Spray Drying: Influence of Developing Drop ... - ACS Publications

May 13, 2000 - Precise experiments were performed in which single drops of coffee or sucrose solutions containing a trace quantity of SF6 were dried...
0 downloads 0 Views 321KB Size
1756

Ind. Eng. Chem. Res. 2000, 39, 1756-1765

Spray Drying: Influence of Developing Drop Morphology on Drying Rates and Retention of Volatile Substances. 1. Single-Drop Experiments John P. Hecht† and C. Judson King* Department of Chemical Engineering, University of California, Berkeley, California 94720

Precise experiments were performed in which single drops of coffee or sucrose solutions containing a trace quantity of SF6 were dried. Simultaneous measurements were made of the drying rate, the loss rate of SF6, the drop temperature, and the physical appearance of the drop. The effects of adding small amounts of carboxymethyl cellulose (CMC) and sodium dodecyl sulfate (SDS) were also explored. Mass-transfer rates of water and SF6 increased during morphological development, or puffing. The enhanced loss of water was caused by liquid-phase convection resulting from stretching of the drop during internal bubble growth and mixing during bubbling. The loss of SF6 during this period was by liquid-phase diffusion into rupturing bubbles. The cooling effect from enhanced water loss during droplet inflation was often sufficient to condense the water vapor in a bubble, thereby leading to inflation-deflation cycles that prevent rupture and remove water selectively. Introduction Spray drying is a common method used to dehydrate foods and beverages. A feed liquid or paste containing dissolved solids is atomized into droplets (10-500 µm) and contacted with hot air.1 The drops dry as they fall through the dryer. The quality of a spray-dried food product often suffers as volatile flavor and aroma components are lost during drying. These losses occur during the atomization process2 and later from individual drops. Once the outside surface of a drop becomes sufficiently dry, the loss of these volatile components diminishes to zero, in accordance with the selectivediffusion theory.3 This theory argues that volatile components are retained, since the diffusion coefficients of these molecules, which are typically larger than water, decrease more rapidly with increasing dryness than does the diffusivity of water. The drier surface of an individual drop gives rise to a reduced drying rate, which results in a rise of the drop temperature. For substances such as coffee extract and starches, drops in a spray dryer often approach and/or reach the boiling temperature, whereby internal vapor bubbles may grow manyfold in size and burst. These bubbles can form and burst repeatedly in a process known as morphological development. Dry particles from industrial spray dryers often show evidence of morphological development; particles are commonly hollow and sometimes contain blowholes, or remnants of burst bubbles.4 Objectives and Methods The purpose of this work was to identify and understand the mass-transfer mechanisms of water and other volatile components occurring within drying drops, with * To whom correspondence should be addressed. † Present address: The Procter & Gamble Company, Corporate Engineering Technologies, 8256 Union Centre Blvd., West Chester, OH 45069.

emphasis on interpreting the causes and effects of different types of morphological development. For these purposes, it is useful to record the physical appearance of and the instantaneous mass-transfer rates from a single drop. The complexity of the spraydrying process makes experimental study of mass transfer from individual drops in a real spray dryer impractical. Therefore, many researchers have studied the drop-drying process by suspending a drop in a fixed position. The drying rates of drops have been measured by several different methods, including suspending a drop from a microbalance and recording the resulting loss of weight,5 measuring the humidity of the gas downstream from the drop,6 using an energy-balance calculation,7 and photographing the reduction of size of the drop with time.8 The loss of trace volatile components from single drops has been studied in a similar manner by adding a volatile component to the drop liquid and analyzing partially or completely dried particles for the remaining quantity.8 All of these methods involve stopping the drying process at different time intervals to make a measurement. The reported rates of loss of water and added volatile components are derived from several or many different drops rather than from one drop during the entire drying process. This approach is not effective for studying the morphological development of drops during drying, since the details of the bubbling and bursting are unique for each drop. A more complete review of previous drop-drying work is given elsewhere.9 To measure the loss of volatile components during morphological development, Verderber and King10 and Sunkel and King11 added sulfur hexafluoride (SF6) to model food solutions and sampled the gas downstream of the drop periodically to analyze it for traces of SF6, using an electron-capture detector. The drying process was not interrupted to make the measurements. They found that significant losses of volatile components resulted from morphological development. The present work is an extension of the work done by Verderber and King10 and Sunkel and King.11 Rather

10.1021/ie9904652 CCC: $19.00 © 2000 American Chemical Society Published on Web 05/13/2000

Ind. Eng. Chem. Res., Vol. 39, No. 6, 2000 1757

Figure 1. Drop-drying apparatus.

Figure 2. Drying column (not to scale).

than periodically sampling the drying gas for SF6 content, the stream was monitored continuously to reveal a more precise loss profile. In addition, the instantaneous drying rate was also measured continuously, using a thermal conductivity detector. At the same time, the drop was dried on a thermocouple junction in some cases, to record the temperature history. A video camera captured changes in the appearance of the drop. These data were taken simultaneously at a high frequency for each drop to elucidate the fundamentals of the drying process. Experimental Apparatus and Procedure A diagram of the experimental apparatus is shown in Figure 1. The feed into the system is prepurified (99.998%, Matheson) nitrogen. Traces of water and oxygen are removed using a carrier gas purifier (Supelco). The feed stream is then split, sending a small fraction to the reference side of the thermal conductivity detector (TCD). The balance of the gas enters the drying column, shown in Figure 2. The gas flows up through the quartz column, where it is heated by a 100 W resistance heater (Omega) and passed over the drop, which dries on the tip of a syringe. The middle portion of the column was made of quartz to facilitate viewing of the drop with a video camera, described below. Below

the viewing section, this piece was formed of concentric chambers, with the annular space isolated and sometimes evacuated for insulation. The gas downstream of the drop then flows through the top piece of the column, which was fabricated from a single piece of Teflon. The top piece has two gas exit streams, which recombine, mix, and then split into three paths. These lead to the thermal conductivity detector (TCD), the electron capture detector (ECD), and a purge. The TCD and the ECD outputs were processed using a Varian Star 3700CX gas chromatograph. Several different types of syringe were used in this work.9 Two of these contained exposed-junction thermocouples (6 in. long, 0.020 in. diameter, Type E, Omega) as plungers. These two syringe types were used in conjunction with the syringe assembly situated atop the drying column. The syringe assembly contained a slide, which was used to form a drop of the desired volume. The first of these syringes was made from a Teflon tube (1/16 in. o.d., 0.031 in. i.d., Valco Instruments). The annular space between the thermocouple and the Teflon tube was filled with 5-minute epoxy (ITW Devcon) to make a seal. The other syringes were made from modified commercial syringes. The syringe barrels were modified by cutting off the tips and grinding the ends into a conical shape, so that the plunger would be in contact with the drying drop. A small O-ring (0.018 in. i.d., 0.042 in. o.d., Silicone, Apple Rubber) was used to provide a seal between the syringe barrel and the thermocouple in the larger of the commercial syringes (50 µL, Hamilton gastight). The original plunger was used with the smaller of the two syringes (10 µL, Hamilton model 710). The thermocouple was connected to a signal box (Omega, Model 199) which gave a dc output of 1 mV/ °F. This signal was fed into a computer (Dell Optiplex GM+ 5133) using an analog-to-digital data acquisition card (National Instruments, AT-MIO-19XE-50) and Labview Software (National Instruments). Data were recorded at five samples per second. The precision of the measurement was within 0.1 °F. Experiments were filmed using a monochrome CCD video camera (Panasonic WV-BP310) fitted with a manual 6:1 zoom iris lens (Toyo 1276-H) and three closeup lenses (+1, +2, and +4; Toyo CL-49). The drops were illuminated from behind using a 150 W floodlight (Sylvania). A white sheet of paper was positioned between the drop and the light source to scatter the light. The signal was recorded at 30 frames per second on S-VHS cassettes (Maxell ST-126) using an S-VHS videocassette recorder (Sony SVO-2100). This VCR was also equipped to write the time code of each frame on the audio track of the videocassette so that individual frames could later be identified (Panasonic Option 422 board SVBK-10 with LTC). The TCD was calibrated by drying a drop of water, measuring the decrease in the image size, and comparing it with the TCD output. The TCD responded to water vapor in nitrogen linearly (0.70 mV/ppm). The ECD response for SF6 was nonlinear.9 Calibrations of the ECD were performed regularly by sending gas mixtures containing 1.0-130 ppb SF6 in nitrogen through the detector. These mixtures were prepared by combining different flow rates of a stream containing 1.5 ppm SF6 in otherwise pure nitrogen with the pure nitrogen feed. The flow rates were set precisely with mass-flow controllers (Brooks, Model 5850). Using this calibration

1758

Ind. Eng. Chem. Res., Vol. 39, No. 6, 2000

Figure 3. Images of 3 µL, 45% sucrose drops drying in nitrogen at 315 °F.

technique, the solubility of SF6 in water was determined by injecting a known volume of SF6-saturated water into the system. The value obtained agreed with the literature value (35 ng/µL at 1.0 atm and 25 °C)12 within 10%.9 Coffee drops were sufficiently opaque so that the images could be digitally analyzed for size. A macro was written in NIH-Image, version 1.6, to direct the VCR to send a single frame of image data to a computer (Macintosh PowerPC 7100/80AV). The program then determined the projected area of the drop image. The area of the syringe by itself was measured separately and subtracted from each area to reveal the image area of the drop alone. The process was repeated automatically for every frame to produce a profile of inflations and deflations. This method provides only semiquantitative information, since only one two-dimensional view of the drop was recorded. Each liquid sample was prepared by weighing appropriate amounts of sucrose (C&H, commercial grade) or coffee (Folgers, Procter & Gamble Co.) and water on a 0.1 mg scale (Mettler). The solution was then divided into 1 mL (approximately) samples, each in its own 4 mL glass vial (Fisher). The vial top consisted of a screwon cap with a Teflon-lined polymeric septum (Fisher, 13 mm). One drop was dried from the contents of each sample vial. SF6 was introduced by filling the headspace of the sample vial with a 10% SF6/90% nitrogen gas mixture (Matheson) by bubbling through the solution. The vial then was shaken overnight in a bath maintained at 25 °C. Drops were introduced into the column by filling a syringe with the desired liquid volume and inserting it in to an opening in the top of the drying column. This opening consisted of a washer to which a round piece of aluminum foil was attached with epoxy resin. The syringe punctured the foil seal and immediately sealed against an O-ring situated directly below the washer. After the syringe was installed, the plunger was depressed and the drop was formed at the syringe tip to dry. Small air bubbles were often present in the drops

as a result of the drop-dispensing process. These served as nuclei for bubble growth. (Air is also commonly entrained into drops during the atomization process in commercial spray drying.) All types of data were taken simultaneously. Several methods were used to align the different measurements on a common time scale. First, an electrical switch was installed to set the thermocouple output to zero degrees and shut off the power to the floodlight simultaneously, thereby connecting the temperature and video measurements. The temperature data and the ECD data were aligned by measuring the residence time of the gas between the drop and the ECD (2.5 s). The difference in residence times between the TCD and the ECD was measured by injecting natural gas into a port upstream of both detectors (shown in Figure 1) and recording the difference in times between the two responses. This difference was about 1 s. More details of the experimental apparatus and procedures are described elsewhere.9 The data shown in this work are examples taken from a larger data set and are representative of the entire body of experiments.9 The trends discussed here were found to be reproducible, although the details of particular drops were different due to natural variability. Experimental Results Morphological Development. The process of morphological development of a sucrose drop is shown in Figure 3. Once the boiling temperature is reached, the drop inflates (see frame at 71 s). The bubble within the inflated drop may burst or shrink due to internal condensation. After either of these events, the drop inflates again, and this process repeats many times. Sucrose drops form single, spherical, and centrally located bubbles. Dried sucrose particles are smooth and solid. The morphological development of a coffee drop is shown in Figure 4. The first inflation occurs after 95 s of drying. The bubble is nonspherical and protrudes

Ind. Eng. Chem. Res., Vol. 39, No. 6, 2000 1759

Figure 4. Images of 5 µL, 45% coffee drops drying in nitrogen at 345 °F.

from the rear of the drop. Upon collapse, the particle assumes a wrinkled appearance. The inflation/deflation cycles persist through the remainder of the drying process. Folds and surface irregularities result, and the final particle is a rough, hollow shell. Similar observations have been reported by Alexander and King,13 Verderber and King,10 Sunkel and King,11 El-Sayed et al.,8 and Wallack et al.14 Sucrose and coffee drops behave differently with respect to bubble geometry and the final particle morphology, stemming from differences in the physical properties of the two substances. The surface tension of sucrose solutions is roughly double that of coffee,9 and the viscosity of concentrated coffee solution is about 3 orders of magnitude higher than that for sucrose,9 owing to the presence of high-molecular-weight components. The bubbles in the sucrose drops remain spherical and centrally located because surface tension serves to minimize the overall interfacial area of the drop. The low viscosity provides little flow resistance to healing flows. Bubbles in drying coffee drops grow toward the rear of the drop, where the drop is wettest and the viscosity is lowest. The gradient in water content results from the gas flow impinging on the drop from below, thereby removing more water from the bottom of the drop than from the top. Dried coffee drops are riddled with craters and folds after drying, since the high viscosity of coffee resists smoothing flows driven by surface tension. The dried drops are also hollow because the high viscosity resists surface tension forces acting to shrink a ruptured shell geometry and also resists implosion due to the vacuum created by the condensation of water vapor in the voids after the particle cools. Mass-Transfer Processes. Figure 5 shows data obtained for a 45 wt % coffee drop drying in nitrogen at 325 °F: the instantaneous rate of loss of SF6, the instantaneous drying rate, the projected area of the drop image, and the drop temperature. Two distinct periods of drying were observed, drying before and after the drop had reached the boiling temperature. The data

Figure 5. Data obtained from drying of a 7 µL, 45 wt % coffee drop in 610 std (25 °C, 1 bar) cm3/min nitrogen at 325 °F.

from this single drop reveal all of the phenomenological results observed in this work. The drying rate initially rose from zero to a maximum and then decreased because of the decreased moisture content, and therefore reduced water activity, of the outer surface and the decrease in outer surface area resulting from shrinkage. After about 1 min of drying, the drying rate began to show peaks, each indicative of inflation. The morphological activity increased the dry-

1760

Ind. Eng. Chem. Res., Vol. 39, No. 6, 2000

Figure 6. Plot of the early mass-transfer rates of SF6 and water from the 45% coffee drop represented in Figure 4. These data are plotted with t-1/2, so as to evaluate the applicability of the penetration theory to the drop-drying problem.

ing rate substantially and continued throughout the drying process. Significant losses of SF6 occur both before and after the onset of morphological development. Before morphological development, the loss occurs as one large, tailing peak. During morphological development, the losses are recorded as many peaks. These will be discussed in more detail below. In the case of a constant diffusion coefficient and in the absence of a water concentration gradient or water flux, the loss of SF6 from the drop before morphological development can be described by the penetration theory,15 given by eq 1.

jSF6 ) FSF6

x

DSF6 πt

(1)

The early SF6 loss data and drying rate are replotted as a function of t-1/2 in Figure 6. Here t-1/2 increases from right to left. If the loss could be described by the penetration theory alone, the upper plot would be a straight line with a slope of FSF6(DSF6/π)-0.5/Adrop, rising from right to left. As can be seen, the slope is not constant and the curve tails off sharply. Furthermore, measurement of the initial slope and calculation of the apparent effective diffusivity of SF6 yield a value of 1.1 × 10-9 m2/s. This is roughly an order of magnitude higher than the diffusivities of water (2.0 × 10-10 m2/s) and acetone (7.1 × 10-11 m2/s) in coffee extract experimentally measured by Thijssen and Rulkens.3 Clearly, the penetration theory fails to predict the loss of SF6 accurately. In reality, SF6 is lost from the drop by liquid-phase convection and diffusion. The convection is caused by the flux of water diffusing toward the outer surface. The diffusive flux of a component in a ternary system also contains a cross-diffusion term, as given by eq 2.

jv1 ) -D11∇F1 - D12∇F2

(2)

Figure 7. Expanded view of data shown in Figure 5.

The diffusivity D11 is the same as DSF6 in eq 1. D12 is called the cross-diffusion coefficient and may have a positive or a negative value.16 The superscript v in eq 1 denotes that the flux is relative to a volume-averaged velocity. The deviation of the SF6-loss data from penetration theory results from the combined effects of convection, cross-diffusion, and diffusion coefficients that are not constant. The bottom plot in Figure 6 reveals that the drying rate reaches a maximum after the maximum in the rate of SF6 loss shown in Figure 5, suggesting that convection may not be the dominant factor. Further evaluation requires numerical computation of the liquidphase concentration profiles. This issue is explored further in Part 2 of this series. The drop image area, shown in Figure 5, decreases steadily before morphological development because of the loss of water. Fluctuations during this period are artifacts of the digitization process. The increases and decreases of the drop size, beginning after about 1 min of drying, are indicative of the inflations and deflations of morphological development. The final particle size is not the minimum size because the particle is hollow. The drop temperature rises steadily from room temperature to the boiling temperature, where it begins to fluctuate because of the inflation cycles. The drop temperature forms a plateau at the boiling temperature and then begins to rise toward the temperature of the drying gas. Mass Transfer during Morphological Development. Figure 7 shows an expanded view of the data shown in Figure 5. Water transport occurs by several different mechanisms during morphological development: by liquid-phase diffusion, by release of water vapor from the bursting of a bubble, by liquid-phase mixing occurring after a bubble has broken, and by the stretching of the liquid during bubble growth. The

Ind. Eng. Chem. Res., Vol. 39, No. 6, 2000 1761

second plot in Figure 7 shows a small rectangle. This rectangle indicates the approximate mass of water present in a typical bubble. Its small size indicates that the water lost from the release of the water vapor inside the bubbles is not a major contributor to the overall rate. The notion of liquid-phase mixing can be addressed by examination of the SF6 releases. Each large release has a similar shape and, unlike the case of the waterloss profile, there is no rise in the baseline between the peaks. This suggests that SF6 is primarily, and perhaps exclusively, lost from the release of the contents of the bubbles themselves during rupture. If significant liquid stirring were resulting from bubbling, then one would expect this to bring SF6-bearing liquid to the surface, resulting in a rise in the SF6-loss baseline between bursts. Further evidence of SF6 loss occurring by bursts is found by the examination of the image-area data. A downward spike in this profile represents a sudden decrease in size of the dropsa burst. Each time this occurs, a corresponding release of SF6 is recorded. The dashed line in Figure 7 shows an example. The stretching of the outer surface of a drop causes a convective flow of liquid toward the surface. As the surface area of the drop increases, more of the drop liquid is closer to the surface than was the case prior to the stretch. The effect of surface stretch on mass transfer has been treated theoretically by Angelo et al.17 They solved the following convective diffusion equation analytically in rectangular coordinates:

∂2Cw ∂Cw ∂Cw + Vy )D 2 ∂t ∂y ∂y

(3)

with the initial and boundary conditions

For all y; t ) 0 Cw ) Cw,∞ Figure 8. Five snapshots of ruptureless inflations. The drop shown initially contained 5 µL of 30 wt % sucrose solution drying in a 594 std cm3/min nitrogen stream at 315 °F. The times given are relative to the first frame shown.

For y f ∞; t > 0 Cw ) Cw,∞ For y ) 0; t > 0 Cw ) Cw,0 Vy is the velocity of the liquid moving toward the surface, induced by stretching. The solution yields the following expression for the liquid-phase mass-transfer coefficient:

xDπ

k)F

S(t)

x∫ S(t) dt t

(4)

2

0

When S(t), the time-dependent surface area, is constant, the expression reduces to the result given by the simple penetration theory, eq 1. The factor involving S(t) accounts for the effect of stretching on the mass-transfer coefficient. If the current value of the surface area exceeds the time-averaged denominator in eq 4, then the surface is stretching, and the value of k is higher than would be predicted by penetration theory alone. The mass-transfer coefficient given by eq 4 cannot be used to predict quantitatively the effect of stretching on the mass transfer of water from the outer drop surface. In reality, this mass-transfer rate must be

calculated numerically because of nonlinear boundary and initial conditions, as well as a nonrectangular coordinate system and a spatially-varying water diffusivity. Instead, this model serves to teach a qualitative understanding of the phenomenon. The importance of the surface-stretch mechanism for the loss of water from the outer surface of a drop is underscored by the discovery of a previously unreported morphological events“ruptureless inflation”. The water loss from the stretching of the drop liquid by internal bubble growth is often sufficient to cool the drop considerably. An internal bubble may cease to grow or begin to shrink, as the water vapor within the bubble condenses. This shrinkage occurs without rupturing the bubble. The smaller drop heats readily as the surfacestretch-enhanced flux of water to the surface diminishes. The rise in drop temperature causes the bubble to grow again. This cycle can repeat many times, so that the drop oscillates in size. Figure 8 shows images from a drop of 30 wt % sucrose solution during ruptureless inflation cycles. It was confirmed that this phenomenon is not a manifestation of fluctuations in drying gas temperature, which were very small. The magnitude of the effect is far larger than could come from that cause,

1762

Ind. Eng. Chem. Res., Vol. 39, No. 6, 2000

Figure 9. SF6 losses from single bubbles and bubbling frequency during drying of a 3.2 µL, 30 wt % sucrose drop drying in 593 std (25 °C, 1 bar) cm3/min nitrogen at 270 °F.

and the frequency of inflation cycling is different and varies over time. An interesting manifestation of the rupturelessinflation mechanism is that water is selectively removed, since SF6 losses occur only by the bursting of bubbles. Figure 5 shows minimal SF6 losses during the first several peaks of water loss. Later in drying, some of the water peaks do not have corresponding SF6 peaks. The magnitudes of the SF6 releases shown in Figure 5 also do not correspond with the magnitudes of water peaks appearing at the same times. A 30 wt % sucrose drop was dried to address the issue of what dominates the magnitude of SF6 lost from a burst. The experimentally measured instantaneous loss of SF6 from this drop is shown in Figure 9. The SF6 losses before morphological development correspond to the first, large peak. This first peak is actually artificially large, since this drop was dried using a Teflon syringe, which absorbed additional SF6 from the sample liquid during loading of the syringe. This additional SF6 desorbed during the drying experiment. This is the only drop described in this paper that has this artifact. The important finding of this experiment comes from the data obtained after morphological development has started. Every peak in Figure 9 represents SF6 lost from the burst of a single bubble, as observed on the videotape. A peculiar pattern of the SF6 loss is that it begins with two high peaks followed by a period of no discernible peaks. Then peaks begin to appear, growing higher and farther apart in time. The bottom plot of Figure 9 shows the bubbling frequency of the drop, as measured by viewing the drying video in slow motion and counting the bursts. The frequency profile shows a maximum frequency shortly after morphological development has started. Toward the end of drying, it was observed that all of the bubbles were roughly the same size and that they formed and burst in succession.

Figure 10. Drying rates and SF6 loss rates from 5 µL coffee drops of different initial concentrations drying at 345 °F.

Figure 9 demonstrates that loss of SF6 into bubbles is a liquid-phase diffusion process. The magnitude of the loss of SF6 from a burst depends on the length of time that a bubble has existed in the drop. As the frequency decreases, each bubble is present in the drop longer, and the bubble contains more SF6 when it ruptures. When the frequency is high, the loss per bubble is low. Under those circumstances the many small peaks combine as a result of axial dispersion in the experimental system to give the rise in the baseline seen between 2.5 and 4 min in Figure 9. The first two peaks are large, since they result from the bursting of bubbles that were present in the drop when it was formed, about 2 min previously. This hypothesis is also supported by the data in Figure 7, which show the largest losses occurring after long periods between peaks. Effects of Initial Concentration and Temperature. Figure 10 shows the rates of loss of SF6 and the drying rates for drops containing different initial concentrations of coffee solutes. In accordance with selective diffusion theory,3 the SF6 losses before morphological development become smaller as the initial coffee concentration increases. The losses during morphological development increased with increasing concentration, so that essentially all of the SF6 originally present in each drop was lost. The results of integrating the SF6 losses from these four drops are shown in Table 1. The solubilities of SF6 in coffee and sucrose solutions were determined in a separate series of experiments described elsewhere.9 The initial drying rate decreased with increasing initial coffee concentration. The drying rate also fell more sharply and earlier for the more concentrated drops. These trends are caused by the reduction of water

Ind. Eng. Chem. Res., Vol. 39, No. 6, 2000 1763 Table 1. Amounts of SF6 Lost from Coffee Drops during Drying Relative to the Total Initial Amount of SF6 Presenta SF6 losses, ng (% of total) init coffee content

before bubbling

during bubbling

total init SF6, ng (from solubility)9

15% 30% 45% 60%

13.1 (81%) 3.7 (32%) 1.6 (17%) 0.83 (10%)

4.2 (26%) 7.7 (66%) 8.3 (89%) 7.7 (91%)

16.2 11.6 9.3 8.5

aThese data were obtained by integrating the SF -loss profiles 6 shown in Figure 10. The total loss for two of the drops above significantly exceeds 100%; this is due to experimental error.

Figure 11. Mass transfer from a 5 µL, 45% coffee drop drying in nitrogen at 315 °F.

Figure 12. Effects of sodium dodecyl sulfate (SDS) and carboxymethyl cellulose (CMC) on drying rates and SF6 loss rates from 45% sucrose drops drying at 315 °F.

activity and water diffusivity with higher coffee concentration. Morphological development also started earlier for the more concentrated drops, since they heated to the boiling point sooner. The data from all four drops show ruptureless inflations. As in Figure 7, these are present early in morphological development and are identifiable by peaks in the water loss with little or no corresponding SF6-loss peaks visible. The first bubble burst causes a large loss of SF6, which occurs when the drop is nearly dry for the 15% and 45% drops. However, once a bubble bursts late in drying, the remainder of SF6 in the drop is lost slowly, as seen from the rise in the baseline. This suggests that the blowhole caused by the burst cannot be sealed from surface-tension forces because of the high viscosity of the drop. The remaining SF6 lost into the bubble may then escape the bubble through this blowhole. The effects of the drying gas temperature on masstransfer rates are shown by comparison of Figure 11 with Figure 10. As expected, the drying rate is higher for the drop dried at 345 °F (Figure 10) than for the drop dried at 315 °F (Figure 11). The prebubbling loss of SF6 is also larger for the drop dried at the higher temperature. Effects of Additives. A surfactant, sodium dodecyl sulfate (SDS, Matheson, Coleman, and Bell), and a thickening agent, carboxymethyl cellulose (CMC, Aldrich), were added in small amounts to sucrose and coffee solutions to study their effects on drying rate and SF6 loss. The effects of SDS and CMC on the drying of sucrose drops are shown in Figure 12. CMC was added in two different concentrations, 0.55 and 1.45 wt %, to increase the viscosity of the samples. Carbon dust was added,

in separate experiments, to sucrose drops containing CMC to determine circulation patterns within drops, caused by the drag of the drying gas. The circulation patterns were dampened slightly for the drops containing 0.55% CMC and almost completely for the drops containing 1.45% CMC. CMC affected the loss of SF6 from drops only by changing the shape of the prebubbling peaks. The amount of SF6 lost during this time, which amounted to about half of the SF6 originally present, was roughly the same for each drop regardless of CMC concentration. This is counterintuitive, since one would expect a drop with increased circulation to lose a higher fraction of SF6. Apparently, the circulation does not have a dominant effect. The centrally located bubble in all of the CMCcontaining drops and the drop without any additives did not burst. The small SF6-loss peaks shown on the 0.55 wt % CMC plot are from tiny bubbles bursting where the drop contacts the syringe. The presence of CMC caused only minor changes to the profiles of the drying rate. As expected, an increase in the CMC concentration led to a slightly lower overall drying rate and longer drying times, presumably due to suppression of convective flow. The effects of 0.55 wt % SDS on the mass-transfer rates of water and SF6 from 45% sucrose drops are also shown in Figure 12. The primary effect of the SDS was that more bubbles burst. Because the drops with 0.55% SDS were less able to sustain large inflations than the drops without SDS, they took longer to dry. Other experiments were conducted with 30% sucrose drops to which 0.34 wt % SDS had been added. These drops burst far less frequently than drops dried without

1764

Ind. Eng. Chem. Res., Vol. 39, No. 6, 2000

SDS. Therefore, SDS can either stabilize or destabilize the liquid shell in a highly inflated drop. These phenomena can be rationalized by considering the fact that the adsorption of a surfactant to an interface lowers the surface tension. Likewise, a higher surface concentration of SDS will lower the surface tension to a greater extent. If a weak spot in the liquid film develops, the rate of stretching of the liquid due to droplet inflation will increase at that spot due to deceased fluid-mechanical resistance to expansion. The local concentration of SDS on the surface will be lower, since the same number of SDS molecules is now spread over a larger area. This SDS surface-concentration gradient gives rise to a surface-tension gradient, causing a lateral flow toward the weak spot. In this way, a surfactant may add stability to a liquid film. However, this mechanism will fail to heal weaknesses if the time required by the lateral flow is longer than the time required for transport of more SDS to the interface. If SDS from the bulk adsorbs too quickly, the surface-tension gradient will disappear, but the weak spot will remain. It will thin further as the bubble continues to expand and cause rupture. This concept can be used to explain why the SDS apparently stabilized the 30% sucrose drops with 0.34% SDS and failed to stabilize the 45% sucrose drops with 0.55% SDS. The 45% drop was more viscous than the 30% drop and contained a higher SDS concentration of SDS, so transport (by diffusion and surface-stretchenhanced convection) of the SDS to the surface may have occurred more quickly than flow driven by the surface-tension gradient. The addition of SDS to 45% sucrose solutions did not stop the internal circulation patterns described above. This was determined by viewing drying drops with added carbon dust. Furthermore, the data in Figure 12 show no evidence of an interfacial resistance, as drops with and without SDS show similar initial drying and SF6-loss rates. Figure 13 shows the effects of CMC and SDS on mass transfer from coffee drops drying at 345 °C. The addition of CMC has little effect on the prebubbling loss of SF6. It is not likely that internal circulation patterns caused by drag from the flowing gas exist in coffee drops, since the viscosity of coffee solutions is much higher. However, the presence of CMC did stifle the action of morphological development, as seen in Figure 14, again at 345 °C. In addition, the CMC appears to have prevented bubbles from bursting until near the end of drying. The effect of 0.55 wt % SDS was to reduce significantly the prebubbling loss of SF6 from coffee. This could be due to the suppression of convective activity from Marangoni effects18 or from the formation of an interfacial resistance.18 However, the addition of 0.055 wt % SDS appears to have had no effect at all. It is possible that a portion of the SDS present adsorbed onto the colloidal solid particles in the coffee solution, thereby lowering the available SDS concentration. Discussion The loss of a trace volatile component (SF6) from suspended drops was found to occur before and after the onset of morphological development. The prebubbling loss could not be predicted using penetration theory alone, probably because of some combination of ternary-diffusion effects, the convective flux of water, and nonconstant values of the diffusion coefficients.

Figure 13. Effects of sodium dodecyl sulfate (SDS) and carboxymethyl cellulose (CMC) on drying rates and SF6 loss rates from 45% coffee drops drying at 345 °F.

Figure 14. Developing morphologies of 5 µL coffee drops without and with 0.55% CMC.

During morphological development, significant losses of SF6 occurred from the release of the contents of ruptured bubbles. For sucrose drops, the losses were observed as sharp peaks due to the rapid closure of the ruptured bubble. For coffee drops, a rupture often led to a peak and then a rise in the baseline, as blowholes in the highly viscous coffee could not seal quickly. The drying rate increased during morphological development because of the stretching of the drop liquid from internal bubble growth and the mixing occurring within drops after the burst of bubbles. It is striking that this liquid-phase mixing apparently increased the drying rate without causing additional loss of the volatile component (SF6). Equally interesting is the finding that sucrose drops with and without CMC lost similar quantities of SF6, despite internal circulation in the drop without CMC. Possible explanations for both observations are contained in Part 2 of this series.

Ind. Eng. Chem. Res., Vol. 39, No. 6, 2000 1765

A new inflation mechanism was discovereds“ruptureless inflation”. The drying rate from the stretching of the drop surface during bubble growth is often sufficient to cool the drop so that the bubble ceases to grow, or shrink. The resulting behavior is a steady or oscillating drop, in which the bubble does not burst, and volatile components contained in the bubble are not released. This mechanism accomplishes the goal of removing water from a drop selectively. Prudent use of surfactants and thickening agents may promote the ruptureless-inflation mechanism. However, surfactants may also encourage ruptures, as seen from Figure 12. A more complete understanding of the drying conditions and/or additives leading to occurrence of ruptureless inflation may prove to be fruitful. Losses of SF6 from drops in this work are nearly total. Substantial retentions of volatile flavor and aroma substances are achievable in commercial spray dryers through appropriate design with varying temperature fields. Drying in the present work, by contrast, was under isothermal conditions. Acknowledgment This research was supported by the National Science Foundation. The authors wish to thank Tom Lawhead and Eric Granlund of the UCB College of Chemistry Glass and Machine Shops, respectively, for the fabrication of the drying column. Nomenclature A ) outer surface area of drop D ) diffusivity Dij ) diffusivity of component i due to a gradient in component j j ) mass flux S(t) ) time-dependent surface area through which mass transfer occurs t ) time Vy ) liquid velocity in the y direction y ) distance coordinate Greek Letters F ) mass density Fi ) mass concentration of component i Subscripts 1 ) component 1: SF6 2 ) component 2: water 0 ) at the drop surface SF6 ) SF6 w ) water

Superscript v ) with respect to the volume-average velocity

Literature Cited (1) Masters, K. Spray Drying Handbook, 4th ed.; Halsted Press: New York, 1985. (2) Kieckbusch, T. G.; King, C. J. Volatiles Loss During Atomization in Spray Drying. AIChE J. 1980, 26 (5), 718. (3) Thijssen, H. A. C.; Rulkens, W. H. Retention of Aromas in Drying Food Liquids. De Ingenieur, JRG 1968, 80, Nr.47. (4) El-Sayed, T. Development of Particle Morphology of Drying Drops. Ph.D. Dissertation, University of California, Berkeley, CA, 1987. (5) Charlesworth, D. H.; Marshall, W. R. Evaporation from Drops Containing Dissolved Solids. AIChE J. 1960, 6 (1), 9. (6) Luyben, K. Ch. A. M.; Olieman, J. J.; Bruin, S. Concentration Dependent Diffusion Coefficients Derived From Experimental Drying Curves. In Drying 80; Mujumdar, A. S., Ed.; McGill University: 1980; pp 243-253. (7) Toei, R.; Furuta, T. Drying of a Droplet in a Non-Supported State. AIChE Symp. Ser. 1982, 78 (218), 111. (8) El-Sayed, T. M.; Wallack, D. A.; King, C. J. Changes in Particle Morphology during Drying of Drops of Carbohydrate Solutions and Food Liquids. 1. Effects of Composition and Drying Conditions. Ind. Eng. Chem. Res. 1990, 29, 2346. (9) Hecht, J. P. Influence of the Development of Drop Morphology on Drying Rates and Loss Rates of Volatile Components during Drying. Ph.D. Dissertation, University of California, Berkeley, CA, 1999. (10) Verderber, P. A.; King, C. J. Measurement of Instantaneous Rates of Loss of Volatile Compounds During Drying of Drops. Drying Technol. 1992, 10, 875. (11) Sunkel, J. M.; King, C. J. Influence of the Development of Particle Morphology upon Rates of Loss of Volatile Solutes during Drying of Drops. Ind. Eng. Chem Res. 1993, 32, 2357. (12) Gerrard, W. Gas Solubility: Widespread Application; Pergamon Press: New York, 1980; Chapter 9, pp 206-216. (13) Alexander, K.; King, C. J. Factors Governing Surface Morphology of Spray-Dried Amorphous Substances. Drying Technol. 1985, 3, 321. (14) Wallack, D. A.; El-Sayed, T. M.; King, C. J. Changes in Particle Morphology during Drying of Drops of Carbohydrate Solutions and Food Liquids. 2. Effects on Drying Rate. Ind. Eng. Chem. Res. 1990, 29, 2346. (15) Sherwood, T. K.; Pigford, R. L.; Wilke, C. R. Mass Transfer; McGraw-Hill: New York, 1975; Chapter 3. (16) Chandrasekaran, S. K.; King, C. J. Multicomponent Diffusion and Vapor-Liquid Equilibria of Dilute Organic Components in Aqueous Sugar Solutions. AIChE J. 1972, 18 (3), 513. (17) Angelo, J. B.; Lightfoot, E. N.; Howard, D. W. Generalization of the Penetration Theory for Surface Stretch: Application to Forming and Oscillating Drops. AIChE J. 1966, 12, 751. (18) Adamson, A. W. Physical Chemistry of Surfaces, 2nd ed.; Interscience: New York, 1967

Received for review June 28, 1999 Revised manuscript received October 12, 1999 Accepted October 15, 1999 IE9904652