# Rollin film Below the [[Lambda_point|lambda point]], [[Liquid_helium|liquid]] [[Helium-4|helium-4]] wets every solid it touches with a [[Superfluidity|superfluid]] film about 30 nm thick, and because the superfluid component of that film has exactly zero [[Viscosity|viscosity]], the film is not a coating but a channel. A beaker lowered into a bath of helium II fills itself; a beaker lifted clear of one empties itself over its own rim, drop by drop, until the two free surfaces are level. It is the most-filmed demonstration in [[Cryogenics|low-temperature]] [[Physics|physics]]. The interesting part is not that it happens — it is that the [[Fluid_dynamics|flow]] rate obeys a law with the wrong variable in it. ## Microsims — three.js <iframe src="https://wikitube-3d-microsims.netlify.app/Rollin_film.html" width="100%" height="620" frameborder="0" loading="lazy" sandbox="allow-scripts allow-same-origin" title="Rollin film — three.js microsim"></iframe> **`Rollin_film` (three.js).** A beaker of helium II hangs clear of its bath in a [[Cryogenics|cryostat]], wetted inside and out by a film the sim draws as a wedge — visibly fat where it leaves the [[Liquid_helium|liquid]] and thin at the rim, because the local thickness really does follow h^(-1/3) — and the beaker drips itself empty. Do three things with the controls, in this order. Drag *Beaker base* upward: the drop now falls three times as far and the [[Velocity|rate]] does not move, because the climb inside the beaker rides with the beaker. Drag *Rim perimeter*: the volumetric rate tracks it exactly while the rate **per centimetre of rim** printed beside it does not budge — that invariant is the whole diagnostic. Drag *Temperature* past 2.1768 K: the film is still drawn, because van der Waals attraction does not switch off at a [[Phase_transition|phase transition]], but rho_s/rho reads zero and everything stops dead. The live HUD quantity is dV/dt = P·d·v_c·(rho_s/rho) together with its per-centimetre reduction, both computed from the running climb height rather than tabulated, and printed against the literature value for comparison; the volume budget across beaker, pendant drop, drops in flight and bath is conserved to round-off. ## What the film is Every wall attracts [[Helium|helium]] atoms by the van der Waals dispersion [[Force|interaction]]. That is not special to helium — it is why anything wets anything, and it is the same [[Molecular_dynamics|intermolecular]] attraction that condenses any [[Noble_gas|noble gas]] onto a cold [[Materials_science|surface]]. What is special is that helium is the only [[Chemical_element|element]] still [[Liquid_helium|liquid]] at temperatures where nothing else has anything left to give — [[Zero-point_energy|zero-point energy]] keeps it from freezing under its own [[Boiling_point|vapour pressure]] — so the adsorbed layer is a *liquid* layer, and below the [[Lambda_point|lambda point]] part of that liquid has no [[Viscosity|viscosity]] at all. The [[Thermodynamic_equilibrium|equilibrium]] condition is a chemical potential balance. An [[Atomic_mass|atom]] in a film of thickness d, sitting at height h above the bulk free surface, carries potential [[Energy|energy]] per unit mass **mu(d, h) = g·h − A / (6·pi·rho·d³)** where the second term is the disjoining-pressure potential of a flat film. Equating mu along the whole wetted wall to its bulk value — the [[Thermodynamics|thermodynamic]] statement that one connected body of [[Fluid_dynamics|fluid]] has one chemical potential — gives the thickness law: **d(h) = ( A / (6·pi·rho·g·h) )^(1/3)** **A** is the van der Waals or Hamaker constant for the helium–[[Materials_science|substrate]] pair — an [[Energy|energy]], of order 10⁻²¹ to 10⁻²⁰ J, and a property of the wall as much as of the [[Fluid_dynamics|fluid]]. **h** is height above the bath, **rho** the liquid [[Density|density]] (145 kg m⁻³), **g** the [[Gravitational_field|gravitational]] acceleration. The cube root is the point. Thickness varies only *weakly* with height: ten times higher and d falls by 10^(1/3) = 2.15, so a film 30 nm thick at one centimetre is still 13.9 nm at ten. **The film does not run out. It only thins.** Duthler and Pollack's 1971 [[Estimation_theory|measurements]] on [[Neon|neon]]-plated [[Silicon_dioxide|glass]] give 2.3·h^(-1/3) × 10⁻⁶ cm — 23 nm at h = 1 cm on that [[Surface_engineering|substrate]], following that exponent exactly. The microsim fixes A = 7.24 × 10⁻²¹ J so d(1 cm) = 30.0 nm, the canonical saturated-film figure, and states the [[Accuracy_and_precision|calibration]] rather than hiding it. ## Why it moves, and why it is not capillarity A 30 nm film of any ordinary [[Fluid_dynamics|liquid]] is immobilised: [[Viscosity|viscous]] drag against the wall scales as the inverse square of the thickness, so at tens of nanometres the [[Reynolds_number|Reynolds number]] vanishes and the [[Thermodynamics|dissipation]] is total. Helium II is the exception because its [[Superfluidity|superfluid]] component has *exactly* zero [[Viscosity|viscosity]]. The film is a free channel, and it runs until the chemical potential is equal at both ends — until the two free surfaces are level. It does not know or care which vessel it started in, and no [[Force|force]] pushes it: this is relaxation to [[Thermodynamic_equilibrium|equilibrium]], not [[Fluid_dynamics|pumping]]. **This is not capillary rise, and the distinction is diagnostic, not pedantic.** Capillary action has a maximum height, set by [[Surface_engineering|surface]] tension against the column's weight; a wick dies there, and so does every ordinary [[Porous_medium|porous]] material. The Rollin film has no such limit — it crosses whatever rim you give it, and lifting the beaker costs almost nothing. The English [[Science|Wikipedia]] article attributes the Onnes effect to "capillary forces dominat[ing] gravity and viscous forces", which inverts the [[Physics|physics]]: it is not that capillary [[Force|forces]] win, it is that no viscous [[Damping|resistance]] is left for anything to win against. Two clauses cover it — van der Waals adsorption puts a film everywhere, [[Superfluidity|superfluidity]] makes it mobile — and neither mentions [[Boiling_point|surface tension]]. ## The transfer law: perimeter, not area The film does not run faster when pushed harder. It saturates at a **critical [[Velocity|velocity]]** v_c, above which quantised vortices — closed [[Torus|toroidal]] loops of circulating [[Superfluid_helium-4|superfluid]] — are shed and the superflow breaks down. So the volumetric rate is a product of three things and no others: **dV/dt = P · d · v_c · (rho_s/rho)** with **P the perimeter of the narrowest place the film has to cross**. Van Alphen, van Haasteren, de Bruyn Ouboter and Taconis established in 1966 that v_c scales as d^(-1/4), with Γ = 1 cm^(5/4) s⁻¹ — so v_c = 24.0 cm/s in a 30 nm film, inside the 20 to 40 cm/s band the literature quotes. Wikipedia's *[[Superfluid_helium-4|Superfluid helium-4]]* gives "about 20 cm/s", its *Superfluid film* "of order 0.1 m/s"; the [[Estimation_theory|spread]] is real, and thinner colder films are faster. One trap: that same article uses 20 cm/s again for the normal-component counterflow speed in [[Heat_transfer|heat transport]], an unrelated [[Velocity|quantity]] that happens to coincide. Three consequences follow, and every one of them is the wrong answer for a pressure-driven [[Fluid_dynamics|flow]]: - **The rate is independent of the level difference driving it.** Halve the head and nothing changes. No [[Reynolds_number|viscous]] channel behaves this way, and no [[Diffusion|diffusive]] [[Heat_transfer|transport]] law does either. - **The rate is nearly independent of the height climbed.** Combining d ∝ h^(-1/3) with v_c ∝ d^(-1/4) makes d·v_c scale as h^(-1/4): a *tenfold* climb costs a factor of 1.78. Duthler and Pollack [[Estimation_theory|measured]] that exponent at 0.26 ± 0.05 — the −1/4 prediction sits dead centre of their error bar. - **The rate is proportional to perimeter, not to area.** Widen the vessel and the volumetric rate rises in step while the rate per centimetre of rim does not move at all — a [[Geometry|geometric]] invariant, testable in an afternoon. That last one is the signature. Smith and Boorse ran it to its logical end in 1955 by polishing and roughening [[Metallurgy|metal]] vessels — [[Aluminium|aluminium]], [[Nickel|nickel]], [[Steel|stainless steel]] and nickel-silver — and concluded that "transport rates adjust to the prevailing microscopic perimeter and do not require the added hypothesis of anomalous flow in surface cracks." The film counts edge, and only edge. A [[Materials_science|rough]] wall transfers faster not because roughness helps the [[Fluid_dynamics|flow]] but because roughness *is* more perimeter. ## The classic number, and what it is a number *of* **7.5 × 10⁻⁵ cm³ s⁻¹ per centimetre of perimeter**, often written 0.75 × 10⁻⁴, is the figure that circulates. Two things about it are routinely got wrong. **First, the units.** cm³ s⁻¹ cm⁻¹ is an *area* per time. 7.5 × 10⁻⁵ cm²/s is 7.5 × 10⁻⁹ m²/s — of order 10⁻⁸, not 10⁻⁷. It is also exactly what the transfer law predicts: 30 nm × 25 cm/s = 3 × 10⁻⁶ cm × 25 cm/s = 7.5 × 10⁻⁵ cm²/s. A thickness times a [[Velocity|velocity]] is an area per time: the [[Accuracy_and_precision|dimensional]] tell that the perimeter law is the right law, which the [[Density|volumetric]] figure alone hides. **Second, the conditions.** Most retellings attach the figure to Daunt and Mendelssohn's 1939 [[Estimation_theory|measurements]]; that attribution is conventional rather than checked, their two Royal Society papers being paywalled and not opened here. What *is* verifiable in print is Dash and Boorse's 7.5 × 10⁻⁵ cm³/cm·s at 1.25 K **for [[Silicon_dioxide|glass]]**, in the same 1951 paper where **etched [[Copper|copper]] gave 51 × 10⁻⁵** — a factor of 6.8 across [[Materials_science|substrates]], one temperature, one [[Experimental_system|apparatus]]. Quoting 7.5 × 10⁻⁵ as *the* transfer rate of helium II without naming the [[Surface_engineering|surface]] is wrong by up to a factor of seven. That paper also kills the obvious alternatives: [[Iron|iron]] transferred identically magnetised or not, a [[Superconductivity|superconductor]] identically superconducting or normal. Duthler and Pollack's fit, σ(h) = 10·(rho_s/rho)·h^-(0.26±0.05) × 10⁻⁵ cm³ per second per cm, puts clean glass at 10 × 10⁻⁵ at h = 1 cm, and a [[Neon|neon]] coating cuts it 16 per cent. The substrate is a parameter, not a detail. Against that, the microsim's [[Mathematical_model|model]] value at h = 1 cm with rho_s/rho = 1 is **7.21 × 10⁻⁵ cm²/s** — four per cent below the glass figure, twenty-eight per cent below Duthler and Pollack's amplitude, and derived from nothing but A, Γ and the two exponents. At default settings (T = 1.60 K, climb 2.45 cm) it reads d = 22.3 nm, v_c = 25.9 cm/s, rho_s/rho = 0.822 and 4.73 × 10⁻⁵ cm²/s, giving 4.26 × 10⁻⁴ cm³/s over a 9 cm rim — one 1.3 mm drop every 2.6 seconds, the once-a-second-ish drip of the classic photographs and the [[Accuracy_and_precision|order-of-magnitude]] check that the [[Physical_system|model]] is the right size. Agreement inside a factor of two, with no fitted parameter beyond A, is as much as the [[Estimation_theory|scatter]] across [[Materials_science|substrates]] lets anyone claim. ## Two men, three groups, and a name Heike Kamerlingh Onnes saw the effect first and did not know what he was looking at. Daunt and Mendelssohn open their 1938 [[Science|letter]] with it: "Kamerlingh Onnes observed in 1922 that the surfaces of two volumes of liquid helium II … in two concentric vessels adjusted themselves automatically to the same level." That quiet equalisation is now the **Onnes effect**, and it is this film doing the work inside one [[Cryogenics|cryostat]] — sixteen years before anyone had the word [[Superfluidity|superfluid]], a decade before the [[Lambda_point|lambda point]] was named. **Bernard V. Rollin**, at the Clarendon Laboratory in Oxford, came at it from the [[Heat_transfer|heat]] side. A vessel of helium II leaked heat far faster than any radiation or conduction budget allowed, and in 1936 he proposed the carrier: a mobile [[Surface_engineering|surface]] film that climbs the wall into the warm region, evaporates, and takes the enthalpy with it. He and Franz Simon published the case as *On the "film" phenomenon of liquid helium II* in 1939 — [[Interwar_period|the same few years]] in which Kapitza, Allen and Misener were establishing that helium II flows without [[Viscosity|viscosity]] at all. **The film is named after Rollin**; it is equally often called the **creeping film**, and the two names are the same object seen from the two ends of its career — a [[Thermal_engineering|thermal]] nuisance and a [[Fluid_dynamics|hydrodynamic]] curiosity. The direct [[Experimental_system|measurements]] are John Daunt and Kurt Mendelssohn's, also at Oxford: *Transfer of Helium II on Glass* in May 1938, *Transfer Effect in Liquid Helium II* that September, and the two-part Royal Society paper of April 1939 that separated the transfer phenomena from the properties of the transfer film. Fairbank and Lane's *Rollin Film Rates in Liquid Helium* of 1949 put the name in a title, pushed the [[Estimation_theory|rate measurements]] from 1.34 K to the lambda point, and found the rate to exceed isothermal [[Fluid_dynamics|flow]] under an equivalent [[Gravitational_field|gravitational]] head, the film [[Distillation|evaporating]] near the parent liquid's [[Boiling_point|temperature]]. Keep the attribution straight: Onnes observed, Rollin diagnosed, Rollin and Simon named, Daunt and Mendelssohn [[Accuracy_and_precision|measured]]. ## It stops at the lambda point, because rho_s does The rate carries the factor rho_s/rho, and the [[Superfluid_helium-4|superfluid density]] goes to zero at **T_lambda = 2.1768 K** at saturated vapour [[Boiling_point|pressure]]. So does the transfer. Say it flatly, because the film itself does *not* vanish there: van der Waals adsorption is indifferent to a [[Phase_transition|phase transition]], and helium I wets the wall just as thoroughly. What disappears is the film's ability to *move* — above the [[Lambda_point|lambda point]] all of it is normal [[Fluid_dynamics|fluid]], pinned by its own [[Viscosity|viscosity]]. The film is there and does nothing. The microsim keeps drawing it above 2.1768 K for that reason: it is the control [[Experimental_system|experiment]], and erasing the film on warming would teach the wrong lesson. The same [[Density|density]] factor governs helium II's other headline effects — [[Second_sound|second sound]] ceases at T_lambda too, because no second [[Velocity|velocity]] field is left to carry it. The [[Helium-3|helium-3]] case is the useful contrast. ³He is a [[Fermion|fermion]], not a [[Boson|boson]], and it has no lambda point near 2 K at all; it stays an ordinary [[Viscosity|viscous]] [[Fluid_dynamics|liquid]] down to about a thousandth of that [[Thermodynamics|temperature]]. A ³He bath wets its walls exactly as a ⁴He bath does, and climbs out of nothing. One further correction, because the [[Science|encyclopedia]] article this page pairs with makes it: the Rollin film is **not** the fountain effect. The fountain (thermomechanical) effect is a *different* phenomenon, driven by a temperature difference across a [[Porous_medium|superleak]] and described by dP = rho·S·dT; it squirts a jet, and it is what drives ³He circulation in a [[Dilution_refrigerator|dilution refrigerator]]. Film creep needs no temperature difference at all, produces no jet, and would happen in a perfectly [[Thermodynamic_equilibrium|isothermal]] [[Cryogenics|cryostat]]. Both fall out of the same two-fluid [[Thermodynamics|thermodynamics]] — one driven by [[Entropy|entropy]], one by [[Gravitational_field|gravity]] — but they are not the same effect. ## Why a party trick is a design constraint **An open vessel of helium II cannot be kept full.** The film climbs the wall, goes over the top, reaches warm [[Metallurgy|metal]], [[Distillation|evaporates]], and the vapour recondenses lower down — the vessel pumps itself dry through a 30 nm channel that cannot be sealed, because the film wets whatever you would seal it with. The [[Energy|energy]] bookkeeping is what makes that expensive: every gram evaporated in the warm region and recondensed on the cold bath deposits its [[Thermodynamics|latent heat]] there, so the [[Fluid_dynamics|circuit]] is an unwanted [[Heat_transfer|heat pipe]] aimed at the coldest part of the machine. In many sub-2 K [[Cryogenics|cryostats]] it is the largest single parasitic [[Thermal_engineering|heat]] load, and a [[Leak|leak]] in the literal sense: mass out, enthalpy down. It never shows on a [[Leak_detection|leak check]], because nothing is broken. Because dV/dt ∝ P and nothing else the [[Design_review|designer]] controls, the fix is always [[Geometry|geometric]]: put the smallest available perimeter in the film's path. Real hardware carries **film-flow constrictions** — knife edges, small orifices, and film burners or breakers that warm the film until it evaporates. In [[Dilution_refrigerator|dilution refrigerator]] practice, ⁴He creep up the still line is a parasitic circulation loading the condenser, and the standard remedy is one line long: film burners or diaphragms alleviate it. The same logic sets the plumbing of [[Superconducting_magnet|superconducting magnet]] baths and 1 K [[Helium_cryogenics|sorption]] stages. The flight case is clearest. NASA Goddard's X-Ray Spectrometer helium dewar was losing [[Superfluidity|superfluid]] through the [[Porous_medium|porous]] plug as film on the vent-tube walls, at about **40 micrograms per second** — roughly 8.7 litres of [[Liquid_helium|liquid]] a year, about equal to the intended evaporation loss, which would have exhausted the [[Helium|helium]] in half the planned mission. Narrowing the vent tube to **0.15 cm inside diameter** cut that to about **9 micrograms per second**, and a [[Copper|copper]] heat exchanger evaporated the rest using the 50 mK by which the emerging film and vapour sit above the bath. The backup was knife edges etched into two [[Silicon|silicon]] wafers: the film stops at each sharp bend, evaporates where it stops, cools the edge, and recondenses downstream — [[Manufacturing|micromachining]] a [[Materials_science|surface]] to make the film's perimeter as short and as hostile as possible. Check the flight number against the law. Nine micrograms per second is 6.2 × 10⁻⁵ cm³/s of liquid; divided by the 0.47 cm perimeter of that tube it is **1.3 × 10⁻⁴ cm²/s** — 1.8 times the [[Silicon_dioxide|glass]] value, comfortably inside the glass-to-etched-[[Copper|copper]] band, on a metal [[Surface_engineering|surface]]. The [[Reliability_engineering|design]] rule, the bench [[Estimation_theory|measurement]] and the [[Aerospace_engineering|spacecraft]] agree. A demonstration that looks like a [[Physics|party trick]] on a lecture bench is, three orders of magnitude away in scale, the [[Engineering|engineering]] constraint that fixes how long an orbiting [[Sensor|instrument]] stays cold. ## Sources Annotated; one clause each on what the source establishes. Bibliography lines are link-light by house rule (§4). **Discovery, naming, and the first measurements** - Rollin, B. V., and Simon, F. (1939). "On the 'film' phenomenon of liquid helium II." *Physica* **6**(2), 219–230. doi:[10.1016/S0031-8914(39)80013-1](https://doi.org/10.1016/S0031-8914(39)80013-1) — the paper the film is named from; title, authors, volume, pages and DOI verified via Crossref. **Paywalled; not opened for this article.** - Rollin, B. V. (1936). *Actes du 7ième Congrès International du Froid* **1**, 187; and Kürti, N., Rollin, B. V., and Simon, F. (1936). *Physica* **3**, 266 — the Oxford anomalous-heat-influx reports that precede the 1939 paper. **[UNVERIFIED]:** cited here as they appear in Balibar's history of the discovery of superfluidity; neither was opened, and the 1936 congress proceedings are not indexed with a DOI. - Daunt, J. G., and Mendelssohn, K. (1938). "Transfer of Helium II on Glass." *Nature* **141**(3577), 911–912. doi:[10.1038/141911a0](https://doi.org/10.1038/141911a0) — the primary source for the Kamerlingh Onnes attribution, verbatim in its opening line: "Kamerlingh Onnes observed in 1922 that the surfaces of two volumes of liquid helium II … in two concentric vessels adjusted themselves automatically to the same level." Abstract read; full text paywalled. - Daunt, J. G., and Mendelssohn, K. (1938). "Transfer Effect in Liquid Helium II." *Nature* **142**, 475. doi:[10.1038/142475a0](https://doi.org/10.1038/142475a0) — the follow-up letter, 1 September 1938. - Daunt, J. G., and Mendelssohn, K. (1939). "The transfer effect in liquid He II. I. The transfer phenomena." *Proceedings of the Royal Society A* **170**(942), 423–439. doi:[10.1098/rspa.1939.0040](https://doi.org/10.1098/rspa.1939.0040); and "II. Properties of the transfer film," *ibid.* 439–450. doi:[10.1098/rspa.1939.0041](https://doi.org/10.1098/rspa.1939.0041) — the systematic transfer-rate measurements, published 3 April 1939. Titles, volume, issue, pagination and DOIs verified via Crossref; **both paywalled and not opened**, which is why the 7.5 × 10⁻⁵ figure is attributed in this article to Dash and Boorse, where it is verifiable, rather than to these papers, where it is merely conventional. - Fairbank, H. A., and Lane, C. T. (1949). "Rollin Film Rates in Liquid Helium." *Physical Review* **76**(8), 1209–1211. doi:[10.1103/PhysRev.76.1209](https://doi.org/10.1103/PhysRev.76.1209) — rates from 1.34 K to the lambda point, and the observation that the transfer rate exceeds isothermal flow under a gravitational potential difference; evidence that the film evaporates near the parent-liquid temperature. - Kürti, N., and Simon, F. (1938). "Heat Transport in Liquid Helium below 1°." *Nature* **142**(3587), 207. doi:[10.1038/142207a0](https://doi.org/10.1038/142207a0) — the Oxford heat-transport context in which the film was first a nuisance rather than a demonstration. **The rate, the exponents, and the substrate** - Dash, J. G., and Boorse, H. A. (1951). "Transport Rates of the Helium II Film Over Various Surfaces." *Physical Review* **82**(6), 851. doi:[10.1103/PhysRev.82.851](https://doi.org/10.1103/PhysRev.82.851) — the load-bearing citation for the classic number: at 1.25 K, rates from 51 × 10⁻⁵ cm³/cm·s for etched copper down to 7.5 × 10⁻⁵ cm³/cm·s for glass, measured with a cylindrical-capacitor depth gauge, and no difference between magnetised and unmagnetised iron or between superconducting and normal states. - Smith, B., and Boorse, H. A. (1955). "Helium II Film Transport. II. The Role of Surface Finish." *Physical Review* **99**, 346. doi:[10.1103/PhysRev.99.346](https://doi.org/10.1103/PhysRev.99.346) — four metals polished and roughened, plus roughened glass; verbatim, "transport rates adjust to the prevailing microscopic perimeter and do not require the added hypothesis of anomalous flow in surface cracks," with the caveat that a large increase in roughness need not imply a large change in the perimeter the film actually sees. - Duthler, C. J., and Pollack, G. L. (1971). "Dependence of the Helium-Film Transfer Rate on Pressure Head, Film Height, and Substrate." *Physical Review A* **3**(1), 191. doi:[10.1103/PhysRevA.3.191](https://doi.org/10.1103/PhysRevA.3.191) — the height law measured directly, σ(h) = 10·(rho_s/rho)·h^-(0.26±0.05) × 10⁻⁵ cm³ per second per cm, which brackets the −1/4 prediction; the level-difference fit σ(z) = 1/[A − B·ln z] with A = (1.80 ± 0.09) × 10⁴ s/cm² at 1.65 K and (1.35 ± 0.02) × 10⁴ s/cm² at 1.28 K; a neon coating reducing the rate 16 per cent; and the neon-substrate thickness 2.3·h^(-1/3) × 10⁻⁶ cm. - van Alphen, W. M., van Haasteren, G. J., de Bruyn Ouboter, R., and Taconis, K. W. (1966). "The dependence of the critical velocity of the superfluid on channel diameter and film thickness." *Physics Letters* **20**(5), 474–475. doi:[10.1016/0031-9163(66)90958-9](https://doi.org/10.1016/0031-9163(66)90958-9) — the v_c ∝ d^(-1/4) relation the microsim uses. Bibliographic record verified via Crossref; **paywalled and not opened.** The value of the constant, Γ = 1 cm^(5/4) s⁻¹, is taken from the secondary source below, which quotes it as the experimental value from this and related work. - Volovik, G. E., and co-workers, [arXiv:0909.1140](https://arxiv.org/pdf/0909.1140) — verbatim: the results "are well described by the temperature independent critical velocity v_c ∝ d^(-1/4)", with a derived Γ = O(1) cm^(5/4) s⁻¹ "in agreement with the experimental value Γ = 1 cm^(5/4) s⁻¹". Used only to pin the numerical constant. **Reference values and the tertiary sources this article corrects** - Donnelly, R. J., and Barenghi, C. F. (1998). "The Observed Properties of Liquid Helium at the Saturated Vapor Pressure." *Journal of Physical and Chemical Reference Data* **27**(6), 1217–1274. doi:[10.1063/1.556028](https://doi.org/10.1063/1.556028) — T_lambda = 2.1768 K on ITS-90 and the tabulated rho_s/rho the microsim's blended fit is checked against. - Wikipedia, "Rollin film" — the source of the 30 nm figure and of the thickness law as usually presented. Two statements in it are corrected above: that "Rollin films are involved in the fountain effect", which conflates film creep with the thermomechanical effect, and that the Onnes effect occurs because "capillary forces dominate gravity and viscous forces", which misattributes the mechanism. Tertiary. - Wikipedia, "Superfluid helium-4" — verbatim, "the flow of the liquid in the layer is not restricted by its viscosity but by a critical velocity which is about 20 cm/s", and separately the normal-component counterflow speed "up to 20 cm/s" in heat transport. The two 20 cm/s figures describe different things and are easy to conflate. Also the source for the fountain effect being driven by a temperature difference across a superleak. Tertiary. - Wikipedia, "Superfluid film" — the critical velocity given as "on the order of 0.1 m/s", the upper end of the quoted band. Tertiary. **The engineering consequence** - NASA Goddard Space Flight Center, Cryogenics and Fluids Branch, "Superfluid Helium Film Killer" (XRS programme). [cryo.gsfc.nasa.gov](https://cryo.gsfc.nasa.gov/XRS/film_kill.html) — the flight-hardware numbers used above: film loss through the porous plug "about 40 micrograms/sec", equal to the expected evaporation loss and therefore halving mission life; a vent tube of "0.15 cm I.D." reducing film flow "to about 9 micrograms/sec"; film and vapour leaving the plug "50 mK higher in temperature than the superfluid helium bath", which the copper heat exchanger uses to evaporate the film; and the backup device, "a series of knife-edges etched into 2 silicon wafers", at which the film stops, evaporates, cools the edge and recondenses downstream. - Heidelberg cryogenics group, *Design and Construction of Dilution Refrigerators*. [emp.kip.uni-heidelberg.de](https://emp.kip.uni-heidelberg.de/cryotools/doku.php?id=wiki:design_construction) — the still-line case: ⁴He film flow creates "a parasitic ⁴He circulation which loads the condensation stage at the refrigerator inlet", and "film burners or diaphragms alleviate this problem." ## Wikipedia : Wikitube **Strict pair:** [Wikipedia](https://en.wikipedia.org/wiki/Rollin_film) : [Wikitube](https://en.wikitube.io/wiki/Rollin_film) ## Previous hub tags Hubs: `Life_Physics`. Portals: [[PORTAL_Helium]], [[PORTAL_Helium-3]].