Sei sulla pagina 1di 17

VOLUME 18

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

MAY 2001

A Continuous-Flow Diffusion Chamber for Airborne Measurements of Ice Nuclei


DAVID C. ROGERS,* PAUL J. DEMOTT, SONIA M. KREIDENWEIS,
AND

YALEI CHEN

Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado (Manuscript received 8 November 1999, in nal form 1 August 2000) ABSTRACT A continuous-ow thermal gradient diffusion chamber was developed for operating in an aircraft and detecting ice nucleating aerosol particles in real time. The chamber volume is the annular space between two vertically oriented concentric cylinders. The surfaces of the chamber are coated with ice and held at different temperatures, thus creating a vapor supersaturation. Upstream of the chamber, all particles in the sample air larger than 2- m diameter are removed with inertial impactors. The air then ows vertically downward through the chamber, where ice crystals nucleate and grow on active ice nuclei to between 3- and 10- m diameter in 310 s of residence time. At the outlet of the chamber, an optical particle counter detects all particles larger than 0.8 m. Those particles larger than 3 m are assumed to be the newly formed ice crystals and comprise the ice nucleus count. This paper describes the principles of operation, hardware and construction, data system, calibration, operational procedures, and performance. Limitations of the technique are presented, and examples of measurements are shown.

1. Introduction Ice nuclei (IN) are aerosol particles that catalyze the formation of ice crystals in clouds (Vali 1985). A wide variety of measurement techniques have been developed over the past 50 years for detecting IN and measuring their characteristics. These techniques include drop freezing, particle capture on supercooled drops, particle capture on lters followed by processing, cloud chambers (static or ow-enhanced diffusion, slow or fast expansion, mixing, sedimenting drops), and others. It is generally agreed that ice will form on nuclei in response to different kinds of thermodynamic forcing, with the primary variables being temperature (T), supersaturation (SS), and the presence of a surface of supercooled liquid water (Vali 1985; Cooper 1974). Four ice nucleation mechanisms are currently recognized: deposition, condensation-freezing, contact-freezing, and immersion-freezing (Vali 1985). In controlled laboratory experiments, it is possible to categorize ice nucleation mechanisms according to whether the waterice transition occurs primarily through vapor or liquid paths.

*Current afliation: Research Aviation Facility, National Center for Atmospheric Research, Broomeld, Colorado. Corresponding author address: David C. Rogers, Research Aviation Facility, National Center for Atmospheric Research, 10802 Airport Ct., Broomeld, CO 80021. E-mail: dcrogers@ucar.edu

In a similar fashion, dominant nucleation mechanisms can sometimes be deduced from eld data by comparing the predicted sizes and numbers of ice crystals with those observed (e.g., Cooper and Vali 1981; Heymseld and Miloshevich 1993). In most cases, there is always a degree of uncertainty, and several mechanisms may be active at the same time. Even after 50 years of ice nuclei studies, a calibration standard ice nucleating material does not exist, and there is no general agreement on a standard technique for measuring IN or on how to interpret IN measurements. Because a variety of processes can lead to ice formation, mechanistic measurement approaches are preferred for improving the understanding of these processes. Atmospheric ice processes are very important in determining the properties of clouds and the development of precipitation, hence there is considerable scientic interest in trying to describe ice formation and growth processes (e.g., Szyrmer and Zawadzki 1997). Scientic review articles of cloud microphysics have consistently called for new ideas, greater versatility, and improved performance of ice nuclei measurement techniques (e.g., Hallett 1983; Cooper 1991; Rasmussen 1995; Baker 1997; and others). Topics of interest include the following. R For subsaturated conditions, detecting and quantifying deposition activity [there is some question about whether deposition occurs on natural IN particles. Its appearance is similar to the freezing of solution drops, although it is a conceptually different mechanism (i.e.,

2001 American Meteorological Society

725

726

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

R R R R R R

vapor transfer direct to ice in the absence of the liquid phase)]; the freezing of activated cloud condensation nuclei (CCN) at slight water supersaturation (0% to 2%); the ice nucleating response of particles at high supersaturations ( 5%); vertical atmospheric proles of IN concentrations; physicochemical properties of IN; geographic regions and natural processes that produce IN; and processes that modify IN activity, such as cycles of condensation and evaporation in clouds, or exposure to pollution gases.

ment and its operating principles, hardware and construction, software, calibration, and performance. Examples of measurements are shown, limitations of the technique are outlined, and suggestions for future development are offered. 2. Description The airborne CFD instrument has been used in several eld projects. It is illustrated in Fig. 1 as installed in a standard 19-in. rack for the National Center for Atmospheric Research (NCAR) Electra aircraft. The entire rack-mounted system is approximately 65 110 130 cm tall, weighs 120 kg, and requires 15 amps of 120 vac power. The mounting arrangement varies for different aircraft. The chamber described here is a descendent of the laboratory instrument described in Rogers (1988, 1994). The lab instrument was used to measure natural aerosol in Wyoming (Rogers 1993), in the 1993 and 1994 Winter Icing and Storms Projects (Rogers et al. 1996), and in North Dakota (DeMott et al. 1995b). It was also used, along with other IN instruments, in studies of silver iodide aerosols (DeMott et al. 1995a). Motivation for developing the airborne version came from generally favorable quantitative comparisons of IN measurements and aircraft observations of ice in clouds in the same region (Rogers et al. 1996). These comparisons were based on bag samples of air collected by the aircraft and delivered to the laboratory for IN measurements. The airborne instrument circumvents the problems associated with collecting air samples in bags and provides the capability for more direct comparisons between aerosol particles and ice formation in clouds. The evolution of the instrument is summarized in Table 1, starting with the laboratory prototype described in Rogers (1988). The airborne chamber is in the middle column, and a newer laboratory chamber is listed in the right column. Important criteria for the airborne version included 1) limiting the size, weight, and power; 2) ensuring the chamber was air tight; 3) recording data at rates 1 Hz or faster; and 4) simplifying and automating many of the operational and analysis procedures. The table shows that only the chamber geometry and diameter have remained the same. The chamber was lengthened to increase the time available for nucleation and growth of ice crystals. Operation of the chamber was considerably simplied with the addition of computerized display and data recording in 1996. These upgrades improved the accuracy of sampling conditions, made them much easier to control, and simplied data analysis. 3. Principles of operation A simplied illustration of the components and conceptual principles of operation in the CFD chamber is shown in Fig. 2. The chamber volume is an annular gap

The variety of different approaches for measuring ice nuclei is diverse. They can be classied according to instrument type (mixing chamber, diffusion chamber, ltration), nucleation mechanism (drop freezing), thermodynamic process (expansion), and so forth. Thus, the same technique can be described several different ways. A common feature of ice nuclei instrumentation is that the nucleation event is detected in association with the appearance of the new phase as: a) growth of new ice crystals to detectable size in a supersaturated region, b) thermal wave from the release of latent heat when a supercooled drop freezes, c) change in opacity of millimeter size droplets when a supercooled drop freezes (gas bubbles dissolved in liquid are created as ice forms), or d) a change in a bulk or ensemble integrated property, such as the dielectric constant. In all of these cases, there is a time delay separating the nucleation event from its detection. Of the measurement approaches available today, only the controlled-expansion cloud chamber can be used to study all four nucleation mechanisms (DeMott 1995). Some techniques try to isolate and emphasize one mechanism; with other techniques, a separation is not possible. Ice nucleation measurement comparison workshops were held in the United States in 1970 (Grant 1971) and 1975 (Vali 1975). In both workshops, experiments were performed with both natural and laboratory-generated aerosols. Measurements from the various instruments did not produce close quantitative agreement at either workshop, but important factors for IN activation were identied, including sensitivity to vapor supersaturation, competition for vapor in lter processing, the increasing activity with particle size, and recognizing the need to identify nucleation mechanisms. Demonstrating the difculty of measuring ice nuclei was a signicant result in itself. For CCN, several intercomparison measurement workshops have taken place since the mid-1970s, but there have been no subsequent workshops for ice nuclei. This paper describes a continuous ow diffusion (CFD) chamber instrument that was developed from an earlier laboratory chamber for airborne use. It has some advantages and some disadvantages, which are discussed here. The following sections describe the instru-

MAY 2001

ROGERS ET AL.

727

FIG. 1. Rack mount arrangement for CFD chamber system as congured for NCAR Electra aircraft. Scales are in.

TABLE 1. Properties and evolution of three ice nuclei CFD chambers. Airborne chamber in middle column. Lab version in right column. Same indicates earlier feature was used again. MCA multi-channel analyzer. Description Chamber geometry Chamber length (cm) Growth region (cm) Cylinder diameters (cm) Annular gap (cm) Thermal control 198288 Vertical axis concentric cylinders 45 30 8, 10 1.1 Two refrigerated bath circulators Same 199698 Airborne Same 150 100 or 150 Same Same Two refrigerated bath circulators Either Climet Same Same Optional Same 660 s particle spectra Same Same Same 8 thermocouples Same Integrated MCA Same Same 1998Lab version

100 50 Same Same One refrigeration compressor and expansion valves Particle detector Climet A6065a Climet 7350A Applying ice on walls Spray water on walls with Flood chamber volume with wachamber open ter Warm wall Inside cylinder Outside cylinder Insulated wall surface (drop evap- Foam insulation on lower 1/3 of Lower 1/3 section plastic PVC oration region) inside cylinder pipe Data logging Strip chart, computer les and Computer written notes Time resolution 10 s to 5 min 0.2 s particle count, 10 s particle spectra (1998) Sample conditions: ow prole, Postexperiment analysis Real-time calculation and dissample location, temperature, play super-saturation, critical ow Airow Rotameters Rotameters and mass ow sensors Air pressure Aneroid gage Strain gage sensor Temperature 4 thermocouples 6 thermocouples Temperature reference Thermocouples 2 precision thermistors Particle size Separate MCA Single threshold (96), integrated MCA (98) Exhaust air Discard Recycle to sheath Real-time data display Temperature, pressure, and parti- Computer-generated plots cles

728

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

FIG. 2. Simplied illustration of CFD chamber components and operating principles. FIG. 3. Temperature and supersaturation conditions at location of sample air as calculated from warm and cold wall temperatures with ow of 10 L min 1 (Rogers 1988). Isotherms ( C) slope from upper left to lower right, and isohumes (% water supersaturation) curve from lower left to upper right.

between two ice-coated cylinders. The cylinders are held at different temperatures, and a supersaturation forms between them. In the rst few centimeters, the sample is smoothly injected into the 1.1-cm annular space between two lamina of particle-free sheath air. The aerosol sample comprises about 10% of the total ow. Droplets and crystals nucleate in the supersaturated region and grow as they pass through the rst two-thirds of the chamber. The last third has no ice on the warm cylinder; instead, it is dry and a poor heat conductor, and this change in boundary conditions creates a region that is below water saturation, forcing water droplets to evaporate (Rogers 1988). This evaporation region helps to exaggerate the size difference between ice crystals ( 3 10 m) from the much smaller residues of cloud droplets ( 0.5 m). At the chambers outlet, the particles are counted with an optical particle counter (OPC); the amplitudes of the OPC voltage pulses correspond to particle size. The pulses are digitized and accumulated in a multichannel analyzer (MCA) and recorded by computer. The detection of ice crystals (or ice nuclei) is based on a particle size exceeding 3 m. Since there are particles 3 m in the air, they could be erroneously counted as crystals. In order to avoid that error, particles larger than 2 m are removed at the inlet of the instrument by an impactor system, as described in section 3b. The wall temperatures and rate of airow determine the temperature and supersaturation that the sample experiences. Sample temperature and supersaturation are independently controllable. Generally speaking, the sample temperature corresponds to the average temperature of the two wall temperatures, whereas the supersaturation corresponds to their difference. Actually, the sample air is not located in the middle, but slightly on the cold side of the midpoint, due to the convective component of the ow. Equations for calculating the time-dependent temperature, supersaturation, and velocity proles were derived in Rogers (1988). Figure 3 shows the relationship between temperatures of the warm and cold walls and the sample conditions at a position four tenths of the distance from the cold wall

to the warm wall (a typical location) after steady-state linear proles of temperature and vapor concentration are achieved. This chart provides guidance for setting the sample conditions. For example, with the walls at 31 and 16 C, the sample is exposed to 25 C and water saturation. It is possible to set the wall temperatures to obtain very high water supersaturations (SS w ) nominally up to 60% and higher, although CCN and IN will activate and grow, thereby limiting the attainable supersaturation. The ultimate limit depends on CCN concentration, temperature, and whether any ice nuclei are present. We estimated the effect on SS w due to vapor competition by droplets and crystals, using equations for a parallel plate CFD chamber from Rogers (1988) and from Plooster (1985). The calculations assumed 1000 cm 3 of CCN particles (activating half at 0.1% and the remainder at 0.5%), 20 C, and ice particle concentrations from Meyers et al. (1992). When only droplets were present (no ice crystals), the calculated supersaturation decreased according to the vapor excess, as follows: at 5% SS w , the droplets had only a slight effect on the vapor eld, reducing it to 4.97%; at 10% SS w , droplets lowered it to 9.8%; at 20%, droplets lowered it to 19%; and by 40% SS w , the presence of droplets reduced the supersaturation to 36%. Ice crystal growth was estimated using a diffusional growth equation with the same condensation and thermal accomodation coefcients as assumed for water (0.04 and 1, respectively). Ice crystals were grown as spheres. Nucleated concentrations of ice ranged from 0.038 cm 3 at 5% SS w to 4.079 cm 3 at 40% SS w . There was no discernable effect of ice crystal growth to reduce SS w beyond the effect from droplets. One may ask why measurements of IN should be made at such high supersaturations when the common understanding among cloud physicists is that the max-

MAY 2001

ROGERS ET AL.

729

FIG. 4. Airow in CFD chamber instrument. Inlet and exhaust min 1 .

1 L min 1 ; sheath

9L

imum rarely exceeds 1% or 2% in natural clouds. There are several reasons. For example, it has been speculated that part of the discrepancy between ice concentrations observed in clouds and measurements of IN could be attributed to an occasional occurrence of very high supersaturations in small pockets in cumulus clouds (e.g., Hobbs and Rangno 1990; Baker 1991; Rogers et al. 1994; Shaw 1998). The response of natural aerosol particles to high SS w can be explored with the CFD. Another reason is that high SS w set points ensure that water saturation is exceeded, even when CCN concentrations are high. With high SS w , all CCN are activated, and droplets will supercool and grow quickly enough to avoid the freezing point depression of concentrated solutions. High supersaturations may also be appropriate for laboratory experiments on particles with hydrophobic surfaces. In measurements of any cloud-active aerosol particles, transient high supersaturations should be avoided. If the sample is saturated and colder than the sheath air, then calculations indicate transient supersaturations can occur in excess of the steady SS values achieved downstream (Fitzgerald 1970; Mahata et al. 1973; Rogers 1988). Several ways are employed in the CFD to avoid such transients. First, 90% of the chambers exhaust air is ltered and used as sheath air. This air contains less moisture, since it is ice-saturated at the cold wall temperature. When sampling at low altitudes in the summer, the air can be quite humid. In this case, a diffusion dryer (e.g., Perma-Pure tube) can be used to reduce the vapor concentration. The second way of avoiding transients is to keep the walls dry initially, while the

temperature is adjusting; in the CFD, the walls are dry for the rst 10 cm. a. Specications Performance goals for the airborne chamber were established from scientic and practical criteria. The scientic goals included sampling in the temperature range 10 to 40 C and humidity from ice saturation to 20% water supersaturation, with sampling rate at least 1 L min 1 . Smooth laminar ow is needed 1) so that the temperature and humidity elds are well dened, allowing mechanistic studies of particular nucleation mechanisms, and 2) to preserve the temporal resolution for real-time sampling. The practical goals were to have the instrument t in standard 19-in. racks in an aircraft cabin as small as that of a Beech King Air ( 127 cm height), to operate at outside pressure in a pressurized cabin, and to use a sufciently small amount of conventional electrical power (115 vac 60-Hz single phase) for laboratory or aircraft operation. The present design meets these goals. b. Airow A sketch of the airow distribution in the CFD system is shown in Fig. 4. The total ow through the CFD chamber is 10 L min 1 , of which the sample is 10%. The other 90% is ltered and returned to the chamber as sheath ow. Electrical sensors measure the pressures, ows, and temperatures. Mass ow meters monitor the air ows, so that the ball-type ow meters are not nec-

730

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

FIG. 5. Inlet impactor removes particles 2 m upstream of CFD chamber. Cut-away view shows construction based on compression tubing tee. Two such impactors are used.

FIG. 6. Efciency of inlet impactor to remove particles. (Dots) Measured efciency for double impactor and 0.8 L min 1 . (Thin line) Theoretical collection efciency based on Ranz and Wong (1952) equations for a single round jet impactor with 0.8 L min 1 and particle density 1.5 g cm 3 .

essary, but they provide convenient visual cues and conrmation of the electronically measured values. The air pump is a diaphragm type and produces oscillations 30 Hz in the airow. These oscillations show up in the mass ow sensors and can produce erroneous estimates of the average ow through aliasing. Check valves and ballast tanks ( 1 L) on the inlet and outlet of the pump are used to damp these oscillations. Rigid stainless steel tubing is used where possible. Otherwise, exible electrically conductive tubing is used on the inlet side of the CFD and silicone tubing on the outlet side. At the system inlet, there are two identical inertial impactors in series to remove particles larger than 2 m. Figure 5 shows the construction of these impactors. The distance from the jet nozzle to the impaction surface is adjusted to equal the nozzle diameter in order to optimize its performance (Marple and Liu 1974). The impacting surface is coated with a thin layer of vacuum grease to increase the likelihood that impacting particles will stick. Measurements of the collection efciency at room pressure ( 850 mb) are shown in Fig. 6 along with theoretical calculations based on Ranz and Wong (1952). The collection efciency, as a function of particle size, was measured by an OPC sampling ambient air with and without the impactor; 200 MCA channels cover the size range 0.6 to 4.8 m. The large variability above 1.5 m is due to the large sampling uncertainty (very few particles got past the impactor). The 90% cutoff occurs at 1.7 m. For particles 2 m, approximately 5% get past the rst stage, but 95% of these are captured in the second stage. With total air ow of 10 L min 1 , the Reynolds number in the chamber is 40, well within the laminar ow regime, so we expected the merging of sample and sheath air to be smooth with little mixing. Flow visualization tests were performed to examine the sample

injection region, and they conrmed that the ows merged smoothly; there was no indication of turbulence or shear induced mixing. Further evidence that the sample remains in a laminar layer and does not spread in this kind of chamber was provided by timing experiments in Rogers (1994). The OPC (Climet 7350A sensor) attaches to the base of the CFD chamber. All of the air ows through the OPC and then through the crystal impactor (described later). The CFD outlet, OPC, and impactor are coaxial tubes, 11 cm long from the chamber to the crystal impaction surface. This region is insulated but not cooled, so crystals could evaporate and/or melt in transit, although there is little time available ( 12 ms). All airow passes through the crystal collector just below the OPC. Beyond the crystal collector, a bypass in the airow can be selected (Fig. 4). This bypass is used for a short time after the icing procedure, when the chamber is ooded with water, to put ice on the walls. The bypass protects the mass ow meter from damage by any residues of water. Not shown in the airow diagram is a parallel branch with a condensation nucleus (CN) counter. The CN measurements are complementary to the IN data and are valuable for estimating particle losses in the inlet piping, for calculating the fraction of total particles that are IN, and, in eld studies, for identifying boundaries between regions with different atmospheric aerosol structure. c. Hardware components and construction The walls of the CFD chamber are constructed of 3 and 4 in. diameter copper tubing, type DVW. Figure 7 shows the arrangement of components and geometry. The inner cylinder consists of the copper pipe with a

MAY 2001

ROGERS ET AL.

731

ice application procedure, it allows water to ow smoothly and cover the entire surface rather than beading up. The outside of the chamber is wrapped in three layers of Reectix insulation, a laminate of aluminum foil and polyethylene bubble pack. It has a thermal R-value of 15 per layer (thermal conductivity 10 3 erg s 1 cm 1 K 1 or 40% of the value for dry air). It is light in weight and unaffected by moisture. The sample and sheath ows are combined at the top section of the chamber in the delrinPVC section. At this point, the gap between the inner and outer walls is 1.74 cm. The sample air passes through a thin slit (0.25 mm) in an annular blade in the middle of this gap. Sheath air ows along the inside and outside of the blade and enfolds the sample air at its tip. The gap then narrows to 1.12 cm for the remainder of the chamber. d. Signals and data system The data system is based on an IBM-type PC computer (486DX-66) with plug-in cards and an electronic interface for signal conditioning. The data systems for the airborne and lab CFD chambers are nearly identical. Interruptions and spikes in electrical power are common on aircraft, so the power for the data system is ltered through an uninterruptible power supply. The collection of signals is illustrated in Fig. 8. Analog electrical signals for temperature, airow, chamber pressure, and refrigerant pressure are measured with an 8-channel 12-bit analog-to-digital card. Inputs for two of the channels are multiplexed to 32 channels with a vendor-provided expansion board; of these 32 channels, 16 are for low-level (microvolt) thermocouple inputs, and the other 16 are for higher level (millivolt) signals. A precision voltage source (AD580J device, 2.500v) is built into the signals interface box. It provides a stable source for the thermistors and a calibration reference check for all voltage conversions. Its voltage is measured every time the data system starts, as a check on consistency and to look for stability and drift of the A/D converter. A digital input/output card communicates with the refrigeration control electronics. Basic calibrations are performed on the electronics before and after eld campaigns. The calibrations include voltage, temperature, air pressure, airow, and aerosol particle size. Temperatures are measured with 2 thermistors and 12 type-T thermocouples (copper-constantan), of which 4 are on the inside of the inner cylinder and 4 are on the outside of the outer copper cylinder (Fig. 7). During construction, the thermocouple junctions were dipped in thermal epoxy adhesive to provide electrical insulation and to prevent oxidation. This epoxy was also used to attach thermocouples to the outside of the outer cylinder. All temperatures are ultimately referenced to a platinum resistance thermometer (PRT). Before the chamber was assembled, the thermistors and thermocouples were

FIG. 7. The CFD chamber is constructed of two concentric cylinders, shown separated here. Overall height 100 cm. Small copper tubing is for refrigerant. Locations of thermocouples (Tc) and thermistors (Ts) indicated.

structural brass collar soldered at the top and a delrin cone at the bottom. Five sections comprise the outer cylinder. From the top down, they are: PVC pipe, delrin, copper pipe, PVC, and a delrin cone. Joints where the outer cylinder components can be separated are sealed with ethylene-propylene o-rings and vacuum grease. The o-rings remain exible at temperatures below 50 C. The refrigerant evaporator for the inner cylinder is 3/8-in. soft copper tubing, coiled to be a snug t inside the 3-in. copper pipe. Thermocouples are supported by the coil and push against the inner surface of the pipe. The inside of the inner cylinder is lled with heat transfer oil to improve the thermal connection between the refrigerant coil and the cylinder wall. Using only oil, however, would give the cylinder a large heat capacity, making it slow to respond when the temperature changed. The amount of oil was reduced by 50% by lling the inside of the inner cylinder with high-density hollow polyethylene balls (20-mm diameter). The refrigerant evaporator coil for the outer cylinder is also 3/8-in. soft copper tubing. It was wrapped tightly around the 4-in. pipe in a spiral with 5-cm spacing between adjacent turns and attached to the pipe with a wide solder joint to enhance the heat transfer. All surfaces of the copper pipe chamber walls were treated with an ebonizing solution to make them wettable and to avoid oxidation of the copper. The ebonizing solution is caustic and reacts with the copper to form a thin surface layer of black cupric sulde crystals. The layer is very wettable (small contact angle). During the

732

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

FIG. 8. Block diagram of signal processing in CFD data system. A/D analog-to-digital converter. CJC cold junction compensation for thermocouples. MCA multichannel analyzer. OPC optical particle counter.

immersed in a temperature-controlled bath circulator and were calibrated against a PRT. Accuracy was within 0.5 C from 20 to 50 C. Thermocouple temperatures are calculated from a third-order polynomial equation that was derived from the manufacturers (Omega ) tabulated values of voltage and temperature from 40 to 60 C. The cubic t is within 0.03 C of tabulated values. Calibration checks have shown that the overall accuracy of the temperature measurements remains within 0.5 C from 20 to 50 C. After the chamber was assembled, it was no longer possible to collocate and calibrate all the temperature sensors simultaneously. Therefore, two precision thermistors (YSI 44003A) were installed in the CFD chamber to serve as independent measurements and built-in calibrations of the thermocouples. Their published accuracy is 0.4 C from 20 to 40 C. They are used in a resistance bridge with metal lm resistors and the precision voltage source. The thermistor beads were also coated with thermal epoxy. One was mounted near a thermocouple in the middle of the outer cylinder, and the other is in a thermowell immersed in the center of the inner cylinder. The temperatures of particular importance are those of the ice surfaces, since they determine the sampling temperature and supersaturation. There is an intrinsic uncertainty or error in these values because our measurements are made not on the surface of the ice but on the copper walls. When temperatures are being changed, this error can be several degrees. When steady conditions have been established, the error is relatively small ( 1 C). This point is illustrated later in section 5 of this paper. On the multiplexer card (Fig. 8), thermocouple signal

wires (copper and constantan) are attached to screw terminal connectors and to tinned copper signal lead wires. Where two different metals connect, a thermocouple junction is formed; hence, each of these connections is another thermocouple and adds a potential error. To correct for this, the connections are made on an isothermal cold junction reference (6-mm-thick aluminum plate covered with 1 cm of thermal insulation). The plate temperature is measured with an imbedded LM-35 sensor that is calibrated against a PRT standard, and the correction is applied in software. Air pressure of the chamber and aircraft cabin are measured with factory-calibrated Motorola MPX4115 piezo-resistive sensors. Additional calibrations are done at our laboratory using a Wallace and Tiernan aneroid gage, which serves as our standard. Total and sheath airows are measured with mass airow sensors (Honeywell Microbridge type AWM5104vn). These were factory-calibrated with dry nitrogen. We performed additional calibrations with air using a Gilibrator (volume measurement) and local pressure and temperature to convert volume ow to mass ow. Coefcients derived from our calibrations are used in the real-time CFD data system. The MCA is a PC plug-in card (Oxford PCA3) of the type designed for nuclear decay measurements. The OPC produces one analog pulse for each particle, with pulse amplitude increasing approximately with the surface area of the particle. The pulses go to both the MCA and a threshold circuit. The MCA sorts pulses into 256 channels and accumulates a pulse height spectrum every 10 s. The threshold circuit (Fig. 8) is independent of the MCA. It generates a digital logic pulse that is counted by the data system when a pulse exceeds the preset

MAY 2001

ROGERS ET AL.

733

FIG. 9. Illustration of (from top) refrigeration valve control, refrigerant pressures, and temperature response on inner (i) and outer (o) walls as measured by thermistors and thermocouples. Ice nuclei sampling 21102145 and 22052245 UTC. Thermal wave at 2150 UTC is from ooding chamber with water to put ice on walls. Line (E) at top is expansion pressure regulator valve used to separate refrigerant ows to inner and outer cylinders.

threshold value corresponding to a 3- m particle and 10 L min 1 ow. e. Temperature control The temperature of the chamber is controlled by cooling or heating the walls with a compressed gas refrigeration system that is similar to a conventional household refrigerator appliance. Cooling occurs as liquid refrigerant (DuPont Suva HP-81) evaporates in coils of copper tubing on the CFD chamber walls. The temperature of the walls is determined by three refrigerant expansion valves that are controlled by the user through the data system computer. In general, lower tempera-

tures are achieved with lower pressures, that is, smaller valve openings. Refrigerant pressures are measured at the low-pressure ends of the evaporator coils. Saturation vapor pressures determine the boiling points of the refrigerant, and these are calculated and displayed by the data system. An example of the temperature control by the refrigeration system is shown in Fig. 9. Initially, the walls were both 25 C. At 2107 UTC, an expansion valve adjustment (line E, top) was made to set the sampling at 20 C and water supersaturation of 5%. The outer wall began warming, and within about 10 min, the sampling conditions were obtained. Rapid warming of the walls can be achieved by routing hot, high-pressure gas from the compressor to the evaporator coils.

734

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

FIG. 10. Three pulse height calibration spectra for 6.1 m oleic acid particles. Note effect of volume ow rate.

This warming is useful for effecting rapid temperature changes and for melting the ice off the chamber walls. f. Particle detection The Climet OPC was calibrated with monodisperse spheres of glycerin (n 1.4746) and oleic acid (n 1.4582), 1.1- to 22- m diameter, that were produced by a vibrating orice aerosol generator (Berglund and Liu 1973). Polystyrene latex spheres 0.47 to 2.0 m (n 1.588), atomized from aqueous dispersions, were also used for calibration. Voltage pulse height distributions from the Climet were measured with the MCA and were used to set the threshold counting circuit voltage to correspond to 3- m particles. The response of this OPC for monodisperse particles is fairly narrow, with resolution 10% of the mean size. For example, Fig. 10 shows MCA spectra for monodisperse 6.1- m oleic acid droplets at 3 different ow rates. At a ow rate of 6.1 L min 1 , the dominant peak is in channel 68 7 (standard deviation). Particles larger than the primary peak are also evident; these form when multiplets of the primary particles agglomerate. At 20 L min 1 , the OPC range is 2 to 22 m. With one stage of manufacturer-supplied amplication, it covers 0.4 to 2 m. The OPC is over sampled in the sense that the MCA provides 256 channels (or more), although the OPC is not capable of resolving more than 21 channels. Our calibrations showed the response is at from 1.5 to 2 m; that is, a 1.6- m particle produces the same amplitude pulse as a 2.0- m particle. Therefore, this OPC was not suitable for calibrating the collection efciency of the inlet impactor (Fig. 6). For that purpose, we used a different OPC (Climet A6065A) that does not have this at response characteristic. Figure 10 also shows that the OPC has a strong sen-

FIG. 11. Calibration of Climet OPC using spheres of oleic acid or polystyrene latex at three different ow rates. Error bars are standard deviation. Curves show simple t, Channel 5 10/Flow Dia 2 for 6, 12, and 20 L min 1 .

sitivity to ow rate. The other peaks were recorded using the same particle source but with different ow rates through the OPC. The explanation for this is that the light detector in the OPC is a PIN photodiode. As a charge integrating device, it is sensitive to the residence time that particles spend in the region illuminated by the laser beam, that is, the reciprocal of the airow rate. At smaller ow rates, particles spend greater time in the beam, and the detector produces larger voltage pulses. This sensitivity is clearly illustrated in Fig. 10, where the peak channel corresponds inversely with ow rate. A series of calibrations was performed to establish the OPCs response to both particle size and ow rate. The results are shown in Fig. 11. The manufacturers specication is 1.0 CFM (28.3 L min 1 ) for this OPC, but we operate it over a range from 5 to 20 L min 1 , most often at 10 L min 1 . The CFD chamber threshold counting voltage is set for 3- m particles at 10 L min 1 . The crystal impactor in Figs. 2 and 4 is an inertial type device for collecting ice crystals that nucleated and grew in the chamber. It consists of a centimeter-size replaceable cartridge; inside it, the airow is divided among as many as ve converging nozzles. The center nozzle is directed at a post that holds a transmission electron microscope (TEM) grid. The cut-off size and efciency of collection depend on ow rate, air density and viscosity, and particle density; the design goal was 50% cutoff at 3- m diameter. To achieve this over a range of ow rates, the airow through the center nozzle can be adjusted by plugging some of the other nozzles with ball bearings. The ice crystals that are collected contain the ice nucleating particles. As the ice evapo-

MAY 2001

ROGERS ET AL.

735

rates, the IN are left on the TEM grid as residuals. Their size, morphology, and composition can be studied with single particle analysis techniques. Tests of the ability of the CFD technique to separate IN from non-IN are described in Kreidenweis et al. (1998). Results using this technique in a eld study are reported in Chen et al. (1998). The strategy for TEM sampling is to keep the chamber temperature and supersaturation steady, so as not to mix nucleation mechanisms, and to estimate the number of particles collected from the OPC count. The goal is to collect at least several hundred particles on each grid, which typically requires sample times of 30 min or more. g. Software Data logging and display are accomplished with a single MS-DOS program. The software was written primarily in Borland Turbo Pascal , with some components in assembler code and linkable drivers provided by the hardware manufacturers. At start up, the program establishes communications with the electronics and loads the calibration values. Then it enters an endless polling loop with a sequential interrogation of the analog and digital cards; one cycle in the loop takes 1 s. At present, analog measurements are made of 23 parameters, and each measurement is an average of 10 A/D conversions. Calculations based on the wall temperatures and airow rates are made in real time to predict the location, temperature, and supersaturation of the air sample lamina using equations from Rogers (1988). The monitor displays the measured and derived values as well as information about data logging and instrument status or errors. The data are recorded as a series of small binary les ( 30 kbytes). A new le is created every 200 records ( 3.3 min) to limit the amount of data loss in case of power failure. The MCA spectra are recorded in separate les. To retrieve the recorded data, the binary records are unpacked and concatenated, derived parameters are calculated, and correction factors are applied. 4. Operating procedures This section describes the chamber preparation and the procedures for sampling. As the chamber and its descendents were developed, these procedures evolved and were modied as needed for eld or lab operations. a. Coating the chamber wall surfaces with ice Ice is applied to the walls by rst removing the OPC and cooling the chamber to 20 C. During development of the airborne chamber, several different methods were explored to put ice on the walls. In the earlier prototype (Rogers 1988), the inner cylinder was removed so that water could be sprayed and smoothed onto the walls;

this procedure was not suitable for airborne applications. One method we tried was to ush the cold chamber with a continuous stream of air saturated with water at 20 C, but the results were unsatisfactory. It produced thick layers of frost near the entrance region with little ice elsewhere. A satisfactory result was obtained by pumping a premeasured volume ( 3 L) of deionized water into the chamber through the outlet hole in the bottom. This volume of water reaches the top of the copper section of the outer cylinder within 12 s. At this point, the pump is switched off; and the water drains through the pump and back into the tank in 25 s, leaving a smooth layer of ice 0.1 mm thick on the inner and outer walls. The thickness was estimated by thawing the ice and measuring the melt water. This rapid ooding of the chamber with water produces a strong thermal wave, as can be seen in Fig. 9 at 2150 UTC. Temperature excursions on the outer wall are larger because the outer cylinder has less mass than the inner cylinder with its heat transfer oil. After applying the ice, the OPC is reattached to the chamber, and a pressure leak test is performed to check that the chamber is air tight. Sample and sheath air are valved off, and the chamber is pumped down to 180 mb. Then the pump is switched off, and the chamber pressure is monitored for 1 min. Pressure increases of 0 to 1 mb min 1 are considered acceptable. After this test, the chamber is returned to ambient pressure, the pump is switched on, and the airow is adjusted to 10 L min 1 . Next, wall temperatures are set to obtain the desired conditions (Fig. 3), and aerosol sampling begins. Because of the gradients in temperature and vapor pressure, water vapor is continuously subliming from the warm wall and depositing on the cold wall. This process eventually leads to the formation of frost crystals growing on the cold wall and ice-free or rough ice patches on the warm wall. After several hours, frost can extend several hundred micrometers into the chamber. We have found that after 23 h, particles 3 m are detected by the OPC even when sampling ltered, particle-free air. We believe that these particles are ice fragments that break off the walls and fall into the air stream. The explanation of this phenomenon is not obvious, but there are two possible sources of particles: frost fragments breaking off the cold wall and residues of ice ngers (spicules) from the warm wall (Oraltay and Hallett 1989). In either case, the problem is solved by ooding the chamber with water to refresh the ice surface. Without reicing, the particle count rate will continue to rise and can reach a few per second, an unacceptably high background rate. b. Temperature and supersaturation control The temperature and vapor elds of the sample are dominated by the walls. Calculations indicate that the incoming air reaches steady-state proles in 2 s (Mahata et al. 1973; Sinnarwalla and Alofs 1973; Rogers

736

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

1988). For setting the sampling conditions, we use wall temperatures averaged from thermocouples, not including those near the top. c. Filtered air There is no way to perform ice nuclei calibrations, since there are no such things as standard ice nuclei. There is, however, a standard to produce particle-free aira HEPA lter. In the CFD system, air can be valved through a HEPA lter (Fig. 4) to remove essentially all particles from the sample (99.97% 0.3 m). Our practice is to use this lter for brief periods ( 1 min) several times per hour to estimate the background (or blank) count; a zero response is usually obtained after ltered air has ushed through the chamber. Since IN are relatively scarce, the crystal count on sample air is usually a small number (few counts per minute), and the effect of the lter is not dramatic. However, its effect is immediately seen on the concentration of small particles in the MCA spectra. That is, the number of particles 2 m falls to 0. 5. Performance and examples of measurements The new airborne chamber has been used on several aircraft, including the University of Wyomings King Air 200T for engineering tests, the National Aeronautics and Space Administrations (NASAs) Ames DC-8 during the SUCCESS project (Rogers et al. 1998), the NCAR Electra during Lake-ICE/Snowband (Kristovich et al. 2000), and the NCAR C-130 in the combined NSFSHEBA/NASA FIRE Arctic Cloud Experiment (Curry et al. 2000). Measurements from one of the Arctic ights are used to illustrate several aspects of the CFDs performance and the types of data it produces. Laboratory versions of the CFD chamber use two temperature-controlled bath circulators in which the refrigeration compressors run constantly and the electrical heaters control the temperature of the uid. The airborne CFD uses only a refrigeration compressor, and we were interested in seeing how its performance compared with bath circulators in terms of the cooling time, the stability after reaching set points, and the uniformity of wall temperatures. The response of wall temperatures to changes in the refrigerant expansion valves was shown in Fig. 9. Notice that the thermocouple temperatures (bottom panel) mimic the thermistor temperatures and show that gradients in the wall temperatures occur during temperature changes. To the extent that thermocouple temperatures accurately represent the ice surfaces, the supersaturations and sampling temperatures will also vary longitudinally in the chamber. These gradients are more obvious when the same data are plotted with the thermocouples referenced to the thermistors, as in Fig. 12. Longitudinal gradients are evident as deviations from horizontal lines. They occur when temperature adjust-

ments are being made and are of short duration ( 1 min). The temperature stability of this system can be estimated from 2130 to 2145 UTC (3 min after the last valve adjustment) in Fig. 12. During this 15-min period, the outer cylinder thermistor measured 9.0 C ( 0.3); the four outer thermocouples differed from the thermistor by 0.7 to 0.1 C and were stable ( 0.05 C). On the inner cylinder, the thermistor was 24.7 C ( 0.2); thermocouples #0 through #3 differed from the thermistor by 0.9 to 0.2 C. Notice that the topmost thermocouples (#0 and #4) are the warmest places on the walls, probably due to heating from the airow, which is injected at the top of the chamber. At the location of the air sample lamina, the calculated temperature was 19.6 C ( 0.2), and the supersaturation was 5.0% ( 0.5). The chamber was reiced at 2150 UTC, and then the wall temperatures were adjusted for sampling at 10 C and 1% SS w . Travel times from the system inlet through the piping and the CFD chamber depend on how the CFD is installed in different aircraft and on the lengths and ventilation rates of tubing. These times are estimated by using an OPC at the outlet and admitting ambient air or particle-free air at various places in the piping network. Typical times needed to ush the piping are 5 10 s, including transport through the chamber to the OPC. Because the sample air is injected into a centrally located lamina, the CFD method can resolve 1-s changes in aerosol properties (Rogers 1994). Typically, low number concentrations of ice nuclei add an additional limitation on temporal resolution, however. In addition to the deliberate removal of particles 2 m by the inlet impactor, there are diffusion and deposition losses in the piping. We estimate the magnitude of these losses with a CN counter by sampling cabin air with and without the tubing, tees, valves, etc. Typically, 30% of the total CN aerosol is lost, and an empirical correction is applied to the CN data. A more accurate assessment of the losses would involve estimates of particle sizes, pressure, temperature, humidity, and electrical charge. For the ice nucleating particles, however, little information is available on their size, and no corrections are applied. This method of measuring ice nuclei relies on identifying ice crystals (containing ice nuclei) solely on the basis of their optically measured size. Size alone differentiates them from cloud droplets, haze droplets, or dry aerosol. Calculations of growth rates suggest there is ample justication for this approach, as the growth rates are approximately proportional to supersaturation. There are uncertainties that can reduce the size of crystals and weaken the power of this method. For example, the locations where ice crystals nucleate may vary. Since this determines the amount of growth time available, crystal sizes will vary accordingly. To illustrate the range of particle sizes measured by the OPC, we examined MCA spectra during the 20 May

MAY 2001

ROGERS ET AL.

737

FIG. 12. Difference between thermocouple and thermistor temperatures for outer cylinder (top) and inner cylinder (bottom) for same time period as Fig. 9. Uppermost thermocouples are 4 and 0.

1998 eld experiment over the Arctic Ocean that was described earlier. From 2111 to 2133 UTC, each MCA spectrum (10 s) had between 0 and 7 ice crystals. Data from 120 spectra were summed in order to increase the number of crystals and produce an average spectrum. Sampling conditions in the CFD chamber during this time were shown in Fig. 13, and Fig. 14 shows the integrated size spectrum. The NCAR C-130 was level at 4-km above mean sea level (MSL), ying a series of open rectangle patterns with 8-km legs ( 1 min). Aerosol concentrations were fairly uniform during this time, as indicated by CN, PCASP, and FSSP-300 measurements. The air was cloud free, ambient temperature 14 C, and relative humidity 23%. The IN were sampled at 23 to 20 C and 15% to 9% SS w . The size distribution in Fig. 14 shows particles from 1 to 13 m. We interpret the slope change at 23 m as dividing ice crystals from other particles. This largesize range of crystals is a feature typical of MCA spectra for the CFD. There were 158 particles 3 m, and the average concentration was 16 per liter. If a different size was used as a threshold, the integrated count would change. For example, the numbers exceeding 3, 4, and 5 m are 158, 106, and 82, respectively. All methods of measuring IN require an evaluation of and accounting for the background or blank count.

This is the number of crystals that are detected when no IN particles are present. For example, with membrane lters, the background is estimated by processing a lter that has sampled no air or that has sampled only particlefree air; the concentration of IN is calculated by subtracting the background count from the total. For the CFD method, the background is estimated by sampling through the HEPA lter for a few minutes every hour (cf. Fig. 4), as mentioned earlier. After accounting for the background, the IN concentration is plotted in Fig. 13 as 10-s average values. The IN concentrations ranged from 0 to 88 L 1 . The maximum would be smaller if the average were taken over longer intervals, and conversely. Figure 13 also illustrates features that are typical of real-time IN measurements made in the eld. The IN data (second panel from bottom) are shown as both the 1-s crystal count and the 10-s average number concentration (plotted upside down for comparing with the counts). There are typically a great number of samples with zero counts. Of the 1169 displayed, there are 1050 s with 0 counts, 102 with counts of 1, 11 with counts of 2, etc.; the largest count in a second was 5. The counting statistics can be approximated with a Poisson or exponential distribution. The last third of the chamber has no ice on the warm

738

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

FIG. 13. Time history of CFD chamber measurements during eld project. Note 10-s average IN concentration (second panel from bottom) is plotted upside down and referenced to ordinate on right side.

FIG. 14. Size distribution of particles 20 May 1998 21222133 UTC, accumulated from 120-MCA spectra; 256 channels were collapsed to 30 bins 0.4 m wide. Same data shown with lled circles (left scale) and open circles on folded (right scale) for clarity 3 m. Crystal count threshold at 3 m.

wall to force droplets to evaporate, as explained earlier in section 3. Calculations by Rogers (1988) indicated it is unlikely that cloud droplets can exceed 3 m at the chamber outlet, even for SS w up to 5%, but those calculations assumed boundary conditions for the ideal evaporation case (warm wall dry with no heat ux), and they assumed crystals nucleated immediately when the vapor became supersaturated. Calculations were made for the geometry and residence time in the new chamber and incorporating ice crystal growth, as described in section 3. They indicate it is unlikely that droplets can exceed 3 m at the chamber outlet for SS w up to 9%. These calculations also indicated ice crystal diameters at the chamber outlet should show clear size separation from liquid droplets. For example, ice crystals should range in size from 8 to 12 m at 5% SS w and 20 C. In spite of the size calculations, the absence of a discernable particle gap in Fig. 14 suggests there is probably a size overlap between ice crystals and other particles. That is, some droplets do not evaporate to subdetectable residues, and all crystals do not nucleate at the same time and grow to the same size. In controlled laboratory experiments, this size information can be

MAY 2001

ROGERS ET AL.

739

FIG. 15. Size distributions of 0.2 m (dry size) ammonium sulfate particles growing in CFD chamber at 20 C and extremely high water supersaturations, as labeled. The size range 1 to 18 m is covered with 256 channels. For convenience, 0 counts are plotted at 0.1 on logarithmic sale.

FIG. 16. Ten-second average IN concentrations showing rapid increase above water saturation. Sample temperature 23 to 20 C.

useful for studying nucleation mechanisms and growth (Chen et al. 2000), The exact value of particle size that determines the threshold counting of ice crystals can occupy a range of values and cannot be precisely dened. It is critical in the sense of avoiding two kinds of errors: counting droplets that have grown to large sizes and failing to detect ice crystals that have not grown large enough. In order to estimate how large droplets might grow in the chamber and, in particular, if they could reach 3 m, laboratory tests were performed with ammonium sulfate particles; these are active CCN but poor IN at temperatures above 45 C (Chen et al. 2000). For our experiment, the CCN particles were atomized from aqueous solution and then dried, and a monodisperse fraction of 0.2- m diameter was extracted with an electrical differential mobility analyzer. The CFD chamber sampling temperature was 20 C. Over a period of 25 min, the supersaturation was increased from 3% to more than 40% with respect to water, and particle size spectra were measured every 10 s. Examples from this series of data are shown in Fig. 15 for SS w of 16% to 40%. At supersaturations below 25%, a few particles 3 m were seen, and as supersaturation increased, the droplet sizes increased dramatically and consistently. For the supersaturations shown, the number of particles exceeding the 3- m threshold was 0, 15, 104, 540, and 3446. The results from this experiment indicate that droplet size can exceed the crystal threshold of 3 m when SS w 20%. With natural CCN particles instead

of these laboratory aerosols, the SS w where droplets exceed 3 m are unknown, but a value of 20% is a reasonable estimate. Therefore, until other methods are developed for differentiating droplets from ice crystals, it appears that peak SS w 20% is an upper limit for this technique. Another example of CFD data was drawn from the same time period of the 20 May 1998 case to show the dependence of IN activity on supersaturation. Figure 16 shows this dependence for the 10-s average IN concentrations. In general, IN activity is also a strong function of temperature, but since the temperature was approximately constant at 20 C, the increasing activity for this dataset is due to supersaturation. Below water saturation, the contribution by deposition (sorption) nucleation, is small. The rapid increase that occurs above water saturation suggests that the strong contribution is due to the onset of another mode, such as condensationfreezing. The fraction of total aerosol particles active as ice nuclei is the ratio of IN and CN concentrations; for these data, the fraction was 0 to 10 5 , on average, 2 10 6 . 6. Qualifying this techniqueAdvantages and disadvantages The advantages of the CFD method are listed here. The aerosol particles are sampled in normal atmospheric concentrations, not supported on a substrate but freely oating in air. The temperature and supersaturation are controllable and are relatively well dened. To the ex-

740

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

tent that temperature and supersaturation determine the nucleation mode, mechanistic studies of ice formation can be done with this technique. For example, ice crystals would form by deposition nucleation below water saturation; above water saturation, nucleation could occur through both deposition and condensation-freezing. The sample can be exposed to extremely high supersaturations that are well dened up to 20%. Temporal resolution 1 s is possible, depending on IN concentration, and measurements are available in real time. The known disadvantages are described here. They include statistical sampling limitations, poor sensitivity for some nucleation mechanisms, lacking capability for unattended operation, and no sampling of particles 2 m. Because of typically low IN concentrations and the small sample rate ( 1 L min 1 ), the sampling statistics are poor. This was described for the measurements in Fig. 14. Although it is possible to derive temperature and supersaturation response spectra for IN, several tens of minutes are needed to obtain such spectra. All IN measurement techniques have a background that needs to be estimated and subtracted. When the IN concentration and the background are both small, short-term estimates of either will contain large uncertainties. Residence time in the airborne CFD chamber is 5 s. Any nucleation modes requiring several seconds or more will not be detected. In particular, this technique is unsuitable for measuring contact freezing nucleation and immersion-freezing. It is difcult to detect homogeneous freezing nucleation with this technique because of slow crystal growth rates at the colder temperatures. To overcome this limitation, a laboratory version CFD chamber has been built (Table 1). It is twice as long as the airborne chamber to provide more growth time, and its optical particle counter is used at higher gains to measure smaller particles. At present, this instrument is not suitable for unattended operation because the ice must be resurfaced approximately every 2 h. Because nearly all particles 2 m are deliberately removed by impactors at the instruments inlet, this technique provides no information about the IN activity of larger particles. The number of particles 2 m that are not sampled can be estimated from typical atmospheric aerosol size distributions, although there will be large variability in the vertical and between different air mass types, etc. For example, Jaenickes (1993) remote continental air has 0.13 cm 3 particles exceeding 2 m, and of these, 54% (Fig. 6), or 70 L 1 , do not get past the inlet impactors. Only a small fraction of these are active as ice nuclei, depending on the temperature and supersaturation. For example, the maximum fraction in our results for Arctic air (Fig. 16) was 10 5 , hence the unmeasured IN may have been 0.0007 L 1 , as a rough estimate. While this seems insignicant, other studies have shown that IN activity generally increases with particle surface area (e.g., Berezinskiy and Stepanov 1986; DeMott 1990). Thus, the removal of particles 2 m may be a signicant limitation for de-

tecting ice nuclei that are active at the warmest temperatures. A related question is the number of non-IN particles that are mistakenly counted because they exceed 3 m. This number can also be estimated from typical size distributions. Thus, for Jaenickes (1993) remote continental air, there are 61 L 1 larger than 3 m, and 0.21% of these can get past the impactors (Fig. 6), leaving 0.13 L 1 to be detected by the OPC. This is an overestimate because particles will be removed by piping upstream and downstream of the impactors. The chamber samples at one set of temperature and supersaturation (T, SS) conditions at a time, as selected by the operator. Choosing the conditions requires careful thought in order to obtain the most useful information. Consideration should be given to the ice forming processes that one wants investigate or to the type of data being sought. For example, when the aircraft is making large changes in altitude, vertical soundings of IN can be obtained. Although it would be interesting to have (T, SS) spectra throughout soundings, spectra require much more time than is available. Therefore, the (T, SS) conditions are xed, and the vertical structure of IN concentration is measured for one set of conditions, with the intention of reducing the already large number of variables that are important for ice formation. Similarly, when TEM impactor samples are being collected, the (T, SS) should be constant so that later analyses may be able to relate IN activity to physical and chemical properties of the aerosol. 7. Conclusions An ice, thermal gradient continuous-ow diffusion chamber was developed for airborne measurements of ice nucleating particles and has been used on several eld projects. Some improvements for future studies with this type of instrument are suggested here. Increasing the airow rate would improve the sampling statistics and the potential for time-resolved sampling. In order to provide adequate time for particle nucleation and growth, greater ow rate would necessitate a larger chamber and/or more precise measurements of particle size. Another possible solution to this problem is to increase the concentration of aerosol particles by virtual impactor methods (e.g., Sioutas et al. 1994). At present, the wall temperatures are set by the operator; we plan to revise this practice and use preset points and automated feedback control. A ow-through detector that could differentiate water from ice at small sizes ( few micrometers) would be a great advantage. This might be possible with a cross-polarizationtype detector (e.g., Turner and Radke 1973; Sassen 1974). Another device that would extend the potential applications of IN measurements is an inline large particle separator, such as a counter-ow virtual impactor. It would put ice crystals (containing nuclei) into a separate airstream at the outlet of the CFD. This stream could feed into a variety of

MAY 2001

ROGERS ET AL.

741

particle characterization instruments, such as a mass spectrometer, to provide real-time measurements of the size and composition of ice nuclei. Acknowledgments. This material is based upon work supported by the National Science Foundation under Grants ATM93-11606, ATM96-32917, and ATM9714177, and by NASA Grant NAG1 2063 and NASA JPL Contract 961353. We also want to acknowledge the technical contributions of Brian Jesse, Charlie Wilkins, Dave Wood, and Bill Stahm. Assistance and advice for installing this instrumentation in research aircraft was provided by the University of Wyoming, the Medium Altitude Missions Branch at NASA Ames Research Center, and the NCAR Research Aviation Facility. Any opinions, ndings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reect the views of the National Science Foundation.
REFERENCES Baker, B. A., 1991: On the nucleation of ice in highly supersaturated regions of clouds. J. Atmos. Sci., 48, 19041907. Baker, M. B., 1997: Cloud microphysics and climate. Science, 276, 10721078. Berezinskiy, N. A., and G. V. Stepanov, 1986: Dependence of the concentration of natural ice-forming nuclei of different size on the temperature and supersaturation. Izv. Atmos. Oceanic Phys., 22, 722727. Berglund, R. N., and Y. H. Liu, 1973: Generation of monodisperse aerosol standards. Environ. Sci. Technol., 7, 147153. Chen, Y., S. M. Kreidenweis, L. M. McInnes, D. C. Rogers, and P. J. DeMott, 1998: Single particle analyses of ice nucleating aerosols in the upper troposphere and lower stratosphere. Geophys. Res. Lett., 25, 13911394. , P. J. DeMott, S. M. Kreidenweis, D. C. Rogers, and D. E. Sherman, 2000: Ice formation by sulfate and sulfuric acid aerosol particles under upper-tropospheric conditions. J. Atmos. Sci., 57, 37523766. Cooper, W. A., 1974: A possible mechanism for contact nucleation. J. Atmos. Sci., 31, 18321837. , 1991: Research in cloud and precipitation physics: Review of U.S. theoretical and observational studies, 19871990. Rev. Geophys., (Suppl.), 6979. , and G. Vali, 1981: The origin of ice in mountain cap clouds. J. Atmos. Sci., 38, 12441259. Curry, J. A., and Coauthors, 2000: FIRE Arctic Clouds Experiment. Bull. Amer. Meteor. Soc., 81, 529. DeMott, P. J., 1990: An exploratory study of ice nucleation by soot aerosols. J. Appl. Meteor., 29, 10721079. , 1995: Quantitative descriptions of ice formation mechanisms of silver iodide-type aerosols. Atmos. Res., 38, 6399. , A. B. Super, G. Langer, D. C. Rogers, and J. T. McPartland, 1995a: Comparative characterizations of the ice nucleus ability of AgI aerosols by three methods. J. Wea. Modif., 27, 116. , J. L. Stith, and D. C. Rogers, 1995b: Measurements of cloudactive aerosols and ice initiation in North Dakota cumuli. Preprints, Conf. on Cloud Physics, Dallas, TX, Amer. Meteor. Soc., 182187. Fitzgerald, J. W., 1970: Non-steady-state supersaturations in thermal diffusion chambers. J. Atmos. Sci., 27, 7072. Grant, L. O., Ed., 1971: The Second International Workshop on Condensation and Ice Nuclei. Department of Atmospheric Science, Colorado State University, Fort Collins, CO, 149 pp.

Hallett, J., 1983: Progress in cloud physics: 19791982. Rev. Geophys. Space Phys., 21, 965983. Heymseld, A. J., and L. M. Miloshevich, 1993: Homogeneous ice nucleation and supercooled liquid water in orographic wave clouds. J. Atmos. Sci., 50, 23352353. Hobbs, P. V., and A. L. Rangno, 1990: Rapid development of high ice particle concentrations in small polar maritime cumuliform clouds. J. Atmos. Sci., 47, 27102722. Jaenicke, R., 1993: Tropospheric aerosols. AerosolCloudClimate Interactions. P. V. Hobbs, Ed., Academic Press, 131. Kreidenweis, S., Y. Chen, D. Rogers, and P. DeMott, 1998: Isolating and identifying atmospheric ice-nucleating aerosols: A new technique. Atmos. Res., 46, 263278. Kristovich, D. A. R., and Coauthors, 2000: The Lake-Induced Convection Experiment (Lake-ICE) and the Snowband Dynamics Project. Bull. Amer. Meteor. Soc., 81, 519542. Mahata, P. C., D. J. Alofs, and A. M. Sinnarwalla, 1973: Supersaturation development in a vertical-ow thermal diffusion chamber. J. Appl. Meteor., 12, 13791383. Marple, V. A., and B. Y. H. Liu, 1974: Characteristics of laminar jet impactors. Environ. Sci. Technol., 8, 648654. Meyers, M. P., P. J. DeMott, and W. R. Cotton, 1992: New primary ice-nucleation parameterizations in an explicit cloud model. J. Appl. Meteor., 31, 708721. Oraltay, R. G., and J. Hallett, 1989: Evaporation and melting of ice crystals: A laboratory study. Atmos. Res., 24, 169189. Plooster, M. N., 1985: Computer models for simulation of experiments in the Atmospheric Cloud Physics Laboratory (ACPL). Report 5-31701, Denver Research Institute, University of Denver, Denver, CO, 78 pp. Ranz, W. E., and J. B. Wong, 1952: Impaction of dust and smoke particles on surface and body collectors. Ind. Eng. Chem., 44, 13721381. Rasmussen, R. M., 1995: A review of theoretical and observational studies in cloud and precipitation physics: 19911994. Rev. Geophys., (Suppl.), 795809. Rogers, D. C., 1988: Development of a continuous ow thermal gradient diffusion chamber for ice nucleation studies. Atmos. Res., 22, 149181. , 1993: Measurements of natural ice nuclei with a continuous ow diffusion chamber. Atmos. Res., 29, 209228. , 1994: Detecting ice nuclei with a continuous-ow diffusion chamberSome exploratory tests of instrument response. J. Atmos. Oceanic Technol., 11, 10421047. , P. J. DeMott, and L. O. Grant, 1994: Concerning primary ice nuclei concentrations and water supersaturations in the atmosphere. Atmos. Res., 33, 151168. , , W. A. Cooper, and R. M. Rasmussen, 1996: Ice formation in wave cloudsComparison of aircraft observations with measurements of ice nuclei. Proc. Second Int. Conf. on Clouds and Precipitation, Zurich, Switzerland, ICCP and IAMAS, 135137. , , S. M. Kreidenweis, and Y. Chen, 1998: Measurements of ice nucleating aerosols during SUCCESS. Geophys. Res. Lett., 25, 13831388. Sassen, K., 1974: Depolarization of laser light backscattered by articial clouds. J. Appl. Meteor., 13, 923933. Shaw, R. A., 1998: A conceptual model of enhanced ice production in turbulence. Preprint, Conf. on Cloud Physics, Everett, WA, Amer. Meteor. Soc., 6467. Sinnarwalla, A. M., and D. J. Alofs, 1973: A cloud nucleus counter with long available growth time. J. Appl. Meteor., 12, 831835. Sioutas C., P. Koutrakis, and B. A. Olson, 1994: Development and evaluation of a low cutpoint virtual impactor. Aerosol Sci. Technol., 21, 223235. Szyrmer, W., and I. Zawadzki, 1997: Biogenic and anthropogenic sources of ice-forming nuclei: A review. Bull. Amer. Meteor. Soc., 78, 209228. Turner, F. M., and L. F. Radke, 1973: The design and evaluation of an airborne optical ice particle counter. J. Appl. Meteor., 12, 13091318. Vali, G., 1975: Ice Nucleation Workshop. Bull. Amer. Meteor. Soc., 56, 11801184. , 1985: Nucleation terminology. J. Aerosol Sci., 16, 575576.

Potrebbero piacerti anche