A hot cup is left in a room. Its temperature begins to fall. After ten minutes, it is much cooler. Does it continue losing the same number of degrees every ten minutes until its temperature becomes zero?
Not under the familiar idealised cooling model.
Instead, the rate of cooling depends on the difference between the object’s temperature and the surrounding temperature.
This is an interesting reason to study Additional Mathematics: functions, logarithms and calculus let us explain why the difference from a surrounding temperature can decrease exponentially even when the measured temperature does not approach zero.
The idea is often introduced through Newton’s law of cooling. It connects a differential equation with a familiar exponential curve and shows how a shift in the vertical axis changes the interpretation.
The examples below use invented temperatures and a simplified constant-environment model. They are educational Mathematics, not a way to determine food safety, healthcare treatment or safe equipment temperatures. Actual cooling can involve changing surroundings, radiation, evaporation, airflow and other effects.
Start with the surrounding temperature
Imagine an idealised object with initial temperature:
T₀ = 85°C.
Its surrounding environment has constant temperature:
T_a = 25°C.
The object begins sixty degrees Celsius warmer than its surroundings.
Define the temperature difference:
D(t) = T(t) – T_a.
At t = 0:
D(0) = 85 – 25 = 60°C.
This difference, rather than the absolute temperature T itself, is the key quantity in the model.
As the object cools towards its surroundings, the temperature difference becomes smaller.
The model does not say that the object must cool towards zero degrees Celsius.
It says the idealised temperature approaches the ambient level.
This shift is important when students compare the cooling model with an ordinary exponential decay curve approaching zero.
The basic Newton cooling relationship
A common idealised cooling model states that the object’s rate of temperature change is proportional to the temperature difference between the object and its surroundings.
Write:
dT/dt = -k(T – T_a),
where:
T is the object’s temperature,
T_a is the constant ambient temperature,
t is time,
and k is a positive constant with inverse-time units.
The minus sign means that when T is greater than T_a, the derivative is negative.
The object cools.
If the object is initially colder than the surroundings, the difference T – T_a is negative.
Then the derivative becomes positive.
The same equation can model warming towards the ambient temperature.
The surrounding temperature acts as the equilibrium level.
Why “proportional to the difference” matters
Suppose the object is much warmer than the room.
Its temperature difference is large.
The ideal model gives a relatively large magnitude of cooling rate.
As the object approaches room temperature, the difference becomes small.
The rate’s magnitude becomes smaller.
This is different from a linear model with a fixed loss such as two degrees every minute.
A fixed-loss model eventually predicts temperatures below the surroundings if extended indefinitely.
The Newton cooling model approaches the ambient level asymptotically under its assumptions.
Students should identify which description the question actually gives.
A-Math helps by turning the verbal statement “rate proportional to the difference” into a precise equation.
Shift the variable to the temperature difference
Define:
D(t) = T(t) – T_a.
Because T_a is constant:
dD/dt = dT/dt.
The cooling equation becomes:
dD/dt = -kD.
This is a simple exponential decay relationship for the temperature difference.
Its solution is:
D(t) = D₀e^(-kt),
where D₀ is the initial difference.
Since:
D₀ = T₀ – T_a,
we obtain:
T(t) – T_a = (T₀ – T_a)e^(-kt).
Therefore:
T(t) = T_a + (T₀ – T_a)e^(-kt).
This equation describes the ideal modelled temperature through time.
It shows why the ambient temperature appears as a vertical shift.
Understand the roles of each parameter
Consider:
T(t) = T_a + (T₀ – T_a)e^(-kt).
The parameter T_a is the equilibrium or ambient level.
The parameter T₀ is the initial temperature.
The difference T₀ – T_a is the initial signed departure from ambient.
The positive coefficient k determines how rapidly that difference decreases in magnitude.
Time t must be measured in units consistent with k.
The exponent -kt is dimensionless.
If k is given per minute, then t should be expressed in minutes or appropriately converted.
The formula is easier to understand when students read each parameter as a physical quantity.
Simply memorising the expression without identifying the ambient shift can cause errors when solving questions.
Build a complete worked cooling example
Suppose:
T_a = 25°C.
T₀ = 85°C.
After ten minutes, the observed value in our idealised data set is:
T(10) = 55°C.
Assume the Newton cooling model fits these values exactly.
The formula is:
T(t) = 25 + 60e^(-kt).
Use the measurement at t = 10:
55 = 25 + 60e^(-10k).
Subtract 25:
30 = 60e^(-10k).
Divide by 60:
1/2 = e^(-10k).
Take natural logarithms:
ln(1/2) = -10k.
Since ln(1/2) = -ln 2:
k = ln 2/10.
Therefore:
k ≈ 0.069315 per minute.
We have determined the model parameter from the stated observation.
Rewrite the model in half-life form
Substitute:
k = ln 2/10.
The temperature is:
T(t) = 25 + 60e^[-(ln 2)t/10].
Using exponential laws:
e^[-(ln 2)t/10] = 2^(-t/10).
Therefore:
T(t) = 25 + 60 × 2^(-t/10).
Notice what halves every ten minutes.
It is not the absolute Celsius temperature.
It is the excess above ambient:
T(t) – 25.
Initially the excess is 60°C.
After ten minutes it is 30°C.
After twenty minutes it is 15°C.
After thirty minutes it is 7.5°C.
This is the correct interpretation of the repeated halving.
Calculate the twenty-minute temperature
Use:
T(t) = 25 + 60 × 2^(-t/10).
At t = 20:
T(20) = 25 + 60 × 2^(-2).
Since 2⁻² = 1/4:
T(20) = 25 + 15.
Therefore:
T(20) = 40°C.
We can check by looking at the excess temperature.
At ten minutes the object is 55°C, thirty degrees warmer than the surroundings.
Another ten minutes halves that excess to fifteen degrees.
Add the ambient 25:
25 + 15 = 40°C.
This check does not require repeating the logarithmic calculation.
It uses the meaning of the fitted model.
Calculate the thirty-minute temperature
At t = 30:
T(30) = 25 + 60 × 2^(-3).
Since:
2⁻³ = 1/8,
the excess is:
60/8 = 7.5°C.
Therefore:
T(30) = 25 + 7.5 = 32.5°C.
Compare the ten-minute cooling intervals.
From 85°C to 55°C, the drop is 30°C.
From 55°C to 40°C, the drop is 15°C.
From 40°C to 32.5°C, the drop is 7.5°C.
The absolute temperature changes become smaller.
The excess temperature halves each interval.
That is why the graph flattens as it approaches room temperature.
Why the object does not reach zero Celsius in this model
Our ambient temperature is 25°C.
The cooling function is:
T(t) = 25 + 60e^(-kt),
with k > 0.
For every finite t, the exponential term is positive.
So:
T(t) > 25.
As t grows without bound, e^(-kt) approaches zero.
Therefore:
T(t) approaches 25°C.
It does not approach zero Celsius.
The horizontal asymptote is:
T = 25.
A student who uses a basic formula T = T₀e^(-kt) without the ambient offset would predict a different limiting temperature.
That is why the model structure must reflect the actual question.
Solve for when a threshold temperature is reached
Suppose we ask:
When does the idealised object reach 35°C?
Use:
25 + 60 × 2^(-t/10) = 35.
Subtract 25:
60 × 2^(-t/10) = 10.
Divide by 60:
2^(-t/10) = 1/6.
Take natural logarithms:
(-t/10)ln 2 = ln(1/6).
Since ln(1/6) = -ln 6:
t = 10 ln 6/ln 2.
Numerically:
t ≈ 25.85 minutes.
The result lies between twenty and thirty minutes.
At twenty minutes the model gives 40°C.
At thirty it gives 32.5°C.
So 35°C must occur between them in the smooth decreasing model.
Why a threshold below ambient cannot be reached
Now ask:
When will the same idealised model reach 20°C?
The ambient temperature is 25°C.
Our equation would require:
25 + 60e^(-kt) = 20.
Subtract 25:
60e^(-kt) = -5.
But the left side is positive for every real t.
It cannot equal -5.
Therefore there is no real time satisfying the equation.
This is not a failure of the logarithm method.
The target is outside the model’s range.
A-Math students should check that a requested output is possible before trying to manipulate logarithms.
Understanding function range can prevent an invalid calculation.
The rate of cooling from the derivative
The temperature model is:
T(t) = 25 + 60e^(-kt).
Differentiate:
T'(t) = -60ke^(-kt).
Since:
T(t) – 25 = 60e^(-kt),
we can write:
T'(t) = -k[T(t) – 25].
This is the original cooling relationship.
At t = 0:
T'(0) = -60k.
With k = ln 2/10:
T'(0) = -6 ln 2.
Numerically:
T'(0) ≈ -4.16°C per minute.
The negative sign indicates a falling temperature.
This is an instantaneous rate at the start, not the exact total fall during the first minute.
Compare the initial and later rates
At ten minutes, the temperature excess is 30°C.
Therefore:
T'(10) = -k(30).
So:
T'(10) = -3ln 2,
approximately:
-2.08°C per minute.
At twenty minutes, the excess is 15°C.
Therefore:
T'(20) = -1.5ln 2,
approximately:
-1.04°C per minute.
The magnitude of the rate halves every ten minutes in this exact model because the temperature difference halves.
The object is still cooling, but more slowly in absolute Celsius per minute.
This is a useful example of a negative derivative becoming less negative without the original function increasing.
Instantaneous rate versus average rate
From zero to ten minutes, the temperature falls:
85 – 55 = 30°C.
The average temperature change per minute is:
(55 – 85)/10 = -3°C per minute.
But the instantaneous rate at t = 0 is approximately:
-4.16°C per minute.
And at t = 10 it is approximately:
-2.08°C per minute.
These are different.
The average compares two separated time points.
The derivative describes local instantaneous change.
A-Math teaches students not to confuse an interval’s average slope with the tangent gradient at one particular time.
Both are useful, but they answer different questions.
Why cooling is fastest at the beginning here
For the model:
T'(t) = -k[T(t) – 25],
the magnitude is:
|T'(t)| = k[T(t) – 25]
while the object remains above ambient.
The excess temperature is largest at the initial time in our model.
Therefore the magnitude of the cooling rate is initially greatest.
As the object approaches ambient, the difference becomes smaller.
So the model predicts slower absolute cooling.
That statement depends on a constant positive k and constant ambient temperature.
If the surrounding conditions or physical process change, the conclusion may not hold in the same form.
Students should state the assumptions behind a general statement about the model’s behaviour.
A colder object can warm using the same equation
Suppose an idealised object begins at:
T₀ = 10°C.
The surroundings remain at:
T_a = 25°C.
Use:
T(t) = 25 + (10 – 25)e^(-kt).
Simplify:
T(t) = 25 – 15e^(-kt).
For k > 0, the exponential factor decreases over time.
Therefore the negative term becomes smaller in magnitude.
The object’s temperature rises towards 25°C.
Differentiate:
T'(t) = 15ke^(-kt).
This is positive.
So the same mathematical law can describe warming towards the surroundings as well as cooling towards them.
The sign of the initial temperature difference determines the direction.
A warming example after one excess half-life
Let k = ln 2/10 again.
Then:
T(t) = 25 – 15 × 2^(-t/10).
At t = 0:
T(0) = 25 – 15 = 10°C.
At t = 10:
T(10) = 25 – 7.5 = 17.5°C.
At t = 20:
T(20) = 25 – 3.75 = 21.25°C.
At t = 30:
T(30) = 25 – 1.875 = 23.125°C.
The object warms by increasingly small absolute amounts over successive ten-minute intervals.
It approaches ambient from below.
The mathematical mechanism is the decay of the signed temperature difference towards zero.
That is why temperature-change models can be understood as shifted exponential functions.
The ambient value is a crucial parameter
Suppose we fitted our original observations using:
T_a = 25°C.
Now imagine the actual surrounding temperature was 30°C.
The model would have a different initial excess:
85 – 30 = 55°C.
The observation at ten minutes would also imply a different k value if we required the same temperature of 55°C at that time.
Thus changing the ambient parameter changes the fitted relationship.
Students should not simply reuse a decay constant obtained under one assumed ambient condition after changing that condition without checking the model.
Parameters are meaningful quantities.
A-Math helps learners see how one measured output constrains several unknowns only when enough other information is supplied.
Solve a cooling parameter from two measurements
Suppose ambient temperature is known to be 20°C.
An object is measured at:
T(0) = 80°C.
After five minutes:
T(5) = 50°C.
Assume the same ideal model.
Its initial excess is:
80 – 20 = 60°C.
Its five-minute excess is:
50 – 20 = 30°C.
The excess has halved in five minutes.
Therefore:
e^(-5k) = 1/2.
So:
k = ln 2/5.
The fitted temperature formula is:
T(t) = 20 + 60e^[-(ln 2)t/5].
Equivalently:
T(t) = 20 + 60 × 2^(-t/5).
This example shows why it is often easier to subtract ambient temperature before calculating the ratio.
A two-point fit requires the model assumption
We used two temperatures and the known ambient level to determine k.
But two observations alone do not prove the real process follows Newton’s cooling law at all other times.
Many functions can pass through the same two data points.
A straight line could connect them.
A more complex curve could also pass through them.
The exponential result is exact only after the function family has been assumed.
A learner should distinguish reconstructing parameters inside a chosen model from validating the model using independent evidence.
This is an important lesson in mathematical modelling.
The calculation is powerful, but the evidence supporting the formula is a separate matter.
Compare exponential and linear predictions
For the original example, T starts at 85°C and reaches 55°C after ten minutes.
A simple linear interpolation is:
L(t) = 85 – 3t,
if t is measured in minutes.
At t = 20, this predicts:
L(20) = 25°C.
At t = 30, it predicts:
L(30) = -5°C.
The exponential cooling model instead predicts:
T(20) = 40°C.
T(30) = 32.5°C.
The two models agree at t = 0 and t = 10.
They diverge afterward.
This shows that matching two points does not establish long-term behaviour.
The mathematical function and its assumptions determine what happens between and beyond the measurements.
What does the graph look like?
For an object warmer than its surroundings, the ideal Newton cooling graph decreases towards a horizontal asymptote at the ambient temperature.
It becomes less steep as time passes.
The graph is not a straight line.
Its first derivative is negative.
Its second derivative is positive.
For:
T(t) = T_a + Ae^(-kt),
with A > 0 and k > 0:
T'(t) = -kAe^(-kt) < 0.
Differentiate again:
T”(t) = k²Ae^(-kt) > 0.
So the curve decreases but is concave upward.
This combination is not contradictory.
A falling function can have a positive second derivative because its negative slope is becoming less negative.
Linearise a cooling model
Let:
D(t) = T(t) – T_a,
for the warmer-than-ambient case where D(t) > 0.
The model is:
D(t) = D₀e^(-kt).
Divide by D₀:
D(t)/D₀ = e^(-kt).
Take natural logarithms:
ln[D(t)/D₀] = -kt.
If we plot the dimensionless logarithmic ratio against t, the exact model produces a straight line through the origin with gradient -k.
That transformation can help analyse exponential relationships.
It does not guarantee that real measurements will lie exactly on a line.
Measurement variability and model limitations may produce deviations.
This is another way A-Math connects logarithms with graph interpretation.
Why the logarithm requires a positive ratio
In the cooling example, D(t) and D₀ are positive because the object begins warmer than ambient and approaches it from above.
Therefore the ratio D(t)/D₀ is positive.
Its logarithm is well defined over the real numbers.
If the object begins colder than ambient, the signed differences are negative, but their ratio remains positive under the ideal model.
We can use a ratio of signed differences that has positive value, or apply the logarithm to the positive magnitude of the difference with a consistent convention.
What we must not do is take a real logarithm of a negative number without changing the mathematical setting.
This illustrates why domain restrictions matter in log-linear transformations.
Cooling times have a useful scaling property
Suppose the initial positive excess is D₀.
We want the time to reduce it to a fraction p of the initial excess, where:
0 < p < 1.
Set:
D₀e^(-kt) = pD₀.
Divide by D₀:
e^(-kt) = p.
Take logarithms:
-kt = ln p.
Therefore:
t = -ln p/k.
The answer depends on the target fraction p and the decay constant k, not directly on D₀.
Under a constant-k model, the time needed to reduce an excess by the same fraction is independent of its starting size.
This is a mathematical property of the exponential function.
It may not hold for a real process whose k changes with temperature or conditions.
A target fraction example
Suppose we want the excess above ambient to become one-quarter of its initial value.
Then:
p = 1/4.
Use:
t = -ln(1/4)/k.
Because:
-ln(1/4) = ln 4 = 2ln 2,
we have:
t = 2ln 2/k.
If the excess half-life is:
H = ln 2/k,
then:
t = 2H.
Thus reducing the excess to one-quarter takes two ideal half-lives.
Reducing it to one-eighth takes three.
This is the same geometric progression idea found in repeated percentage changes.
But the important quantity is excess relative to ambient, not the absolute temperature.
A common mistake: halving the Celsius reading
In the original cooling model:
T(0) = 85°C.
T(10) = 55°C.
A student sees “half-life of ten minutes” for the temperature difference and predicts:
T(20) = 55/2 = 27.5°C.
That is wrong.
The model halves the difference from 25°C.
At ten minutes the difference is:
55 – 25 = 30°C.
Another ten minutes gives:
30/2 = 15°C.
Add the ambient level:
25 + 15 = 40°C.
The correct prediction is 40°C.
The error arose from applying the exponential factor to the whole Celsius reading instead of the excess above equilibrium.
A second mistake: subtracting the first drop repeatedly
The temperature dropped by 30°C during the first ten minutes.
A student predicts the next ten minutes will cause another 30°C drop.
That would give:
55 – 30 = 25°C.
But our exponential model predicts:
40°C.
The absolute temperature loss decreases as the excess decreases.
Subtracting the same number repeatedly would describe a linear cooling model, not the stated proportional-rate relationship.
The repair is to identify whether the question gives a constant absolute change or a constant fractional change in the excess.
Additional Mathematics helps students choose the correct function family.
A third mistake: using a negative decay constant
The model is:
T(t) = T_a + (T₀ – T_a)e^(-kt),
with k > 0.
If a student instead uses e^(kt) with a positive k for a warmer object, the temperature difference grows without bound.
The result describes movement away from ambient, not cooling towards it.
The sign convention can be changed if the coefficient itself is declared negative, but the meaning must be consistent.
A simple check is to evaluate the model at a later time.
Does it move the temperature towards the stated ambient level?
If not, inspect the sign of the exponent.
A fourth mistake: confusing a derivative with a finite change
The initial cooling rate in the example is approximately:
-4.16°C per minute.
That does not mean every following minute produces a drop of 4.16°C.
The derivative changes with time.
The exact drop across an interval is found from the difference between the temperature function’s endpoint values.
The derivative supplies a local approximation for a sufficiently small interval.
The repair is to distinguish the instantaneous rate from the average rate and exact finite change.
This is a general calculus skill, not something specific to temperature.
Real cooling may need more complicated functions
A real object can have more than one internal temperature.
Its surface may cool at a different rate from its centre.
The ambient temperature may vary.
Evaporation, phase change, conduction through supports or air movement may matter.
The effective coefficient k may not remain constant.
Under those conditions, one simple exponential curve may not provide an accurate description across the full process.
The formula remains a useful teaching model.
But applying it to a real scientific or safety decision requires evidence that its assumptions are appropriate.
The OpenStax explanation of Newton’s law of cooling gives a mathematical treatment of the proportional-temperature-difference model.
Ambient temperature can change with time
Suppose the surroundings are not constant.
Instead of one fixed T_a, we might have:
T_a(t).
The idealised equation could be written:
T'(t) = -k[T(t) – T_a(t)],
under additional assumptions.
This no longer leads automatically to the simple shifted exponential formula with a constant ambient offset.
The solution depends on the time-varying surroundings.
This is a more advanced differential-equation problem.
For G3 Additional Mathematics, the useful connection is recognising why a parameter that was treated as constant in the original formula matters.
Changing that assumption changes the mathematical problem itself.
Students should not reuse the old solution unchanged.
A piecewise ambient-temperature model
Imagine an illustrative experiment in which the surroundings remain at 25°C for an initial period and then change to 30°C.
We could model each interval using its own constant-ambient exponential relationship, provided the simplifying assumptions hold within each interval.
At the switch, the object’s temperature should ordinarily be continuous.
The temperature at the end of the first interval becomes the initial condition for the second interval.
The formula’s constant changes because the equilibrium level changes.
This is an enrichment link to piecewise functions.
It shows how a simple model can be assembled from different regimes.
The correct calculation begins by identifying the switch time and preserving the state at that boundary.
A five-session study plan
Session one: identify the equilibrium level
Start with a hot object at 85°C in a room at 25°C.
Calculate its excess temperature.
Distinguish the object temperature from its difference above ambient.
Session two: fit an exponential model
Use T(10) = 55°C.
Derive k = ln 2/10.
Rewrite the formula using the half-life of the temperature excess.
Session three: solve a threshold
Find when the model reaches 35°C.
Use logarithms and check the answer between the twenty- and thirty-minute values.
Session four: read the derivative
Calculate T'(0), T'(10) and T'(20).
Explain why the derivative is negative but its magnitude decreases.
Compare an instantaneous rate with a ten-minute average.
Session five: test model limits
Compare a straight-line interpolation with exponential cooling.
Investigate why a target below ambient is impossible under the warmer-object model.
Discuss what would change if ambient temperature varied.
The aim is to connect functions, calculus and interpretation.
What parents can ask during revision
Ask what temperature the object is approaching.
Ask whether the absolute Celsius reading or the excess above room temperature is halving.
Ask why the first ten-minute drop is larger than later drops.
Ask what the derivative tells us at one instant.
Ask why a logarithmic equation has no real solution for a target below the model’s minimum range.
Ask what would happen to the mathematical assumptions if the room became warmer.
These questions encourage explanation rather than memorisation.
A learner who can correctly identify the ambient offset is developing useful mathematical model awareness.
Singapore Additional Mathematics syllabus context
The official 2027 Singapore-Cambridge Secondary Education Certificate G3 Additional Mathematics syllabus K341 includes exponential and logarithmic functions and their use as models, differentiation and rates of change.
The Newton cooling example connects those mathematical ideas. Solving differential equations from first principles, changing ambient temperature and more advanced physical modelling are enrichment applications rather than separate asserted G3 examination methods.
Students and families can consult the SEAB 2027 G3 Additional Mathematics syllabus and the 2027 G3 syllabus directory for precise requirements.
For the scientific model’s mathematical derivation, the OpenStax calculus explanation is useful further reading.
How a three-student tutorial can help
One learner may use exponent laws accurately but forget the ambient offset.
Another may solve the logarithmic equation and still choose a target time outside the model’s range.
A third may differentiate correctly but confuse the instantaneous cooling rate with the total temperature decrease during an interval.
These students need different explanations.
In a carefully managed three-student Additional Mathematics tutorial, a tutor can inspect the function, the algebra and the interpretation separately.
The immutable Secondary 1 Mathematics small-group tutorial in Clementi illustrates why strong functions, percentages and symbolic reasoning matter before more advanced exponential applications.
Students who are already progressing effectively independently may not need tuition. Extra support is most useful when it addresses a specific recurring difficulty.
Frequently asked questions
What does Newton’s law of cooling say mathematically?
In an ideal constant-environment model, temperature changes at a rate proportional to the difference between the object’s temperature and the surrounding temperature.
Why does temperature not approach zero Celsius?
Because the exponential model approaches the ambient temperature T_a, which need not be zero.
What exactly halves in a cooling half-life?
Under a fixed exponential model, the magnitude of the temperature difference from ambient halves in each half-life interval, not necessarily the absolute Celsius temperature.
Can the same equation describe warming?
Yes. When an object begins colder than its surroundings, the signed difference is negative and the model predicts warming towards ambient.
Does the object reach ambient at a finite time?
In the ideal exponential model with a non-zero initial difference, it approaches ambient asymptotically and does not equal it at a finite time.
Why use logarithms?
They allow us to solve for unknown time or the decay coefficient when those variables appear in exponential expressions.
Can the formula be used for real-world temperature safety decisions?
Not on its own. Actual conditions, measurements and relevant safety standards must be considered beyond the simplified school model.
The lasting reason to understand a shifted exponential
Additional Mathematics teaches students that a familiar exponential curve can describe more than growth towards infinity or decay towards zero.
By subtracting the ambient temperature, a cooling process becomes a simple exponential decay of the difference from equilibrium.
Logarithms recover times and coefficients.
Derivatives explain why the temperature changes rapidly at first and more slowly later.
Graph interpretation reveals the horizontal asymptote.
And modelling checks remind us that the equation is meaningful only when its assumptions fit the situation.
That combination of calculation and interpretation is a valuable reason to study Additional Mathematics.
For connected reading, explore Why Study Additional Mathematics | Understand Half-Life Through Exponential Decay, Why Study Additional Mathematics | Compare Discrete Changes with Continuous Rates, Why Study Additional Mathematics | Recognise When a Mathematical Model Stops Being Useful, and the eduKateSG Additional Mathematics project library.