A Review of Some Experimental Spray Methods for Marine Cloud Brightening ()
1. Introduction
This paper summarizes the results of a number of research experiments that we have carried out to produce a very fine spray of seawater that might be suitable for the Marine Cloud Brightening (MCB) geo-engineering method [1-3]. The technique proposes to counteract the thermal (and regrettably, only the thermal) effects of climate disruption by increasing the world-wide cloud albedo of marine boundary layer clouds by introducing artificially formed salt nuclei into these clouds.
The goal was to identify a spray system that could conceptually provide roughly 1017 Cloud Condensation Nuclei (CCN) per second from a single seafaring vessel. Assuming no coalescence, and no significant disintegration of droplets once formed, this would mean the initial creation of an equal number of droplets per second. The formation of such fine sprays implies, at its minimum, creation of the surface energy associated with the increase in surface area when the liquid is dispersed. Most methods require considerably more energy than this minimum, but because over its lifespan a cloud droplet reflects many million times the energy needed to create it, the MCB method remains very energy efficient. While we attempted to minimize the required energy in each approach, the primary research goal was to develop a working method suitable for scientific cloud experimentation, regardless of energy cost.
As envisioned by Salter et al. [3], this method would spray from each vessel about 30 L/s of sea water in droplets 0.8 µm in diameter, yielding 1017 nuclei per second. Worldwide, 4.5 × 104 kg/s of seawater would need to be dispersed. Preferably, to be most effective, the nuclei formed should be monodisperse or of narrow size distribution [3].
It is not the amount of sprayed water that is important, but the number of suitable nuclei that are formed. Each suitable nucleus, above a certain minimal salt mass will create a droplet many times its original size when converted in the cloud. The critical requirement is that the salt mass, ms, be high enough that it can convert into cloud droplets at supersaturations S occurring in marine stratocumulus clouds. S depends on updraft speed, and the properties of the air mass. Cloud modeling has provided values of this critical mass for a variety of relevant scenarios. They show that significant droplet formation and its associated cloud albedo increase can occur for salt masses down to about 5 to 7 × 10-20 kg [4]. This would correspond to dry salt cubes 28 to 32 nm on edge (35 to 40 nm equivalent sphere diameter), and to sprayed seawater droplets on the order of 140 to 160 nm in diameter (a factor of four reduction in diameter occurs when seawater droplets of 3.6% salt concentration evaporate).
For maximum energy efficiency, it is of course advantageous to make the droplets as close as possible to this lower acceptable limit, thus limiting the amount of liquid that needs to be sprayed and the associated spray production energy.
There is believed to be a significant secondary benefit. The smaller the amount of liquid sprayed, the smaller the negative buoyancy created by the evaporation of sprayed droplets, which might possibly inhibit their vertical ascent. The intense concentrated spray, i.e. 30 L/s envisioned in some MCB implementations, and its associated concentrated cooling, may inhibit the buoyancy required for the nuclei to reach the marine boundary layer even when using a considerable forced air launch (10 m/s). Artificial cloud formation using water spray over land has been a commercial goal for a long time for various special effects in movies, TV commercials, etc., but it has never been achieved [5]. Even when substantial updrafts are used, invariably the result is a low-hanging fog, a consequence of the intense localized evaporative cooling. Simple engineering calculations confirm these effects. While the negative buoyancy created by evaporative cooling is not exactly proof that the approach will necessary fail over the ocean, where a different set of updraft and wind conditions exist, we have eventually concentrated on methods that do not suffer, or at least suffer much less, from this potential difficulty, mainly because in general we spray much less water to get the same number of particles.
For completeness we will describe all of the techniques that have been investigated, including those that failed to achieve the desired result. Currently we are investigating the spraying of saltwater at or near critical conditions, with initial results that are very promising. These results will be presented in a different publication.
2. Results and Discussion
2.1. Use of Commercial Nozzle Sprayers
At the outset of the investigation we investigated, as a starting point, the performance of a number of standard commercial nozzles (Mee Industries, Bete, Steinen) used in fogging systems. The purpose was to see if these nozzles and the particle size distributions they produced could be scaled down by micro-machining. A Malvern Spraytec model RTS5114 particle size analyzer (courtesy of Mee Industries, Inc.), calibrated using lithographically defined targets (courtesy of D. Hirleman, Purdue University), was used to analyze the spray. The nozzles were fed from a high pressure vessel (Parr Instruments 4601) pressurized with nitrogen. This pressurization produces a significant amount of dissolved gas at high pressure, with subsequent expansion upon spraying. However, it is believed that the nozzle operation was not significantly affected by these differences, although gaseous desolvation, if used properly, can reduce the particle size distribution of sprays (as observed with effervescent spray nozzles).
Impact spray nozzles supplied to us by Mee Industries, by far the best-documented fog nozzles, operated at 6.9 MPa (1000 psi), produced droplets with a Sauter Mean Diameter of 15 µm (somewhat larger than that reported by Chaker) [6]. The Sauter Mean Diameter (SMD, d32) is defined as the diameter of a drop having the same volume to surface ratio as the entire spray. These nozzles direct a high speed jet (150 µm diameter) toward an impact pin, where a conical spray cone is formed. Figure 1 shows the results for a Mee Industries nozzle measured with the Malvern Spraytec. It is typical of the results for all commercial nozzles tested. A further increase in pressure produces only a marginal decrease in the droplet size.
These nozzles form rapidly moving, thin, conical liquid sheets, whose interaction with air leads to a pressure-driven radial instability creating broken up flat rings [7,8]. The breakup mechanism described is as follows: any perpendicular perturbation on the sheet creates a
Figure 1. Particle size distribution from a Mee Industries nozzle as measured with the Malvern Spraytec showing both particle size and the cumulative mass distribution.
pressure excursion (as in an airplane wing), thereby amplifying itself and creating breakup into flat rings at some optimum wavelength. Surface tension then contracts these flat rings into thin toroids. The toroid subsequently breaks up longitudinally according to the Rayleigh jet criterion [9]. In contrast to Rayleigh jet breakup, the size distributions formed are far from monodisperse, so that the theory is at best a rough approximation. Further drop breakup may occur through turbulent air interaction, depending on the velocity of the spray.
If the liquid sheet was as thin as 1 µm, one would end up with droplets as large as 10 µm if no further turbulent breakup occurs [7]. In order to produce the desired droplet size, extreme miniaturization would be needed.
More elaborate nozzles, such as the effervescent spray nozzle [10], which use a mixture of fluid and compressed air, are capable of producing sprays with 2.5 µm diameter droplets. This would require the spraying of almost 900 kg of water per second to achieve the required nuclei count and would produce significant air cooling.
2.2. Toroidal Cone Sprayer with Electrical Charging
Inspired by a design for a very high power water-cooled X-ray lithography tube [11], a special conical nozzle with a toroidal cone (see Figure 2) was designed. The purpose was to create a liquid sheet in a form that can be highly charged so that additional disintegration of the droplets may be accomplished electrically by Coulombic fission (the electrodes, not shown in Figure 2, surround the cone exit).
In this device a jet is projected onto a curved conical cusp. The curved cusp produces very high centripetal acceleration (approaching 400,000 g), pinning and thinning the liquid along the cone surface. Because of this guiding the liquid can thin without premature break-up. At the cone exit, electrical charging can be applied to the emerging liquid sheet, with the aim of producing further
Figure 2. Toroidal Sprayer before assembly of the electrodes. The liquid thins along the cone, pinned by centrifugal forces, producing a very thin exit sheet.
droplet disintegration during evaporation. In principle the emerging thin flat liquid sheet lends itself very effecttively to induction charging from planar conical electrodes.
The toroid was designed to gradually thin an initial liquid annulus 100 µm in thickness to 1 µm thickness at the cone exit. Experiments showed that the sheet slowed down significantly before reaching the cone exit. Modeling the flow along the toroid showed that as the liquid sheet thickness approached the boundary layer thickness, excessive friction losses would overtake and slow the flow down.
The use of graphite air bearing technology for the cone, to reduce friction, was considered. But because uncharged liquid sheets even at 1 µm thickness will produce 10 µm droplets, redesign for exit thickness greater than the boundary layer, or using boundary layer thinning techniques was not pursued. As will be shown, even charging would not produce the desired result.
If the emerging droplets are highly charged and evaporate, they could potentially reach the Rayleigh droplet stability limit where electrical pressure overcomes surface tension [12,13]. Mother droplets will then eject charged daughter droplets through the emergence of two opposing Taylor cone-jets, until the electrical pressure is significantly less than the surface tension for the parent droplet [14]. The ejection carries a relatively large amount of charge, but relatively little mass, and is self extinguishing [15,16]. By a fortuitous coincidence, given the high conductivity of seawater, these daughter droplets are very small, close to the desired range, with a size independent of that of the generating droplet [17]. The process repeats itself as the evaporation of the parent droplet proceeds. In principle this would go on until all droplets are completely dispersed (or solidified).
Unfortunately, the amount of charging that can be produced is not sufficient to create instant Rayleigh disintegration of the droplets and produce daughter nuclei by droplet fission. Hence, we expect that we would need to rely on substantial evaporation before this could happen. Analysis shows that charging a 1-µm thick liquid seawater sheet (on both sides) with fields approaching electrical breakdown in air at the electrodes, does not produce enough charge to ever produce droplet fission. At the level of charge that can be imposed, and as the droplets contain 3.6% salt, droplets would crystallize after shrinking by a factor of four in radius before ever reaching the Rayleigh limit. Hence splitting could not occur unless some other process took place.
2.3. Colliding-Jets Spraying
Suggestions have been made by various authors [3,18] that the head-on collision of two jets might produce more intense and finer atomization. Intuitively, a symmetric collision might be expected to channel more of the energy of the incoming jets into disruptive energy (being a center of mass system) and might produce droplets of smaller diameter.
What is observed experimentally is that if the two identical emerging jets have not broken up before collision, they fuse to form a perpendicular moving radial liquid sheet, just as if each one of them were to hit a solid surface [19]. This free-standing radial moving liquid sheet itself then breaks up as described previously for the liquid sheets from fogging nozzles, but as it has twice the sheet thickness, it actually ends up producing larger droplets. Figure 3 shows a picture of the colliding jet experiment. It was found that the jets need to be slightly angled to each other to produce a stable intersection, as the intersection is otherwise indeterminate.
It might be surmised that if the jets were fully broken up before collision, individual droplets might be ex-
Figure 3. Colliding Jet Experiment. Under a pressure of 70 bar the jets are emanating at a velocity of 60 m/s, and are 1.4 mm in diameter. The radially moving sheet can (barely) be seen at the intersection of the jets.
pected to collide with the desired effect. However the jets do not disperse rapidly, in spite of the fact that they are saturated with dissolved nitrogen at 6.9 MPa from the driving pressure. In an attempt to produce explosive desolvation of the absorbed gases in the jets, high intensity ultrasonic energy at 30 kHz from the horn of a cell disrupter (Branson Sonifier 450) was applied to the exit nozzles. This created a marked change in the exit appearance of the jets, but the sprayed droplet distribution of the colliding jets was not measurably improved. Figure 4 shows a typical Spraytec result for the droplet distribution of the colliding jets of Figure 3. These distributions are unsuitable for MCB work.
2.4. Ultra-High Pressure Jet Spraying
Increasing the driving pressure of a jet and decreasing surface tension can in principle decrease the droplet distribution size. Experimentally it is well established that some diesel injectors operating at 13.8 MPa (2000 psi) produce 2-µm droplets [20].
The average particle size emerging from a diesel injector nozzle follows the following scaling laws of Equations (1) and (2) [20]:
(1)
where:
(2)
SMD = Sauter mean diameter
σl = surface tension of the liquid
μl = dynamic liquid viscosity
ρl = liquid density Vl = injection velocity
Figure 4. Spraytec results from colliding jets: pressure 6.9 MPa (1000 psi), 1.4 mm jet diameter, jet velocity 60 m/s.
ρg = gas density Pl = liquid pressure Pg = gas pressure Cv = orifice flow coefficient Briefly, reducing the liquid’s surface tension, increasing the driving pressure and, to a smaller degree, reducing the liquid’s viscosity and density all reduce the droplet size. The exponents for density and viscosity dependence are small (0.1, 0.5). The formula suggests an almost inverse linear dependence on the driving pressure.
Compact, mass-produced and relatively inexpensive pumps for water-jet cutting machines can now achieve over 600 MPa, and have flow rates on the order of 90 mL/s. Thus, if the sprayed droplet size distribution were to be well into the submicron range, only a modest number of jets per ship would be required.
We tested the plume of a jet emanating from a waterjet cutting machine, operating at 344.7 MPa (50,000 psi) without the cutting grit present, at the Monterey Bay Aquarium Research Institute (MBARI) [21]. Because the jet is oriented vertically for cutting purposes and could not be reoriented horizontally for the experiment, it was deflected off a flat carbide tool oriented at 45˚. The carbide tool was positioned close enough to the jet so that there was no visible jet breakup before bouncing off the tool.
Experimentally, very broad distributions (10 to 250 µm) were observed in the plume of the jet, rather than the small particles that had been hoped for. Figure 5 shows the MBARI experiment in operation and Figure 6 shows the Spraytec results. It can be seen that the plume, given its 45˚ launching rises inside the warehouse, but levels off and stops rising before full evaporation.
There are very few, if any, submicron particles to be seen in the measured distribution. (It should be noted that even if they had been present, such particles could not be measured with the instrument in use.) However, if they had been present in copious amounts one might have expected a bulge in the lower end of the spectrum. Such a signal was, however, not observed in any of the histograms.
No surfactant was used. It would have been advantageous to introduce the surfactant in the port where cutting grit is normally supplied thereby surrounding the surface of the emerging jet, but this was not performed.
The nozzle for this water jet cutting device is, of course, specifically designed for jet formation, not dispersion. A more appropriate design might have produced results more in line with expectations from the theory.
2.5. Rayleigh-Jet Spraying through Small Apertures
In terms of droplet uniformity, and also from an energy view point, the most efficient way to produce narrow
Figure 5. MBARI jet experiment with a 344.7-MPa jet.
Figure 6. Spraytec results obtained at MBARI from a 344.7-MPa (50,000 psi) jet.
particle size distributions is to create an array of sprays using acoustically controlled Rayleigh-mode jet breakup [9]. Provided all the jets are of the same size, and with suitable precautions such as perpendicular airflow and appropriate charging to inhibit the coalescence of the resulting drops, very narrow distributions with only a few percent droplet coalescence can be obtained [22].
To generate the 0.8-µm diameter droplets as proposed by Salter et al. [3] would require apertures on the order of 0.5 µm in diameter. Historically, suitable small holemaking was the difficult part of the technology. In the last decade there has been great technological progress in the fabrication of small holes in silicon using the Bosch etching process, achieving spectacular hole aspect ratios (>50). Because of the need for through-silicon wafer “vias” (TSV), even standard etchers have lately achieved remarkable results, and this technology has entered not just micromachining, but mainstream silicon manufacturing.
The main problem with this approach is therefore no longer the formation of small holes, but keeping them open while spraying over extended periods without clogging. Clogging is a poorly understood process and may have both chemical (reactive film deposition) and physiccal (particulate obstruction) components to it. A generally accepted rule of thumb in spraying is that no particles larger than one tenth of the spray aperture diameter should be present in the fluid to be sprayed. Techniques such as periodic backflow may be used to flush out debris and extend the lifetime [3].
The holes should have minimum flow impedance for minimum clogging and low power requirements. Optimum holes have a roughly conical shape, with the apex at the exit, so as to produce a significant vena contracta, further reducing the exit jet diameter and requiring minimal pressure across the aperture for a given jet diameter.
To produce flow through small holes over an extended period, we have found it strictly necessary to employ the fluid filtering system shown in Figure 7.
The fluid to be sprayed is first circulated and filtered continuously in its closed loop for an extended period of time (12 - 24 h) before spraying.
We were able to spray continuously for 140 h (20 L) through square, 5-µm, anisotropically etched holes in silicon, using a 0.5-µm filter followed by a 0.2-µm final filter just in front the apertures. Without the pre-filtering operation, clogging was disappointingly fast (a matter of seconds). No back flushing (to remove accumulated debris) of the aperture was used in this experiment. While the target 0.5-µm hole is 10 times smaller in diameter (100 times in cross-section), this result encouraged us to attempt the fabrication of much smaller holes.
An ideal substrate for spray aperture fabrication would
Figure 7. Saltwater filtering system using continuous recirculation before spray.
be a titanium foil, 12 µm thick. Nearly corrosion-free and not brittle, Ti lends itself both to wet and dry etching. The foil can be formed into a spherical shape, thereby providing angularly divergent jets, which should help to avoid coalescence of the sprayed particles. Entrainment and pumping of air causes collinear jets in dense arrays to collapse towards each other, resulting in significant coalescence.
Through cooperation with, and a grant from, the Stanford Nanofabrication Facility the fabrication of 0.5-µm holes in 12-µm thick titanium membranes was undertaken. Ti membranes supported by silicon wafers were patterned both with e-beam and contact projection lithography. However, wet etching (producing hemispherical etch pits) could not achieve the desired resolution. Although dry etching could very likely achieve the desired geometries, the Si etching equipment at the facility could not be used with Ti because of the ensuing silicon contamination issues in the etcher. The fabrication of these Ti holes was therefore given lower priority until the issue of 0.5-µm hole-clogging could be resolved.
Hence it was decided to use as a substitute the holes in commercial radiation-track filters, well known in biological research. These filters have holes in polycarbonate membranes, created by alpha particle tracks that are subsequently chemically etched, are remarkably uniform and well controlled and are therefore used as absolute filters (see Figure 8). They are inexpensive and commercially available from a variety of manufacturers (e.g., Whatman Corp.).
The cylindrical hole shape provides for a large viscous resistance, and over 3 MPa of pressure would be required to produce the jet velocity required for Rayleigh spraying. This is well outside the operating parameter region of our present filter system. Because of the thinness of these unsupported flexible membranes (8 - 10 µm), and the high pressure required by the cylindrical holes, fully formed free jets could not be drawn out of this arrangement.
Therefore we could only operate the system near its maximum pressure, producing only a seeping flow through the holes, which was measured over time, presumably giving an estimate of the clogging rate of the holes. The seepage rate decreased by an order of magnitude over 350 h, indicating prompt clogging of the 0.6-µm holes at low flow rates.
Figure 9 illustrates the measured inverse flow rate (right) and total flow (left) as a function of operating time. The seawater had first been circulation-filtered for 8 h through a 0.05-µm filter to remove all possible debris before actual “spraying” was started.