Ma · Mathematics

How to Prove Inequalities for Infima and Suprema

Published By the 9to5Science editorial team

At a glance

  1. 1. Check boundedness

    Work with a nonempty domain and functions whose infima and suprema are finite.

  2. 2. Compare pointwise

    Use a pointwise bound to relate the functions before taking infima or suprema.

  3. 3. Take extrema carefully

    An inequality holding everywhere gives the corresponding inequality between infima or suprema.

  4. 4. Apply the bound

    For two functions, bound their infimum difference by the supremum of their pointwise difference.

The title does not state which inequality needs proving. A standard result that fits the topic is this: if ff and gg are bounded real-valued functions on a nonempty set XX, then the difference between their infima is controlled by the supremum of their pointwise absolute differences. The same argument also gives useful inequalities for sums and suprema.

The basic inequality

For bounded functions f,g:X→Rf,g:X\to\mathbb{R}, define

δ=sup⁡x∈X∣f(x)−g(x)∣.\delta=\sup_{x\in X}|f(x)-g(x)|.

Then

∣inf⁡x∈Xf(x)−inf⁡x∈Xg(x)∣≤δ.|\inf_{x\in X}f(x)-\inf_{x\in X}g(x)|\leq \delta.

In words, if ff and gg never differ by more than δ\delta, their infima cannot differ by more than δ\delta. The corresponding statement for suprema is

∣sup⁡x∈Xf(x)−sup⁡x∈Xg(x)∣≤δ.|\sup_{x\in X}f(x)-\sup_{x\in X}g(x)|\leq \delta.

The assumptions ensure that the infima and suprema are finite real numbers. If you work with extended real values, additional care is needed when subtracting infinities.

Proof for infima

By definition of δ\delta, for every x∈Xx\in X,

∣f(x)−g(x)∣≤δ.|f(x)-g(x)|\leq\delta.

Thus f(x)≥g(x)−δf(x)\geq g(x)-\delta for every xx. Since g(x)≥inf⁡Xgg(x)\geq\inf_X g for every xx, it follows that

f(x)≥inf⁡Xg−δfor every x∈X.f(x)\geq\inf_X g-\delta\qquad\text{for every }x\in X.

The number inf⁡Xg−δ\inf_X g-\delta is therefore a lower bound for the values of ff. The greatest lower bound of those values must be at least as large, so

inf⁡Xf≥inf⁡Xg−δ.\inf_X f\geq\inf_X g-\delta.

Interchanging ff and gg gives inf⁡Xg≥inf⁡Xf−δ\inf_X g\geq\inf_X f-\delta. Together, these inequalities say that each infimum is no more than δ\delta below the other, which is exactly

∣inf⁡Xf−inf⁡Xg∣≤δ.|\inf_X f-\inf_X g|\leq\delta.

For the supremum version, use f(x)≤g(x)+δf(x)\leq g(x)+\delta and then reverse the roles of the functions. This yields sup⁡Xf≤sup⁡Xg+δ\sup_X f\leq\sup_X g+\delta and sup⁡Xg≤sup⁡Xf+δ\sup_X g\leq\sup_X f+\delta.

Worked example

Take X=[0,1]X=[0,1], f(x)=xf(x)=x, and g(x)=x+0.4g(x)=x+0.4. For every xx,

∣f(x)−g(x)∣=0.4,|f(x)-g(x)|=0.4,

so δ=0.4\delta=0.4. The infima are inf⁡Xf=0\inf_X f=0 and inf⁡Xg=0.4\inf_X g=0.4. Hence

∣inf⁡Xf−inf⁡Xg∣=∣0−0.4∣=0.4=δ.|\inf_X f-\inf_X g|=|0-0.4|=0.4=\delta.

The bound is attained in this example. The suprema are 11 and 1.41.4, and their absolute difference is also 0.40.4.

The same pointwise reasoning gives standard bounds for two bounded functions on the same nonempty set:

inf⁡Xf+inf⁡Xg≤inf⁡X(f+g),\inf_X f+\inf_X g\leq\inf_X(f+g),

and

sup⁡X(f+g)≤sup⁡Xf+sup⁡Xg.\sup_X(f+g)\leq\sup_X f+\sup_X g.

For the first inequality, every xx satisfies f(x)≥inf⁡Xff(x)\geq\inf_X f and g(x)≥inf⁡Xgg(x)\geq\inf_X g. Adding gives f(x)+g(x)≥inf⁡Xf+inf⁡Xgf(x)+g(x)\geq\inf_X f+\inf_X g; take the infimum over xx. For the second, add the pointwise upper bounds and take the supremum.

These need not be equalities. For example, on X={1,2}X=\{1,2\}, let f(1)=0f(1)=0, f(2)=1f(2)=1, g(1)=1g(1)=1, and g(2)=0g(2)=0. Then inf⁡Xf+inf⁡Xg=0\inf_X f+\inf_X g=0, while f(x)+g(x)=1f(x)+g(x)=1 at both points, so inf⁡X(f+g)=1\inf_X(f+g)=1. The two functions attain their minima at different points.

Common mistakes

  • Treating an infimum as a minimum. An infimum need not be attained. The proof uses lower-bound properties, not a point where the function reaches its infimum.
  • Reversing an inequality when taking an infimum. If a(x)≤b(x)a(x)\leq b(x) for every xx, then inf⁡Xa≤inf⁡Xb\inf_X a\leq\inf_X b. For suprema, the same pointwise order gives sup⁡Xa≤sup⁡Xb\sup_X a\leq\sup_X b.
  • Assuming the sum inequality is always an equality. Equality can fail when the functions reach their extrema at different inputs.
  • Forgetting the domain. Infima and suprema here are taken over the same nonempty set XX. If the domains differ, first specify how the functions are being compared.

In a metric-space setting, the same result says that the map taking a bounded function to its infimum changes by at most the uniform distance between functions. In particular, if the uniform distance is defined by d∞(f,g)=sup⁡x∈X∣f(x)−g(x)∣d_\infty(f,g)=\sup_{x\in X}|f(x)-g(x)|, then ∣inf⁡Xf−inf⁡Xg∣≤d∞(f,g)|\inf_X f-\inf_X g|\leq d_\infty(f,g). No monotonicity assumption is needed. If a different inequality was intended, its exact expression is necessary to prove that particular statement.

Spotted an error? Send a correction. Read how we write and check articles in our editorial policy.