Journal of Controlled Release 88 (2003) 7183 www.elsevier.

com / locate / jconrel

Wetting mechanisms of gel-based controlled-release fertilizers

U. Shavit*, M. Reiss, A. Shaviv
Faculty of Agricultural Engineering, Technion, Haifa 32000, Israel Received 12 July 2002; accepted 20 November 2002

Abstract The release mechanism of gel-based controlled release fertilizers (CRFs) involves water penetration into dry mixtures of fertilizers and gel forming polymers. Water penetration provides an upper limit to the whole release process. Where wetting prediction is often based on models that describe the ow of the liquid phase, vapor motion may become signicant when a sharp wetting front exists. In this study we examine the role of vapor and uid ows in the wetting process of CRFs consisting of urea or KNO 3 mixed with polyacrylamide (PAM). Vapor adsorption isotherms were obtained for typical fertilizerPAM mixtures. Wetting and release experiments were conducted by dividing the CRFs into regions alternately lled with a pure fertilizer and mixtures of PAM and fertilizer. The experiments were designed in such a way that when the wetting front reaches a mixtures interface, its motion depends on the gradient imposed by the difference in osmotic potential (OP). The coupled equations of vapor and liquid ow in initially dry conditions were solved numerically to demonstrate the conceptual understanding gained by the experiments. The results show that wetting front motion is affected by transport and adsorption of vapor. It was also shown that the release rate is different when wetting is governed by vapor ow or by liquid ow. The release pattern from a multi-regions device was consistent with the wetting pattern, demonstrating the possibility to tailor the release according to periods of peak demand. 2002 Elsevier Science B.V. All rights reserved.
Keywords: Control released fertilizers (CRF); Wetting; Vapor; Experimental and numerical; Hydrogel

1. Introduction Controlled release fertilizers are used to meet crop needs while signicantly reducing environmental pollution [1,2]. Water-swellable polymers are commonly implemented in controlled release products [3,4]. Brazel and Peppas [5] stress the importance of controlling release pattern and achieving linear release rate (Case II transport). Shaviv [6] and Shoji
*Corresponding author. Tel.: 1972-4-829-3568; fax: 1972-4822-1529. E-mail address: aguri@tx.technion.ac.il (U. Shavit).

and Kano [7] demonstrated the importance of synchronizing the release with plant demand using CRFs with linear or sigmoidal release patterns. Release rate control can be successfully achieved with hydrogel-based controlled release devices in which the wetting and swelling of the hydrophilic polymer [5,8] and / or the dissolution rate of the active ingredient [4] play an important role. Highly water-swellable polymers were also developed for controlled release of agrochemicals and nutrients in agricultural applications [912]. In order to improve release characteristics, a delivery system, based on gel-forming polymers, was examined [13]. The

0168-3659 / 02 / $ see front matter 2002 Elsevier Science B.V. All rights reserved. doi:10.1016 / S0168-3659(02)00455-8

72

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183

fertilizer delivery system consists of a tubular small container with at least one opening lled with a mixture of a highly water-swellable polymer (e.g., partially hydrolyzed polyacrylamide, PAM) and fertilizers (e.g., urea or KNO 3 ). The temporal release rate and pattern were adjusted to meet the pattern of nutrient uptake by an appropriate tuning of the design parameters. Release is inuenced by the formation of the hydrogel matrix (i.e., wetting and conned swelling), the internal dissolution of the fertilizer, and the advectivediffusive transport of the dissolved fertilizer [14]. Among these mechanisms, water penetration is the primary mechanism that affects the release rate. The research objective of the current paper is to study the physical mechanisms of water penetration and to distinguish between the role of the liquid phase and the role of the vapor phase.

1.1. Transport processes inside the CRF device

The fertilizer release is characterized by three

different stages. An initial and short stage in which water penetrates quickly (burst effect) and creates, almost immediately, a small wetted volume. The sharp front that separates the dry zone from the wetted zone was denoted as the wetting front. As the second stage starts, a second front, named the solute front, appears between the wetting front and the opening of the CRF device. Both fronts have sharp and clear boundaries and their penetration rate is related to the release rate. As shown in Fig. 1, the two fronts dene three zones. The transparent zone between the opening and the solute front is the solute zone, which consists of water, dissolved solutes, and a hydrogelswollen polymer. In between the solute and wetting fronts is the middle zone, which includes an undissolved fertilizer, a hydrogelswollen polymer and a saturated solute solution. The zone behind the wetting front is denoted the dry zone. The third releasing period starts when the wetting front reaches the far end of the container. The propagation rate of the fronts varies as a function of mixture composition, density of the

Fig. 1. The experimental controlled-release device and schematic proles of porosity (water or air-lled), water and vapor content, solute concentration, and solid fertilizer.

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183

73

packed dry mixture, and device geometry [14]. During this process, water ows inwards causing a convective ux of solutes. This convective ux opposes the release of solutes driven by the high concentration gradient. A saturated solute solution exists within the middle zone preventing further dissolution of the solid undissolved fertilizer reservoir. Dissolution takes place, therefore, at the solute front alone. The schematic proles in Fig. 1 illustrate up-to-date understanding [13,14] of the assumed spatial variations of porosity (water- or air-lled), water and vapor content, solute concentration, and solid fertilizer. In a preliminary experiment, a pH indicator was added to the dry mixture allowing observation of tiny bluish wet spots in the dry zone. These spots appeared after a few days or weeks from the beginning of the experiment, implying that the pH indicator was activated by some kind of slow wetting process. Based on these observations it was hypothesized that water vapor is generated at the wetting front and diffuses into the dry zone. The vapor then condenses on the dry powder of the polymer fertilizer mixture. The formation of bluish wet spots was due to the reaction of water with the pH indicator under a basic environment induced by the PAM. The number of wet spots and the area covered by them increased with time. It was noticed that more wet spots were accumulated near the wetting front and less near the far end of the device. It was thus hypothesized that propagation of the wetting front is affected by the coupling of two simultaneous processes. The rst is the transport of the liquid phase. The second is a ow of vapor from the wetting front into the dry zone. Vapor is driven by the hydrophilicity of the dry hydrogel and the solubility of the fertilizer. The vapor adsorbed by the dry matrix increases its water content thus facilitating penetration through a gradual increased diffusivity. Grismer [15] showed that under relatively dry conditions of a porous matrix, vapor motion is dominant. However, an experimental distinction between liquid ux and vapor ux is nearly impossible [16]. In the following, we present the mass balance equations that were used to simulate these two uxes. We examine the dry region, downstream from the wetting front. Water exchange between the

liquid phase and gas phase takes place by evaporation and condensation. For simplicity, we assume that the void space near the wetting front is saturated with vapor. With these assumptions, the mass balance equation of the water liquid [17] is, ](uw ( rw 1 c)) 5 2=? (( rw 1 c)qw ) 1 G t (1)

uw is the volumetric water content (volume of liquid / total volume) and qw is the liquid volume ux (volume of liquid / total area per time). The mass of the solution is calculated using both the salt concentration, c (mass of salt / volume of water) and the water density rw . The source term G represents the gain of water by condensation (mass / total volume per time). Assuming that the water solution is saturated (c 5 c s ), Eq. (1) may be simplied,
](uw ) 5 2=? (qw ) 1 G /( rw 1 c s ) t (2)

The water ux, qw , is described by the modied Dracy law, qw 5 2 k w (Fm / uw )=uw , with k w as the solution hydraulic conductivity and Fm as the matric potential. The simulation of a sharp wetting front was obtained by the appropriate diffusivity coefcient [18,19], Dw 5k w (Fm / uw )5Dw0 e b(u 2 u c ) , where Dw0 , b, and uc are parameters. The sink term was dened as G /( rw 1 c s ) 5 arv (uw s 2 uw ), where a is a rate constant, rv is the vapor density (mass of vapor / volume of void space), and uw s is saturation water content. Vapor transport was formulated using the simple mass balance equation as follows, ](n a rv ) 5 2=? (qv ) 2 G, t (3)

where n a is the void space. The vapor mass ux, qv (mass of vapor / total area per time) is dened as qv 5 2 k v=Fo where k v is the vapor conductivity and Fo is the osmotic potential (OP). It is common to simulate the vapor transport using a constant diffusivity coefcient, Dv 5 k v (Fo / rv ), and hence, qv 5 2 Dv=rv [20]. Philip and de Vries [21] have shown that multiple liquidvapor exchanges increase the net vapor ux. This was related to a fast condensationevaporation process, which occurs at the gasliquid interface. Tortuosity becomes less important due to short-cuts generated by the combi-

74

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183

nation of molecular diffusion through the void space and condensationevaporation process that crosses the volume occupied by liquid and wet solid, increasing the rate of vapor ow [22].

2. Experimental and numerical procedures The wetting mechanism was investigated through several experiments including vapor adsorption experiments, viscosity measurements, dissolution tests and wetting and release experiments. The wetting and release experiments were conducted in a similar fashion to experiments that were reported by Shavit et al. [13,14]. In these previous experiments the delivery system was lled with a homogeneous mixture. In the current study the delivery system was divided into regions, which were lled with distinct mixtures. The wetting front moved initially across the rst mixture, then into and through the second and other mixtures. The experiments were designed in such a way that the wetting front encounters several possible potential gradients when it comes across an interface between two mixtures. If vaporow governs the wetting front, it is expected that the front will slow down (or even pause) when the potential gradient is positive and will accelerate when the gradient is negative. It is understood, however, that both liquid and vapor uxes may affect the front propagation and a clear distinction between the two is difcult.

the desiccator [23]. All experiments were conducted at 2161 8C, in four replicates. When no further weight changes were detected, the water content of the equilibrated samples was determined gravimetrically. The osmotic potential was calculated by, Fo 5 2 (RT /V ) Ln( p /p0 ),where p and p0 represent the partial and saturated vapor pressure, respectively. The ratio p /p0 is dened as the water vapor activity (relative humidity) and is linearly related to the vapor density within the void space. Unlike the solid matrix inside the controlled release device, samples of the vapor adsorption experiments are unconned and free to relax potential swelling. This difference is in favor of the approximation that the vapor adsorption experiment reects mainly the osmotic potential rather than the total potential.

2.2. Viscosity and dissolution experiments

Shear viscosity of mixtures containing PAM, urea, and KNO 3 was determined using a Brookeld viscometer. Solutions containing 1 g PAM l 21 dissolved in a 2 mol urea solution and in a 0.1 mol KNO 3 solution were prepared while a shear rate of 0.65 s 21 was applied. The solubility of KNO 3 was determined in saturated urea solutions containing 1, 1.5, or 2% PAM by weight. The KNO 3 was added to a 1-l solution of 1.12 kg urea, at 25 8C. The solubility of KNO 3 was determined by the total weight of added nitrate when precipitation was rst observed.

2.1. Vapor adsorption experiments

The osmotic potential (OP) prevailing in different parts of the device (due to different composition and water content) was evaluated by the vapor adsorption method [23]. It covers a range of tens of atmospheres and is considered appropriate for measuring the OP around the wetting front and the dry zone [24]. Dry samples (20 mg) were taken from a variety of gelfertilizer mixtures and put in a sealed desiccator containing H 2 SO 4 solution in large excess. The relation between H 2 SO 4 concentration and relative humidity in the air above the solution can be found in Fasullo [25]. Each sample adsorbed water until a steady state was reached and its activity was equal to the activity of the internal space inside

2.3. Water penetration rate experiments

Fig. 2 shows the controlled release experimental device. It consists of a glass tube, 70 mm long and 10 mm inner diameter, with one end (device exit) covered by a 1.5-mm porous plate (0.1 mm mesh size) to prevent a spill out of the dry mixture. The opposite end of the tube was used for rell and was kept sealed during the experiments. A second glass tube was attached, perpendicular, to the main tube and was left open to prevent pressure build up due to entrapped air. Dry mixtures of 10% by weight gelforming polymer (sodium polyacrylamide [Na-PAM], Complete Green Company, CA, technical grade) and 90% fertilizer (KNO 3 or urea) were prepared. The mixtures were grinded and sieved thorough a 250-

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183

75

Fig. 2. Scheme of the experimental devices: (a) containing two different mixtures; (b) containing alternating regions of pure fertilizer and fertilizerPAM mixtures.

mm mesh and then weighted and mixed. Then, they were carefully packed, applying a uniform pressure while lling the devices. Fig. 2 shows the regions into which different mixtures (or pure materials) were packed. Four different cases with the following compositions were studied (Table 1). Region A in case 1 (Fig. 2a) was lled with a PAMurea mixture. Region B in case 1 was lled with ne sand. Regions A and B in case 2 were lled with PAMurea and
Table 1 The initial arrangement in cases 14 Case 1 Region size (mm) 020 201 Initial content Urea1PAM Fine sand Case 2 Region size (mm) 020 2040 401 Initial content Urea1PAM KNO 3 Fine sand

KNO 3 (Fig. 2a), respectively. As listed in Table 1, cases 3 and 4 consist of a series of alternate mixtures, PAMurea and urea in case 3 and PAM KNO 3 and KNO 3 in case 4 (Fig. 2b). In all cases, the farthest region was lled with ne sand. Each of the experiments started when the device was submerged in a 3-l container containing distilled water. The penetration rate of the wetting and solute fronts was measured daily using a ruler that was attached to

Case 3 Region size (mm) 015 1525 2535 3545 4555 Initial content Urea1PAM Urea Urea1PAM Urea Urea1PAM

Case 4 Region size (mm) 010 1017 1724 2431 3139 3945 4554 Initial content KNO 3 1PAM KNO 3 KNO 3 1PAM KNO 3 KNO 3 1PAM KNO 3 KNO 3 1PAM

76 Table 2 The simulation parameters Parameter Dw0 Dv b na uc uw s a

a

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183

Units cm day cm 2 day 21 day 21

2 21

Value 0.1 1 6 0.3 0.8 1 1a

a 50 in x5224 cm and x5628 cm.

the device. The maximum error in reading the location of the fronts was 60.5 mm; however, vapor motion causes higher uncertainty when prolonged release experiments were tested. Potassium concentration was measured in case 4 using ICP-AS, allowing calculation of the cumulative release. Longitudinal distribution of water content was obtained by visualization and image analysis during the experiment and gravimetrically at its end (for a single replicate). The study was conducted in a temperature-controlled environment (2161 8C) and the results are based on four replicates.

Fig. 3. Water adsorption isotherms of KNO 3, urea and their mixtures with PAM. The plot represents an average of four replicates. Maximum deviation from mean values was 3%.

2.4. Numerical simulation

The solution of the coupled Eqs. (2) and (3) was obtained for a one-dimensional conguration using FEMLAB (Comsol), a nite element code developed for the MATLAB environment. Table 2 shows the value of parameters used in the simulation. A mesh of 480 elements using a bilinear weight function generated a stable and accurate solution. The numerical solution was accurate when validated against a variety of simple known problems such as a constant diffusivity problem.

3. Results and discussion

3.1. Osmotic potential ( OP)

Four adsorption isotherms are presented in Fig. 3. Each isotherm is an average of four replicates. The maximum deviation from the average was 3%. The isotherms of KNO 3 and PAMKNO 3 mixtures show a sharp decrease in water content within an osmotic

potential range of 5 to 15 MPa. The decrease in water content of the PAMKNO 3 mixture occurs at a higher osmotic potential than that of pure KNO 3 (a mixture of solid salt with its saturated solution). The isotherms of urea and PAMurea mixtures show a decrease in water content within a range of 550 MPa. As in the KNO 3 mixtures, the decrease in water content of the PAMurea mixture below 45% occurs at higher osmotic potential than that of pure urea. Within the range of nearly 080% water content, the osmotic potential of the urea mixtures is signicantly higher than that of the KNO 3 mixtures. Given that the average water content in the middle zone near the wetting front is approximately 30% [24], the osmotic potential of the PAMKNO 3 mixture is estimated around 10 MPa. The osmotic potential that the PAMurea mixture applies under such conditions is about 35 MPa. Although, a complete equilibrium at the wetting front is not necessarily reached, the slow temporal changes allow a local equilibrium assumption. We assume that the deviation from Fo (u ) is small and data from the adsorption isotherms can be used to design the water penetration experiments and to explain the wetting process in the vicinity of the wetting front.

3.2. Water penetration experiments 3.2.1. Case 1 The penetration of water through a PAMurea

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183

77

region and then the transport of the PAMurea solution into ne sand were tested in case 1. The particle size distribution and porosity of the ne sand are similar to the size distribution and porosity of the dry PAMurea mixture. Fig. 4a shows that the wetting rate is high in region A and low in region B. With the assumption of local equilibrium, vapor ow is not expected to cross interface a due to a positive potential gradient. The small penetration rate of the wetting front in region B implies that the capillary driving force is small compared to the forces driving the wetting front in region A. Fig. 4a shows that wetting of region B has started only after the arrival of the solute front to interface a. As long as the middle zone exists in region A, the volumetric water content is about 0.3 [24]. The arrival of the solute front and the termination of the middle zone generate a sharp increase in water content (to about u 5 0.6) and an increase in the driving gradient, which enable wetting of the ne sand.

3.2.2. Case 2 The expected differences in OP near the wetting front were used to design the rest of the experiments. The mixture arrangement was designed such that the driving force due to OP gradient will become small or cease to exist at the interface between the appropriate adjacent mixtures. A positive gradient was obtained in case 2 by using a PAMurea mixture in region A and pure KNO 3 in region B (Fig. 2a). If

transport of vapor by OP gradient is the wetting most signicant mechanism, it is expected that when the wetting front reaches interface a, an OP of 35 MPa in the PAMurea (see Fig. 3) will keep the vapor content in the pure KNO 3 near zero, resulting in a pronounced delay in the wetting front propagation process. It was expected that a possible penetration rate would be similar to that of the penetration through the ne sand as in case 1. After a few days, when the wetting front was only half way through to interface a, wetting spots were observed at the interface in all the replicates. Such an accumulation of water could be explained by a fast ux of water vapor from the wetting front into the dry zone. The ability to visualize this phenomenon was due to the wetting of KNO 3 particles, which were in contact with moistened PAMurea powder (the wetting of KNO 3 crystals is signicantly more obvious than that of PAM or its mixtures). Despite the expectations and the evidence of wet spots at interface a, Fig. 4b shows that the wetting front did not slow down at the interface. Instead, together with the arrival of the solute front, the wetting front accelerated its penetration rate in region B. Since the rate of wetting increased in region B, it was reasonable to look for unexpected changes in this region. While taking apart the devices, it was found that a viscous liquid solution was formed in region B. As no dry PAM was initially packed in this region, it was concluded that a viscous solution containing PAM and urea was moving through

Fig. 4. Wetting and solute fronts in devices of: (a) PAMurea mixture, and ne sand (case 1); and (b) PAMurea and KNO 3 regions (case 2). Plots represent an average of four replicates. Maximum deviation from mean values was 10%.

78

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183

interface a, and into region B. It may be assumed that the more mobile part of the gelviscous solution moved into region B. Following the unexpected fast motion of the wetting front, the effect of the gel viscous solution on solubility and viscosity in region B were examined. The solubility test was conducted to check the ability of the saturated PAMurea solution to dissolve solid KNO 3 in region B. It was found that, at a temperature of 25 8C, 224 g (610 g) of KNO 3 is dissolved in 1 l of saturated urea solution which contained 1, 1.5, or 2% PAM. When the wetting front penetrates into region B, a signicance mass of KNO 3 is dissolved. Such dissolution did not occur in case 1. The increase of porosity due to dissolution and the reduced PAM concentration may explain the fast penetration rate in case 2. Furthermore, shear viscosity measurements have shown that the viscosity of 1% PAM2 M urea solution is 200 times higher than the shear viscosity of 1% PAM 0.1 M KNO 3 solution. As the PAMurea solution penetrates into region B and large quantities of dry KNO 3 are dissolved, a PAMureaKNO 3 solution is formed. The viscosity experiment shows that a concentration of 0.1 M KNO 3 is enough to signicantly reduce the viscosity of the penetrating solution [26]. The remarkable decrease in viscosity further increase the penetration rate of the wetting front. As long as the middle zone exists, solid fertilizer surrounded by saturated solute solution is maintained behind the wetting front. This was indeed the case when the wetting front propagates through region A. However, when the solute front crosses the interface, no solid urea is left. As a result, the urea concentration and urea diffusion ux decrease with time. Fig. 4b shows that the solute front propagation speed was reduced in region B. The solubility of urea is about 3.8 (by volume) times higher than that of KNO 3 at 20 8C [27]. Hence, the gradual decrease of concentration gradient of urea and the low solubility of KNO 3 reduce the rate of solute release and limit the penetration rate of the solute front. It is also possible that the increase in wetting speed has changed the balance between convective and diffusive uxes leading to a further delay in the rate of release and the motion of the solute front.

3.2.3. Cases 3 and 4 The modications of viscosity and solubility in

case 2 were avoided by applying either urea (case 3) or KNO 3 (case 4) and not a combination of the two. Furthermore, the conguration of cases 1 and 2 was expanded to form a series of alternate regions lled with PAMfertilizer and fertilizer mixtures as shown in Fig. 2b. The length of the different regions and their specic initial content are listed in Table 1. Water vapor that ows through the system is expected to condense where the OP is lower. Since regions with PAM are characterized by a lower OP (Fig. 3), growing water content is expected to develop in these regions even when their location is far downstream from the wetting front. A photograph of one of the four replicates (of case 4), 22 days after the beginning of the experiment is shown in Fig. 5. The location of the wetting and solute fronts is at x516.5 mm and x514 mm, respectively. The solute and middle zones are clearly shown to the left of the wetting front. The small distance (|1 mm) between the wetting front and the next PAMKNO 3 region is also visible. Color change in the PAMKNO 3 regions was rst observed about 10 days after the beginning of the experiment. At day 22, the PAM KNO 3 regions shown in Fig. 5 appeared wet and the KNO 3 regions appeared dry. The image grayscale level served as one measure of the water content. The result is marked as a line along the device in Fig. 5. It shows that water content is high inside the PAMKNO 3 regions and low in the KNO 3 regions. The device was dismantled and dissected for gravimetric measurement on the same day. The water content was then obtained from samples taken in each region, weighted, dried in a 90 8C oven for 48 h and than weighted again. The visualization, the grayscale measure, and the gravimetric results are in good agreement (Fig. 5). Water content in the dry KNO 3 regions is less than 10 23 (by weight). Water content in the PAMKNO 3 regions decreases from above 1% to about 0.6%. Fig. 6 shows the wetting front location for case 3 (urea) and case 4 (KNO 3 ). The front propagation rate in case 3 is faster than the propagation rate in case 4. The propagation rate ratio is approximately equal to the solubility ratio between urea and KNO 3 [27]. A delay was observed whenever the wetting front reaches an interface. These delays are longer in case 4. The time of delay in case 3 is nearly un-noticeable at the rst interface and short thereafter. Despite these differences, the general behavior of urea-based

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183

79

Fig. 5. Visualized (upper) and measured moisture content (lower) in a device with alternating regions of KNO 3 PAM mixtures and pure KNO 3 .

mixtures and KNO 3 -based mixtures is identical. In both cases the wetting rate through the PAM regions is higher than in the pure fertilizer regions. Fig. 5 provides an insight into the wetting process. The results indicate that during the process, PAM KNO 3 regions adsorb water vapor and increase their water content. The viscous solution of PAM and fertilizer moves slowly as a liquid bulk across the fertilizer regions while at the same time water vapor diffuses into deeper parts of the device. When the wetting front reaches an already wet PAMfertilizer region, wetting becomes almost instantaneous. This is our explanation to the clear high propagation

speed through the second PAMKNO 3 region in Fig. 6. Tracing the wetting front, at this stage, is a challenge as it becomes progressively ill dened. Wetting by vapor motion is very different from wetting through motion of liquid bulk. The inuence of these two types of water penetration on solute transport and release is highly signicant. When transport of liquid bulk is the only mechanism, saturated solution is moving through the solid matrix of the dry and solute zones. Dissolution of solid fertilizer occurs at the solute front alone, and the solute front becomes the governing release process. A different behavior and thus a different release

80

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183

Fig. 6. The wetting front in case 3 (PAM1urea) and in case 4 (PAM1KNO 3 ). Horizontal lines mark the mixture interfaces. The curves represent one representative replicate in each case.

both water and fertilizer. Fig. 7 demonstrates this phenomenon. The wetting front, the solute front, the accumulated fertilizer release and rate of release are plotted for a single replicate. Following the arrival of the solute front (day 15) and the fast penetration of the wetting front through the already wet region, a sharp increase in the release is observed. This sharp increase supports the conclusion that the wetting mechanism was governed by vapor transport. Another observation which is explained by differences in solubility is the different distance between the wetting and solute fronts. This distance is larger with the PAMKNO 3 mixture as compared with the PAMurea mixture. Fig. 7 shows that the wetting front crosses the KNO 3 regions only when the distance between the wetting and solute fronts is minimized. It also shows that the rate of motion of the solute front does not change much while crossing the interface.

pattern is expected when vapor motion controls the wetting mechanism. As OP drives the process, wetting of internal regions results in a local dissolution of the solid fertilizer, deep inside the device. Such a mechanism might accelerate the release process whenever the wetting front reaches a gel fertilizer region that earlier became saturated with

3.3. Numerical solution

To illustrate our conceptual understanding, we have solved numerically the coupled liquid and vapor ow problem (Eqs. (2) and (3)). For numerical compatibility, the vapor equation (Eq. (3)) was

Fig. 7. The release rate, cumulative release, wetting front and solute front of a representative replicate in case 4 (same replicate as shown in Fig. 6).

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183

81

normalized with respect to water density, rw . Boundary conditions for both water liquid and vapor were saturation at one end (x50) and insulation at the other end. Initial condition was zero content for both phases. The value of a, the rate constant for vapor condensation (Table 2), was 0 in regions containing fertilizer (24 and 68 cm), and 1 in regions containing PAMfertilizer mixture (02, 46 and 810 cm). For simplicity, void space was assumed constant and the sum of water density and solute concentration was approximated as rw 1 c s rw . Hillel [28] shows that diffusivity varies with wetness such that at low water content vapor diffusivity becomes larger then liquid diffusivity. As no specic data is available, the choice of diffusivity values (as shown in Table 2) should be considered as rst approximation only. To avoid the task of solving a moving boundary problem, we did not move the location of the vapor boundary condition together with the wetting front. The resulting error is relatively small as vapor diffusion in the dry regions is high and its density increases in a higher rate than

the rate of propagation of the wetting front. This is demonstrated in the inset of Fig. 8, which shows the vapor longitudinal prole for two extreme cases. In the rst case, regions x5022 cm, x5426 cm and x58210 cm contained gelfertilizer mixture where in the second case this mixture was introduced only in regions x5224 cm, and x5628 cm. The plot shows very little difference between the two cases. The numerical solution of water saturation as a function of time and space is shown in Fig. 8. Water ows through the rst gelfertilizer region as a liquid, forming a sharp wetting front. At the same time, vapor diffuses through the gelfertilizer and fertilizer regions but condenses solely inside the gelfertilizer regions. The negative slope of the water prole in these regions reects the negative slope of the vapor prole. On the other hand the positive slope of the water prole represents a water liquid ux in the opposite (negative) direction lling the dry fertilizer regions with water. Fig. 8 provides an explanation to the difculty in tracing the wetting front towards the end of the experiments. This is

Fig. 8. Simulated distribution of water and vapor (insert) content in a multi-layered device packed with fertilizer and fertilizergel mixtures calculated for 5, 10, 20, 40 and 60 days of water penetration and vapor motion.

82

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183 [2] A. Shaviv, Advances in controlled release fertilizers, Adv. Agron. 71 (2001) 149. [3] N.A. Peppas, R. Gurny, E. Doelker, P. Buri, Modeling of drug diffusion through swellable polymeric systems, J. Membr. Sci. 7 (1980) 241254. [4] P. Colombo, R. Bettini, P. Santi, A. De Ascentiis, N.A. Peppas, Analysis of the swelling and release mechanisms from drug delivery systems with emphasis on drug solubility and water transport, J. Control. Release 39 (1996) 231237. [5] C.S. Brazel, N.A. Peppas, Mechanisms of solute and drug transport in relaxing, swellable, hydrophilic glassy polymers, Polymer 40 (1999) 33833398. [6] A. Shaviv, Plant response and environmental aspects as affected by rate and pattern of nitrogen release from controlled release N fertilizers, in: O. Van Cleemput, G. Hofman, A. Vermoesen (Eds.), Progress in Nitrogen Cycling Studies, Kluwer Academic Publishers, Dordrecht, The Netherlands, 1996, pp. 285291. [7] S. Shoji, H. Kanno, Use of polyolen-coated fertilizers for increasing fertilizer efciency and reducing nitrate leaching and nitrous oxide emissions, Fertil. Res. 39 (1994) 147152. [8] D.S. G Hu, K.J.N. Chou, Kinetics of water swelling and development of porous structure in ionic poly(acrylonitrileacrylamide-acrylic acid) hydrogels, Polymer 37 (6) (1996) 10191025. [9] M.A. Teixeira, J.P. William, J.D. Edward, L. Qiangliang, K.H. Brian, F.A. Mattheus, Assessment of chitosan gels for the controlled release of agrochemicals, Ind. Eng. Chem. Res. 29 (1990) 12051209. [10] J.D. Smith, C.H. Hellen, Evaluation of polymers for controlled-release properties when incorporated with nitrogen fertilizer solutions, Commun. Soil Sci. Plant Anal. 22 (1991) 559573. [11] R.L. Mikkelsen, Using hydrophilic polymers to control nutrient release, Fertil. Res. 38 (1994) 5359. [12] G.K. Chatzoudis, F. Rigas, Macroreticular hydrogel effects on dissolution rate of controlled-release fertilizers, J. Agric. Food Chem. 46 (1998) 28302833. [13] U. Shavit, A. Shaviv, D. Zaslavsky, Solute diffusion coefcient in the internal medium of a new gel based controlledrelease fertilizer, J. Control. Release 37 (1995) 2132. [14] U. Shavit, A. Shaviv, G. Shalit, D. Zaslavsky, Release characteristics of a new controlled release fertilizer, J. Control. Release 43 (1997) 131138. [15] M.E. Grismer, Vapor transport during solution displacement in soils, Soil Sci. 146 (4) (1988) 215220. [16] J.Y. Parlange, Movement of salt and water in relatively dry soils, Soil Sci. 116 (4) (1973) 249255. [17] L. Masaro, M. Ousalem, W.E. Baille, D. Lessard, X.X. Zhu, Self-diffusion studies of water and poly(ethylene glycol) in solutions and gels of selected hydrophilic polymers, Macromolecules 32 (13) (1999) 43754382. [18] J. Crank, in: The Mathematics of Diffusion, 2nd Edition, Clarendon Press, Oxford, 1975. [19] R.W. Zimmerman, G.S. Bodvarsson, E.M. kwicklis, Absorption of water into porous blocks of various shapes and sizes, Water Resourc. Res. 26 (11) (1990) 27972806. [20] R.D. Jackson, Water vapor diffusion in relatively dry soil: I.

shown by the gradual increase of water content inside the fertilizer regions and the simultaneous ux from both sides.

4. Conclusions Visualization, release experiments, wetting measurements, and numerical analysis show that wetting and release of gel based CRFs are governed by the simultaneous transport of liquid and vapor. The separation between the inuence of liquid ow and vapor ow was obtained by dividing the CRF device into fertilizer and PAMfertilizer regions characterized by different chemo-physical properties. The experiments were designed in such a way that when the wetting front reaches a region interface, its motion depends on the gradient imposed by the difference in osmotic potential (OP). As a rule, it was found that wetting of PAMfertilizer regions is signicantly faster than regions containing fertilizer only. This is explained by the higher ability of the PAMfertilizer to adsorb the penetrated vapor. Differences in the wetting process alter the release of CRFs. When wetting is dominated by liquid ow, fertilizer dissolution is limited to the vicinity of solute front. When wetting is accomplished by vapor transport, fertilizer dissolution is being added deep inside the device. This distinction was demonstrated by the fast fertilizer release observed during the fast water penetration through PAMfertilizer regions. Vapor motion increases the water content in deep parts of the CRF, improving, in turn, the liquid diffusivity. A combination of these two mechanisms may result in a variety of wetting patterns, from Fickian to linear, CASE II and super CASE II [29]. Construction of layered devices provides the possibility to tailor periods of alternating release rates. This option may be of a great agronomic and environmental benet in systems with periods of peak fertilizers demand.

References
[1] A. Shaviv, R.L. Mikkelsen, Slow release fertilizers for a safer environment maintaining high agronomic use efciency, a review, Fertil. Res. 35 (1993) 112.

U. Shavit et al. / Journal of Controlled Release 88 (2003) 7183 Theoretical consideration and sorption experiments, Soil Sci. Soc. Am. Proc. 28 (1964) 172176. J.R. Philip, D.A. De Vries, Moisture movement in porous materials under temperature gradients, Trans. Am. Geophys. Union 38 (1957) 222232. W.A. Jury, W.R. Gardner, W.H. Gardner, in: Soil Physics, Vol. V, Wiley, New York, 1991. V.A. Parsegian, R.P. Rand, N.L. Fuller, D.C. Rau, Osmotic stress for the direct measurement of intermolecular forces, Methods Enzymol. 127 (1986) 400416. M. Reiss, Nutrient Controlled Release From Gel Based Devices. MSc Thesis, Faculty of Agricultural Engineering, Technion (in Hebrew; Tables, Figs. and Refs. in English), 1996.

83

[21]

[22] [23]

[24]

[25] O.T. Fasullo, in: Sulfuric Acid, Use and Handling, McGrawHill, New York, 1965, p. 283. [26] G. Muller, J.C. Fenjo, J.P. Laine, Polyacrylamides in the assisted recovery of petroleum, Inf. Chim. 169 (1977) 223 226. [27] R.H. Perry, D. Green, in: Perrys Chemical Engineers Handbook, Vol. VI, McGraw-Hill, New York, 1984, pp. 399. [28] D. Hillel, in: Environmental Soil Physics, Academic Press, New York, 1998. [29] A.H. Windle, Case II sorption, in: J. Comyn (Ed.), Polymer Permeability, Elsevier Applied Science, New York, 1985, Chapter 3.