Method line M17 · When and why do we use it?

Step-by-Step Adjustment

In the Chinese book: 逐步调整法

Freeze all but two variables and move those two to improve, until you cannot.

📖 Chapter 17 of 《解题方法与策略》

One idea, one pictureTap two bars, then tap “Closer”. Does the sum change? What happens to the sum of squares? Keep adjusting until nothing more can be done.

Loading figure…

In one sentence

With many variables, thinking about all of them at once is too hard. Step-by-step adjustment says: freeze the other variables, move just two, and see which move makes the target better.

This is because and . As long as two numbers are “not close enough”, you can improve one more step; the state where no further adjustment is possible is the optimal one.

Predict first

positive integers add up to . What is the smallest possible sum of their squares?

Choose an answer first; then you can see the result.

Example: x₁ + x₂ + x₃ + x₄ = 20 (positive integers). Find the minimum of the sum of squares

Step 1 / 4
Loading figure…
  1. Move only two

    Pick the largest, , and the smallest, , and move each one step toward the middle: , . The sum does not change.

Stop 1 · Proofs of the mean inequalities (Vol. 6)

Adjustment gives one of the most intuitive proofs of the mean inequality. The key is to make the process end after finitely many steps.

Example: positive numbers x₁, …, xₙ have sum nA. Then x₁x₂⋯xₙ ≤ Aⁿ

Step 1 / 4
  1. If they are not all equal to A

    Then the largest and the smallest .

Stop 2 · Extreme values: splitting a number for the largest product (Vol. 6)

When the number of terms is not fixed, an adjustment is not only “move two numbers closer”: you can also “split” or “merge”.

Loading figure…
Four positive integers with sum : moving them closer makes the product larger, up to . But if the number of terms is free, can it get even larger?

Example: write 20 as a sum of positive integers. What is the largest possible product?

Step 1 / 4
  1. Merge any 1

    : merging a into another number makes the product larger. So the optimum has no .

Stop 3 · Trigonometric inequalities (Vol. 3)

One problem, several solutions

In , prove that .

Step-by-step adjustment

: with fixed, the smaller is, the larger the sum. If the three angles are not all , then (after relabeling) . Replace by and : the sum of the angles does not change, and , so increases. Now one angle is and the other two add up to ; one more adjustment makes them all . So the maximum is reached by the equilateral triangle: .

CostUses only sum-to-product, and reaches the equilateral triangle in at most two steps.

Jensen's inequality

is concave on , so .

CostOne line, but you must first know that sin x is concave on (0, π).

Stop 4 · Ordering and swapping (Vol. 1)

“Take the best arrangement; if a swap could still improve it…”: adjustment and the extremal principle often appear together.

Example: a₁ ≤ a₂ ≤ ⋯ ≤ aₙ and b₁ ≤ ⋯ ≤ bₙ. Over all permutations of the b’s, when is Σ aᵢb_{σ(i)} largest?

Step 1 / 4
  1. Take a largest one

    There are only finitely many permutations, so some makes the sum largest.

Practice

Exercise 1Basic

positive integers add up to . What is the smallest possible sum of their squares?

Exercise 2Basic

positive integers add up to . What is the smallest possible sum of their squares?

Exercise 3Advanced

Some positive integers (any number of them) add up to . What is the largest possible product?

Exercise 4Challenge

Some positive integers add up to . When their product is as large as possible, it equals . Find .

Mistake Museum · Exhibits from this stop

Before you leave, look at the traps other people fell into here.See every exhibit in the museum →

Exhibit

An adjustment with no end

The wrong step: To prove that positive numbers with a fixed sum have the largest product when they are all equal: each time, replace the largest and the smallest by their average. The product grows, and if you keep going they all become equal.
What went wrong? Think first, then open

Diagnosis With real numbers, “replace by the average” may never finish. For example becomes in one step, but becomes , then again… and every step leaves new unequal numbers. Whether they are equal after infinitely many steps needs a limit argument; you cannot just claim it.

The right step: Instead, make one number exactly equal to the average at each step: replace the largest and the smallest by and . Each step adds at least one more number equal to , so it ends after at most steps.

“An adjustment must be sure to end after finitely many steps”

Exhibit

Pushing integers toward “all equal” too

The wrong step: Write as a sum of positive integers. The sum of squares is smallest when all three equal , so the minimum is .
What went wrong? Think first, then open

Diagnosis The variables must be integers, so is not allowed. The end point of the adjustment is “any two numbers differ by at most 1”, not “all equal”.

The right step: Adjust to : any two differ by at most , and moving them closer only swaps them. The minimum is .

“The constraints decide the end point; for integers it is “differ by at most 1””

Stops on this line

The same idea, used once in each of several places.

  1. Proofs of the Mean InequalitiesVol. 6 The Mean Inequalities and the Cauchy Inequality, §1.2
  2. Extrema with AM–GMVol. 6 The Mean Inequalities and the Cauchy Inequality, §2.2
  3. Trigonometric InequalitiesVol. 3 Trigonometric Functions, ch. 5
  4. The Extremal PrincipleVol. 1 Sets, ch. 8