JC1 H3 Mathematics Tutor Bukit Timah | Bernoulli’s Inequality and Inductive Bounds. Did you know? You can prove that a very large power exceeds a simple straight-line expression without expanding any of its many terms. One general argument establishes the result for every positive integer exponent.
This guide is for JC1 learners in Bukit Timah, Sixth Avenue and Beauty World who want a strong H2 Mathematics foundation for possible later H3 study. Bernoulli’s inequality offers a friendly but rigorous way to practise induction, nonnegative factors, equality conditions and a theorem’s domain. The student’s junior college determines H3 subject eligibility; this article does not assert a dedicated JC1 H3 class is running.
The official 2027 Singapore–Cambridge H3 Mathematics 9820 syllabus includes proof by induction, non-routine problem solving and additional inequalities. Bernoulli’s inequality is used here as enrichment demonstrating those named capabilities, not claimed as a separately named syllabus subtopic. The worked questions are original teaching examples.
Quick answer: the theorem and its domain
- For every positive integer n and real x≥−1, Bernoulli’s inequality states (1+x)ⁿ≥1+nx.
- The expression 1+nx is a lower bound, not an exact calculation of the power in general.
- For n=1, equality holds for every eligible x. For n≥2, equality occurs at x=0.
- The inequality proof relies on 1+x≥0 so multiplication preserves the inequality direction.
- The n=0 case can be added when 1+x>0; using n≥1 throughout avoids a convention-sensitive 0⁰ at x=−1.
The full inductive proof
Base case. At n=1 the claim reads 1+x≥1+x, true as equality for every real x≥−1. This gives a clear starting point without asking a learner to interpret 0⁰ at the endpoint x=−1.
Induction hypothesis. Fix a real x≥−1. Assume the inequality holds at an arbitrary positive integer k: (1+x)ᵏ≥1+kx. The index k changes during the induction; the chosen x remains fixed.
Inductive step. Since 1+x≥0, multiply the hypothesis by 1+x without reversing the inequality. Then (1+x)ᵏ⁺¹≥(1+kx)(1+x)=1+(k+1)x+kx². The term kx² is nonnegative, hence (1+x)ᵏ⁺¹≥1+(k+1)x.
Conclusion. A verified base case and a valid transition prove the result for every positive integer n and every real x≥−1. The proof is independent of calculator approximations, and the domain condition is essential to the multiplication step.
Worked examples: understand what the inequality really guarantees
Worked example 1: The base exponent n=1
For n=1, (1+x)¹ and 1+1×x are identical, so equality holds everywhere in the permitted domain. This is a useful first check but also an actual complete induction base case.
A proof must show the proposition for its starting index, not merely announce that it is obvious. Choosing n=1 avoids an unnecessary debate about the expression 0⁰ that can occur at n=0 and x=−1.
Worked example 2: The special case n=2
Expand (1+x)²=1+2x+x². Since x²≥0 for every real x, the expression is at least 1+2x. Equality occurs exactly when x=0.
This simple calculation gives students intuition for the general proof. The extra square cannot lower the left side. It also illustrates how to determine an equality case rather than just checking a numerical inequality.
Worked example 3: The positive-input case
Take x=0.1,n=4. The theorem gives (1.1)⁴≥1.4, while direct calculation gives 1.4641. Both statements are compatible because 1.4 is a lower bound, not the exact result.
A learner who replaces the ≥ symbol by = has changed a true theorem into a false numerical claim. The role of each comparison sign is part of mathematical communication, not a formatting detail.
Worked example 4: A two-percent increase
Let x=0.02,n=5. Bernoulli yields (1.02)⁵≥1.10. The exact power is approximately 1.1040808, strictly larger than the lower bound.
This is a mathematical comparison of repeated factors. It does not predict a real financial outcome because it assumes the same multiplicative factor in every period, with no fees or other changes.
Worked example 5: One hundred factors
For x=1/100,n=100, the expression (101/100)¹⁰⁰ is at least 1+100(1/100)=2. The inequality obtains this comparison without expanding a hundred factors.
This is why a general proof has more value than numerical testing: a result verified for arbitrary n can be applied immediately to large instances. The conclusion is a guaranteed lower bound, not equality.
Worked example 6: A negative but permitted change
For x=−0.1,n=5, Bernoulli gives (0.9)⁵≥1−0.5=0.5. Direct calculation gives 0.59049, so the inequality holds.
Negative x is not automatically disallowed. What matters is x≥−1, keeping the base 1+x nonnegative. However, a lower bound may become loose when many negative changes accumulate.
Worked example 7: The endpoint x=−1
At x=−1 and n≥1, the left side is zero while the right side is 1−n. For n=1 both are zero; for n≥2 the right side is negative, so the inequality is strict.
This is an important boundary check. It is valid for the positive-integer theorem, but using n=0 at x=−1 would introduce 0⁰, so the article deliberately states the clean n≥1 formulation.
Worked example 8: When a bound becomes uninformative
Take x=−0.2,n=10. The lower bound is 1−2=−1. Since (0.8)¹⁰ is nonnegative, the theorem is correct yet less useful than the obvious bound of zero.
A mathematically valid bound need not approximate the actual value closely. Advanced problem solving requires students to judge both the truth and the usefulness of an inequality.
Worked example 9: A fraction rather than a decimal
Choose x=1/4,n=4. The exact left side is (5/4)⁴=625/256, while the right side equals 2=512/256. Since 625≥512, the theorem is confirmed.
Exact fractions preserve mathematical information and avoid unnecessary rounding. The example is a check, not the proof for all n and x, which is supplied by induction.
Worked example 10: A numerical example outside the domain
Suppose someone claims the theorem is valid for all real x, including x<−1. Take x=−4 and n=3: the left side is (−3)³=−27 while the proposed bound is 1−12=−11.
Since −27 is not at least −11, the unrestricted claim fails. The counterexample does not contradict the correctly stated theorem; it demonstrates why x≥−1 must not be removed without a new argument.
Worked example 11: Why multiplication needs a sign check
The inductive hypothesis is an inequality. Multiplying both sides by 1+x keeps the direction only when 1+x≥0, exactly the stated domain condition.
If 1+x is negative, the inequality reverses and the original induction chain is invalid. Many inequality proofs go wrong at this step, even when their algebraic expansion is completely accurate.
Worked example 12: How to recognise the leftover square
Multiplication yields (1+kx)(1+x)=1+(k+1)x+kx². The term kx² is nonnegative for positive integer k and real x.
The inductive result follows by dropping this positive-or-zero term from a lower bound. Students who lose the kx² term without explaining its sign have an incomplete proof, despite writing the correct final inequality.
Worked example 13: Equality for n≥2
At n=2 the difference between the two sides is x², so equality holds only at x=0. The induction step also preserves strictness away from zero, with x=−1 handled directly.
Consequently for positive integer n≥2, the equality point on x≥−1 is x=0. This is stronger information than merely knowing the lower bound. It matters in optimisation questions.
Worked example 14: Turn the theorem into an optimisation result
For a fixed n≥2, define D(x)=(1+x)ⁿ−1−nx on x≥−1. Bernoulli says D(x)≥0, and at x=0 the difference is exactly zero.
Thus D has minimum value zero, attained at x=0. A bound becomes an actual minimum only when a permitted equality witness has been identified. Do not confuse a lower bound with an automatically attained optimum.
Worked example 15: A simple binomial expansion on nonnegative x
For x≥0 and integer n≥2, expand (1+x)ⁿ=1+nx+C(n,2)x²+… . Every remaining term is nonnegative, so (1+x)ⁿ≥1+nx+C(n,2)x².
This is a refinement of Bernoulli’s inequality on a smaller domain. When x is negative, the signs of higher powers can alternate, so the same term-dropping reasoning must not be transferred carelessly.
Worked example 16: The five-term quadratic refinement
For x=0.02 and n=5, the quadratic lower bound is 1+5(0.02)+10(0.02)²=1.104. The exact power is approximately 1.1040808.
The refined bound is substantially closer to the exact result than 1.10. This shows how retaining a further nonnegative binomial term can improve a rigorous estimate.
Worked example 17: An upper bound on a restricted domain
For 0≤x≤1, every power xʳ with integer r≥1 satisfies xʳ≤x. Therefore the binomial expansion yields (1+x)ⁿ≤1+(2ⁿ−1)x.
Combined with Bernoulli, we get 1+nx≤(1+x)ⁿ≤1+(2ⁿ−1)x. The upper estimate requires x≤1, an additional restriction. At x=1, that upper bound is attained.
Worked example 18: A product of two nonnegative increments
For x,y≥0, (1+x)(1+y)=1+x+y+xy≥1+x+y. The extra product xy is nonnegative.
This is the two-factor version of keeping only linear terms. If x and y have opposite signs, xy is negative and the argument fails. Correct proof-writing keeps the sign condition visible.
Worked example 19: Extend to many nonnegative increments
For x₁,…,xₙ≥0, expanding ∏(1+xᵢ) gives one plus the sum of individual xᵢ and many products of at least two such variables.
Every omitted product is nonnegative, so ∏(1+xᵢ)≥1+∑xᵢ. This is a useful general product bound; its proof depends on nonnegative inputs rather than a numerical coincidence.
Worked example 20: A false fractional-exponent extension
For x=3 and exponent r=1/2, the proposed inequality (1+x)ʳ≥1+rx would say 2≥2.5, which is false.
Bernoulli’s positive-integer version cannot be applied to arbitrary fractional exponents. For 0
Extensions, rigorous comparisons and proof diagnostics
Extended worked example 1: A lower bound for eight periods
If the assumed multiplier is 1.05 for eight periods, Bernoulli gives (1.05)⁸≥1.40. Therefore it is certainly greater than 1.35 without expanding the eighth power. This is a productive application of a lower bound: the lower bound itself already exceeds the threshold that the question requires us to establish.
Extended worked example 2: When a lower bound is too weak
For (1.02)⁵, Bernoulli yields only 1.10. That cannot settle whether the actual power exceeds 1.103. The stronger binomial bound for x≥0 gives 1+5(0.02)+10(0.02)²=1.104, which does settle it. The lesson is to choose a bound strong enough for the actual decision.
Extended worked example 3: A variable fraction inside the power
For t≥0 and n≥1, set x=t/n in Bernoulli. Then (1+t/n)ⁿ≥1+t. The substitution works because t/n≥0, which lies inside the permitted domain. This yields a family of comparisons in two variables without calculating a single complicated power. State explicitly that n is positive before dividing by it.
Extended worked example 4: A connection with the exponential limit
For fixed t and sufficiently large n, the familiar limit of (1+t/n)ⁿ equals eᵗ. For those n the Bernoulli lower bound is at least 1+t. Passing to the limit gives eᵗ≥1+t. This last step uses a separate established exponential-limit theorem; the integer inequality alone does not define the exponential.
Extended worked example 5: An independent calculus proof
Define g(t)=eᵗ−1−t. We have g′(t)=eᵗ−1, which is negative for t<0, zero at t=0, and positive for t>0. Thus g has global minimum zero at t=0. Hence eᵗ≥1+t for all real t. Comparing this proof with induction reveals two distinct paths to related bounds.
Extended worked example 6: An exponential upper bound
For real x>−1, the well-known inequality ln(1+x)≤x implies n ln(1+x)≤nx when n≥1. Exponentiating gives (1+x)ⁿ≤eⁿˣ. Combined with Bernoulli, we obtain 1+nx≤(1+x)ⁿ≤eⁿˣ. The logarithm requires 1+x>0, so the endpoint x=−1 is excluded from this upper-bound proof.
Extended worked example 7: Prove the logarithm comparison
Let h(x)=x−ln(1+x) for x>−1. Its derivative is x/(1+x), negative for negative x and positive for positive x. Therefore h is minimised at x=0, with h(0)=0. This proves ln(1+x)≤x. Naming the domain and checking the derivative signs make the result self-contained.
Extended worked example 8: Compare three numerical quantities
For x=0.01,n=10, the lower bound is 1.10; the exact power is about 1.104622; and e^0.1 is about 1.105171. The values appear in increasing order, matching the proved lower and upper bounds. A calculator helps illustrate the theorem but should not be mistaken for the proof that the ordering holds.
Extended worked example 9: Real exponents above one
For real r≥1 and x>−1, f(x)=(1+x)ʳ has second derivative r(r−1)(1+x)ʳ⁻²≥0. The function is convex, so it lies above its tangent line at x=0, namely 1+rx. Thus the bound extends to real exponents at least one, by a new calculus argument rather than the original integer induction.
Extended worked example 10: Fractional exponents between zero and one
When 0
Extended worked example 11: What fails below x=−1
The induction proof multiplies by 1+x, which becomes negative if x<−1. That can reverse the inequality. At x=−4,n=3, the left side is −27 and the right side is −11, so the proposed ≥ comparison is false. One admissible counterexample to the expanded domain defeats the careless universal statement.
Extended worked example 12: A large power with a tiny increment
For n=1000,x=0.001, Bernoulli gives (1.001)¹⁰⁰⁰≥2. The strength of the method is that the inequality is valid for the entire domain, including huge powers. A direct decimal calculation may be interesting, but it cannot replace the general proof or alter the guarantee.
Extended worked example 13: Scale a variable power inequality
For t>0 and integer n≥1, put x=1/t. Bernoulli yields (1+1/t)ⁿ≥1+n/t. Multiplying both sides by positive tⁿ gives (t+1)ⁿ≥tⁿ+ntⁿ⁻¹. At t=0 the substitution is not legal; check that boundary separately. Dividing by a parameter always requires examining its zero case.
Extended worked example 14: A difference quotient derived from a lower bound
Starting from (1+x)ⁿ≥1+nx for positive integer n and x≥−1, subtract one and divide by the positive n. The result is [(1+x)ⁿ−1]/n≥x. The divisor’s sign matters: an inequality may be safely divided only after its sign and nonzero nature are known.
Extended worked example 15: A more informative quadratic bound
For x≥0 and n≥2, the binomial theorem gives (1+x)ⁿ≥1+nx+[n(n−1)/2]x². Every omitted higher-degree term is nonnegative. The additional quadratic term makes this bound stronger than Bernoulli. But when x is negative, odd powers may be negative, so the term-dropping proof needs a different argument.
Extended worked example 16: Check the quadratic bound at a small rate
With n=5,x=0.02, the quadratic lower bound equals 1+0.10+10(0.0004)=1.104. Direct evaluation gives approximately 1.1040808. The small remaining gap is consistent with dropping positive cubic and higher-degree terms. This demonstrates a meaningful improvement without pretending the truncated polynomial is an exact power.
Extended worked example 17: An upper estimate for x inside the unit interval
For 0≤x≤1 and every integer k≥1, we have xᵏ≤x. Therefore the binomial expansion yields (1+x)ⁿ≤1+(2ⁿ−1)x. Combine it with Bernoulli to get 1+nx≤(1+x)ⁿ≤1+(2ⁿ−1)x. The upper estimate is not valid for every positive x; the restriction x≤1 does real work.
Extended worked example 18: Sharpness at the interval endpoints
For the upper bound with 0≤x≤1, equality holds at x=0 and x=1. At x=1 both sides equal 2ⁿ because the sum of the positive-degree binomial coefficients is 2ⁿ−1. Identifying equality points shows a bound is sharp, rather than merely correct.
Extended worked example 19: A product bound from nonnegative increments
For x,y≥0, (1+x)(1+y)=1+x+y+xy≥1+x+y. A product of several factors (1+xᵢ) with xᵢ≥0 similarly exceeds or equals one plus the sum of the individual increments, because all higher-degree mixed products are nonnegative. This generalises the lower-bound idea beyond one repeated increment.
Extended worked example 20: A proof of the general product bound
Expand the product ∏(1+xᵢ), with every xᵢ≥0. Every term is a product of some selected xᵢ values, and all such products are nonnegative. Keeping only the empty product one and the single-variable terms gives ∏(1+xᵢ)≥1+∑xᵢ. The combinatorial interpretation of the expansion makes the reasoning transparent.
Extended worked example 21: The right meaning of a lower bound
An inequality such as (1+x)ⁿ≥1+nx says the power can never fall below the given comparison under its assumptions. It does not say the power is close to the comparison, or that the comparison is nonnegative. A useful bound is judged by whether it proves the question being asked.
Extended worked example 22: Combining Bernoulli with nonnegativity
For −1≤x<0, the base 1+x lies between zero and one, so its positive integer power is nonnegative. If 1+nx is negative, the stronger lower bound is max(0,1+nx). Both component bounds are valid; choosing their maximum gives more information than either one alone.
Extended worked example 23: A minimum of the difference function
Fix n≥2 and define D(x)=(1+x)ⁿ−1−nx for real x≥−1. Bernoulli proves D(x)≥0, and the permitted point x=0 yields D(0)=0. Hence D has minimum zero at x=0. This is a complete optimisation proof: a bound plus an attaining input.
Extended worked example 24: Increasing the exponent with positive x
For x≥0, the factor 1+x≥1. Therefore (1+x)ⁿ⁺¹=(1+x)ⁿ(1+x)≥(1+x)ⁿ for n≥1. The powers are nondecreasing in the integer exponent. For −1 For integer n≥2 and x>0, the binomial expansion has positive quadratic and higher contributions. Thus the lower bound 1+nx is strictly less than (1+x)ⁿ. At x=0 the omitted terms vanish. This explains both strictness and the equality case without relying on a numerical pattern. The base case shows the theorem holds for n=1. The inductive step shows that whenever it holds for k it holds for k+1. Either component alone is insufficient. A student who skips the base has not begun the chain; a student who checks only the base has proved only one exponent. If an induction or inequality proof multiplies both sides by a variable expression without checking its sign, the direction may change or the expression may be zero. In Bernoulli, the condition 1+x≥0 is the reason the step works. Include that sentence at the actual multiplication, not only in the opening statement. The identity (1+kx)(1+x)=1+(k+1)x+kx² is an equality. Dropping kx² produces a smaller expression when kx²≥0, so the next relation is ≥, not =. A proof that writes equality after deleting a nonzero term is false even if its final bound is correct. For n=3, the difference (1+x)³−(1+3x)=3x²+x³=x²(3+x). On x≥−1, both factors are nonnegative, and 3+x≥2. This provides an independent proof at that exponent. It is useful for understanding the theorem, but does not replace induction for all n. For n=4, the difference is 6x²+4x³+x⁴=x²(x²+4x+6). The quadratic factor equals (x+2)²+2>0, so the difference is nonnegative for all real x. This special even-exponent fact is stronger in domain than the general theorem; it must not be used to overgeneralise odd exponents. At n=1, every eligible x gives equality; at n≥2, x=0 is the equality input. A proof that mistakenly asserts strict inequality for all x≥−1 fails immediately at x=0. Checking where the omitted correction term vanishes is an efficient way to establish equality conditions. A hypothetical quantity increasing by a fixed r≥0 per period has model factor (1+r)ⁿ≥1+nr. The inequality follows rigorously from the mathematical model. It does not establish that a real system has that rate, no additional costs, or the same mechanism in every period. State assumptions before interpreting a numeric result. At n=3,x=1/2, the exact power is (3/2)³=27/8, whereas the lower bound is 1+3/2=5/2=20/8. The gap is 7/8>0. The calculation verifies one pair and shows why the inequality is strict away from zero, but the general proof supplies the universal claim. Suppose the task is to prove (1.02)⁵>1.103. Bernoulli gives only ≥1.10, which is insufficient. The quadratic refinement for x≥0 gives ≥1.104, which is enough. Recognising when a known bound cannot prove a requested threshold is as important as knowing the theorem itself. If a question asks for the theorem for every positive integer n, induction is a natural method. If it asks for n=2, an expansion suffices. If it asks whether a fixed input violates the theorem outside the domain, a counterexample may be decisive. Choose the proof strategy based on the quantifiers and task wording rather than automatic procedure. In a finished solution, distinguish = for exact equality, ≥ for a bound, and ≈ for a rounded approximation. Name x≥−1 and n≥1, justify multiplying by 1+x, and identify equality. An answer can have the correct final expression while still missing the steps that make it a sound proof. Prove n=2 using an exact square, establish (1.05)⁸>1.35 without expansion, and decide whether the simple lower bound proves (1.02)⁵>1.103. Answers: the first difference is x²≥0; the second follows from the lower bound 1.40; the third cannot be settled from 1.10 alone, but the quadratic lower bound 1.104 proves it. For the equality question, n=1 permits equality at every x≥−1; n≥2 permits equality at x=0. For the domain question, x=−4 and n=3 show why an unrestricted x claim is false. Write these as complete explanatory statements rather than isolated numbers. The 2027 H3 Mathematics syllabus lists induction, mathematical proof and additional inequalities. Bernoulli’s inequality is an illustrative enrichment of those competencies, not claimed as an individually specified syllabus heading. Yes, when −1≤x<0 and n is a positive integer. The lower bound might be weak, but the standard inequality still holds. Do not remove the restriction x≥−1 without a new proof. No. The elementary induction proof uses integer exponents. For 0 You need an admissible equality point. For D(x)=(1+x)ⁿ−1−nx at n≥2, x=0 shows that the lower bound zero is attained. Use the SEAB 2027 A-Level syllabus directory, the induction guide, the AM–GM guide and extremal bounds for further practice. The unchanged eduKateSG small-group Mathematics reference describes its three-student approach near Sixth Avenue MRT. It does not confirm a current JC1 H3 specialist or open class. Enquire directly about appropriate JC Mathematics support, available tutors, schedules and fees. Did you know? The most useful part of a general inequality is not only the number it bounds. It is the assurance that the result holds throughout its entire domain, with no unseen exceptions. Suppose an unfamiliar task asks whether a fixed positive integer n≥2 permits a real x≥−1 for which (1+x)ⁿ is exactly 1+nx. The answer is yes: x=0 is a witness. But proving that this is the only possible x requires more than writing down one successful value. The n=2 difference is x², strictly positive away from zero, and the induction proof’s nonnegative correction carries the same equality information through larger exponents. This contrasts existence with uniqueness and demonstrates how a named inequality can support a fresh proof question. A tutor should ask the student to identify both obligations explicitly rather than moving from one example to a universal claim.Extended worked example 25: The linear term is not the whole binomial
Extended worked example 26: The induction chain needs two separate arguments
Extended worked example 27: Multiplying by an unknown sign is invalid
Extended worked example 28: An omitted square needs an inequality sign
Extended worked example 29: A direct factorisation for exponent three
Extended worked example 30: A direct factorisation for exponent four
Extended worked example 31: A weaker condition can still permit equality
Extended worked example 32: Distinguish a theorem from an assumed physical rate
Extended worked example 33: A small exact fraction comparison
Extended worked example 34: A threshold that a bound cannot settle
Extended worked example 35: The meaning of an appropriate proof method
Extended worked example 36: A final audit of mathematical language
The six-question check a JC tutor can use
Practice and answers
Frequently asked questions
Is Bernoulli’s inequality a separately named H3 syllabus topic?
Can I use a negative x?
Does the proof work for every positive real exponent?
How do I know a bound is a minimum?
Official syllabus, further learning and nearby tuition enquiries
Why the equality condition matters in a new question