Diffusion & Osmosis Lab · Guide · glidepathscience.org
GuideDiffusion & Osmosis Lab
Contents
This Guide explains the ideas and equations behind every Walkthrough step, Concept Challenge question and Predict round, in the order a cell biology text would take them. Part 1 prepares the Intro track and Part 2 the Advanced track; each takes about fifteen minutes to read. Each section ends with a worked example that is not one of the sim's own problems.
Part 1 · The essentials
Every molecule wanders
Over the short distances in and around cells, molecules move mostly by diffusion. Water molecules strike a dissolved molecule from every side, trillions of times a second, and each blow sends it a tiny distance in a direction unrelated to the last: its path is a random walk.
Nothing pulls the molecule toward empty water or pushes it out of a crowd. Yet a drop of dye spreads, always from more to less, by counting, not force: near any line through the solution each molecule is as likely to step across one way as the other, but the crowded side has more molecules to take that step.
Random steps also go nowhere fast. After $N$ steps of length $a$ along a line, the typical distance from the start, the root-mean-square $x_{rms}$ over many walkers, is only
$x_{rms} = a\sqrt N$
because steps forward and back mostly cancel.
The usual mistake is to give molecules intentions: they "seek" open space, or stop once spread out. They never stop; in an even solution only the average stops changing.
Two-way traffic, and equilibrium that never stops
Whatever crosses the membrane decides what a cell gets. If a membrane lets a solute through, molecules on each side wander across at random, and the number crossing from a side each second is proportional to its concentration. Both flows run at once, and the flow down a gradient is their difference:
$\text{net into the inside} = N_{in} - N_{out}$
Here $N_{in}$ and $N_{out}$ count the crossings into and out of the inside in some interval. Each crossing is a random event, so the net carries a counting noise of about $\sqrt{N_{in} + N_{out}}$.
When the two sides reach the same concentration, molecules arrive at the membrane equally often from both sides and the net is zero on average, yet they still cross both ways in large numbers. This is dynamic equilibrium: nothing changes on average because two equal and opposite flows continue.
Mistaking equilibrium for stillness is the common slip. What sets the size of the net flow comes next.
What sets the rate: Fick’s law in words
In 1855 Adolf Fick found the rule for how fast a substance arrives. The net rate of diffusion across a barrier grows with the area $A$ open to it and with the concentration difference $\Delta c$ across it, and shrinks as the barrier thickens:
$\text{rate} \propto \dfrac{A\,\Delta c}{\text{thickness}}$
More area gives more places to cross, a larger $\Delta c$ a larger imbalance between the two flows, and a thicker barrier a longer walk. How fast each molecule wanders is measured by its diffusion coefficient $D$ (m²/s), near $10^{-9}$ m²/s for small molecules in water and larger for smaller molecules in warmer, thinner liquids: roughly proportional to the absolute temperature over the viscosity η (Part 2).
A larger difference does not make molecules move faster; more of them cross. And thickness costs more than you might guess.
Twice as far, four times as long
A random walk's spread grows as the square root of time, so the time diffusion needs grows as the square of the distance. The time for the spread along one direction to reach a distance $x$ is about
$t \approx \dfrac{x^2}{2D}$
It is a typical time, not an arrival time: some molecules get there sooner and some later. Twice the distance takes four times as long, ten times the distance a hundred times as long.
The mistake is to scale the time with the distance, as if diffusion had a speed. What is quick across a cell is hopeless across an organ, which is why animals pump blood, and one reason cells are small.
Why cells are small: surface and volume
Everything a cell takes in crosses its surface, and everything it uses is consumed in its volume, which grows faster. A cube of side $s$ has surface $6s^2$ and volume $s^3$; a sphere of radius $r$ has $4\pi r^2$ and $\tfrac43\pi r^3$. The surface per volume is
$SA/V = 6/s$ (cube) $A/V = 3/r$ (sphere)
Double the size and each unit of volume has half the membrane to supply it, while its middle is twice as far from the surface, at four times the diffusion time. Large cells change shape instead: a red blood cell is a flattened disk, gut cells fold their surface into microvilli, and a nerve cell is long and thin.
A larger cell has more surface, but not enough: surface goes as the size squared, volume as its cube. What the surface lets through comes next.
What crosses the bilayer
A cell membrane is a lipid bilayer about 5 nm thick, with water-loving heads outward and an oily core of lipid tails. It has no holes: a molecule crosses by dissolving in the core, diffusing across and dissolving back into water. So the bilayer sorts molecules mainly by how well they dissolve in oil, then by size. Small nonpolar molecules (O₂, CO₂, N₂) and steroid hormones cross easily; small polar ones (water, urea, ethanol) slowly; glucose and every ion practically never, since an ion would have to shed its shell of water. A membrane's ease of passage for one molecule is its permeability $P$, in m/s, and the flux, the amount crossing each square meter each second, is
$J = P\,\Delta c$
Bilayer permeabilities span about twelve powers of ten, following oil solubility: Overton's rule, from the 1890s.
The usual mistake is to picture the bilayer as a sieve. For what the lipid stops, cells use proteins.
Channels and carriers
Membrane proteins give polar molecules and ions their own routes. A channel is a selective pore that opens and closes: a potassium channel passes K⁺ about a thousand times more readily than the smaller Na⁺, some 10⁷ to 10⁸ ions a second when open; aquaporins are channels for water. A carrier binds one molecule, changes shape and releases it on the other side, a cycle per molecule, so it manages only some 10² to 10⁴ a second.
Both are facilitated diffusion: the protein provides the route and the concentration difference the drive. Neither uses energy or can move anything uphill; that takes a pump spending ATP, which is active transport. A protein moving $k$ molecules a second moves
$n = k\,t$
in a time $t$, as long as the gradient lasts.
Read "facilitated" as "given a route", never "given energy". A carrier's cycle also gives it a limit a channel lacks.
Carriers saturate
At low concentration a carrier spends most of its time empty, and its rate rises in proportion to the concentration. At high concentration every carrier is nearly always busy, and the rate levels off at a ceiling, $J_{max}$, that only more carriers could raise. With the inside kept near zero, the rate is
$J = J_{max}\,\dfrac{c}{K + c}$
where $c$ is the concentration outside and the half-saturation constant $K$ is the concentration at which the carrier runs at half its maximum.
The mistake is to expect every route to act like a channel, where doubling the concentration doubles the flow; a flattening curve is a carrier's signature. Water has its own routes.
Osmosis: water follows dissolved particles
Suppose a membrane passes water but not sucrose, and one side has more sucrose. The sucrose cannot even out, so the water does: it moves toward the side with more dissolved particles and dilutes it. This is osmosis. Water crosses through aquaporins and, more slowly, the lipid, and it crosses both ways; the net flow is the difference.
What matters is the number of dissolved particles. A sugar molecule counts once, but NaCl splits into Na⁺ and Cl⁻ and counts twice. The particle concentration, the osmolarity, is
$\text{osmolarity} = iM$
where $M$ is the molarity in mol/L and $i$ the number of particles per formula unit (1 for sugars, 2 for NaCl, ideally). Water moves toward the higher osmolarity.
Many texts say water diffuses down its own concentration gradient: a shortcut that predicts the direction but is not the mechanism (Part 2). The error to avoid is sending water toward less solute, or moving the solute instead.
Hypotonic, isotonic, hypertonic
A bath's tonicity describes what it does to a cell, comparing the particles outside that stay outside, $c_{out}$, with the cell's own, $c_{in}$:
$c_{out} < c_{in}$: hypotonic $c_{out} = c_{in}$: isotonic $c_{out} > c_{in}$: hypertonic
Water moves in, not at all, or out. A red blood cell has no wall: hypotonic, it swells toward a sphere and bursts (lyses) at the sphere its area allows; hypertonic, it shrinks and puckers (crenates). A plant cell has a stiff wall: hypotonic, it swells until the wall pushes back and stops the inflow (turgid, the state that holds a plant up); isotonic, it is limp (flaccid); hypertonic, its contents pull away from the wall (plasmolysis).
A plant cell in pure water does not burst; its wall is the difference. To put numbers on this, botanists use water potential.
Water potential
The water potential Ψ measures water's tendency to move, in megapascals or bars (10 bar = 1 MPa); pure water at atmospheric pressure has Ψ = 0. Dissolved particles lower it (a negative solute potential $\Psi_s$) and pressure raises it (a pressure potential $\Psi_p$, positive in a turgid cell). Water moves from higher Ψ to lower:
$\Psi = \Psi_s + \Psi_p$ $\Psi_s = -iCRT$
with $i$ the particles per formula unit, $C$ in mol/L, $T$ in kelvins and $R$ = 0.0831 L·bar/(mol·K), giving bars. At equilibrium a cell's Ψ equals the bath's; a tissue's isotonic point is the bath in which it neither gains nor loses water.
The slips are a positive $\Psi_s$, a temperature in °C, and forgetting $i$ for a salt. Part 2 returns to these topics with the physics behind them.
Part 2 · Going further
The random walk in numbers
Averaged over many molecules, the displacement from the start is zero, since no direction is preferred, but the mean-square displacement grows steadily with time:
$⟨x^2⟩ = 2Dt$ per axis $⟨r^2⟩ = 4Dt$ on a plane $⟨r^2⟩ = 6Dt$ in space
Plotted against time, the mean-square displacement along one axis rises with slope $2D$; for steps of length $a$ every $\tau$, $D = a^2/2\tau$. Many molecules released at one point spread into a Gaussian of variance $2Dt$, the solution of the diffusion equation (Fick's second law):
$\dfrac{\partial c}{\partial t} = D\,\dfrac{\partial^2 c}{\partial x^2}$
Its area counts the molecules and stays fixed, so as it widens as $\sqrt t$ its peak falls: as $1/\sqrt t$ along a line, $1/t$ on a plane, $1/t^{3/2}$ in space.
The usual mistake is to use $2Dt$ for a track on a plane, which doubles $D$; and a mean-square displacement growing as $t^2$ is drift, not diffusion. What sets $D$, Einstein answered in 1905.
Stokes–Einstein: what sets D
Thermal kicks against viscous drag give, for a sphere of radius $r$ in a liquid of viscosity η at absolute temperature $T$,
$D = \dfrac{k_B T}{6\pi\eta r}$
with Boltzmann's constant $k_B$ = 1.380649 × 10⁻²³ J/K. It treats the liquid as smooth, so it holds for particles much larger than water molecules (about 0.3 nm), such as beads and proteins, and only roughly for small molecules. Around 1908 Jean Perrin measured $D$ for beads under a microscope and, knowing their size, found $k_B$ and Avogadro's number, $N_A = R/k_B$.
The classic slip is the diameter in place of the radius, which halves $D$. Measuring from a few hundred dots raises a question: how much do counts scatter?
Counting noise
When events happen at random and independently, as when molecules land in a region or cross a membrane, a count that averages $N$ scatters by about $\sqrt N$, so its relative scatter is
$\dfrac{\sqrt N}{N} = \dfrac{1}{\sqrt N}$
Four times the counts halve it. Each Sandbox dot stands for millions of molecules, so a real solution is thousands of times smoother; but in a bacterium a protein at a few nanomolar numbers only a handful of copies, and neighboring cells differ noticeably.
The slip is to think more molecules leave the relative noise unchanged. With many, counts become a smooth concentration.
Fick’s first law
Where the concentration changes along $x$, more molecules sit on one side of any small step than the other, and the imbalance drives a net flux down the slope. Fick's first law makes it exact: the flux $J$, in mol m⁻² s⁻¹, is
$J = -D\,\dfrac{dc}{dx}$
The minus sign sends the flux from high concentration to low; in SI units $c$ is in mol/m³, and 1 mM is exactly 1 mol/m³.
Nearly every wrong answer comes from units: micrometers left unconverted, or millimolar taken as 10⁻³ mol/m³. A membrane is described instead by one number.
Permeability, and how fast a cell fills
Across a membrane, $J = P\,\Delta c$. For a bilayer, a molecule must dissolve in the oily core and diffuse across it, so
$P = \dfrac{K D_m}{L}$
with $K$ the partition coefficient (the molecule's solubility in oil over that in water), $D_m$ its diffusion coefficient in the membrane and $L$ the thickness: Overton's rule made quantitative. A cell of volume $V$ and area $A$ in a large bath at $c_{out}$, well mixed inside, fills as $V\,dc_{in}/dt = PA(c_{out} - c_{in})$: the gap closes exponentially with time constant $V/(PA)$, which for a sphere ($A/V = 3/r$) is
$\tau = r/3P$
After one τ the gap is 1/e (37 %) of its start, after three τ 5 %. "Well mixed" needs $Pr/D \ll 1$: the inside mixes faster than the membrane passes molecules.
The usual slip is $A/V = 1/r$, which triples τ. Whether the membrane is really the slow step depends on the water around it.
Membrane-limited or diffusion-limited?
A molecule crossing a membrane also crosses the still water beside it, the unstirred layer. The two are resistances in series, $1/P$ for the membrane and $L/D$ for a layer of thickness $L$, and their ratio says which sets the rate:
$\dfrac{PL}{D}$
When $PL/D \ll 1$ the membrane is the bottleneck: the water on each side stays nearly uniform and the concentration steps sharply across the membrane. When $PL/D \gg 1$ the concentration slopes through the water instead, and stirring speeds the flux. Together, the effective permeability is $P/(1 + PL/D)$.
The mistake is to assume the membrane always sets the rate: for O₂, with its enormous $P$, the water does. For carriers, the protein's own cycle does.
The carrier cycle
A carrier is empty or loaded, facing out or in. An empty carrier binds a molecule on the side it faces; it flips, loaded or not; a loaded carrier that flips releases its molecule. If it flips equally fast in every state and binding is fast compared with flipping, it faces each way half the time, loaded in proportion to $c/(K + c)$ there, and the net flux per carrier is
$J = J_{max}\left(\dfrac{c_o}{K + c_o} - \dfrac{c_i}{K + c_i}\right)$
where $c_o$ and $c_i$ are the concentrations outside and inside and $J_{max}$ is half the flip rate. Molecules inside ride out too, so the net changes sign with the gradient: a carrier works both ways, always downhill.
That is the classic error: $J_{max}\,\Delta c/(K + \Delta c)$ has no basis, because each side saturates on its own. Water has no carrier, but it has a pressure that measures its drive.
Osmotic pressure
Seal a solution behind a membrane that stops its solute, with pure water outside and a pressure gauge. Water flows in until the pressure stops it: the osmotic pressure Π. For an ideal, dilute solution, van 't Hoff found it is the solute particles' ideal-gas pressure:
$\Pi = iMRT$
with $iM$ the particle concentration in mol/m³ (1 mol/L = 1,000 mol/m³), $R$ = 8.314 J/(mol·K) and $T$ in kelvins, giving pascals. "Dilute" means the particles are few compared with the water molecules and hardly interact; the first correction is an osmotic coefficient φ multiplying the formula, near 0.93 for NaCl at body-fluid strength.
The slips are leaving $M$ in mol/L, a thousand times too small, and using °C. Pressures this large raise a question: what actually pushes the water?
What really drives osmosis
The shortcut that water "diffuses down its own concentration gradient" fails a test. A solute lowers water's own concentration (55.3 mol/L) by the room its particles take: a sucrose molecule displaces about a dozen water molecules, an Na⁺ ion very little. Yet at the same osmolarity they drive the same flow: the drive counts particles.
The drive is a pressure. A solute particle that cannot pass is turned back at the membrane and pushes on it, as gas molecules push on a wall, with a total push of $RTc$ (the contact theorem). That lowers the water's pressure at the membrane on the solute's side, and water flows down the pressure difference through the pores as a column. In an aquaporin water moves in single file, as a train, so the osmotic permeability $P_f$ from the net flow exceeds the diffusional $P_d$ from tracing labeled water, by about the number of molecules in the file. Through the lipid, water diffuses one molecule at a time, and $P_f \approx P_d$. Kedem and Katchalsky wrote the volume flow as
$J_v = L_p A\,(\Delta P - \sigma RT\,\Delta c)$
with each Δ taken inside minus outside and $J_v$ counting outward flow, $L_p$ the water conductance and σ the reflection coefficient, 1 for a solute the membrane stops (two sections on).
The mistakes are solute "attracting" water, which a formula blind to chemistry rules out, and water molecules diffusing independently. A whole cell is an osmometer.
The cell as an osmometer
In a bath of a solute that stays out, a red cell gains or loses water until its fixed load of particles is as concentrated as the bath's, so its water volume goes as one over the bath's osmolarity. Part of the cell, mostly hemoglobin, is not water and does not change: the osmotically inactive volume $V_b$. This is Boyle–van 't Hoff:
$V = V_b + (V_0 - V_b)\,\dfrac{\Pi_0}{\Pi}$
with $V_0$ the volume at the cell's own osmolarity $\Pi_0$: against $\Pi_0/\Pi$, a straight line with intercept $V_b$, assuming only water moves and ideal solutions. A red cell's fixed area caps its volume at the sphere of that area, $V_l$, and $V = V_l$ gives the bath at which it lyses, $\Pi = \Pi_0(V_0 - V_b)/(V_l - V_b)$.
The usual slip is scaling the whole cell, forgetting $V_b$. And the solute must stay out.
Isosmotic is not isotonic
Solutions of the same osmolarity are isosmotic; a bath is isotonic only if it leaves a cell's volume alone, which depends on what crosses. A leaking solute holds water back less: the reflection coefficient σ is 1 for a solute the membrane stops and less for one that slips through, and the effective osmotic pressure is
$\Pi_{eff} = \sigma RT\,\Delta c$
Over time, any solute that crosses evens out and then counts for nothing: in the end only particles that stay out matter. Urea crosses red cells in seconds, through urea channels, glycerol more slowly; this sim's model gives urea σ = 1 and lets it in through its permeability.
The mistake is judging tonicity by osmolarity alone. Ions add one more twist: charge.
An equilibrium with a gradient: the Nernst potential
Give a membrane channels for K⁺ only, with more K⁺ inside. Each K⁺ that leaks out takes a positive charge with it, so the inside turns negative and pulls K⁺ back; the net flow stops at the Nernst potential:
$E = \dfrac{RT}{zF}\ln\dfrac{c_o}{c_i} = \dfrac{s}{z}\log_{10}\dfrac{c_o}{c_i}$ $s = RT\ln 10/F$
$E$ is the inside's potential relative to the outside, $z$ the charge number (+1 for K⁺, −1 for Cl⁻, +2 for Ca²⁺), $F$ = 96,485 C/mol and $s$ the slope per factor of ten, as for a pH electrode; concentrations are treated as ideal. K⁺ still crosses both ways, equally, and little charge is needed, since the membrane is a capacitor of about 1 µF/cm².
Watch the sign (inside relative to outside, $\ln(c_o/c_i)$; an anion's $z = -1$ flips it), and keep $RT/zF$ with ln and $s$ with log₁₀. The RC Circuit Lab shows how such a capacitor charges.
Reference
Common mistakes
- Giving molecules a purpose. They wander at random; a gradient flows because more start on the crowded side.
- Mistaking equilibrium for stillness, or scaling time with distance. Molecules keep crossing both ways; diffusion time grows as $x^2$.
- Reading "facilitated" as "uses energy". Channels and carriers are passive, and work only downhill.
- Expecting a carrier to keep up. It saturates, and with both sides each saturates on its own.
- Sending water toward less solute. Water moves toward more particles, lower Ψ; NaCl counts twice; $\Psi_s$ is negative, and $T$ is in kelvins.
- Counting axes or radii wrongly. $2Dt$ per axis, $4Dt$ on a plane, $6Dt$ in space; the radius in Stokes–Einstein.
- Scaling the whole cell. Only $V_0 - V_b$ follows 1/Π, and only particles that stay out decide tonicity.
Equation sheet
| Part 1 | |
|---|---|
| random walk | $x_{rms} = a\sqrt N$; net = $N_{in} - N_{out}$, noise $\sqrt{N_{in} + N_{out}}$ |
| Fick in words | $\text{rate} \propto A\,\Delta c/\text{thickness}$ |
| distance | $t \approx x^2/2D$ |
| size | $SA/V = 6/s$ (cube); $A/V = 3/r$ (sphere) |
| bilayer | $J = P\,\Delta c$ |
| carrier rate | $J_{max}\,c/(K + c)$ |
| particles | osmolarity = $iM$ |
| water potential | $\Psi = \Psi_s + \Psi_p$, $\Psi_s = -iCRT$ ($R$ = 0.0831 L·bar/(mol·K); 10 bar = 1 MPa); water moves toward lower $\Psi$ |
| Part 2 | |
|---|---|
| spreading | $⟨x^2⟩ = 2Dt$ per axis ($4Dt$ plane, $6Dt$ space); $\partial c/\partial t = D\,\partial^2 c/\partial x^2$ |
| Stokes–Einstein | $D = k_B T/(6\pi\eta r)$; $N_A = R/k_B$ |
| flux | $J = -D\,dc/dx$, $J = P\,\Delta c$, $P = KD_m/L$, $\tau = r/3P$ (sphere) |
| carrier | $J = J_{max}[c_o/(K + c_o) - c_i/(K + c_i)]$ |
| osmosis | $\Pi = iMRT$; $J_v = L_p A(\Delta P - \sigma RT\Delta c)$ |
| osmometer | $V = V_b + (V_0 - V_b)\Pi_0/\Pi$ |
| Nernst | $E = (RT/zF)\ln(c_o/c_i) = (s/z)\log_{10}(c_o/c_i)$, $s = RT\ln 10/F$ |
Units: 1 mM = 1 mol/m³ · $D$ in m²/s · $P$ in m/s · $J$ in mol m⁻² s⁻¹ (per carrier, /s) · $R$ = 8.314 J/(mol·K) = 0.0831 L·bar/(mol·K) · $T$ in kelvins · $s$ = 59.16 mV at 25 °C.
Further reading
OpenStax, Biology 2e, §5.1–5.2 and §30.5 (CC BY-NC-SA 4.0), whose order and notation Part 1 follows; H. C. Berg, Random Walks in Biology; E. M. Kramer and D. R. Myers, “Five popular misconceptions about osmosis,” Am. J. Phys. 80, 694 (2012). The Field Manual lists every source.
Diffusion & Osmosis Lab v1.4 · © 2026 Dr. David Meier · CC BY-NC-SA 4.0 · glidepathscience.org