A rational function can have a range that is difficult to read from its original numerator and denominator. An effective Additional Mathematics technique is to call its output y, rearrange y=f(x) into a quadratic equation in x, and require that quadratic to have at least one real, permitted solution. The discriminant then becomes an inequality in y rather than x.
For example, consider f(x)=(x²+2x+2)/(x²+1). Rearranging y=f(x) gives (y−1)x²−2x+(y−2)=0. The discriminant condition yields −y²+3y−1≥0, so the exact range is [(3−√5)/2,(3+√5)/2]. The endpoints are actually attained and the value y=1 is allowed, although the equation in x becomes linear in that special case.
Did you know that a non-negative discriminant is not always the whole answer? If the original denominator can vanish, an apparent real root of the rearranged equation may be forbidden. This guide explains the method, checks its boundary cases, and shows how a rational graph can have two disconnected output ranges or a removable hole.
Range means outputs actually reached by allowed inputs
The domain of f is the collection of x-values for which its original formula is defined. The range is the collection of real y-values that f(x) attains for at least one such x.
The question “does y=k belong to the range?” is equivalent to “does the equation f(x)=k have a real solution in the domain?”
This viewpoint changes a graph-reading problem into an equation-solvability problem. For many rational functions, cross-multiplication makes the equation quadratic in x.
The discriminant is then useful because a real quadratic has real roots precisely when its discriminant is non-negative. However, the original denominator and any special linear cases must still be checked before declaring an output attainable.
Check the domain before rearranging
Our first example is f(x)=(x²+2x+2)/(x²+1). Its denominator x²+1 is strictly positive for every real x.
Therefore the natural real domain is all of ℝ. Multiplying by x²+1 never introduces a forbidden input in this example.
This is a particularly clean setting for the discriminant method. Every real x obtained from the rearranged equation will belong to the original function’s domain.
Other rational functions may have excluded inputs where their denominators vanish. In those cases, multiplying through can create algebraic candidates that are not valid solutions of the original rational equation. Always establish the exclusions first.
Introduce y as the proposed function output
Write y=(x²+2x+2)/(x²+1). Treat y temporarily as an arbitrary real number whose attainability we want to test.
Multiply both sides by x²+1: y(x²+1)=x²+2x+2.
Expand: yx²+y=x²+2x+2. Bring all terms to one side to obtain (y−1)x²−2x+(y−2)=0.
For each fixed y, this is an equation in x. The y-values that allow at least one real x are precisely the range of the original function, provided the domain and any degenerate cases are respected.
Identify the quadratic coefficients correctly
In (y−1)x²−2x+(y−2)=0, regard x as the unknown variable.
The coefficients are A=y−1, B=−2 and C=y−2. The letters A, B and C are temporary to avoid confusing a coefficient with the original function.
Do not treat y as the quadratic unknown when calculating the discriminant. We are asking whether x can be real for a proposed output y.
The leading coefficient A depends on y and becomes zero at y=1. This requires a special check because the equation stops being quadratic there. We will handle it explicitly rather than silently assuming A≠0 for every output.
Build the discriminant inequality
For a genuine quadratic Ax²+Bx+C=0, its discriminant is D=B²−4AC.
Substitute A=y−1, B=−2 and C=y−2. Then D=(−2)²−4(y−1)(y−2).
Expand the product: (y−1)(y−2)=y²−3y+2. Hence D=4−4y²+12y−8.
Simplify to D=−4y²+12y−4=4(−y²+3y−1).
To have a real solution x, require D≥0, giving −y²+3y−1≥0. This inequality is in the output variable y, which is exactly what we need for the range.
Solve the resulting quadratic inequality in y
Rewrite −y²+3y−1≥0 as y²−3y+1≤0 by multiplying through by negative one and reversing the inequality.
The boundary equation y²−3y+1=0 has roots y=[3±√(9−4)]/2=(3±√5)/2.
Because y²−3y+1 opens upward, it is non-positive between its two roots, including the boundaries.
Therefore the candidate output interval is (3−√5)/2≤y≤(3+√5)/2.
The approximate bounds are 0.382 and 2.618. Those numbers describe function outputs, not x-inputs. The exact surd forms should be preserved in the final mathematical answer.
Check the special value y equals one
At y=1, the coefficient of x² in the rearranged equation vanishes, leaving −2x−1=0.
This linear equation has real solution x=−1/2. The original denominator is (−1/2)²+1=5/4>0, so the input is valid.
Direct substitution confirms f(−1/2)=[1/4−1+2]/[1/4+1]=(5/4)/(5/4)=1.
Thus y=1 belongs to the range. It lies inside the candidate interval, but the ordinary quadratic discriminant argument alone did not formally cover it because the leading coefficient became zero. The explicit linear check closes that logical gap.
Why both range endpoints are included
At either endpoint y=(3±√5)/2, the discriminant is zero. A zero discriminant means the rearranged genuine quadratic has one repeated real root.
These endpoints are not simply limiting values approached without attainment; there is an actual real x where each occurs.
The original denominator is always positive, so no candidate x is excluded. Therefore both boundary outputs belong to the range, and square brackets are correct.
This is a general principle: when a range endpoint arises from discriminant equality, check that the repeated root is in the original domain. If so, the boundary is attained and included.
Find the inputs where the endpoint values occur
When D=0 and A=y−1≠0, the repeated x-root is x=−B/(2A)=1/(y−1).
For y_high=(3+√5)/2, the value y_high−1=(1+√5)/2. Its reciprocal is (√5−1)/2.
Thus the maximum occurs at x=(−1+√5)/2, approximately 0.618.
For y_low=(3−√5)/2, the reciprocal calculation gives x=(−1−√5)/2, approximately −1.618.
These are exact real inputs. They confirm that the extreme outputs are not artificial results of manipulating an inequality.
Check the extrema with differentiation
Differentiate f(x)=(x²+2x+2)/(x²+1) using the quotient rule.
The numerator of f′ is (2x+2)(x²+1)−2x(x²+2x+2), which simplifies to −2(x²+x−1).
Therefore f′(x)=−2(x²+x−1)/(x²+1)².
The stationary inputs solve x²+x−1=0, giving x=(−1±√5)/2. These match the discriminant boundary inputs exactly.
The derivative is negative outside those roots and positive between them, so the left stationary point is a minimum and the right is a maximum. This is an independent calculus verification of the exact range.
The graph’s horizontal asymptote lies inside its range
Divide numerator and denominator by x² in the expression f(x), or compare their leading coefficients. As x→±∞, f(x) approaches 1.
Thus the graph has horizontal asymptote y=1. But y=1 is also actually attained at x=−1/2.
A horizontal asymptote is not necessarily an excluded output. Rational graphs can cross their horizontal asymptotes at finite inputs.
The example reinforces why range cannot be inferred solely from an asymptote line. The original equation f(x)=1 and its domain determine whether the asymptote value is attained.
A useful simpler fraction form
Rewrite the first example as f(x)=1+(2x+1)/(x²+1).
This makes the horizontal asymptote y=1 visible because the second fraction tends to zero for large |x|.
It also makes the crossing of that asymptote easy to find: f(x)=1 exactly when 2x+1=0, giving x=−1/2.
But the exact global minimum and maximum are not immediately obvious from this form. The discriminant and derivative methods supply those values.
Different algebraic forms reveal different properties of the same function. A good learner chooses the representation that matches the quantity requested rather than expecting one formula to expose everything.
A rational function can have a disconnected range
Consider g(x)=(x²+2)/(x²−1), with original domain x≠−1 and x≠1.
The squared terms make the graph symmetric about the y-axis, but the denominator has real zeros that divide its domain into three parts.
Rewrite g(x)=1+3/(x²−1). For |x|>1, the denominator is positive and the output is greater than one.
For |x|<1, the denominator is negative and the output is at most −2, with equality at x=0. We will derive these output regions algebraically to demonstrate how a range may contain a real gap.
Rearrange the second example into an x-squared condition
Write y=(x²+2)/(x²−1). Multiply by x²−1 for allowed x, giving yx²−y=x²+2.
Rearrange to (y−1)x²=y+2.
If y≠1, obtain x²=(y+2)/(y−1). Since x²≥0 for real x, require (y+2)/(y−1)≥0.
Solving this rational sign inequality gives y≤−2 or y>1. The value y=1 is impossible because it would require 0=3.
Thus the range is (−∞,−2]∪(1,∞). The gap (−2,1] is not attained. This is a range with two disconnected output components.
Check the second example’s excluded x-values
The original function g is undefined at x=±1. We multiplied by x²−1, so every proposed solution must respect those exclusions.
The rearrangement x²=(y+2)/(y−1) does not produce x²=1 for any finite admissible y, because that equation would imply y+2=y−1, or 2=−1.
Therefore the real solutions found for the allowed y-ranges do not reintroduce the poles x=±1.
This check supports the derived range. It shows that denominator restrictions can sometimes be automatically avoided by the final algebra, but they should never be ignored merely because the sign inequality looks correct.
Why the value y equals minus two is attained
In the second example, choose x=0. The denominator is 0²−1=−1, so the input is valid.
Calculate g(0)=(0+2)/(0−1)=−2. Thus the upper endpoint of the negative output branch is attained.
For inputs approaching ±1 from inside the middle domain section, the denominator tends to zero from below and the function tends toward negative infinity.
Hence the negative range is (−∞,−2], with −2 included but no finite lower bound. The endpoint check explains the square bracket and the unbounded side of the interval.
Why the value y equals one is not attained
For g(x)=1+3/(x²−1), the correction term is never zero at a finite permitted input because its numerator is the nonzero constant three.
Thus g(x) never equals one, even though it approaches one as |x| tends to infinity.
This differs from the first example, whose horizontal asymptote y=1 was crossed at x=−1/2.
These two examples make an important distinction. Whether a horizontal asymptote is part of a function’s range depends on the exact rational formula, not on a universal rule that asymptotes can or cannot be crossed.
A third example with symmetric bounded outputs
Let h(x)=x/(x²+1), defined for all real x. Set y=h(x) and multiply by the strictly positive denominator.
The equation becomes yx²−x+y=0. For y≠0, this is a quadratic in x whose discriminant is 1−4y².
Require 1−4y²≥0, giving −1/2≤y≤1/2. At y=0, the equation becomes −x=0, which has valid input x=0.
The endpoint values are attained: h(1)=1/2 and h(−1)=−1/2. Therefore its exact range is [−1/2,1/2]. The result is again verified by a special-case check when the apparent leading quadratic coefficient vanishes.
Why the third example has symmetric extrema
The function h(x)=x/(x²+1) is odd because h(−x)=−h(x).
Therefore its graph has symmetry through the origin, and its range must be symmetric about zero.
The derived range [−1/2,1/2] has exactly that symmetry. The maximum at x=1 and minimum at x=−1 form a matching pair.
This kind of structural check is useful when a discriminant inequality yields a complicated expression. If the function is clearly odd but the proposed range is not symmetric, revisit the algebra or the interpretation of endpoints.
A cancelled denominator can create a hidden hole
Consider p(x)=(x²−1)/(x−1), with original domain x≠1.
Factorise the numerator as (x−1)(x+1). For every permitted input, p(x)=x+1, but x=1 remains excluded.
Therefore the graph follows a straight line with a hole at (1,2). Its range is all real numbers except y=2, because the only input yielding output two would be x=1.
The function looks simple after cancellation, yet the original domain matters. This example shows why a purely algebraic range calculation can be misleading if it forgets which inputs were forbidden before simplifying.
See how the discriminant can give a false candidate
Set y=(x²−1)/(x−1) and cross-multiply to obtain y(x−1)=x²−1.
Rearrange to x²−yx+(y−1)=0. Its discriminant is y²−4(y−1)=(y−2)², non-negative for every real y.
A careless solution might therefore claim that the range is all real numbers. But at y=2 the quadratic becomes (x−1)²=0.
Its only real solution is x=1, precisely the excluded denominator root. Therefore y=2 is not attained. The correct range is ℝ excluding 2, proving that the discriminant is necessary but not always sufficient after cross-multiplication.
The domain check has to occur after solving for x
For a proposed output y, a non-negative discriminant establishes the existence of real roots of the cross-multiplied polynomial.
It does not establish that one of those roots belongs to the original rational function’s domain. If every candidate root makes the original denominator zero, the proposed output is not attainable.
In the removable-hole example, the issue occurs at y=2, where both quadratic roots coincide at the excluded input x=1.
A reliable procedure therefore includes two tests: reality of algebraic roots and admissibility of at least one root. This is a more precise statement of the discriminant method than “D≥0 automatically equals the range.”
When a rearranged equation becomes linear
In the first example, the quadratic coefficient is y−1 and vanishes at y=1. In the third example, it is y and vanishes at y=0.
At these output levels, the cross-multiplied equation becomes linear rather than quadratic. A quadratic discriminant formula is not formally the right solvability test at that point.
Solve the reduced equation directly and check its root against the original domain. Both first and third examples admit their special output values.
In other functions, the degenerate equation may instead become an impossibility such as 0=3. That happens at y=1 in the second example. The special-case check can therefore change whether an endpoint or asymptote level belongs to the range.
Range endpoints from discriminant equality have geometric meaning
For a smooth rational function with positive denominator, an extreme attainable output often corresponds to a horizontal line tangent to the graph.
At such a level y=c, solving f(x)=c gives a repeated real root in x. That is why discriminant equality appears at many range boundaries.
In the first example, the two extreme y-values create repeated-root equations, and the corresponding inputs are precisely where f′(x)=0.
This connection is useful but not universal without care: endpoint restrictions, holes, vertical boundaries and non-quadratic rearrangements can create range boundaries for other reasons. The derivative and discriminant methods should be interpreted alongside the actual domain.
A general template for rational quadratics
Suppose f(x)=(Ax²+Bx+C)/(Dx²+Ex+F), with a real domain excluding zeros of the denominator.
Set y=f(x). Rearranging gives (Dy−A)x²+(Ey−B)x+(Fy−C)=0.
For output values where Dy−A≠0, require the discriminant Δ(y)=(Ey−B)²−4(Dy−A)(Fy−C)≥0.
Then check whether at least one real root x avoids the original denominator zeros. Handle values Dy−A=0 by solving the resulting linear or constant equation directly.
This provides a reusable framework for many quadratic-over-quadratic rational functions. The precise range still depends on the polynomial inequality and domain exclusions of the particular example.
Do not square inequalities involving the denominator without sign checks
A different way to analyse rational functions uses inequalities such as f(x)≤k. But multiplying such an inequality by a denominator whose sign is unknown can reverse its direction.
For example, x²−1 is positive for |x|>1 and negative for |x|<1. Treating it as always positive would corrupt the second example's range.
The equation-solvability approach avoids inequality direction changes during cross-multiplication of an equality, but still requires denominator restrictions.
If using direct inequalities instead, split the x-domain into sign-consistent regions or use a properly constructed rational sign chart. Method safety matters as much as speed.
Why graphing alone may hide a missing output
A graphing calculator may show a rational curve approaching an asymptote or passing close to a removable hole.
At ordinary display resolution, a hole at a single coordinate can be difficult to distinguish from a graph point that is present. A plotted line y=x+1 may appear complete even when the original rational function excludes x=1.
Exact domain analysis reveals whether the output y=2 is genuinely attained. A numerical graph can support understanding but should not replace the algebraic check for a missing point.
This is a useful lesson about mathematical precision: a range can differ by just one excluded value, and a picture may not make that distinction obvious.
Compare the three range patterns
The first function (x²+2x+2)/(x²+1) has a single closed bounded range [(3−√5)/2,(3+√5)/2].
The second function (x²+2)/(x²−1) has two unbounded pieces, (−∞,−2]∪(1,∞), separated by a gap.
The third function x/(x²+1) has the symmetric closed range [−1/2,1/2]. The cancelled rational example (x²−1)/(x−1) has all real outputs except one missing value.
These contrasting outcomes show why there is no universal interval shape for rational-function ranges. The discriminant, special coefficients, original domain and limiting behaviour determine the answer together.
A five-stage examination routine
First write the original domain restrictions clearly, especially any forbidden denominator zeros. Second set y=f(x) and cross-multiply to obtain a polynomial equation in x.
Third, when it is genuinely quadratic in x, impose discriminant ≥0 and solve the resulting inequality in y. Fourth handle every y-value at which the quadratic leading coefficient vanishes.
Fifth check that at least one real solution x belongs to the original domain. Verify range endpoints through direct substitution or a derivative calculation when helpful.
This sequence gives a reasoned range rather than a guess based on asymptotes. The final answer should use interval brackets and unions that reflect exactly which outputs are attained.
Practice questions with checked answers
Question one: find the range of (x²+2x+2)/(x²+1). Answer: [(3−√5)/2,(3+√5)/2].
Question two: explain why y=1 belongs to Question one’s range. Answer: x=−1/2 is a valid input giving output one.
Question three: find the range of (x²+2)/(x²−1). Answer: (−∞,−2]∪(1,∞), with x≠±1.
Question four: find the range of x/(x²+1). Answer: [−1/2,1/2].
Question five: find the range of (x²−1)/(x−1), with x≠1. Answer: all real numbers except 2.
Question six: why is discriminant ≥0 insufficient in Question five? At y=2 the only algebraic root is x=1, which is forbidden.
Question seven: for the first function, find the maximum-output input. Answer: x=(−1+√5)/2.
Show the original domain before reporting the output range in every question.
Syllabus alignment and connected study
Finding rational-function ranges through discriminants links quadratic solvability, inequalities, domain restrictions, function graphs and differentiation. Some quadratic-over-quadratic range questions can serve as enrichment for learners studying beyond routine secondary examples.
For Singapore, consult the appropriate SEAB syllabus resources for 2026 O-Level Additional Mathematics and 2027 SEC G3, according to the school’s taught sequence.
Related eduKateSG guides include How to Invert a Fractional Linear Function Without Losing Its Domain, How to Distinguish a Hole from a Vertical Asymptote, How to Find the Range of a Quadratic on Disconnected Intervals and the Additional Mathematics archive.
Families may enquire about suitable eduKate Singapore small-group Mathematics tutorials near Sixth Avenue MRT. The enduring skill is to treat every proposed output as an existence question in x, then check that the resulting real input genuinely belongs to the original function. Properly taught kids shine a bright light into the future.