Fluid Mechanics · Unit 1 · Lesson 8 of 13
If Water Is Made Of Molecules, Why Treat It As Smooth?
Water is made of discrete molecules that are always jostling, so at the smallest scales its density flickers from place to place. How can water have one density value "at a point"?
1. Real-World Problem
A pipe designer uses one density for water, 998 kg/m³, and draws a smooth velocity profile across the pipe. A hard disk designer, working on an air gap about 10 nanometres thick, cannot do that. Why does the same physics need two different descriptions?
2. Predict
Make your guess before you try the experiment. You will see the answer after the next step.
Predict
3. Experiment
Drag the sampling cube from 10 μm down to 1 nm and watch the molecule count and the measured density. In air the reading starts to flicker below about 100 nm; in water it stays steady to much smaller sizes.
In Engineering mode, try the Knudsen check for a car, two micro channels and a hard disk head gap, and see which flow regime each one is in.
4. Understand
Matter is made of discrete molecules in constant motion. If you measure density in a very small volume, the answer flickers as molecules move in and out. In a larger volume the flicker averages away and the reading settles to a steady value. This is scale-dependent averaging.
The continuum hypothesis says: when the important length scales of a flow are much larger than this averaging size, we can treat the fluid as continuous. Density, pressure, velocity and temperature are then defined at every mathematical point, by assigning to each point the average over a small representative volume around it. They become smooth fields, such as ρ(x, y, z, t), that calculus can handle.
It works because even tiny volumes hold huge numbers of molecules. Air at 20 °C and sea level has about 2.5 × 10²⁵ molecules per cubic metre. A cube just 1 μm on each side (10⁻¹⁸ m³) still holds about 2.5 × 10⁷ of them, so its average flickers by only about 0.02%. Liquid water is far more densely packed (about 3.3 × 10²⁸ molecules per m³), so it averages well at much smaller sizes.
The density at a point is written as a limit that stops at a small but not molecular volume, δV*:
\rho = \lim_{\delta V \to \delta V^{*}} \frac{\delta m}{\delta V}δm is the mass inside the small volume δV. δV* is not a universal constant: it depends on the fluid, its state, and how much fluctuation is acceptable. As an order of magnitude for gases and liquids at standard conditions, δV* is about 10⁻¹⁸ m³ (10⁻⁹ mm³); for liquid water a much smaller volume already averages well.
When does the continuum model fail for a gas? When the flow's size is not much bigger than the distance a molecule travels between collisions, the mean free path λ (about 68 nm for air at sea level). The check is the Knudsen number:
Kn = \frac{\lambda}{L}L is a characteristic length of the flow (m), such as a channel width or the size of a body.
| Knudsen number | Flow regime | Continuum model |
|---|---|---|
| Below 0.01 | Continuum flow | Works |
| 0.01 to 0.1 | Slip flow | Works with corrections at walls |
| 0.1 to 10 | Transition flow | Breaks down |
| Above 10 | Free molecular flow | Does not apply; track molecules |
5. Derive
The core idea is the Knudsen check in Understand. The counting statistics behind the simulation are in the Deep Dive below.
Step 1. Expected number of molecules in a cube of side L, with number density n:
\[\bar{N} = n\,L^3\]
Step 2. For molecules at random positions (a gas), the count fluctuates with a relative spread of about:
\[\frac{\Delta N}{\bar{N}} \approx \frac{1}{\sqrt{\bar{N}}}\]
Step 3. Choose the fluctuation you accept, then solve for the cube size. For 1% in air: N̄ = 10⁴, so L³ = 10⁴ ÷ (2.5 × 10²⁵) = 4 × 10⁻²² m³, giving L ≈ 74 nm. For 0.1%: N̄ = 10⁶ and L ≈ 0.34 μm. The averaging size depends on the accuracy you need; that is what scale-dependent means.
Counting is not the whole test. Having sufficiently many molecules in a sampling region reduces statistical fluctuations, but continuum validity also requires the molecular mean free path to be small compared with the length scale over which the flow changes. At 74 nm, air gives 1% counting fluctuations, yet its mean free path (68 nm) is about the same size: a flow that changed over 74 nm would have Kn ≈ 0.9 and would not be a continuum flow. Both conditions must hold: enough molecules to average, and Kn below about 0.01.
6. Calculate
Knudsen numbers for air (λ = 68 nm):
- Car, L = 4 m: Kn = 68 × 10⁻⁹ ÷ 4 = 1.7 × 10⁻⁸. Continuum, easily.
- Micro channel, L = 10 μm: Kn = 0.0068. Continuum, just.
- Micro channel, L = 1 μm: Kn = 0.068. Slip flow: continuum with wall corrections.
- Hard disk head gap, L = 10 nm: Kn = 6.8. Transition flow: the continuum model breaks down.
7. Apply
- A car at sea level: continuum is perfect.
- A hard disk read head flies on an air gap of about 10 nanometres or less, so engineers use molecular (kinetic theory) models for that gap.
- Spacecraft in the very thin upper atmosphere: λ becomes centimetres to metres, so the continuum model fails.
- Micro channels in lab-on-a-chip devices sit near the slip-flow limit.
- Vacuum pumps and semiconductor equipment work in the transition and free molecular regimes.
8. Validate
In the simulation set air and L = 74 nm: N̄ should read about 10,000 and the shaded band about ±1%. Set L = 1 μm: about 2.5 × 10⁷ molecules and ±0.02%. In the Knudsen panel the four situations must show the four values above. Where the continuum model does not apply: Kn above about 0.1 in gases, and channels only a few molecules wide (about 1 nm) in liquids.
Easy To Mix Up
- The continuum hypothesis does not claim molecules do not exist. It is a model that is accurate at the sizes engineers usually work with.
- A "point" in the equations is a true mathematical point. The value it carries is an average over a small representative volume around it.
- δV* is not one fixed number; it depends on the fluid and on how much fluctuation is acceptable.
- Liquid water is not "mostly empty space". Its molecules are densely packed; the issue at small scales is discreteness and fluctuation.
9. Quiz
Four quick questions: a concept, a calculation and an interpretation. Each answer comes with an explanation. Score 3 or more to mark this lesson completed.
Check Your Understanding
Four questions: a concept, a calculation and an interpretation. Every answer comes with an explanation. Score 3 of 4 to mark the lesson completed.
If the fluid is smooth, how do we keep track of one tiny piece of it as it moves?
Go to Lesson 9: How Do We Follow A Tiny Piece Of Fluid? →Read More
- White, Fluid Mechanics, chapter 1 (the fluid as a continuum); Karniadakis, Beskok and Aluru, Microflows and Nanoflows (Knudsen number regimes).
- NPTEL, Introduction to Fluid Mechanics (Prof. Suman Chakraborty, IIT Kharagpur): Week 1.