Droplet and Ice-Crystal Growth: Bergeron–Findeisen and Coalescence
CSIR NET Part C meteorology: diffusional droplet growth, fall speeds, collision–coalescence, ice-crystal habits, Bergeron–Findeisen process, riming and hail.
- 21 min
On this page · 16 sections
- Prerequisites
- The Hook: The Factor of a Million
- Growth by Condensation (Diffusion of Vapour)
- Fall Speeds
- Collision–Coalescence (the Warm-Rain Process)
- Ice-Crystal Growth by Deposition
- The Bergeron–Findeisen Process
- Growth by Collection in Ice and Mixed-Phase Clouds
- Warm Rain versus Cold Rain
- Memory Tricks
- Don’t Confuse
- Common Mistakes
- Practice
- Quick Revision Table
- Key Takeaways
- What’s Next
Prerequisites
- Köhler theory, CCN and ice nuclei from Cloud Classification and Condensation Nuclei: growth starts where activation ends.
- Saturation vapour pressure over water and over ice, from Thermodynamics of Dry and Moist Air.
- Updraft speeds and the depth of convection from Vertical Stability and Tropical Convection.
The Hook: The Factor of a Million
Growth by Condensation (Diffusion of Vapour)
The growth equation
An activated droplet in supersaturated air gains mass by diffusion of vapour toward it. Two fluxes are coupled:
- Vapour diffusion inward: \( \dfrac{dm}{dt} = 4\pi r D(\rho_{v,\infty} - \rho_{v,r}) \), where \( D \) is the diffusivity of water vapour.
- Heat conduction outward: condensation releases latent heat, warming the droplet surface and raising its vapour pressure, so the heat \( L\,dm/dt \) must be conducted away at rate \( 4\pi r K(T_r - T_\infty) \), with \( K \) the thermal conductivity of air.
Eliminating the surface temperature (using Clausius–Clapeyron) gives
$$\boxed{r\frac{dr}{dt} = G\,(S - 1)}, \qquad G = \frac{1}{F_k + F_d}$$with
$$F_k = \left(\frac{L}{R_vT} - 1\right)\frac{L\rho_w}{KT}, \qquad F_d = \frac{\rho_w R_v T}{D\,e_s(T)}$$\( F_k \) is the heat-conduction term and \( F_d \) the vapour-diffusion term. With curvature and solute effects added, \( S - 1 \) becomes \( S - 1 - a/r + b/r^3 \), but for droplets of a few micrometres or more the simple form is enough.
\( G \) depends on temperature and pressure. Calculated values:
| Conditions | \( G \) |
|---|---|
| 10 °C, 800 hPa | about \( 1.0\times 10^{-10} \) m\(^2\) s\(^{-1}\) (100 μm\(^2\) s\(^{-1}\)) |
| 0 °C, 800 hPa | about \( 0.7\times 10^{-10} \) m\(^2\) s\(^{-1}\) |
| −10 °C, 700 hPa | about \( 0.4\times 10^{-10} \) m\(^2\) s\(^{-1}\) |
Growth is slower in cold air because \( e_s \) is small, so \( F_d \) dominates.
Parabolic growth and its consequences
At constant supersaturation \( s = S - 1 \),
$$\boxed{r^2 = r_0^2 + 2Gst}$$Since \( dr/dt = Gs/r \), small droplets grow faster than large ones (in radius). Two consequences follow:
- Diffusional growth narrows the size spectrum. Droplets that started at different sizes converge.
- Diffusional growth slows down as the droplet grows. It is fast to 10 μm and hopeless beyond about 20–30 μm.
With \( G = 100 \) μm\(^2\) s\(^{-1}\) and \( s = 0.5\% \) (so \( 2Gs = 1 \) μm\(^2\) s\(^{-1}\)):
| Growth | Time |
|---|---|
| 1 → 10 μm | about 100 s |
| 10 → 20 μm | about 5 min |
| 20 → 30 μm | about 8 min |
| 10 → 100 μm | about 2.8 h |
| 10 μm → 1 mm | about 280 h (about 12 days) |
Real cloud supersaturations are usually smaller than 0.5%, making these times longer still.
Fall Speeds
A drop reaches terminal velocity when drag balances weight. Three regimes are used (the coefficients are the approximations given in Rogers and Yau, for near-surface air):
| Regime | Radius range | Law | Coefficient |
|---|---|---|---|
| Stokes (viscous) | below about 30 μm | \( v = k_1 r^2 \) | \( k_1 \approx 1.2\times 10^{8} \) m\(^{-1}\) s\(^{-1}\) |
| Intermediate | about 40 μm to 0.6 mm | \( v = k_3 r \) | \( k_3 \approx 8\times 10^{3} \) s\(^{-1}\) |
| Large drops | about 0.6 to 2 mm | \( v = k_2 r^{1/2} \) | \( k_2 \approx 201 \) m\(^{1/2}\) s\(^{-1}\) |
The Stokes coefficient follows from \( v = \dfrac{2g\rho_w r^2}{9\eta} \) with \( \eta \approx 1.8\times 10^{-5} \) Pa s: about 1.2 cm s\(^{-1}\) for a 10 μm droplet and 4.8 cm s\(^{-1}\) for 20 μm. A 1 mm-radius raindrop falls at about 6.5 m s\(^{-1}\). Large drops flatten into a “hamburger” shape (flat base, rounded top), their speed levels off at roughly 9 m s\(^{-1}\), and the largest break up. A cloud droplet falling at 1 cm s\(^{-1}\) is carried up by almost any updraft; this is why clouds float.
Collision–Coalescence (the Warm-Rain Process)
Collection efficiency
A large drop (collector, radius \( R \)) falling through smaller droplets (radius \( r \)) sweeps out a cylinder of cross-section \( \pi(R + r)^2 \), but not every droplet in that cylinder is hit:
- Collision efficiency \( E_c \): small droplets follow the airflow around the collector and are deflected. \( E_c \) is very small when the collector is below about 20 μm radius, and when the collected droplet is very small relative to the collector.
- Coalescence efficiency \( E_{co} \): colliding drops may bounce (an air film between them) or separate again after temporary merging.
- Collection efficiency \( E = E_c\,E_{co} \).
The practical threshold: a cloud needs some droplets larger than about 20 μm radius before collection becomes efficient. That is why the broadening mechanisms above matter.
The continuous collection model
Treat the collector as falling through a uniform cloud of liquid water content \( M \) (kg m\(^{-3}\)), with the small droplets’ fall speed neglected:
$$\frac{dm}{dt} = E\,\pi R^2\,V(R)\,M$$With \( m = \tfrac{4}{3}\pi R^3\rho_w \):
$$\boxed{\frac{dR}{dt} = \frac{E\,M\,V(R)}{4\rho_w}}$$In the intermediate regime \( V = k_3R \), so growth is exponential:
$$R(t) = R_0\exp\left(\frac{EMk_3}{4\rho_w}t\right)$$With \( E = 0.8 \), \( M = 1 \) g m\(^{-3}\) and \( k_3 = 8000 \) s\(^{-1}\), the e-folding time is \( 4\rho_w/(EMk_3) = 625 \) s, and growth from 40 to 400 μm takes \( 625\ln 10 \approx 1440 \) s, about 24 minutes. Doubling the liquid water content halves the time.
Beyond the continuous model
- The stochastic collection model recognises that collisions are discrete, random events: a few lucky droplets that collide early race ahead. Warm rain develops faster than the continuous model predicts.
- Updraft effect: in an updraft, a growing drop first rises (its fall speed is smaller than the updraft), sweeping droplets on the way up, then falls back through the cloud once it is heavy enough. A deeper cloud and stronger updraft give a longer path and bigger drops.
- Breakup: large drops break up by collisions and by hydrodynamic instability. The fragments are themselves large enough to collect efficiently, so the number of raindrops can multiply (the “Langmuir chain reaction” idea). Breakup limits raindrop size and shapes the drop-size distribution described on the radar page.
When warm rain is efficient
| Favourable | Unfavourable |
|---|---|
| maritime air: few CCN, larger droplets | polluted continental air: many small droplets |
| giant sea-salt nuclei | narrow droplet spectrum |
| deep warm layer (warm cloud base, high freezing level) | shallow cloud |
| high liquid water content | low liquid water content |
| long residence time in strong but not extreme updrafts | very short-lived cloud |
Warm rain is common in the tropics, from maritime cumulus and from shallow orographic clouds; radar often sees rain from clouds whose tops never reach the freezing level.
Ice-Crystal Growth by Deposition
The growth law
Ice grows by vapour deposition according to an equation of the same form as for droplets, with two changes: the driving force is the supersaturation over ice \( S_i - 1 \), and the radius is replaced by a capacitance \( C \) that depends on crystal shape (borrowed from electrostatics, because the vapour field obeys the same Laplace equation as the potential around a charged conductor):
$$\frac{dm}{dt} = \frac{4\pi C\,(S_i - 1)}{F_k' + F_d'}$$For a sphere \( C = r \); for a thin disc of radius \( r \), \( C = 2r/\pi \). \( F_k' \) and \( F_d' \) use \( L_s \) and \( e_{s,i} \).
Habits: shape set by temperature
Laboratory and field observations show that the basic habit (plate-like or column-like) is controlled mainly by temperature, and the degree of elaboration (solid, hollow, dendritic) by supersaturation:
| Temperature range | Basic habit | At high supersaturation |
|---|---|---|
| 0 to about −4 °C | thin hexagonal plates | |
| about −4 to −10 °C | prisms: needles (near −5 °C) and columns | hollow columns, needles |
| about −10 to −22 °C | plates | sectors and dendrites near −12 to −16 °C |
| below about −22 °C | columns, bullet rosettes |
The habit switches plate → column → plate → column as temperature falls. Dendrites, the classic snowflake form, grow fastest of all, in the same temperature band where the vapour-pressure difference between water and ice is greatest.
The Bergeron–Findeisen Process
The physics in one inequality
Below 0 °C, ice molecules are more tightly bound than liquid molecules, so
$$\boxed{e_{s,i}(T) < e_{s,w}(T) \quad \text{for } T < 0\ ^\circ\text{C}}$$Integrating Clausius–Clapeyron for both phases from the triple point,
$$\ln\frac{e_{s,w}}{e_{s,i}} \approx \frac{L_f}{R_v}\left(\frac{1}{T} - \frac{1}{T_0}\right), \qquad L_f = L_s - L_v$$Air at water saturation is therefore supersaturated with respect to ice, by about 10% at −10 °C and about 22% at −20 °C. Calculated values (Magnus-type formulas):
| \( T \) | \( e_{s,w} \) (hPa) | \( e_{s,i} \) (hPa) | \( e_{s,w} - e_{s,i} \) (hPa) | \( e_{s,w}/e_{s,i} \) |
|---|---|---|---|---|
| −5 °C | 4.22 | 4.02 | 0.20 | 1.05 |
| −10 °C | 2.87 | 2.60 | 0.27 | 1.10 |
| −12 °C | 2.45 | 2.17 | 0.27 (maximum) | 1.13 |
| −15 °C | 1.92 | 1.65 | 0.26 | 1.16 |
| −20 °C | 1.26 | 1.03 | 0.22 | 1.22 |
| −30 °C | 0.51 | 0.38 | 0.13 | 1.34 |
The ratio keeps increasing as it gets colder, but the difference, which sets the vapour-density gradient that drives deposition, peaks near −12 °C. Both \( e_s \) values shrink at lower temperatures, so the difference eventually falls.
The sequence in a mixed-phase cloud
graph TD
A[Supercooled cloud: many droplets at water saturation] --> B[A few ice crystals form on INPs]
B --> C[Air is supersaturated over ice: crystals grow by deposition]
C --> D[Vapour drops below water saturation: droplets evaporate]
D --> C
C --> E[Crystals heavy enough to fall]
E --> F[Riming: collect supercooled droplets, graupel and hail]
E --> G[Aggregation: crystals stick together, snowflakes]
F --> H{Air below cloud warmer than 0 °C?}
G --> H
H -->|yes| I[Melt: rain]
H -->|no| J[Snow, graupel at ground]Why it works so well:
- Few crystals, many droplets. With about one ice crystal per \( 10^5 \) droplets (see the nucleation page), the water of thousands of droplets ends up on each crystal.
- Ice supersaturation of 10–20% is far larger than the fraction of a percent available for droplet growth, so deposition is fast.
- The crystals start to fall and can then grow by collection, which accelerates.
The process dominates in cold clouds: mid- and high-latitude stratiform precipitation (most rain there begins as snow) and the upper, mixed-phase parts of deep convection everywhere. It needs supercooled water and a shortage of ice. A cloud that is already glaciated, or too warm for any ice to form, cannot use it.
Growth by Collection in Ice and Mixed-Phase Clouds
| Process | What collides | Product |
|---|---|---|
| Riming (accretion) | ice crystal collects supercooled droplets, which freeze on contact | rimed crystals; heavily rimed: graupel (soft, opaque pellets about 2–5 mm) |
| Aggregation | ice crystals stick together | snowflakes; most efficient near 0 °C (sticky, wet surfaces) and when dendrites interlock |
| Accretion of rain by ice / freezing of drops | supercooled raindrops freeze or are captured | frozen drops, hail embryos |
Hail
Hail is ice of diameter 5 mm or more, grown in cumulonimbus with strong updrafts and large supercooled liquid water content:
- Embryos (graupel or frozen drops) are carried up into the region of high supercooled water.
- Dry growth: when the stone’s surface stays below 0 °C, collected droplets freeze at once, trapping air: opaque, white ice.
- Wet growth: at high accretion rates, the released latent heat of freezing cannot be removed fast enough; the surface warms to 0 °C, part of the water stays liquid and soaks into the stone, giving clear ice and allowing some water to be shed.
- Alternating clear and opaque layers record changes in growth regime as the stone moves through regions of different temperature and liquid water content.
- The largest stones need updrafts strong enough to support them for a long time, which is why severe hail comes from supercells.
The same supercooled water that feeds hail is targeted in hail suppression. Riming and graupel–crystal collisions are also central to thunderstorm electrification, through charge separation in the mixed-phase region.
Warm Rain versus Cold Rain
| Collision–coalescence (warm rain) | Bergeron–Findeisen (cold rain) | |
|---|---|---|
| Cloud temperature | can operate entirely above 0 °C | needs a layer between 0 and −38 °C with supercooled water |
| Key physics | different fall speeds of different-sized drops | \( e_{s,i} < e_{s,w} \) |
| Needs | broad droplet spectrum, drops above about 20 μm | INPs, but few of them; supercooled liquid |
| Favoured by | maritime air, giant CCN, high liquid water | deep cold clouds, mid- and high latitudes |
| Typical precipitation | tropical showers, drizzle from stratocumulus | stratiform rain and snow, upper parts of thunderstorms |
| Radar signature | no bright band from the warm process | bright band below the 0 °C level in stratiform rain |
The two are not exclusive. In tropical deep convection both act: warm rain in the lower cloud, ice processes above. In a raindrop reaching the ground, the history is usually ambiguous.
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| \( e_{s,w} - e_{s,i} \): maximum near −12 °C | \( e_{s,w}/e_{s,i} \): increases steadily as \( T \) falls | |
| Diffusional growth narrows the spectrum | collection broadens it | |
| Collision efficiency: geometry and airflow | coalescence efficiency: do the drops merge after touching | |
| Riming: ice + supercooled droplets → graupel | aggregation: ice + ice → snowflakes | |
| Graupel: soft, about 2–5 mm, from riming | hail: 5 mm or more, layered, from Cb | |
| Habit: set mainly by temperature | elaboration (dendrites, hollow columns): set by supersaturation | |
| Dry growth of hail: opaque ice | wet growth: clear ice |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper.
Level 1: Foundation
- Write the diffusional growth law for a cloud droplet at constant supersaturation.
- Near what temperature is \( e_{s,w} - e_{s,i} \) largest?
- Which crystal habit is typical at −15 °C and high supersaturation?
Answers
- \( r\,dr/dt = G(S - 1) \), so \( r^2 = r_0^2 + 2Gst \).
- About −12 °C.
- Dendrites (or sector plates): the plate regime of −10 to −22 °C, elaborated by high supersaturation.
Level 2: Part B Style
Practice Q1. With \( G = 100 \) μm\(^2\) s\(^{-1}\) and a constant supersaturation of 0.5%, the time for a droplet to grow by diffusion from 10 μm to 20 μm radius is closest to:
(a) 1 min (b) 5 min (c) 15 min (d) 1 h
Answer: (b) 5 min
\( t = (r^2 - r_0^2)/(2Gs) = (400 - 100)/(2\times 100\times 0.005) = 300/1 = 300 \) s.
Practice Q2. Using Stokes’ law with \( k_1 = 1.2\times 10^8 \) m\(^{-1}\) s\(^{-1}\), the terminal velocity of a 20 μm radius droplet is about:
(a) 0.5 cm s\(^{-1}\) (b) 1.2 cm s\(^{-1}\) (c) 4.8 cm s\(^{-1}\) (d) 24 cm s\(^{-1}\)
Answer: (c) 4.8 cm/s
\( v = 1.2\times 10^8 \times (2\times 10^{-5})^2 = 0.048 \) m s\(^{-1}\). Option (b) is the value for 10 μm: doubling the radius quadruples the speed.
Practice Q3. Using \( \ln(e_{s,w}/e_{s,i}) = (L_f/R_v)(1/T - 1/273.15) \) with \( L_f = 3.34\times 10^5 \) J kg\(^{-1}\) and \( R_v = 461.5 \) J kg\(^{-1}\) K\(^{-1}\), the supersaturation with respect to ice of air that is just saturated with respect to water at −15 °C is closest to:
(a) 5% (b) 10% (c) 17% (d) 25%
Answer: (c) 17%
\( L_f/R_v = 723.7 \) K; \( 1/258.15 - 1/273.15 = 2.127\times 10^{-4} \) K\(^{-1}\); product 0.154; \( e^{0.154} = 1.166 \), i.e. about 17%.
Practice Q4. Hailstones with alternating clear and opaque layers indicate that the stone:
(a) grew only by vapour deposition (b) alternated between wet and dry growth regimes (c) formed entirely below −38 °C (d) grew by aggregation of dendrites
Answer: (b)
Clear ice forms in wet growth (surface at 0 °C, water soaks in, no trapped air); opaque ice in dry growth (instant freezing traps air). Layers record changes in temperature and liquid water along the trajectory.
Level 3: Part C Style
Practice Q5. A collector drop falls through a cloud of liquid water content 1 g m\(^{-3}\), with collection efficiency 0.8. Its fall speed is \( V = k_3R \), \( k_3 = 8000 \) s\(^{-1}\), and the cloud droplets’ fall speed is neglected. Which statements are correct?
- The radius grows exponentially with an e-folding time of 625 s.
- Growth from 40 μm to 400 μm takes about 24 min.
- If the liquid water content were 2 g m\(^{-3}\) with \( E = 1 \), the same growth would take under 10 min.
- The growth rate \( dR/dt \) at \( R = 100 \) μm is about 2 μm s\(^{-1}\).
(a) 1 and 2 only (b) 1, 2 and 3 only (c) 1, 2, 3 and 4 (d) 2 and 4 only
Answer: (b) 1, 2 and 3 only
\( dR/dt = EMk_3R/(4\rho_w) \), rate constant \( EMk_3/4\rho_w \).
- 1 true: \( 0.8\times 10^{-3}\times 8000/4000 = 1.6\times 10^{-3} \) s\(^{-1}\), so e-folding time 625 s.
- 2 true: \( 625\ln 10 = 1439 \) s, about 24 min.
- 3 true: the rate constant becomes \( 1\times 2\times 10^{-3}\times 8000/4000 = 4\times 10^{-3} \) s\(^{-1}\); \( 250\ln 10 = 576 \) s, about 9.6 min.
- 4 false: \( 1.6\times 10^{-3}\times 100 \) μm \( = 0.16 \) μm s\(^{-1}\). Option values like 2 μm s\(^{-1}\) come from slipping a factor of 10 in the units of \( M \).
Practice Q6. Use the table of \( e_{s,w} \) and \( e_{s,i} \) on this page. A mixed-phase cloud layer is at water saturation. Which are correct?
- The vapour-density excess available for deposition on ice is larger at −12 °C than at −25 °C.
- Ice supersaturation at water saturation is larger at −25 °C than at −12 °C.
- Dendritic snowflakes are expected near −15 °C because that is where the habit regime is plate-like and the excess vapour is near its maximum.
- At −2 °C the Bergeron–Findeisen process is as effective as at −12 °C because both are below 0 °C.
(a) 1, 2 and 3 only (b) 1 and 4 only (c) 2 and 3 only (d) 1, 2, 3 and 4
Answer: (a) 1, 2 and 3 only
- 1 true: the difference is about 0.27 hPa at −12 °C, and about 0.18 hPa at −25 °C (it lies between the tabulated 0.22 at −20 °C and 0.13 hPa at −30 °C).
- 2 true: the ratio grows steadily with cooling (about 1.13 at −12 °C, about 1.28 at −25 °C).
- 3 true: plate regime (−10 to −22 °C) plus high supersaturation gives dendrites near −12 to −16 °C.
- 4 false: at −2 °C the difference is small (about 0.1 hPa) and very few INPs are active, so the process is weak.
Practice Q7. Radar and aircraft observations of three clouds over a tropical coast:
| Cloud | Top temperature | Droplet concentration | Rain at ground |
|---|---|---|---|
| A | +8 °C | 60 cm\(^{-3}\) | yes, within 20 min |
| B | +8 °C | 900 cm\(^{-3}\) | no |
| C | −20 °C | 900 cm\(^{-3}\) | yes, after about 40 min, with a bright band in the decaying stage |
Which interpretations are consistent with the data?
- Rain from A formed by collision–coalescence.
- B did not rain because its many small droplets coalesce inefficiently.
- C’s rain must have formed by collision–coalescence, since its droplet concentration equals B’s.
- The bright band in C indicates melting snow, so ice processes contributed to C’s rain.
(a) 1, 2 and 4 only (b) 1 and 3 only (c) 2, 3 and 4 only (d) 1, 2, 3 and 4
Answer: (a) 1, 2 and 4 only
- 1 true: A is entirely above 0 °C, so only the warm process is available; low droplet number means larger droplets.
- 2 true: same depth as A but 15 times more droplets: smaller drops, narrow spectrum, little collection.
- 3 false: C reaches −20 °C, so the ice route is available, and statement 4 shows it acted.
- 4 true: a bright band is the radar signature of a melting layer of snow and ice particles.
Quick Revision Table
| Idea | Fact or formula |
|---|---|
| Diffusional growth | \( r\,dr/dt = G(S-1) \); \( G \approx 100 \) μm\(^2\) s\(^{-1}\) at 10 °C |
| Parabolic law | \( r^2 = r_0^2 + 2Gst \); narrows spectrum; 10 μm → 1 mm takes days |
| Fall speeds | \( k_1r^2 \) (below 30 μm), \( k_3r \) (40 μm–0.6 mm), \( k_2r^{1/2} \) (0.6–2 mm) |
| Values | 10 μm: 1.2 cm s\(^{-1}\); 1 mm: about 6.5 m s\(^{-1}\); maximum about 9 m s\(^{-1}\) |
| Collection | \( dR/dt = EMV/4\rho_w \); exponential when \( V = k_3R \) |
| Collector threshold | radius above about 20 μm |
| Ice vs water | \( e_{s,i} < e_{s,w} \); ratio 1.10 (−10 °C), 1.22 (−20 °C); difference maximum near −12 °C |
| Habits | plates (0 to −4), columns/needles (−4 to −10), plates/dendrites (−10 to −22), columns (below −22) °C |
| Ice growth | \( dm/dt \propto C(S_i - 1) \); capacitance \( C \) |
| Riming → graupel; aggregation → snow | hail at least 5 mm; wet growth clear, dry growth opaque |
| Warm rain | maritime, giant CCN, deep warm cloud |
| Cold rain | supercooled water + few ice crystals; bright band on radar |
Key Takeaways
What’s Next
- Artificial Precipitation, Hail Suppression and Fog Dissipation: how seeding tries to exploit these processes.
- Weather Radar and Rainfall Measurement: drop-size distributions and how radar sees the bright band.
- Clouds, Precipitation and General Circulation: the Part B summary of the same mechanisms.
- Back to the chapter overview.