market index of any country (published by the exchange or a central bank), or use these illustrative values: $+18\%, -6\%, +25\%, +3\%, -12\%, +30\%, +8\%, -20\%, +15\%, +10\%$.
- Convert them to growth factors ($+18\%\to1.18$). Find the arithmetic mean $A$ and the geometric mean $G$ of the factors.
- Start with $\text{₹}\,1$ lakh and compute the final value year by year. Compare it with $1\times A^{10}$ and $1\times G^{10}$ lakh. Which one is exact?
- Compute the variance $s^2$ of the factors (mean of the squared deviations from $A$) and compare $A-G$ with $\tfrac{s^2}{2}$. How close are they?
Tools: A spreadsheet or calculator.
Hand in: A table of the yearly factors and values, $A$, $G$, the three final values, and two sentences explaining to a first-time investor why the advertised "average return" can be misleading.
Going further. Two funds have the same arithmetic mean return, but one swings much more from year to year. Using A.M. $\ge$ G.M. and its equality case, explain which one grows more in the long run.
🧪 Self-check quiz
Q1. The A.M. of two positive numbers is $25$ and their G.M. is $7$. The numbers are
Q2. For $x\gt0$, the least value of $x+\dfrac1x$ is
Q3. $1^2+2^2+\dots+10^2$ equals
Q4. $\displaystyle\sum_{k=1}^{99}\frac{1}{k(k+1)}$ equals
Q5. For every $n\in\mathbb{N}$, $1^3+2^3+\dots+n^3$ equals