Why an Air Purifier Is a Mathematics Lesson
An air purifier seems simple: air enters, passes through a filter and returns to the room. Yet a sensible choice or experiment depends on quantities that are easy to confuse. Clean air delivery rate, room volume, air changes per hour, removal efficiency, electrical power, sound level and particle concentration do not describe the same thing. Mathematics connects them without turning any single label into a guarantee.
This article explains why mathematics is important in everyday life through air purifiers. It is about quantitative reasoning, not medical advice. A portable cleaner can be one part of an indoor-air strategy, but it does not replace source control, suitable ventilation, building maintenance or professional guidance when a hazard is present.
Use manufacturer data, classroom simulations or teacher-provided measurements. Do not generate smoke, aerosols, mould, dust or other pollutants for a home experiment. Do not open, modify or obstruct a powered appliance. The safest investigations use published numbers, clean-room air, paper models and spreadsheets.
Quick Reading Routes
- CADR, room volume and air changes
- Particle decay and exponential models
- Filters, pressure drop and operating points
- Evidence, uncertainty and fair comparisons
- Student investigations
- Parents, misconceptions and FAQs
CADR, Room Volume and Air Changes
CADR is a rate of equivalent clean air
Clean air delivery rate, usually shortened to CADR, combines airflow with removal performance for a stated particle category and test method. It is commonly reported in cubic feet per minute or cubic metres per hour. A unit of volume per time is the clue: CADR is a rate, not a percentage and not a room area.
The US Environmental Protection Agency guide to air cleaners advises choosing a portable unit with a CADR large enough for the room. The AHAM Verifide directory lists independently verified smoke, dust and pollen CADR values for participating products. These sources support comparison, but they do not make every real room identical to a test chamber.
Suppose a cleaner has a smoke CADR of 240 cubic metres per hour. This means its tested particle-removal effect is equivalent to supplying 240 cubic metres of clean air each hour under the rating conditions. It does not mean the machine physically stores that volume, nor that every particle in a 240-cubic-metre room disappears in one hour.
Room area must become room volume
A floor area alone is not enough for an air-change calculation. A room 5 m by 4 m with a 2.6 m ceiling has volume 5 × 4 × 2.6 = 52 cubic metres. A room with the same floor area and a 3.2 m ceiling has volume 64 cubic metres, about 23% larger.
If the 240 m³/h CADR unit serves the 52 m³ room, the idealised clean-air-change rate contributed by the device is 240 ÷ 52 = 4.62 per hour. In the taller 64 m³ room, it is 240 ÷ 64 = 3.75 per hour. The appliance has not changed; the denominator has.
This is a useful example of mathematical literacy. A “recommended room size” may assume a ceiling height, target air-change rate and test convention. A careful reader finds those assumptions instead of comparing square metres as if height were irrelevant.
Converting units without losing meaning
One cubic foot is approximately 0.0283168 cubic metres. Therefore 150 cubic feet per minute is 150 × 0.0283168 × 60 ≈ 254.9 cubic metres per hour. The factor of 60 converts minutes to hours.
A common error is multiplying by the volume conversion but forgetting the time conversion. Another is comparing a m³/h number directly with a ft³/min number. A simple unit chain prevents both mistakes:
150 ft³/min × 0.0283168 m³/ft³ × 60 min/h ≈ 255 m³/h.
Units are not decorations. They show which operations make sense. Dividing m³/h by m³ leaves 1/h, the unit of an air-change rate. Adding a sound level in decibels to a CADR would have no physical meaning.
Particle Decay and Exponential Models
Why concentration does not usually fall in a straight line
In a simplified well-mixed room, the cleaner removes a fraction of the remaining airborne particles during each small time interval. When less remains, less is removed per minute. That produces exponential decay rather than a constant straight-line decrease.
A basic model is C(t)=C₀e^(-kt), where C₀ is the starting concentration, C(t) is the concentration after time t and k is the total first-order removal-rate constant. If the purifier is the only process in the ideal model, k can be approximated by CADR divided by room volume, using compatible units.
For CADR 240 m³/h in a 52 m³ room, k≈4.615 h⁻¹. After 30 minutes, or 0.5 h, the fraction remaining is e^(-4.615×0.5)≈0.099. The ideal model predicts about 9.9% remaining, or roughly a 90.1% reduction from the starting value.
Half-life translates a rate into an intuitive time
For exponential decay, the half-life is ln(2)/k. With k=4.615 h⁻¹, the modelled half-life is 0.693/4.615=0.150 h, or about 9.0 minutes. After two half-lives about one-quarter remains; after three, about one-eighth remains.
The half-life is not a fixed property of “dust.” It belongs to the specified model and conditions. If the door opens, a source continues, the fan speed changes or air does not mix well, the observed curve can differ.
Students should state “the modelled half-life under these assumptions” rather than “the purifier clears the room in nine minutes.” The first is a defensible mathematical statement; the second hides the meaning of “clear.”
Multiple processes can be added as rates in a simple model
Natural deposition, ventilation and filtration may all remove particles. If they behave approximately as independent first-order processes, their rate constants can be added: k_total=k_filter+k_ventilation+k_deposition.
Suppose filtration contributes 4.6 h⁻¹, ventilation 0.5 h⁻¹ and deposition 0.2 h⁻¹. The simple total is 5.3 h⁻¹. The predicted fraction remaining after 20 minutes, or one-third of an hour, is e^(-5.3/3)≈0.171.
This addition is a model, not a universal law for every room. Ventilation can also introduce outdoor particles. Deposition can later be disturbed. Some particles and surfaces behave differently. Mathematics is strongest when the assumptions are visible.
A continuing source changes the equation
If particles enter at a steady rate S while removal occurs, a simple mass-balance model is dC/dt=S/V-kC. The long-run concentration approaches S/(Vk) rather than zero.
Imagine a source adds 5,000 arbitrary particle units per hour to a 50 m³ room, and k=4 h⁻¹. The steady model gives 5,000/(50×4)=25 units per cubic metre. Doubling k to 8 h⁻¹ halves this theoretical steady value to 12.5.
That calculation shows why source control matters. Increasing removal can reduce concentration, but eliminating or reducing the source changes the numerator. The EPA’s broader guidance treats filtration as part of an indoor-air strategy, not a licence to ignore sources.
Filters, Pressure Drop and Operating Points
A filter does not act independently of the fan
Air must pass through pre-filters, fine filters, grilles and internal passages. Each adds resistance. A fan produces a pressure rise, and the system consumes that rise through losses. The airflow settles where the fan’s pressure-flow curve intersects the system’s resistance curve.
If a filter becomes loaded, the resistance may rise. The operating point can move to lower airflow. A nameplate fan speed does not guarantee unchanged CADR throughout every filter condition.
This is why “higher efficiency filter” is not a complete system claim. The media’s capture performance, its pressure drop, the fan and the sealing all matter. Air bypassing the filter is not treated by the media at all.
A toy pressure-drop model
For a safe paper exercise, let system pressure loss be Δp=RQ², where Q is airflow and R is a fitted resistance coefficient. If R=0.020 Pa/(m³/h)² and Q=100 m³/h, the model loss is 200 Pa.
If filter loading raises R by 50% to 0.030, holding 100 m³/h would require 300 Pa. A real fan may not supply that extra pressure at the same flow, so Q falls until fan and system curves meet.
The quadratic form is illustrative. Real components can show more complicated behaviour, and units of R depend on the chosen units for Q. Students should not copy a coefficient from one setup into another without evidence.
Filter replacement is a threshold decision
A replacement reminder based only on calendar time ignores usage and conditions. A pressure sensor, runtime counter or airflow estimate may support a more informative threshold, but each needs calibration.
Suppose a clean-state differential pressure is 90 Pa and a service threshold is 180 Pa. A reading of 170 Pa is not “94% used” unless pressure rise is known to track usable life linearly. It may simply be 80 Pa above the clean baseline and 10 Pa below a declared threshold.
Thresholds should come from the manufacturer or a validated maintenance method. A student model can compare rules, but it should not invent a safety or health limit.
Leakage can dominate a good material
Suppose 90% of airflow crosses media that captures 95% of a target particle, while 10% bypasses the filter. Of 100 particle units entering, 90 go through media and 4.5 of those pass; 10 bypass. Total passing is 14.5, so system removal is 85.5%, not 95%.
If bypass rises to 20%, media captures 0.95×80=76 units and 24 pass. System removal falls to 76%. The calculation explains why construction and fit matter as well as a media label.
This example is deliberately simplified. Particle size, leakage paths, fan behaviour and repeated passes affect real results. Its value is conceptual: a component rating and a whole-device outcome are different quantities.
Sound, Energy and Usability
Decibels use a logarithmic scale
Sound-pressure level is logarithmic. Two equal independent sound sources do not normally add by ordinary arithmetic. Combining two 50 dB sources gives about 53 dB, not 100 dB, under the standard energy-sum model.
The combination formula is L_total=10 log₁₀(10^(L₁/10)+10^(L₂/10)). For 45 dB and 50 dB, the result is about 51.2 dB.
Room reflections, distance, frequency weighting and measurement method influence a reading. The arithmetic does not tell whether someone will find a sound acceptable during sleep or study. Usability includes context and preference.
Energy is power multiplied by time
A 45 W purifier running for 10 hours consumes 45×10=450 Wh, or 0.45 kWh. Running every day for 30 days gives 13.5 kWh in this simple calculation.
A 70 W high setting used for four hours consumes 0.28 kWh. A 35 W setting used for eight hours also consumes 0.28 kWh. Equal energy does not imply equal air cleaning because CADR may differ by setting and the concentration curve depends on timing.
The useful comparison might be energy per cubic metre of clean air: input power divided by CADR. At 45 W and 240 m³/h, the ratio is 0.1875 Wh per cubic metre of rated clean air. It is a comparative indicator under rating conditions, not a complete environmental assessment.
Scheduling is a control problem
Because early exponential decay can be rapid, running before occupancy may reduce an accumulated concentration in the model. A continuing source, open window or changing room use can alter the best schedule.
Consider two equal-energy plans: 60 W for two hours or 30 W for four hours. If CADR is proportional to power, their ideal integrated clean-air delivery may be similar. If fan efficiency and CADR are nonlinear, they can differ. Published performance at each setting is needed.
This is a useful optimisation lesson: objective, constraints and response curve must be known. “Run less” and “run at maximum” are not mathematical solutions until the goal is stated.
Evidence, Uncertainty and Fair Comparisons
Test methods make numbers comparable
CADR values are meaningful because they refer to a defined method. The public AHAM filtration standards page explains that separate CADR values are reported for common particle categories. A student should not silently replace one category with another.
The EPA technical summary on residential air cleaners discusses performance and limitations. A real room may have furniture, leakage, irregular mixing, doors and sources that are absent from a chamber.
A fair comparison therefore records the test basis, fan setting, filter state and units. “Model A has a higher smoke CADR than Model B under the listed rating” is stronger than “Model A cleans every room better.”
Sensor readings have noise and bias
Low-cost particle sensors can respond to humidity, particle composition, placement and airflow. Repeated readings show variation but do not by themselves reveal systematic bias.
Suppose five readings are 12, 15, 13, 16 and 14 units. Their mean is 14, and their range is 4. Repeating a condition can estimate scatter. Checking against a reference or known procedure is needed to assess accuracy.
Plot raw data, mark the start time and retain units. Smoothing may reveal a trend, but it can also hide a short peak. Report both the chosen method and its effect.
Baselines and controls prevent false credit
If concentration falls naturally even when the purifier is off, a before-and-after purifier run cannot attribute the whole decline to the device. A matched off-condition baseline is needed.
Suppose the on-condition falls from 100 to 30 in 30 minutes, while the off-condition falls from 100 to 70. The purifier is associated with an additional decline in that setup, but subtracting percentages directly is not a complete dynamic model.
Alternating conditions, using the same room and waiting for comparable starts can improve the design. Safety and ethics still come first: do not create pollution to make the experiment dramatic.
Uncertainty belongs in the conclusion
If repeated fitted rate constants are 3.8, 4.2, 4.0, 4.1 and 3.9 h⁻¹, the mean is 4.0 h⁻¹ and the observed range is 3.8–4.2. Reporting only 4.000 suggests unjustified precision.
If two settings give means of 4.0 and 4.1 with similar run-to-run variation, the data may not separate them convincingly. “The runs overlapped under this method” is a useful result.
Mathematics does not force every comparison to produce a winner. It can show when evidence is too weak for a strong ranking.
A Worked Room Example
Define the room and cleaner
Consider a fictional study room 4.8 m long, 3.5 m wide and 2.7 m high. Its volume is 45.36 m³. A purifier has rated smoke CADR of 200 m³/h on high and 120 m³/h on medium.
The ideal purifier-only rate constants are 200/45.36=4.41 h⁻¹ and 120/45.36=2.65 h⁻¹. Corresponding half-lives are 0.693/4.41=9.43 minutes and 0.693/2.65=15.7 minutes.
After 30 minutes, the ideal fractions remaining are e^(-4.41×0.5)=0.110 on high and e^(-2.65×0.5)=0.266 on medium. These are model outputs, not health promises.
Add a background removal rate
Assume a teacher-provided background rate of 0.4 h⁻¹. Total rates become 4.81 and 3.05 h⁻¹. After 30 minutes, fractions become about 0.090 and 0.218.
If outdoor particles enter through ventilation, the source term must also be included. Treating ventilation only as removal would be incomplete. The direction and amount depend on indoor and outdoor concentrations.
The example shows why a single “air changes” number cannot describe every situation. It is one parameter inside a model.
Compare energy for a planned session
Suppose high draws 55 W and medium 32 W. High for 30 minutes uses 27.5 Wh; medium uses 16 Wh. High gives greater modelled reduction, while medium uses less energy and may be quieter.
A decision needs priorities: target reduction, available time, sound tolerance and energy. A student can make a table instead of pretending there is one universally best setting.
Do not infer a medical threshold from the model. If air quality presents a health concern, follow authoritative public-health or professional advice.
Fifteen Safe Student Investigations
1. Convert room labels into a common basis
Collect fictional or teacher-selected room dimensions and compute volumes. Convert recommended floor areas into implied volumes under several ceiling heights.
Explain why the same square-metre claim yields different air-change rates in rooms with different heights. State every assumed height.
2. Build a unit-conversion audit
Convert three CADR values between ft³/min and m³/h. Write the units beside every factor and reverse one calculation as a check.
Compare rounding to the nearest whole unit with retaining one decimal place. Decide which precision is useful for a consumer table.
3. Plot ideal exponential decay
Choose C₀=100 and k values of 2, 4 and 6 h⁻¹. Calculate C at five-minute intervals for one hour.
Plot all three curves on the same axes. Describe why equal time steps do not produce equal absolute decreases.
4. Compare half-life and time-to-90%-reduction
For each k, calculate half-life ln(2)/k and t₉₀=ln(10)/k. Notice that 90% reduction takes about 3.32 half-lives.
Explain why neither time means “zero particles.”
5. Explore room-size sensitivity
Hold CADR at 180 m³/h while varying room volume from 30 to 90 m³. Compute k and half-life.
Graph half-life against volume. Identify whether the relationship is linear under the simple model.
6. Test a continuing-source model in a spreadsheet
Use a recurrence C_next=C_now+Δt(S/V-kC_now). Choose a small Δt and teacher-supplied fictional values.
Change S while holding k fixed. Observe how the long-run level changes.
7. Investigate filter bypass
Create a table of media efficiency from 80% to 99% and bypass fraction from 0% to 20%. Calculate whole-system removal for each pair.
Find a case where improving fit matters more than improving the media percentage.
8. Draw fan and system curves
Use fictional fan pressure Δp_fan=400-0.02Q² and system loss Δp_system=0.01Q². Find their approximate intersection.
Increase the system coefficient to represent loading and find the new point. Keep the exercise on paper.
9. Compare energy-normalised delivery
For fictional settings, divide watts by m³/h CADR. Rank the settings by this indicator.
Then explain why the ratio omits sound, room mixing, filter life and manufacturing impacts.
10. Combine decibel levels
Use the logarithmic energy-sum formula to combine 35 dB background with purifier levels of 35, 40 and 45 dB.
Compare results with ordinary addition and explain why the latter is invalid.
11. Model a door-opening event
Plot decay for 20 minutes, then add a sudden concentration increase and continue the curve. Label the event.
Discuss why fitting one exponential across the whole record would be misleading.
12. Compare geometric and arithmetic summaries
For concentration ratios across repeated intervals, compare an arithmetic average of percentage reductions with a product-based overall ratio.
Explain which summary matches the sequence and why.
13. Design a fair published-data comparison
Choose two certified products with the same CADR particle category. Record rating, power, room-size basis and setting.
Write one defensible sentence and one tempting but unsupported sentence. Explain the difference.
14. Audit significant figures
Start with room dimensions measured to the nearest 0.1 m. Calculate volume and air-change rate.
Compare a result reported as 4.376921 h⁻¹ with 4.4 h⁻¹. Explain why more digits do not create more information.
15. Write a model-limit statement
Take the worked room example and list five assumptions: mixing, constant CADR, no new source, stable filter and fixed room volume.
Change one assumption and predict the direction of error. This is mathematical reasoning even before new data are collected.
Misconceptions That Good Mathematics Corrects
“One air change replaces every molecule”
An air-change rate is a flow-to-volume ratio. In a well-mixed exponential model, one time constant leaves e⁻¹, about 36.8%, of the initial concentration. Real mixing can be less uniform.
“A HEPA label tells the whole device CADR”
A media property and a device delivery rate are different. Airflow, sealing and system design affect the whole-device result.
“The biggest recommended room is always the strongest”
Recommended area can be based on different target air-change rates and ceiling assumptions. Compare CADR and the declared basis.
“A sensor reading proves health protection”
A reading is evidence about the sensor’s response under conditions. Health interpretation requires appropriate pollutants, methods and authoritative guidance.
“Running at maximum is always optimal”
Maximum may reduce concentration faster in a model, but sound, energy, occupancy, source timing and nonlinear performance shape the decision.
How Students Can Transfer the Mathematics
Primary learners can work with rectangular volume, multiplication and rates. Secondary learners can add unit conversion, exponentials, logarithms, graph interpretation and uncertainty. Older students can study differential equations, fan curves, experimental design and optimisation.
The same habits appear in ventilation engineering, environmental monitoring, filtration, public health, building services and data science. Mathematics alone does not guarantee entry to any course or career. Current official admissions requirements, subject expectations and programme details should be checked separately.
Students who want a neighbouring airflow example can read Why Mathematics? | Vacuum Cleaners, Pressure Difference, Airflow and Filter Resistance. For measurement thinking, Why Mathematics? | Comparing Percentages Fairly is useful when labels turn test outcomes into claims.
Parents can ask: What does the number measure? What is the denominator? Which conditions are assumed? What evidence would change the conclusion? Those questions make mathematics practical and calm.
Questions Parents and Students Often Ask
Is CADR the same as airflow?
No. CADR represents equivalent clean-air delivery for a stated particle category and method. Raw airflow does not include the removal effect in the same way.
Why are there several CADR values?
Different particle categories are used in the rating system. Compare like with like and read the current rating label or certified directory.
Can CADR divided by volume predict exact real-room performance?
It creates a useful ideal rate under a well-mixed, constant-performance model. Doors, sources, furniture, placement, leakage and filter condition can make reality differ.
Does a larger purifier remove gases?
Particle CADR does not establish gas removal. Gas-phase performance depends on sorbent type, amount, contact time, contaminant and test evidence.
Can we test a purifier with incense?
Do not deliberately create indoor pollution for a home experiment. Use published data, safe clean-air observations or teacher-designed simulations.
How should a purifier be placed?
Follow the manufacturer’s instructions and keep inlets and outlets unobstructed. A mathematical model cannot override safety, electrical or placement guidance.
Does filtration replace ventilation?
No universal substitution follows from the arithmetic. Ventilation, source control and filtration have different roles, and outdoor conditions matter.
What is the most important mathematical habit?
Keep the chain visible: rated quantity, units, room volume, model, assumptions, uncertainty and the exact claim being made.
Continue Reading
Four Extended Air-Cleaning Case Studies
Case 1: The classroom that is not one well-mixed box
Imagine a rectangular classroom 9 m by 7 m by 3 m, giving 189 m³. Two 300 m³/h purifiers provide a nominal 600/189=3.17 equivalent clean-air changes per hour. The quick arithmetic looks encouraging, but placement determines whether air near every desk joins the same circulation.
Divide the room conceptually into a front zone of 70 m³ and a rear zone of 119 m³. If one unit mostly recirculates the front and the other reaches the rear, the zones may have different effective rates. A two-box model uses separate concentrations C₁ and C₂ plus an exchange flow between them.
Suppose the front receives 250 m³/h clean-air delivery, the rear 350 m³/h and 120 m³/h exchanges between zones. The model becomes two coupled equations rather than one exponential. A spreadsheet can update each zone every minute using mass balance.
If a source appears at the rear, the front sensor may remain low for several minutes. One sensor cannot represent the whole room. This is a measurement-location issue, not proof that the device has failed.
Students can compare the one-box and two-box predictions. The purpose is not to find a perfect classroom number but to see when added complexity changes the decision.
Case 2: A filter that loads during an event
Consider a fictional purifier with clean-filter CADR 260 m³/h in a 52 m³ room. During a long dusty event, CADR falls linearly to 190 m³/h over four hours. The instantaneous rate k(t)=CADR(t)/52 therefore changes with time.
Using the starting CADR for the whole event would overestimate total delivery. The average of a linear decline is (260+190)/2=225 m³/h, giving integrated clean-air volume 900 m³ over four hours.
If a new source is absent and the well-mixed assumptions hold, the decay exponent uses the integral of k over time. Integrated k is 900/52=17.31, leaving e^-17.31 of the initial concentration—numerically tiny in the ideal model.
That result shows another model limit. Real concentrations do not fall indefinitely below background measurement noise, and infiltration or sources usually matter. The mathematically tiny output should prompt a floor term, not a boast about perfect air.
Maintenance decisions can track the observed CADR proxy, differential pressure or scheduled inspection. A student must not infer a replacement threshold from this invented curve.
Case 3: Outdoor air reverses a simple ventilation story
Let indoor concentration be 20 units and outdoor concentration 80. Ventilation at 0.6 h⁻¹ introduces outdoor air as well as removing indoor air. The ventilation term is λ(C_out-C_in), positive while outdoors is higher.
With purifier rate 4 h⁻¹ and no indoor source, steady concentration in the simple model is λC_out/(λ+k)=0.6×80/4.6≈10.4 units. Without the purifier, the room would approach 80.
If outdoors later falls to 5 units, ventilation becomes beneficial and the new combined steady level is 0.6×5/4.6≈0.65. The same ventilation rate participates in opposite concentration changes because the boundary condition changed.
This is why “open a window” or “close every window” cannot be derived from air-change arithmetic alone. Outdoor conditions, pollutant type, heat and official advice matter.
Students can plot the steady indoor value against outdoor concentration. The straight line’s slope is λ/(λ+k), revealing how filtration changes sensitivity without erasing the outside world.
Case 4: Comparing a large quiet unit with a small fast unit
Purifier A supplies 250 m³/h at 38 dB and 40 W. Purifier B supplies 320 m³/h at 52 dB and 58 W. In a 50 m³ room, ideal rates are 5.0 and 6.4 h⁻¹.
Time to 90% reduction under a purifier-only model is ln(10)/k. A needs 0.461 h, or 27.6 minutes; B needs 0.360 h, or 21.6 minutes. The difference is six minutes.
Energy to that endpoint is about 18.4 Wh for A and 20.9 Wh for B. B is faster, while A is quieter and uses slightly less modelled endpoint energy.
Combining the purifier with a 35 dB background gives total levels of about 39.8 dB for A and 52.1 dB for B under the independent-source formula. Perceived sound and frequency content still require human context.
A family could prefer A for study and B for rapid unoccupied cleaning. Mathematics clarifies the trade-off; it does not impose one preference.
Air-purifier mathematics begins with a rate divided by a volume, but it quickly becomes a rich study of exponentials, mass balance, logarithms, system curves and evidence. That progression is a good answer to “why mathematics?”: the formulas help, but the deeper benefit is learning to name what is known and what remains uncertain.
The EPA’s Guide to Air Cleaners in the Home and the AHAM Verifide air-cleaner directory are useful starting points for current product-independent interpretation. Read the actual manufacturer manual for operation and maintenance.
For a broader study route, continue with eduKateSG’s verified mathematics articles and the Bukit Timah Tutor Mathematics Learning Library. The aim is not to memorise one purifier formula. It is to become the person who can examine a quantitative claim, reconstruct its assumptions and make a proportionate decision.
That habit remains useful whenever a room, a rate and a changing concentration appear together.