Mathematics: Applications & Interpretation HL
Mathematical exploration preparation: dynamic model evaluation
- 1.
For exploration preparation, a hypothetical model is N(t)=1000/(1+9 exp(-kt)), with t in days. Compare k=0.2 and k=0.25 per day by finding the time to reach N=500 and N(10) in each case. Calculate the percentage decrease in the time to reach 500 when k increases. Explain why changing k does not change the carrying capacity.
[6 marks]Worked explanation
- Set the denominator to 2: 9 exp(-kt)=1, so t=ln(9)/k. The two times are about 10.99 and 8.79 days.
- At day 10 the denominators are 1+9exp(-2) and 1+9exp(-2.5), producing about 450.853 and 575.121 respectively.
- The ratio of the times is 0.2/0.25=0.8, so the decrease is 20%. A 25% rate increase does not mean a 25% time decrease.
- Both models start at 100 and approach 1000. The comparison explores sensitivity of a hypothetical model, not measured growth.
Common mistake: Using the new time as the percentage denominator answers the reverse percentage-change question.
Marking points
- Derives t=ln(9)/k when N=500.
- Finds times 10.9861... days and 8.78890... days for k=0.2 and 0.25 respectively.
- Finds N(10)=450.853... for k=0.2, approximately 451.
- Finds N(10)=575.121... for k=0.25, approximately 575.
- Calculates percentage time decrease as 20%, relative to the k=0.2 time.
- Explains that both positive k values make the exponential tend to zero, so the limiting carrying capacity remains 1000; only the growth timescale changes.
Examiner tip: Using the new time as the percentage denominator answers the reverse percentage-change question.
- 2.
An exploration-preparation model tracks two transport choices A and B. Use column state vectors and T=[[0.8,0.3],[0.2,0.7]], where entry T_ij is the probability of moving from choice j to choice i in one week. Starting with s_0=(100,0), calculate s_1 and s_2. Solve for a steady state containing 100 people. If a hypothetical later week has 75 people choosing A, explain why this alone cannot validate or refute the steady-state model, and identify one assumption to investigate.
[6 marks]Worked explanation
- Columns sum to 1, so this convention preserves total population under multiplication T s. It gives (80,20) then (0.8(80)+0.3(20),0.2(80)+0.7(20))=(70,30).
- Steady state satisfies 0.2a=0.3b, so a:b=3:2. Together with total 100 this gives (60,40).
- The deterministic expected A counts approach 60 from the initial 100; they need not equal 60 after a finite number of weeks. Observed counts can also fluctuate.
- One un-timed observation is insufficient to decide model fit. A proposed follow-up would compare transitions across several weeks and check whether the population stays closed, without claiming that those observations exist.
Common mistake: Multiplying a row state by this column-stochastic matrix changes the specified probability convention and can fail to preserve the total.
Marking points
- Uses s_1=T s_0=(80,20).
- Obtains s_2=T s_1=(70,30).
- Writes a=0.8a+0.3b and a+b=100 for a steady state.
- Solves a=60,b=40 and verifies T(60,40)=(60,40).
- Explains that steady state is a long-run expected distribution, not an exact prediction for every finite week; the unspecified later time and random variation prevent a decisive conclusion from 75 alone.
- Identifies a specific assumption to investigate, such as constant transition probabilities across weeks or a closed population of 100 people.
Examiner tip: Multiplying a row state by this column-stochastic matrix changes the specified probability convention and can fail to preserve the total.
- 3.
For exploration preparation, hypothetical training observations are (t,N)=(0,10),(2,40), with t in days. Fit both N_L=a+bt and N_E=A exp(kt) to these two points. A held-out hypothetical observation is N=70 at t=4. Calculate both predictions there and the absolute prediction errors. Explain which model this one validation point favours, why it does not establish a universally better growth law, and propose a useful next check.
[6 marks]Worked explanation
- A line through the two training points has intercept 10 and slope (40-10)/2=15 per day.
- The exponential has A=10 and exp(2k)=4, so k=ln(2) per day. Both models fit the two training points exactly.
- At day 4 their predictions differ: 70 versus 10(2^4)=160. Against the held-out value 70, absolute errors are 0 and 90.
- The linear fit wins this one comparison. More independent observations, including later times where capacity limits might matter, are needed to evaluate generalisation; no empirical investigation is claimed.
Common mistake: Perfect agreement with the training points cannot distinguish models that were each fitted exactly to those same two points.
Marking points
- Fits N_L=10+15t.
- Fits N_E=10exp((ln(2))t), using exp(2k)=4.
- Predicts N_L(4)=70 and N_E(4)=160.
- Finds absolute held-out errors 0 and 90 respectively.
- Concludes that this validation point favours the linear model but a single point cannot establish behaviour at other times or account for measurement uncertainty.
- Proposes testing further independent observations over a stated wider time interval and examining residual patterns or plausible growth constraints.
Examiner tip: Perfect agreement with the training points cannot distinguish models that were each fitted exactly to those same two points.
- 4.
For written exploration preparation, a hypothetical amount follows dN/dt=-0.5N, N(0)=100, with t in hours. Euler's rule is N_next=N_current+h(-0.5N_current). Estimate N(2) using h=1 and then h=0.5. Compare both with the supplied exact solution N(t)=100exp(-0.5t), calculating absolute errors to two decimal places. Explain what this comparison can and cannot establish about the model.
[5 marks]Syllabus placement evidence: Mathematics AI guide, first assessment 2021: AHL 5.16 Euler's method; internal assessment, Criterion D: Reflection. AI HL numerical differential-equation work with an explicit update and exact comparator. The application tests discretisation rather than a biological law; written exploration preparation only, not an SL requirement or assessed IA.
Worked explanation
- Each one-hour step multiplies by 0.5; each half-hour step multiplies by 0.75. Two and four steps respectively reach the same endpoint t=2.
- Compare both estimates against 36.787944..., not against each other as a substitute for an exact benchmark. Reduced discretisation error here says nothing about whether the differential equation describes empirical observations.
Common mistake: Halving the step without doubling the number of steps compares predictions at different times.
Marking points
- For h=1 the sequence is 100,50,25, so N(2)=25.
- For h=0.5 the sequence is 100,75,56.25,42.1875,31.640625.
- Exact N(2)=100exp(-1)=36.787944... .
- Absolute errors are 11.79 and 5.15 respectively; the smaller step is more accurate here.
- This checks numerical approximation to the assumed equation, not agreement with a real decay process.
Examiner tip: Halving the step without doubling the number of steps compares predictions at different times.
- 5.
In a hypothetical exploration-preparation model, x_(n+1)=60-0.5x_n and x_0=20, where x is a weekly resource index. Calculate x_1, x_2 and x_3, and solve for the equilibrium x*. Defining e_n=x_n-x*, derive the recurrence for e_n and use it to explain the approach to equilibrium. Explain why this mathematical convergence does not by itself validate the weekly rule.
[5 marks]Syllabus placement evidence: Mathematics AI guide, first assessment 2021: SL 1.3 geometric sequences; SL 2.6 modelling skills; internal assessment, Criterion D: Reflection. Supplied recurrence reduces to geometric deviations with ratio -0.5; no general dynamical-systems theorem is assumed. This bounded AI HL exploration-preparation application does not certify a real model or a completed exploration.
Worked explanation
- Iterate the rule in order, then solve its fixed-point equation: 1.5x*=60. The values cross 40 rather than approaching monotonically.
- Subtract 40 from the update: x_(n+1)-40=-0.5(x_n-40). Thus e_n=(-0.5)^n(-20), tending to zero. This is a conditional property of the formula, not an empirical test of its assumptions.
Common mistake: A negative multiplier need not mean divergence; its absolute value governs whether deviations shrink.
Marking points
- x_1=50, x_2=35, x_3=42.5.
- x*=60-0.5x* gives x*=40.
- e_(n+1)=-0.5e_n.
- The negative factor alternates the side of equilibrium and its magnitude below 1 halves the deviation each week.
- A stable recurrence may still have incorrect coefficients or omit real influences; independent weekly observations are needed.
Examiner tip: A negative multiplier need not mean divergence; its absolute value governs whether deviations shrink.
- 6.
For written dynamic-model preparation, use column states s_(n+1)=T s_n with T=[[0.7,0.2],[0.3,0.8]] and s_0=(100,0)^T. The entries are expected numbers in choices A and B after n weeks, in a closed population. Find the two eigenvalues and a corresponding eigenvector for each. Give matrices S and D such that T=S D S^(-1). Hence derive an expression for s_n, calculate s_3, and state its limiting expected state. Do not interpret a fractional expected count as a fraction of a person observed.
[7 marks]Syllabus placement evidence: Mathematics AI guide, first assessment 2021, August 2025 update: AHL 1.15, printed p.33; AHL 4.19, printed pp.62-63. The inspected bank mentions eigenvalues but never calculates eigenpairs or diagonalisation. The August 2025 edition confirms the cited placement and column-state convention. Written preparation, not teacher approval or evidence validating a real transition model.
Worked explanation
- Expand the determinant (0.7-lambda)(0.8-lambda)-0.06. Factoring gives (lambda-1)(lambda-0.5), two distinct real eigenvalues.
- For eigenvalue 1, -0.3a+0.2b=0 gives a:b=2:3. For 0.5, 0.2a+0.2b=0 gives b=-a. Put the two eigenvectors into S as columns in the same order as D.
- S^(-1)s_0=(20,60)^T, so s_0=20(2,3)^T+60(1,-1)^T. Repeated multiplication leaves the first component fixed and multiplies the second by (0.5)^n, giving the stated formula.
- At n=3, the deviation is 60/8=7.5; s_3=(47.5,52.5)^T. At n=0 the formula returns (100,0)^T and at n=1 it returns (70,30)^T, independent checks of the expression.
- The decaying eigencomponent yields expected state (40,60)^T. Random observed counts can differ, and a fixed transition matrix remains an assumption about behaviour.
Common mistake: Swapping eigenvectors without swapping their corresponding diagonal entries changes the matrix reconstructed by S D S^(-1).
Marking points
- Solves det(T-lambda I)=lambda^2-1.5lambda+0.5=0 to obtain eigenvalues 1 and 0.5.
- For lambda=1, obtains an eigenvector proportional to (2,3)^T.
- For lambda=0.5, obtains an eigenvector proportional to (1,-1)^T.
- One valid choice is S=[[2,1],[3,-1]], D=diag(1,0.5), S^(-1)=[[0.2,0.2],[0.6,-0.4]], with matched eigenvector/eigenvalue order.
- Uses T^n=S D^n S^(-1), giving s_n=(40+60(0.5)^n,60-60(0.5)^n)^T.
- Calculates s_3=(47.5,52.5)^T expected people; the total remains 100.
- Since (0.5)^n tends to zero, the limiting expected state is (40,60)^T, not an exact observed weekly allocation.
Examiner tip: Swapping eigenvectors without swapping their corresponding diagonal entries changes the matrix reconstructed by S D S^(-1).
- 7.
For written dynamic-model preparation, points in the complex plane obey z_(n+1)=q z_n, where q=(sqrt(3)+i)/2, z_0=4 and n is a nonnegative integer. Write q in polar form with principal argument in radians. Obtain a polar expression for z_n, calculate z_4 in Cartesian form, and find the smallest positive n for which z_n=z_0. Explain geometrically why this sequence does not converge even though every term has the same modulus.
[6 marks]Syllabus placement evidence: Mathematics AI guide, first assessment 2021, August 2025 update: AHL 1.12 and AHL 1.13, printed pp.31-32. No inspected AI HL original operates on complex numbers. The August 2025 edition confirms this placement. Nonconvergence follows from two explicit repeating values, not an additional recalled theorem. Written preparation only; independent teacher review remains required.
Worked explanation
- The real and imaginary parts of q are sqrt(3)/2 and 1/2, so its squared modulus is 3/4+1/4=1 and its quadrant-I argument is pi/6.
- Iterating multiplication gives z_n=q^n z_0. Multiplication adds arguments and multiplies moduli, producing a rotation sequence on the circle of radius 4.
- For n=4 the argument is 2pi/3: real part 4(-1/2)=-2 and imaginary part 4(sqrt(3)/2)=2sqrt(3).
- Twelve steps give one full turn. At indices 12k the point is 4, while at indices 12k+6 it is -4; these remain separated by distance 8, so the sequence has no single limit.
Common mistake: Constant modulus controls distance from the origin, not distance from a fixed limiting point.
Marking points
- Finds |q|=1 and arg(q)=pi/6, so q=cos(pi/6)+i sin(pi/6).
- Uses integer powers to obtain z_n=4[cos(n pi/6)+i sin(n pi/6)].
- At n=4, obtains z_4=-2+2sqrt(3)i.
- Every term has modulus 4; each update rotates by pi/6 counterclockwise with no change of scale.
- A return requires n pi/6=2pi k, so the smallest positive return index is n=12.
- The sequence repeats distinct points, including 4 and -4 infinitely often; constant radius does not imply convergence to a single point.
Examiner tip: Constant modulus controls distance from the origin, not distance from a fixed limiting point.
- 8.
For written complex-model preparation, let z=(5+5i)/(1-2i). Find z in Cartesian form and in polar form with principal argument in (-pi,pi]. Describe the change of modulus and argument caused by dividing 5+5i by 1-2i. Verify your Cartesian result by multiplying back. Angles are in radians.
[6 marks]Syllabus placement evidence: Mathematics AI guide, first assessment 2021, August 2025 update: AHL 1.12-1.13, printed pp.31-32. Adds an actual complex quotient and its geometric interpretation; the preserved rotation task uses powers of a unit-modulus multiplier instead.
Worked explanation
- Conjugate multiplication yields numerator 5+10i+5i+10i^2=-5+15i and denominator 1+4=5.
- The negative real and positive imaginary components locate the result in quadrant II. The modulus is sqrt(1+9), and pi-arctan 3 selects the principal argument.
- The divisor has modulus sqrt(5) and argument -arctan 2. Inverting it divides lengths by sqrt(5) and reverses its angle; apply that change to the original complex point, not to the origin as a translation.
- Multiplying by the original divisor returns 5+5i, checking signs independently of the polar calculation.
Common mistake: Dividing complex numbers subtracts the divisor's argument; here subtracting its negative argument produces a positive rotation.
Marking points
- Multiplies numerator and denominator by 1+2i, obtaining denominator 5.
- Simplifies to z=-1+3i.
- Finds |z|=sqrt(10).
- Uses quadrant II to obtain arg(z)=pi-arctan(3), approximately 1.89, and gives z=sqrt(10)[cos(arg z)+i sin(arg z)].
- Division scales modulus by 1/sqrt(5) and adds arctan(2) to the argument modulo 2pi, a counterclockwise rotation.
- Checks (-1+3i)(1-2i)=5+5i.
Examiner tip: Dividing complex numbers subtracts the divisor's argument; here subtracting its negative argument produces a positive rotation.
- 9.
For written geometric-model preparation, a point p=(x,y)^T is transformed by first stretching its displacement from c=(1,1)^T horizontally by factor 2, then rotating that displacement 90 degrees counterclockwise, then adding c back. The supplied matrices are H=[[2,0],[0,1]] and R=[[0,-1],[1,0]]. Write p'=M p+b, finding M and b. Transform B(3,1) and C(1,4), noting that A=c remains fixed. Use det(M) to find the area of the image of triangle ABC, checking it by its image base and height.
[6 marks] · no calculatorSyllabus placement evidence: Mathematics AI guide, first assessment 2021, August 2025 update: AHL 3.9, printed p.46; AHL 1.14, printed p.33. Adds composed point transformations and determinant-based area, not just existing matrix arithmetic. The matrices and centre are supplied; no general affine-transformation theory is assumed.
Worked explanation
- The rightmost matrix acts first: p'=R H(p-c)+c. Expanding separates the linear part M from the translation b; multiplying absolute coordinates by M alone would use the wrong centre.
- Using the final coordinate rule gives B'=(1,5) and C'=(-2,1). Applying it to c gives c again, a check on b.
- The determinant's absolute value supplies the area multiplier. The original right triangle has perpendicular legs 2 and 3, and the image has legs 4 and 3, verifying both the determinant and the geometric answer.
Common mistake: Reversing H and R stretches a different direction; omitting b rotates and stretches about the origin.
Marking points
- Uses the ordered product M=R H=[[0,-1],[2,0]].
- Finds b=c-Mc=(2,-1)^T, so p'=(-y+2,2x-1)^T.
- Obtains B'=(1,5).
- Obtains C'=(-2,1), while A'=(1,1).
- Calculates det(M)=2, so image area is twice original area.
- Original area is (1/2)(2)(3)=3 square units, hence image area 6; the perpendicular image sides have lengths 4 and 3, also giving 6.
Examiner tip: Reversing H and R stretches a different direction; omitting b rotates and stretches about the origin.
- 10.
For written dynamic-model preparation, dimensionless quantities satisfy x'=-2x+y and y'=x-2y, with time in hours, x(0)=3, y(0)=1. Using the supplied substitutions u=x+y and v=x-y, derive uncoupled equations and solve for x(t),y(t). Then use simultaneous Euler updates with step h=0.25 hours to estimate both at t=0.5, showing both steps. State the limiting exact state as t tends to infinity. Each Euler slope must use both values from the start of its step.
[6 marks]Syllabus placement evidence: Mathematics AI guide, first assessment 2021, August 2025 update: AHL 5.14, printed p.69; AHL 5.16-5.18, printed pp.69-70. Adds coupled rather than scalar differential equations. The supplied substitution selects distinct real modes, within the guide's exact-solution boundary; no general solution for complex eigenvalue systems is required.
Worked explanation
- Add and subtract the original equations: x'+y'=-(x+y), while x'-y'=-3(x-y). The given initial state fixes u0=4 and v0=2.
- Solve each separable decay equation. Since x=(u+v)/2 and y=(u-v)/2, obtain the two exact expressions; at zero they recover 3 and 1.
- At (3,1), both slopes are computed before updating: (-5,1). Add 0.25 times these to get (1.75,1.25). Recompute both slopes there, then update again to (1.1875,1.0625).
- The modes have distinct negative real rates -1 and -3. Both exact coordinates tend to zero. For comparison the exact t=0.5 state is approximately (1.43619,0.989931), not the Euler pair.
Common mistake: Updating y with the newly updated x in the same step changes the specified simultaneous Euler method.
Marking points
- Derives u'=-u and v'=-3v, with u(0)=4 and v(0)=2.
- Solves u=4exp(-t), v=2exp(-3t).
- Recovers x=2exp(-t)+exp(-3t) and y=2exp(-t)-exp(-3t).
- First simultaneous Euler step gives (x,y)=(1.75,1.25) at t=0.25.
- At that state the slopes are (-2.25,-0.75), giving (x,y)=(1.1875,1.0625) at t=0.5.
- Both exponential modes decay, so the exact state tends to (0,0); the approximate finite-time Euler state is not the exact solution.
Examiner tip: Updating y with the newly updated x in the same step changes the specified simultaneous Euler method.