Experiment 15: Build a Water Filter — Separation and Pollution

The water going in is brown. The water coming out is nearly clear, and every layer of the filter earned part of that change by a *different* mechanism. This experiment takes the filter apart — testing layers singly and in sequence — and makes one safety point that matters more than all the chemistry: clear is not the same as safe.

The science behind it

Dirt in muddy water is a mixture of systems. Coarse grit is stopped by straining — particles physically too big for the gaps. Finer sediment needs the slow, deep bed of filtration through sand, where particles lodge in the maze of pores. Dissolved colour and smells are molecules, far too small to strain; they are removed by adsorption onto activated charcoal, whose internal surface is enormous — around 1,000 m² of it in a single gram — so passing molecules stick and stay. And some things no kitchen filter touches: dissolved salts sail through every layer (they are single ions, smaller than any pore), and bacteria, at about a thousandth of a millimetre, pass through sand and gravel too. Turbidity is measured in NTU (nephelometric turbidity units): muddy water here starts near 500 NTU; treated drinking water is usually below 1 NTU — and it is also *disinfected*, a step this filter does not have.

Variables. Independent variable: filter construction (which layers, in which test). Dependent variable: clarity/turbidity of the output. Controlled variables: dirty-water recipe, volume filtered, pour rate, cup and depth of each layer.

Safety first

• Never drink the filtered water, however clear it looks. It still contains bacteria and dissolved substances. This is the lesson, not a technicality.
• Label the cups clearly ("DIRTY", "FILTERED — DO NOT DRINK") so nobody confuses them later.
• An adult cuts the bottle. Wash hands after handling dirty water, and rinse charcoal before use — its dust blacks the first water through.

Doing the investigation

Hands-on. Build the filter in a cut bottle, tip down: coffee-filter paper, then charcoal, fine sand, coarse sand, gravel on top. Prime it with clean water and discard that first flush. Then pour 500 mL of muddy water through slowly, rating the output's clarity (1–5) and smell. Separately, filter identical dirty water through single layers (cotton only, gravel only, sand only, charcoal only) to isolate each layer's contribution.

In the virtual lab below (Clean Water Challenge) tick the filter layers you want, choose a pollutant (muddy water 500 NTU, food colouring 200 NTU, cooking oil 350 NTU), press Build & filter, and read the input and output turbidity and the percentage removed, with the stage-by-stage chart showing what each layer caught. Follow its guided procedure: gravel alone, the full stack, then colouring with and without charcoal. Log the stage numbers in your observation log.

Worked investigation: what each layer actually did

Full stack on muddy water, stage by stage:

And the single-layer tests explain the chart. Cotton alone took 500 → 310 NTU (38% removed) — a strainer, nothing more. For food colouring (200 NTU), the story flips: sand barely touched it (200 → 190), but the charcoal stage dragged it down to 35 NTU, removing about 82% of the colour that reached it. The largest single NTU drop for mud was the coarse-sand stage (170 NTU).

Total removal, full stack = (500 - 25)/500 x 100 = 95%
Charcoal stage on colour  = (190 - 35)/190 x 100 = about 82%
Yet the output is 25 NTU - drinking water is < 1 NTU *and* disinfected

Conclusion. Each layer owns a particle size range and a mechanism: straining for the big, deep-bed filtration for the fine, adsorption for the dissolved. Order matters — gravel first protects the sand from clogging with grit. The stack removes 95% of turbidity and most colour and smell, and it is *still* not drinking water, because the hazards you cannot see (bacteria, dissolved chemicals) are exactly the ones these mechanisms miss.

Uncertainty. Hands-on clarity ratings are a 1–5 eyeball scale, ±1 at least — two observers will disagree. A cheap sharpening trick: read printed text through the sample (a mini Secchi test) and record the smallest readable font, or photograph samples against a white card under fixed lighting. The virtual lab's NTU values are exact, which is precisely why pairing the two matters.

Evaluation: how good is this evidence?

• Channelling. Poured too fast, water carves one highway through the layers and bypasses most of the bed. Improvement: pour slowly onto a "splash guard" (a small piece of filter paper on the gravel) to spread the flow.
• Layers mix during assembly, blurring which layer did what. Improvement: separate layers with filter-paper discs; dismantle afterwards and inspect each layer's trapped load.
• First-flush contamination. Loose charcoal dust and sand fines wash into the first output. Improvement: discard the first 50 mL, as done in the priming step.
• The dirty water itself varies between batches (how much soil, how well stirred). Improvement: make one large batch, keep it stirred, and draw all tests from it.

Common mistakes

• Drinking, or letting anyone drink, the output because it "looks clean".
• Concluding the filter "removed the dirt" without asking what it cannot remove (salt, bacteria).
• Building the stack in reverse order (fine layers first clog instantly under raw grit).
• Judging by one quick glance instead of a consistent clarity scale or turbidity reading.

Exam Experience

Question 1 — calculation and analysis [4 marks]. A filter processes muddy water: in at 500 NTU, after gravel 380 NTU, after coarse sand 210 NTU, after fine sand 85 NTU, after charcoal 40 NTU. (a) Calculate the percentage of turbidity removed by the whole filter. (b) Which layer removed the most turbidity in absolute NTU, and how much?

Scoring guide:

• Total removed = 500 − 40 = 460 NTU; 460 ÷ 500 = 92% [1]
• Gravel removed 500 − 380 = 120 NTU [1]
• Coarse sand removed 380 − 210 = 170 NTU — the largest absolute drop [1]
• (Fine sand 125 NTU and charcoal 45 NTU for comparison; percentage and absolute contributions must not be confused) [1]

Question 2 — evaluation and design [4 marks]. A classmate says: "Our filtered water came out clear, so it is safe to drink." Evaluate the claim, and describe one addition a real treatment works makes that this filter lacks.

Scoring guide:

• Claim rejected: clarity measures suspended particles only [1]
• Bacteria/viruses and dissolved substances pass through these layers largely untouched [1]
• Missing step identified: disinfection (chlorination, UV or ozone) after filtration [1]
• Supporting evidence suggested, e.g. output turbidity ~25 NTU vs <1 NTU standard, or bacterial testing would be needed before any safety claim [1]

Multiple choice 1. Dissolved salt passes through the entire filter because:

• A. The layers were built in the wrong order
• B. Salt ions are far smaller than the pores in every layer, and are not meaningfully adsorbed
• C. Salt is heavier than water
• D. Charcoal only works on Tuesdays

Answer: B. Filtration and adsorption act on particles and molecules above a certain scale; individual dissolved ions are below it.

Multiple choice 2. Activated charcoal's special contribution is adsorption, which works because:

• A. It is black, so it absorbs colour visually
• B. Its structure gives it an enormous internal surface area (about 1,000 m² per gram) where dissolved molecules stick
• C. It reacts with mud to make it dissolve
• D. It magnetically attracts dirt

Answer: B. Adsorption is surface capture; activated charcoal is engineered to maximise surface, which is why it strips dissolved colour and odour that straining cannot touch.

Going deeper: how a treatment works actually does it

Your bottle filter is the back half of a municipal treatment works, in miniature. The full sequence runs: screening (bars strip out the big debris — your gravel's coarse cousin), coagulation and flocculation (dosed chemicals neutralise the charges on fine clay particles — Experiment 14's isoelectric trick — so they clump into fluffy *flocs* big enough to settle), sedimentation (hours in quiet tanks while flocs sink), filtration (through deep sand beds, your fine-sand stage at building scale), and finally disinfection (chlorine, ozone or ultraviolet light) — the step your filter lacks and the reason its output is not drinkable.

Two details reward a closer look. The classic slow sand filter grows a living skin on its surface — the *schmutzdecke*, a biofilm of microbes that grazes on bacteria and fine particles passing through, turning a strainer into an ecosystem; it is filtration plus biology, and it must mature for days before the water is trusted. And emergency ceramic pot filters, used worldwide where treatment works do not reach, fire clay with sawdust that burns out to leave fine pores, then coat the ceramic with colloidal silver — the silver doing, chemically, the disinfecting job your stack cannot, so the output *can* be drunk. Line the three up — bottle filter, slow sand filter, ceramic-and-silver filter — and you have a history of water engineering as a series of answers to one question: the particles are gone, so what is still in the water, and what removes *that*?

What a turbidity number feels like

NTU is measured by shining light through the sample and detecting how much the suspended particles scatter sideways — a nephelometer — calibrated against a standard suspension called formazin. To anchor the scale: treated tap water sits below 1 NTU and looks perfectly clear; your stack's 25 NTU output looks *slightly* hazy against a bright window; a river in flood can run past 1,000 NTU and look like milky coffee. The gap between 25 and 1 is why treatment works do not stop at filtration — and why "it looks clear" is a statement about human eyes (poor nephelometers, easily fooled at dusk) rather than about the water.