Singapore School Mathematics Operating Manual · Chapter 7
Many wrong solutions are almost correct everywhere except at the boundary. A strict inequality includes the right region but accidentally includes equality. A geometric rule works for an ordinary triangle but fails when three points become collinear. A maximum is guessed from a graph without checking whether an endpoint is larger. A formula is simplified correctly for positive values and then applied at zero, where a denominator disappears.
Boundary cases deserve deliberate attention because they are the places where mathematical behaviour often changes. The boundary may be a number such as zero, an endpoint of an interval, a capacity threshold, a domain restriction, a repeated root, a turning point, or the limiting configuration between two geometric cases.
This chapter develops a practical habit: after solving the interior of a problem, inspect the edges of what is allowed. The aim is not to add ceremonial checking to every question. It is to know when equality, zero, endpoints or extreme configurations can change the answer.
Inequality boundaries · Zero as a structural boundary · Geometric boundaries · Extreme values · Practice · Worked answers
1. Strict and inclusive inequalities encode different boundaries
Solve 2x + 1 < 9. The result is x < 4. The value x = 4 is not allowed because it makes the original left side equal to 9 rather than less than 9.
Replace the sign with ≤ and the solution becomes x ≤ 4. The algebraic manipulation is almost identical; the endpoint classification is different.
On a number line, this difference is represented by an open marker for x < 4 and a closed marker for x ≤ 4. The marker is not decoration. It stores the truth value of the equation at the boundary.
A quick endpoint test is powerful: substitute the boundary back into the original relation. If equality satisfies the condition, include it. If not, exclude it.
2. Multiplying an inequality by zero destroys ordering information
If a < b and c is positive, then ac < bc. If c is negative, the inequality reverses: ac > bc. If c = 0, both products equal zero and the strict inequality disappears entirely.
This is why an inequality cannot be multiplied or divided by an expression of unknown sign without case analysis. The special boundary c = 0 is structurally different from the positive and negative regions.
For example, solving x(x − 3) > 0 requires the critical values 0 and 3. The product is positive when both factors are positive or both are negative, giving x > 3 or x < 0. At x = 0 or 3, the product equals zero, so both endpoints are excluded.
Change the relation to ≥ 0 and both endpoints become included. The sign chart is the same; the boundary decision changes.
3. Zero is not just another number in division
The expression 5/x is defined for every real x except zero. A learner who solves 5/x = 1 by multiplying by x obtains x = 5, which is valid. But the excluded boundary x = 0 still belongs to the domain record.
The expression x/x looks equal to 1, but only for x ≠ 0. At zero, the original ratio is undefined. Cancelling x does not fill in that missing point.
Likewise, the statement ab = 0 has special structure. It implies a = 0 or b = 0. That conclusion depends on zero’s absorbing role in multiplication. A non-zero product equation such as ab = 12 does not split into “a = 12 or b = 12”.
Whenever a variable appears in a denominator, a cancelled factor or a zero-product equation, zero deserves explicit attention.
4. A square reaches its boundary at zero
For every real x, (x − 3)² ≥ 0. The smallest possible value of the square is zero, attained when x = 3.
Therefore f(x) = (x − 3)² + 5 has minimum value 5. The boundary is not an endpoint of a finite interval; it is the lower bound imposed by the square structure itself.
If the domain is changed to x > 3, then f(x) is always greater than 5, but there is no minimum value. Values can approach 5 as closely as desired without attaining it.
This distinction between a lower bound and an attained minimum is a boundary issue. Whether equality is allowed determines whether the extremum exists as an actual value.
5. Endpoint checks are essential on closed intervals
Suppose f(x) = x² − 4x + 7 on 0 ≤ x ≤ 5. Completing the square gives f(x) = (x − 2)² + 3, so the interior minimum is 3 at x = 2.
For a maximum on the closed interval, inspect the endpoints. f(0) = 7 and f(5) = 12. Therefore the maximum is 12 at x = 5.
A learner who differentiates and checks only stationary points would find x = 2 and miss the maximum entirely. On a closed interval, global extrema can occur at stationary points or endpoints.
This is not a calculus-only habit. In any finite allowed range, the edge values can compete with the interior.
6. Whole-number boundaries behave differently from real-number boundaries
Find the greatest integer n satisfying n < 5. The answer is 4.
Find the greatest real number x satisfying x < 5. There is no greatest real number. Any candidate below 5 can be replaced by a slightly larger allowed value.
The same symbolic boundary gives different conclusions because the domains are different. Discrete sets can have a nearest allowed point below a boundary; continuous intervals often do not.
This distinction is developed further in Discrete and Continuous Models.
7. Capacity questions often turn on one boundary value
Suppose a vehicle holds at most 40 passengers. For 40 passengers, one vehicle is enough. For 41, two are required. The entire decision changes at one additional passenger.
Now consider 80 passengers. Exactly two vehicles are enough. At 81, three are required.
When checking a whole-number capacity model, test the values immediately below, at and above the threshold. If a proposed formula says one vehicle is needed for 41 passengers, the failure becomes obvious.
Threshold tests are efficient because they target the point where the output can jump.
8. Collinearity is a boundary between triangle and no triangle
Three positive lengths a, b and c form a non-degenerate triangle only if each side is less than the sum of the other two. For ordered sides with c largest, the crucial condition is c < a + b.
If c = a + b, the points can lie on one straight line. The enclosed area collapses to zero. This is a degenerate boundary case, not an ordinary triangle.
If c > a + b, the segments cannot close to form a triangle.
Thus the equality boundary separates feasible and infeasible geometry. A student who remembers only “largest side not greater than the other two” may accidentally admit the degenerate case when the task requires a genuine triangle.
9. Tangency is a boundary between two intersections and none
A line can cut a circle at two points, touch it at one point, or miss it completely. Tangency is the boundary case between the first and third situations.
Algebraically, substituting the line equation into the circle can produce a quadratic. Two distinct real roots correspond to two intersections. One repeated real root corresponds to tangency. No real roots correspond to no intersection.
For example, consider the circle x² + y² = 25 and horizontal line y = k. Substitution gives x² + k² = 25, so x² = 25 − k².
If |k| < 5, there are two x-values. If |k| = 5, there is one repeated x-value x = 0: tangency. If |k| > 5, there are no real intersections.
The equality case is geometrically special and algebraically visible.
10. Repeated roots are algebraic boundary signals
For the quadratic x² − 6x + k = 0, the discriminant is Δ = 36 − 4k.
If Δ > 0, there are two distinct real roots. If Δ = 0, there is one repeated real root. If Δ < 0, there are no real roots.
The boundary k = 9 produces Δ = 0 and the repeated root x = 3. This is the algebraic transition between two-root and no-real-root regimes.
Parameter questions often ask for the values that create tangency, repeated roots, one solution or a change in the number of intersections. The correct answer is frequently found by identifying the boundary condition rather than solving many separate examples.
11. Rounding intervals have included and excluded endpoints
If a positive measurement is 8.6 correct to the nearest 0.1 under the usual half-up convention, then 8.55 ≤ L < 8.65.
The lower endpoint is included because 8.55 rounds to 8.6. The upper endpoint is excluded because 8.65 rounds to 8.7.
That asymmetry can matter in strict comparisons. L < 8.65 is guaranteed; L ≤ 8.64 is not guaranteed because values such as 8.649 are allowed.
The companion chapter on Rounded Data, Thresholds and Guaranteed Conclusions develops this uncertainty framework in detail.
12. Probability boundaries are zero and one
A probability must satisfy 0 ≤ P(E) ≤ 1. Values outside this interval indicate an error in the model or calculation.
The boundary P(E) = 0 means the event is impossible under the stated model. The boundary P(E) = 1 means it is certain under that model.
A probability extremely close to zero is not the same as impossible. Likewise, a probability extremely close to one is not the same as certain. The endpoints have exact logical meanings.
This distinction matters when interpreting numerical output. Rounding 0.9996 to 1.000 does not turn a non-certain event into a mathematically certain one.
13. A local extreme is not automatically the global extreme
A function can have several turning points. A local maximum is greater than nearby values; a global maximum is greatest over the entire stated domain.
On an interval, compare all relevant candidates: stationary points, endpoints, and any points where the function is not differentiable but remains defined.
For example, f(x) = x³ − 3x on −2 ≤ x ≤ 2 has derivative 3x² − 3, so stationary points at x = ±1. Evaluate f at −2, −1, 1 and 2: f(−2) = −2, f(−1) = 2, f(1) = −2, f(2) = 2.
The global maximum is 2, attained at x = −1 and x = 2. One maximum occurs at a stationary point and another at an endpoint. A method that excludes either class is incomplete.
14. A denominator approaching zero can signal unbounded behaviour
Consider f(x) = 1/x. As x approaches zero from the positive side, f(x) grows without bound. From the negative side, it decreases without bound. The function is not defined at x = 0.
The graph’s vertical asymptote is tied to the domain boundary. Substituting x = 0 is impossible, but examining values near zero reveals the behaviour around the missing point.
This is a later-stage example of why a boundary can matter even when it is not part of the domain. Excluded points can organise the behaviour of nearby values.
15. Boundary cases are excellent debugging tools
Suppose a student proposes the formula n(n − 1)/2 for the number of handshakes if every pair among n people shakes hands once. Test n = 1: the formula gives zero, which makes sense. Test n = 2: it gives one. Test n = 0 in an abstract counting model: it gives zero.
These small boundary cases do not prove the formula for all n, but they can expose obvious structural errors quickly.
For a geometric formula, test a square, a symmetric case or a limiting configuration. For a probability formula, test an impossible case and a certain case. For a percentage formula, test no change. For a recurrence, test its initial value.
A boundary test is especially useful when the general derivation is long enough that a hidden sign or off-by-one error could survive unnoticed.
16. The boundary itself may need a separate case
Some problems have three natural regions: below the threshold, exactly at it, and above it.
For example, define the sign of x. If x > 0, the sign is positive. If x < 0, it is negative. At x = 0, neither description applies. Zero is its own case.
Likewise, a quadratic discriminant has three regimes: positive, zero and negative. A capacity threshold may have “below capacity”, “exactly full” and “over capacity”.
Do not force the equality case into one neighbouring region when the mathematics treats it differently.
17. Independent practice
1. Solve x(x − 6) ≥ 0.
2. Find the greatest integer n satisfying n < 12.3. Then explain whether a greatest real x satisfying x < 12.3 exists.
3. For f(x) = (x − 4)² + 2 with domain x > 4, does f have a minimum? State its lower bound.
4. A vehicle holds 30 passengers. Find the minimum number of vehicles for 30, 31, 60 and 61 passengers.
5. Decide whether side lengths 4, 7 and 11 form a non-degenerate triangle.
6. Find the values of k for which x² − 4x + k = 0 has a repeated real root.
7. The circle x² + y² = 16 is intersected by y = k. For what values of k are there two intersections, one intersection, or none?
8. Find the maximum and minimum of f(x) = x² − 2x on 0 ≤ x ≤ 3.
9. Explain why probability 1.000 after rounding does not necessarily prove certainty.
10. Test the formula n(n + 1)/2 for 1 + 2 + … + n at the boundary values n = 1 and n = 0.
18. Worked answers
1. Critical values are 0 and 6. The product is non-negative outside the interval and zero at the endpoints. Therefore x ≤ 0 or x ≥ 6.
2. The greatest integer is 12. There is no greatest real number below 12.3 because any allowed real value can be increased slightly while remaining below 12.3.
3. Since x > 4, the squared term is positive rather than zero. Hence f(x) > 2. The lower bound is 2, but no minimum is attained.
4. For 30 passengers: 1 vehicle. For 31: 2. For 60: 2. For 61: 3. The decision changes immediately after each multiple of 30.
5. No. The largest side equals the sum of the other two: 11 = 4 + 7. The configuration is degenerate and has zero area, not a non-degenerate triangle.
6. Discriminant Δ = 16 − 4k. Repeated root requires Δ = 0, so k = 4.
7. Substitution gives x² = 16 − k². Two intersections when |k| < 4, one tangent intersection when |k| = 4, and none when |k| > 4.
8. Complete the square: f(x) = (x − 1)² − 1. Minimum −1 at x = 1. Check endpoints for maximum: f(0) = 0 and f(3) = 3. Maximum 3 at x = 3.
9. A value such as 0.9996 may round to 1.000 while remaining strictly below 1. Certainty requires exact probability 1 under the model, not merely a rounded display.
10. At n = 1, the formula gives 1, matching the sum. At n = 0, the empty sum is 0 and the formula gives 0. These checks support consistency at the edge but do not by themselves prove the formula for every n.
19. Continue through the operating manual
Use Case Splitting and Piecewise Reasoning when the boundary separates different formulas or branches. Use Reversible Steps and Extraneous Solutions when transformations behave differently at zero or another excluded value. Continue to Discrete and Continuous Models when the domain determines whether a boundary is attainable.
Return to the BTT Mathematics Hub for Batch 02.