Intuition The one core idea
An average is a single number that stands in for a whole list — but there is more than one honest way to "stand in for" numbers, depending on whether you care about their sum , their product , or their rates . This page builds every symbol you need — sums, products, roots, reciprocals, subscripts, and the "greater-than-or-equal" sign itself — so that when the parent proof lines these means up in order, not one squiggle is a stranger.
Everything below is a prerequisite for the AM–GM–HM proofs . We start from nothing.
Definition Positive real, and the symbol
> 0
A real number is any point on the number line — whole numbers, fractions, decimals, 2 , everything. We write a > 0 to mean "a sits strictly to the right of zero" — it is genuinely positive, not zero, not negative.
Why does the whole topic insist on this? Because we are about to take square roots and reciprocals (1/ a ). A square root of a negative number is not a point on our line, and 1/0 does not exist. So the ticket to enter this topic is: all numbers strictly positive . Every proof quietly leans on this.
Worked example What figure s01 shows
The number line below marks 0 and paints the region to its right in mint — that is the "allowed" zone a > 0 where every operation in this topic is legal. The coral region to the left is forbidden: there live the negative inputs that break square roots, and 1/0 that breaks reciprocals. Whenever you feel unsure "am I allowed to do this?", picture landing in the mint zone.
Definition Addition, and subtraction as its inverse
Addition p + q means "start at p on the number line and step q units to the right ." Subtraction p − q is the inverse (the "undo") of addition: it steps q units to the left , answering "what number, when you add q back, returns to p ?" Symbolically: ( p − q ) + q = p .
We call this out early because the master trick of the whole topic (Section 8) starts from a subtraction, a − b . The topic needs a solid grip on "− undoes + " before that step makes sense.
Definition Multiplication, division, and the power notation
x n
Multiplication p × q is repeated addition ("q copies of p added up"). Its inverse is division p ÷ q = q p , answering "what number, times q , gives p ?"
The little raised number in x n — the exponent — is a counter for repeated multiplication :
x n = n copies x × x × ⋯ × x .
So x 2 = x × x ("x squared"), x 3 = x × x × x , and 2 n means 2 multiplied by itself n times.
Intuition Why the raised counter is worth its own symbol
Writing x × x × x × x × x is clumsy; x 5 says the same thing and makes the counter (5 ) visible, so we can reason about it. Just as × is a shorthand for repeated + , the exponent x n is a shorthand for repeated × . Each layer up the tower (+ → × → power ) has its own dedicated undo: subtraction, division, and — coming in Section 5 — the root .
Worked example Warm-up on powers
p 2 = p × p ; and 2 3 = 2 × 2 × 2 = 8 , so doubling a list of 3 numbers scales their product by 2 3 = 8 .
a i and the ellipsis …
When we have several numbers and don't want to invent new letters (a , b , c , d , … runs out fast), we reuse one letter a and tag it with a little number below: a 1 ("a-one"), a 2 ("a-two"), and so on. The tag i is a counter : a placeholder that walks through 1 , 2 , 3 , …
The three dots … (the ellipsis ) mean "continue the obvious pattern, one step at a time, up to the stated last term" — so a 1 , a 2 , … , a n is the complete list from box 1 to box n , with every box in between present even though we don't write them all. It is a space-saver, never a vague "and so on forever": the pattern and its final stopping point n are both pinned down.
a 1 = the first number, a 2 = the second , a n = the last .
n = how many numbers there are in total.
a i = "the i -th one, whichever that is."
i and the total n are positive whole numbers
Both i and n are positive integers (1 , 2 , 3 , … ) — you cannot have a "box number 2.5 " or a "box number 0 ". In this topic we always take n ≥ 1 (at least one number in the list), and for the means to be interesting, usually n ≥ 2 . An empty list (n = 0 ) is deliberately excluded: with no numbers there is nothing to average, and — as the next two sections note — an empty ∑ and an empty ∏ have special conventional values that we simply never need here.
Think of a row of labelled boxes. Each box holds a number; the label under the box is its index. "a i " points at box number i without committing to a specific one — it lets one formula speak about all boxes at once.
The topic needs this because AM, GM, HM are all statements about a whole list a 1 , … , a n . Without subscripts you could only ever talk about two or three specific numbers.
Definition Sigma notation
The tall symbol i = 1 ∑ n a i is shorthand for
∑ i = 1 n a i = a 1 + a 2 + ⋯ + a n ,
where the ⋯ is the same ellipsis as in Section 3: every term from a 1 up to a n , added, with the in-between ones understood.
Read it: "start the counter i at 1 , add a i , bump i up by one, repeat until i = n ."
The ∑ (Greek capital S , for S um) is just a machine that runs the counter and adds. The bottom (i = 1 ) says where to start, the top (n ) says where to stop. Since n ≥ 1 (Section 3), the machine always runs at least once — we never meet an "empty sum" here.
If the list is a 1 = 3 , a 2 = 5 , a 3 = 4 then n = 3 and
∑ i = 1 3 a i = 3 + 5 + 4 = 12.
This is the raw material of the Arithmetic Mean : sum everything, then share it out. It also directly connects to Arithmetic Progressions , where you first met sums of listed terms.
The tall symbol i = 1 ∏ n a i means multiply instead of add:
∏ i = 1 n a i = a 1 × a 2 × ⋯ × a n ,
again the ⋯ standing for every factor from a 1 to a n , all present.
∏ is Greek capital P (for P roduct) — the multiplication twin of ∑ . Same counter, same start/stop, but the operation is × not + . Again n ≥ 1 , so there is always at least one factor — no "empty product" to puzzle over.
With a 1 = 3 , a 2 = 5 , a 3 = 4 : i = 1 ∏ 3 a i = 3 × 5 × 4 = 60.
This is the raw material of the Geometric Mean , and it is the natural language of Geometric Progressions , where terms grow by multiplying.
∑ and ∏ are not the same shape of idea
Why it confuses: both are "combine the whole list into one number."
The catch: ∑ scales gently (double every number → double the sum), but ∏ scales violently (double every one of n numbers → multiply the product by 2 n , using the power notation from Section 2). That difference is exactly why AM and GM feel the numbers differently.
n -th root
The square root x answers: "what positive number, multiplied by itself, gives x ?" — it undoes squaring (x 2 ). More generally the n -th root, written x 1/ n or n x , answers: "what positive number, multiplied by itself n times, gives x ?" — it undoes the power x n from Section 2:
( x 1/ n ) n = x .
The raised 1/ n is a fractional exponent: it is defined to be precisely this "undo the n -th power" operation.
root is the right undo for a product
The GM multiplies n numbers together, giving something roughly "n copies big." To bring it back down to the size of a single number, we must undo "n copies of multiplication" — and that is exactly what the n -th root does. It sits at the top of the tower of undos: subtraction undoes addition, division undoes multiplication, and the root undoes the power.
Worked example What figure s02 shows
The picture traces the GM pipeline left to right on the sample list 3 , 5 , 4 : three lavender number-boxes flow into a "multiply (∏ )" arrow producing the butter-coloured product 60 , which then flows through a "cube root" arrow into the mint GM box ≈ 3.91 . Read it as a factory: product first, root second — the root is the final step that shrinks the big product back down to one number's size.
The GM formula ( a 1 ⋯ a n ) 1/ n is literally: take the product, then take the n -th root . And notice — the root needs its input to be positive (Section 0), or we'd be asking for the square root of a negative.
The reciprocal of a is a 1 : the number that multiplies with a to give 1 . Big numbers have tiny reciprocals; tiny numbers have huge reciprocals.
Intuition The seesaw picture
Imagine the number line with 1 marked. Reciprocating is a seesaw pivoting at 1 : numbers above 1 swap to below 1 , and vice versa. 2 ↔ 2 1 , 5 ↔ 5 1 , 3 1 ↔ 3 . This "flip" is the entire personality of the Harmonic Mean .
Worked example What figure s03 shows
The curve y = 1/ x is drawn in lavender, with dotted guide-lines at the pivot x = 1 , y = 1 . Two coloured pairs make the "flip" visible: the coral marks show 2 (above 1 ) landing at 2 1 (below 1 ), and the mint marks show 2 1 landing back at 2 . The picture is here so you can see , not just be told, that reciprocating swaps big and small — which is exactly why it reverses inequalities in the next callout.
The HM is built by: flip every number, average the flips, then flip back. That's why HM lives in the world of Harmonic Progression and rate problems (speed, work, resistance) — anything measured per unit of something.
Definition Reading the crocodile
a ≥ b : "a is greater than or equal to b " — a is at least as big.
a ≤ b : "a is less than or equal to b ."
iff = "if and only if" — a two-way street. "P iff Q " means P forces Q and Q forces P .
The whole topic is an inequality: AM ≥ GM ≥ HM . The word iff shows up in the equality condition ("equality iff all numbers are equal") — meaning the means coincide exactly when the numbers are all identical, and only then.
Definition A square is never negative
For any real number t , t 2 ≥ 0 (recall t 2 = t × t from Section 2). Squaring wipes out the sign: ( − 3 ) 2 = 9 = 3 2 . The result is zero only when t = 0 .
Intuition Why this humble fact is the engine of the whole topic
Every "≥ " in the proofs is secretly a disguised "( something ) 2 ≥ 0 ." For two numbers, the trick is ( a − b ) 2 ≥ 0 : it is obviously true, and when we expand it we get exactly AM ≥ GM.
plus and minus, times and power
subscripts a_i and count n
Once these are solid you are ready for Weighted Means (Power Mean Inequality) , Cauchy-Schwarz Inequality , Jensen's Inequality , and Optimization using inequalities — all of which reuse exactly these bricks.
Cover the right-hand side and test yourself.
What does a > 0 guarantee we're allowed to do? Take square roots and reciprocals safely — no undefined operations.
What operation undoes addition, and what undoes multiplication? Subtraction undoes addition; division undoes multiplication.
What does the exponent in x n count? The number of copies of x multiplied together: x n = x × x × ⋯ × x (n times).
What precisely does the ellipsis … in a 1 , … , a n mean? Continue the obvious pattern one step at a time, every box present, up to the stated last term a n — not "forever".
What kind of numbers are the counter i and total n , and what is ruled out? Positive integers (1 , 2 , 3 , … ); the empty list n = 0 is excluded — nothing to average.
What does the subscript a i let one formula do? Talk about a whole list of numbers at once, using a counter i .
Expand i = 1 ∑ n a i in words. a 1 + a 2 + ⋯ + a n — add every term from the first to the last.
Expand i = 1 ∏ n a i in words. a 1 × a 2 × ⋯ × a n — multiply every term.
What question does x 1/ n answer, and which operation does it undo? Which positive number multiplied by itself n times gives x ? — it undoes the power x n .
Why is a b = ab for positive a , b ? Squaring
a b gives
ab , and it is positive, so it must be
ab .
On the reciprocal seesaw, where does 2 land? At 2 1 — above 1 flips to below 1 .
If a ≤ b (both positive), what happens to a 1 vs b 1 ? The inequality flips: a 1 ≥ b 1 .
How does ( p − q ) 2 expand, and why? p 2 − 2 pq + q 2 , from multiplying ( p − q ) × ( p − q ) term by term.
Expand ( a − b ) 2 ≥ 0 — what inequality pops out? a − 2 ab + b ≥ 0 , i.e.
a + b ≥ 2 ab (that's AM
≥ GM).
What does "AM = GM iff all equal" claim in both directions? All numbers equal forces the means equal, AND the means equal forces the numbers equal.
Which single fact powers every ≥ in the proofs? A square is never negative: t 2 ≥ 0 , with equality only at t = 0 .