Knowing the definition of a supremum is only half the battle. The examination challenge in CSIR NET, GATE, and IIT JAM is to actually compute it and then prove your answer rigorously. This post gives you a complete, systematic toolkit: three standard techniques covering virtually every set you will encounter in a real analysis course.
Definitions and the ε-Characterisation
Let $S \subseteq \mathbb{R}$, $S \neq \emptyset$, bounded above. A number $M \in \mathbb{R}$ is the supremum of $S$, written $M = \sup S$, if:
(i) $s \leq M$ for all $s \in S$, and
(ii) if $\gamma < M$ then there exists $s \in S$ with $\gamma < s$.
$m \in \mathbb{R}$ is the infimum of $S$, written $m = \inf S$, if:
(i) $m \leq s$ for all $s \in S$, and
(ii) if $\gamma > m$ then $\gamma$ is not a lower bound of $S$.
Key relation: $\inf S = -\sup(-S)$ where $-S = \{-s : s \in S\}$.
$M = \sup S$ if and only if:
(i) $s \leq M$ for all $s \in S$, and
(ii) for every $\varepsilon > 0$, there exists $s_\varepsilon \in S$ such that $s_\varepsilon > M - \varepsilon$.
Dually, $m = \inf S$ iff $m \leq s$ for all $s \in S$, and for every $\varepsilon > 0$ there exists $s_\varepsilon \in S$ with $s_\varepsilon < m + \varepsilon$.
Rudin, Principles of Mathematical Analysis, 3rd Ed., Ch. 1, §1.7–§1.11
Bartle & Sherbert, Introduction to Real Analysis, 4th Ed., Ch. 2, §2.3
→ Bounds and Extrema of Sets — upper/lower bounds and the formal definition of sup/inf
→ Order Relations and Ordered Sets — ordered fields and the completeness axiom
→ Foundations of Real Numbers — the Archimedean property used in proofs
Three Standard Techniques & Historical Context
There are three standard strategies for computing and proving $\sup S$ or $\inf S$. Mastering these three covers virtually every problem in the CSIR NET and IIT JAM syllabi.
When to use: The set is defined by a simple formula and the supremum is a boundary value not in $S$.
- Manipulate the defining inequality to identify the candidate $M$.
- Verify $M$ is an upper bound ($s \leq M$ for all $s \in S$).
- For each $\varepsilon > 0$, explicitly construct $s_\varepsilon \in S$ with $s_\varepsilon > M - \varepsilon$ (use the Archimedean property to find $n_0$).
When to use: $S$ is the range of a monotone sequence on $\mathbb{N}$ and the sup/inf is a limit.
- Show the sequence is monotone and bounded.
- Identify the candidate limit $L$ by taking $n \to \infty$.
- Show $L$ is an upper bound and that elements get arbitrarily close to $L$.
When to use: All upper bounds can be characterised explicitly.
- Show $M$ is an upper bound.
- Suppose $M' < M$. Derive a contradiction by exhibiting $s \in S$ with $s > M'$, showing $M'$ is not an upper bound.
"The creation of the concept of irrational numbers was necessary to fill the gaps left by the rational numbers on the number line."
— Richard Dedekind, Continuity and Irrational Numbers, 1872
Dedekind's Insight (1872). The least-upper-bound property was the key gap Dedekind identified when he formalised $\mathbb{R}$. In $\mathbb{Q}$, the set $S = \{q \in \mathbb{Q} : q^2 < 2\}$ is bounded above (e.g., by 2) yet has no supremum in $\mathbb{Q}$ — the missing least upper bound is $\sqrt{2} \notin \mathbb{Q}$. By formally adjoining all such "cuts," Dedekind produced the complete ordered field $\mathbb{R}$ where every bounded non-empty set has a supremum. This is the axiomatic bedrock on which all three techniques above rest.
Why it matters for CSIR NET / IIT JAM. The ε-characterisation of supremum appears in nearly every convergence proof, Riemann sum argument, and continuity theorem in analysis. Exam questions routinely require both identifying the sup/inf and providing a rigorous proof.
A bounded set $S \subset \mathbb{R}$: $\inf S$ is attained (solid dot); $\sup S = M$ is not attained (open circle). The point $s_\varepsilon \in S$ satisfying $s_\varepsilon > M - \varepsilon$ demonstrates the ε-characterisation.
Solved Examples
1Easy — Technique 1
Find $\sup S$ and $\inf S$ for $S = \left\{\dfrac{n}{n+1} : n \in \mathbb{N}\right\}$.
Step 1 — Candidate. Write $\dfrac{n}{n+1} = 1 - \dfrac{1}{n+1}$. As $n \to \infty$, values increase toward $1$. So conjecture $\sup S = 1$. The minimum is at $n=1$: $\frac{1}{2}$, so $\inf S = \frac{1}{2}$.
Step 2 — Prove $\sup S = 1$. For all $n \in \mathbb{N}$, $\frac{n}{n+1} < 1$, so $1$ is an upper bound. Given $\varepsilon > 0$, choose $n_0 \in \mathbb{N}$ with $n_0 > \frac{1}{\varepsilon} - 1$ (Archimedean). Then $\frac{n_0}{n_0+1} = 1 - \frac{1}{n_0+1} > 1 - \varepsilon$. Hence $\sup S = 1 \notin S$.
Step 3 — $\inf S = \frac{1}{2}$. The sequence is strictly increasing, so $\frac{1}{2} = \min S \in S$.
2Medium — Technique 1
Find $\sup S$ for $S = \{x \in \mathbb{R} : x^2 < 5\}$.
Step 1. $x^2 < 5 \Leftrightarrow -\sqrt{5} < x < \sqrt{5}$, so $S = (-\sqrt{5}, \sqrt{5})$. Conjecture $\sup S = \sqrt{5}$.
Step 2 — Upper bound. For all $x \in S$, $x < \sqrt{5}$. ✓
Step 3 — ε-characterisation. Given $\varepsilon > 0$ (take $\varepsilon < 2\sqrt{5}$), set $s_\varepsilon = \sqrt{5} - \frac{\varepsilon}{2}$. Then $s_\varepsilon^2 = 5 - \sqrt{5}\,\varepsilon + \frac{\varepsilon^2}{4} < 5$, so $s_\varepsilon \in S$, and $s_\varepsilon > \sqrt{5} - \varepsilon$. ✓
3Medium–Hard — Technique 3 (Contrapositive)
Prove $\sup\!\left\{\dfrac{2n-1}{n} : n \in \mathbb{N}\right\} = 2$.
Upper bound. $\dfrac{2n-1}{n} = 2 - \dfrac{1}{n} < 2$ for all $n \in \mathbb{N}$. So $2$ is an upper bound.
Contrapositive. Suppose $M' < 2$, say $M' = 2 - \delta$, $\delta > 0$. By the Archimedean property, choose $n_0 \in \mathbb{N}$ with $n_0 > 1/\delta$. Then $\dfrac{2n_0-1}{n_0} = 2 - \dfrac{1}{n_0} > 2 - \delta = M'$. So $M'$ is not an upper bound. No number less than $2$ is an upper bound.
4Hard — CSIR NET / IIT JAM Level
Let $A = \left\{\dfrac{(-1)^n \cdot n}{n+2} : n \in \mathbb{N}\right\}$. Find $\sup A$ and $\inf A$.
Step 1 — Separate even and odd terms. For even $n = 2k$: $a_{2k} = \dfrac{2k}{2k+2} = \dfrac{k}{k+1}$. For odd $n = 2k-1$: $a_{2k-1} = -\dfrac{2k-1}{2k+1}$.
Step 2 — Even subsequence. $\dfrac{k}{k+1} = 1 - \dfrac{1}{k+1} \nearrow 1$ as $k \to \infty$. By the same ε-argument as Example 1, $\sup\{a_{2k}\} = 1$ (not attained).
Step 3 — Odd subsequence. $-\dfrac{2k-1}{2k+1} = -1 + \dfrac{2}{2k+1} \searrow -1$ as $k \to \infty$. So $\inf\{a_{2k-1}\} = -1$ (not attained). By symmetry with the ε-argument: for any $\varepsilon > 0$, choose $k_0$ with $\frac{2}{2k_0+1} < \varepsilon$, giving $a_{2k_0-1} < -1 + \varepsilon$.
Quick Revision Cards
A — The ε-Test
$M = \sup S$ iff $M$ is UB and $\forall\varepsilon>0,\;\exists\,s\in S: s > M-\varepsilon$
$m = \inf S$ iff $m$ is LB and $\forall\varepsilon>0,\;\exists\,s\in S: s < m+\varepsilon$
Use Archimedean property to find $n_0 > \frac{1}{\varepsilon}$
B — Algebraic Candidates
Rewrite $s_n$ as $L \pm f(n)$ where $f(n)\to 0$
$L$ is the candidate sup (if $-f$) or inf (if $+f$)
Check monotonicity to confirm min/max
$\inf S = -\sup(-S)$
C — Attainment & Exam Tips
🔵 CSIR NET: ε-proof always required
🟢 GATE: identify technique quickly
🟠 IIT JAM: state whether sup/inf attained
🔴 B.Sc.: distinguish sup from max
Open intervals: sup and inf not in $S$
Closed bounded intervals: both attained
Common Mistakes
1. Confusing limit with supremum. A sequence may approach $L$ without $L$ being the supremum of its range. Always verify the upper bound condition and the ε-condition separately — both are required.
2. Skipping the upper-bound verification. The ε-condition alone does not prove $M = \sup S$. You must also show $s \leq M$ for all $s \in S$.
3. Assuming $\sup S \in S$. The supremum may not belong to $S$ (e.g., $\sup(0,1) = 1 \notin (0,1)$). Always explicitly state whether the supremum is attained.
4. Wrong Archimedean choice. When constructing $n_0$, ensure the strict inequality $\frac{1}{n_0} < \varepsilon$ holds — choose $n_0 > \frac{1}{\varepsilon}$ explicitly.
5. Sign error with $-S$. $\sup(-S) = -\inf S$ and $\inf(-S) = -\sup S$. Forgetting the negation is a frequent error in exam proofs when using the infimum from a supremum.
Real-World Applications
Every numerical solver seeks the $\sup$ or $\inf$ of an objective function. The Extreme Value Theorem guarantees a continuous function on $[a,b]$ attains its supremum.
Darboux upper and lower sums are suprema and infima of $f$ over subintervals. The Riemann integral is defined via $\inf_P U(f,P) = \sup_P L(f,P)$.
The operator norm $\|T\| = \sup\{\|Tx\| : \|x\|=1\}$ is a supremum construction central to Banach space theory and quantum mechanics.
IEEE 754 floating-point overflow triggers when a value exceeds the supremum of representable numbers. Algorithm analysis uses infima of running times.
Summary Table
| Technique | Best For | Key Tool | Attainment |
|---|---|---|---|
| 1 — ε-Characterisation | Algebraic/formula sets; open boundary | Archimedean property; solve $s_\varepsilon > M-\varepsilon$ | Usually $\notin S$ |
| 2 — Monotone Limit | Sequences; range of monotone function | Monotone Convergence; limit computation | Limit $\notin$ range |
| 3 — Contrapositive | Any $M$ where UBs are characterisable | Suppose $M' < M$; exhibit $s > M'$ | Depends on set |
| Direct Minimum | Finite sets; closed intervals | Show $m \in S$ and $m \leq s$ for all $s$ | Yes — $m \in S$ |
Cross-References
→ Bounds and Extrema of Sets — the prerequisite post covering upper/lower bounds and the formal definitions of sup/inf
→ Order Relations and Ordered Sets — the ordered field structure and the completeness axiom
→ Foundations of Real Numbers — introduces $\mathbb{R}$ and the Archimedean property
→ Absolute Value and the Real Line — the $\lvert x\rvert \leq M$ boundedness criterion
→ Intervals and the Real Line — bounded intervals as concrete examples of bounded sets
Have a question, doubt, or thought about this post? Choose how you want to join the discussion below.
💬 Comment on Telegram — No account neededRequires a free GitHub account — takes 30 seconds to create with just an email address.