Professional Documents
Culture Documents
ScienceDirect
Research Paper
article info
Urea, when applied to crops is vulnerable to losses from volatilisation and leaching.
Article history: Controlled release urea enhances nitrogen use efficiency by plants which not only in-
Received 25 May 2014 creases the crop yields but also contributes towards environmental pollution control in
Received in revised form terms of the alleviation of hazardous gaseous emissions and water eutrophication. This
7 October 2014 study proposes a multi-diffusion model that simulates nitrogen release from coated urea
Accepted 9 December 2014 particles for both the constant and decay release stages of urea. It is developed for
Published online 25 December 2014 multilayer included the coating and water zone, and integrates a Finite-element method
(FEM) with 2D geometry to enhance the accuracy of simulation by introducing urea
Keywords: diffusivity in water domain as a function of its concentration. Results from the simulations
Controlled-release fertiliser agreed with preliminary experimental data to a standard error of estimate (SEE) that
Finite-element method ranged from 0.0159 to 0.0567 in which the model successfully simulated and predicted
FEM nitrogen release from hours to days from smaller to larger particles. A more thorough
Multi-diffusion model investigation was then conducted for which the model not only predicted nitrogen release
Nitrogen release from coated urea but also described the internal release mechanism of urea from the core
to urea-coated interface and into aqueous environment.
© 2014 IAgrE. Published by Elsevier Ltd. All rights reserved.
* Corresponding author. Department of Chemical Engineering, Universiti Teknologi PETRONAS, Bandar Seri Iskandar, 31750 Tronoh,
Perak, Malaysia. Tel.: þ60 1 07700361.
** Corresponding author. Tel.: þ60 1 94107181.
E-mail addresses: onalone2000@gmail.com, ththanh@hcmut.edu.vn (T.H. Trinh), kuzilati_kushaari@petronas.com.my (K. Kushaari).
http://dx.doi.org/10.1016/j.biosystemseng.2014.12.004
1537-5110/© 2014 IAgrE. Published by Elsevier Ltd. All rights reserved.
b i o s y s t e m s e n g i n e e r i n g 1 3 0 ( 2 0 1 5 ) 3 4 e4 2 35
Fig. 1 e a) Agrium® coated urea used in release experiment; b) SEM of Agrium® coated urea (magnified 180£).
The development of controlled-release fertilisers (CRF) is a A second approach was proposed in 1980 by Jarrell and
green technology that not only reduces fertiliser loss caused Boersma (1980) who developed a mathematical model of
by volatilisation and leaching but also alters the kinetics of urea release from sulphur coated urea (SCU) particle. Glasser,
nutrient release, which, in turn, provides nutrients to plants at Stajer, Vidensky, Svandova, and Knor (1987) employed a one-
a pace that is more compatible with their metabolic needs. dimensional coordinate diffusion system to predict the
Urea is chosen for many CRF applications because of its high release of nutrient from polymer-coated granules in which a
nitrogen content, low-cost and economical availability (Kent, nonlinear release was produced by assuming that the diffu-
2007). It is believed that the initial world class study of the sion coefficient (D) was time dependent. Lu and Lee (1992)
application of CRF was done by Ortil in 1962 (Shaviv, Raban, & applied Fick's law as a spherical coordinate for the release of
Zaidel, 2003b). CRFs are generally divided into three groups: nitrogen from latex coated urea (LCU). Shavit, Shaviv, Shalit,
reservoir, chemically controlled release, and matrix type (Al- and Zaslavsky (1997) proposed a new type of CRF using a
Zahrani, 1999). In reservoir formulations, the fertiliser core is mixture of soluble fertiliser with a thickener as the coating
encapsulated within an inert carrier and these are regarded as material. He used the simple model introduced by Lee (1985) to
coated fertilisers where nutrient release is controlled by the compare experimental data where ai (equal to 0.5) was
diffusion through the coating. Sulphur-coated urea (SCU) and considered the Fick release factor. Zhang, Nyborg, and
polyethylene-coated urea are examples of this type of Izaurralde (1997) introduced a soil retardation factor (Sb) as a
formulation. Chemically controlled release fertilisers are diffusion resistance parameter to simulate the release of ni-
synthetically produced organic nitrogen, low solubility com- trogen from polymer coated urea (PCU) with which he evalu-
pounds that include condensation products from urea and ated the effect of the coating layer on the release rate. Al-
acetaldehyde such as urea-formaldehyde. Matrix (or mono- Zahrani (1999) modelled the unsteady state of the polymeric
lithic) CRFs hold active ingredients that are dispersed within membrane particle by assuming a well-mixed condition
the matrix where they are initially dissolved and then diffused within the spherical particle. Later, Ni, Liu, Lü, Xie, & Wang
across the matrix continuum to the external surface (Al- (2011) compared experimental results with this model. In
Zahrani, 1999). most of the cited modelling studies, the reigning hypothesis
Nutrient release from coated CRFs is generally controlled was that the release of nutrients from coated fertiliser was
by diffusion through the coated layer (Plimmer et al., 2003; controlled by a single solute diffusion mechanism.
Shaviv et al., 2003b) and numerous recent studies have Based on Raban's experiments, Shaviv et al. (2003b) posited
focused on predicting nutrient release behaviour. The that the release profile from a single polymer coated granule
simplest approach is regression modelling reported and used had a sigmoidal shape consisting of three stages: an initial
by Kochba, Gambash, and Avnimelech (1990), Gandeza, Shoji, stage during which no release is observed (lag period); a stage
and Yamada (1991), Medina, Sartain, and Obreza (2009), of constant or steady state release; and finally, a stage of
Wilson, Rosen, and Moncrief (2009) and Zheng, Liang, Ye, and gradual decay of release. Shaviv, Raban, and Zaidel (2003a)
He (2009). Kochba et al. (1990) used a semi-empirical model also developed a statistical model for a population of gran-
where the release of CRF was considered a first-order process. ules. Lu et al., 2007 proposed a mathematical model for urea
However, they ignored effects from geometry and size, and release from a scoop of PCU taking into account the popula-
also failed to account for the lag period. Gandeza et al. (1991) tion effect. His model was based on the mass balance equation
developed a semi-empirical model to study the effect of soil of a pseudo-steady state according to Fick's law. He neglected
temperature on nutrient release from CNR-polyolefin-coated the first stage of the release process and used a “trial and
urea using a quadric equation: CNR ¼ a þ b(CT) þ c(CT)2. error” method to estimate the diffusion coefficient. Ito,
Wang, Alva, Li, and Zhang (2011) also studied the effect of Golman, and Shinohara (2003) used a 'random walk' method
temperature on the release rate by using a regression model to estimate the effective diffusivity (De) of coating material for
that reduced experimental time from days to hours. However, different starch fractions.
each of these models only related to a specific coating The finite-element method (FEM) has been applied since
material. 2003 to study the release of CRF (Shavit, Reiss, & Shaviv, 2003).
36 b i o s y s t e m s e n g i n e e r i n g 1 3 0 ( 2 0 1 5 ) 3 4 e4 2
Fig. 2 e (a) Dissolution model of a spherical urea particle in water environment; (b) Geometry and mesh generation of a urea
dissolution model for coated urea in Comsol.
However, the geometry used in his model was not a particle layer. This model assumed that the coating layer was satu-
type specific. Trinh, KuShaari, Basit, Azeem, and Shuib (2014) rated with water at the time (t0) of initial release. Water within
used 2D geometry and the FEM model to simulate urea release the core begins dissolving the solid urea where urea concen-
from coated particle but this model only served the second tration is maintained at a saturated level as long as solid urea
stage of release. Therefore, the objective of this study aims to exists in the core. Nitrogen, in the form of urea, begins its
extend our prior study (Trinh, KuShaari, et al. 2014) to the release through the coating layer by diffusion with a constant
decay release rate stage. By applying the multi-diffusion rate. Hence, it is called “constant release” stage. When solid
mechanism in this model, a sigmoidal release profile was urea in the core is totally dissolved, urea concentration de-
obtained and validated by a release test with Agrium® coated creases, and a “decay release” stage begins and then continues
urea. The model was also compared with experimental data to the end of the process.
from literature to confirm its reliability and robustness and Based on mass transport equation for porous medium, the
further showed the advantages of FEM in the study of urea transient stage of urea diffusion through the coating is rep-
release behaviour by a thorough investigation of the release resented by Eq. (1) (Lu et al., 2007).
mechanism within the coated particle. 2
v Cðr; tÞ 2 vCðr; tÞ vCðr; tÞ
De þ ¼ε (1)
vr2 r vr vt
2. Experiments for model verification Since the urea particle is motionless, it is assumed that
urea flux from the coating interface to the liquid environment
2.1. Material is controlled by the diffusion of urea in liquid. Hence, diffusion
is calculated on the basis of mass transport equation of urea in
We used Agrium® coated urea to study its release profile. All water as follows:
the urea particles under investigation were weighed on an
individual basis and coating thickness range from 59 to 72 mm v2 Cðr; tÞ 2Durea vCðr; tÞ vCðr; tÞ
Durea þ ¼ (2)
as determined by TM3030 scanning electron microscope (SEM, vr2 r vr vt
Hitachi) as shown in Fig. 1. Within the water zone, urea diffusivity in liquid, Durea,
varies according to urea concentration displacement as
2.2. Nitrogen release experiment described by Eq. (3) (Cussler, 2009).
The release test was performed in a beaker containing 250 ml Durea ðr; tÞ ¼ 1:380 0:0782Cðr; tÞ þ 0:00464Cðr; tÞ2 105 cm2 s1
distilled water at 27 C. For each interval of 2e5 days, urea (3)
concentration was measured with a U-2010 spectrometer
(Hitachi) at a wavelength of 210 nm. We constructed the ni- Finite element method (FEM) and 2D geometry was
trogen release profile from these results. More explicit employed for our urea release model. Comsol Multiphysics
experimental details are found elsewhere (Trinh, Dao, 4.3a (FNL license) was used to investigate urea release
KuShaari, Ismail, & Shuib, 2014). behaviour. This model is specifically developed for single
coated particle and because this particle is spherical, the
model's geometry was based on one quarter of a sphere. Ge-
3. Model development ometry and meshing steps were done for single urea particle
(Fig. 2b) where the fluid zone diameter, i.e. the distance from
Figure 2a illustrates the dissolution model of a spherical urea the centre of the urea particle centre to the water boundary
particle surrounded by water defined as the fluid zone. Urea (R1), was set at twenty-two times of the core radius (R0) (Trinh,
particles comprise two parts: the urea core and the coating Shaari, Shuib, & Ismail, 2013). A mesh convergence analysis
b i o s y s t e m s e n g i n e e r i n g 1 3 0 ( 2 0 1 5 ) 3 4 e4 2 37
where
CDR(r, t1) is the initial urea concentration of decay release
2 1 stage, mol m3,
Jurea is urea diffusive flux, mol m s
CCR(r, t1) is urea concentration at the end of constant
Adiff is urea particle surface, m2
release stage, mol m3.
Acore is urea core surface, m2.
3.1. “Constant release” stage The urea concentration profile is determined by CDR ðr; tÞ
within the particle. Urea concentration at the urea core at any
As mentioned above, we assumed that water penetration had time t was calculated by Eq. (13):
finished before urea release occurred; an interval called the
mcore ðtÞ
lag time (t0). The lag time was not a subject in this study and CDR ðR0 ; tÞ ¼ ðt t1 Þ (13)
Vcore
was obtained from other related experiments. From time t0
therefore, if the urea solution at the surface of the urea core is As mentioned by Shaviv (2005), a good quality CRF releases
saturated and the concentration at the outer layer of the fluid more than 90% of its urea after the stated time. Additionally,
field is zero, then the boundary conditions are as follows the release rate slows considerably when mrelease(t) reaches
(Trinh, KuShaari, et al. 2014): mtotal urea. However, as this increases calculation time and re-
sources and is unnecessary, we composed the end condition
CCR ðR0 ; tÞ ¼ Csaturated ðt0 t t1 Þ (7) for this stage as.
mrelease ðt2 Þ
CCR ðR1 ; tÞ ¼ 0 ðt0 t t1 Þ (8) percent release ¼ 100% ¼ 95% (14)
mtotal urea
and the initial condition is:
3.3. Model verification and comparison with other
CCR ðr; t0 Þ ¼ 0 ðR0 r R1 Þ (9)
literature
The amount of urea released (mrelease) and the urea mass
within the core (mcore) were calculated using Eqs. (5) and (6). In Prior to validation process, a mesh convergence analysis was
these simulations, urea release time depended on the amount performed to estimate the accuracy of simulation. Percent
of urea (size and shape) and the coating material's properties released of nitrogen at transition time between constant
(thickness, diffusivity, porosity). This stage ended when the release and decay release stage was used to determine the
38 b i o s y s t e m s e n g i n e e r i n g 1 3 0 ( 2 0 1 5 ) 3 4 e4 2
estimation accuracy of simulation, and chosed an appropriate 4.1. Model verification through the release of nitrogen
mesh size. Secondly, model verification was conducted with from Agrium® coated urea
our experimental data from the release test of Agrium® coated
urea. The model was set up for a single granule based on Mesh convergence analysis was performed to estimate the
specific parameters as shown in Table 1. accuracy of simulation as in Table 3. Percent released
Standard error of the estimate (SEE), was used to determine converged to the second decimal place for mesh size factors
the accuracy between simulation and experiment. from 0.7 to 1.1 which is sufficient to the accuracy that can be
obtained from experiments. Error between mesh size factor of
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P 2 0.7 and 1.0 was 0.00125%, hence the factor of 1.0 was chosen
y yest
SEE ¼ (15) for the model.
n
The simulation result is plotted in Fig. 3 and corresponded
where well with our experimental data. The standard error of esti-
mate (SEE) was 0.0300, which is acceptable for the prediction
y is the experimental data of nitrogen release. The minor discrepancy between the
yest is the simulation results experimental and simulation results may be attributed to
n is number of given data differences in particle shape, the particles population effect,
and asymmetry of coating thickness. Predicted data is also
Thirdly, in order to further validate the model reliability presented as a straight line in the release curve from the
and robustness, a comparison was conducted with other beginning of the release and shows a decrease in release rate
experimental data obtained from literature. Simulation trials during the second stage. Both predicted and experimental
were done for three types of coating material (included small data presented similar sigmoidal release trends in agreement
and large particles) of Cellulose Acetate Propionate (CAP); with the cited literature (Ito et al., 2003; Lu et al., 2007; Lu &
Modified Polyolefin (MPO); and Latex Film in which release Lee, 1992; Shaviv et al., 2003b).
time ranged from hours to days. Geometry and effective
diffusivity parameters used in these trials are summarised in
Table 2. 4.2. Comparison with experimental data from other
studies
Table 4 e Comparison of simulation data with results from cited literature (Lu et al., 2007; Lu & Lee, 1992; Shaviv et al.,
2003b).
Name t0 t1 t2 SEE R0 (mm) l (mm) Deff 1014 (m2s1)
CAP150R1 0h 7.75 h 20 h 0.0567 0.8925 0.0711 70.0
CAP300R1 0h 18.47 h 90 h 0.0377 0.8925 0.2489
MPO 1 0.5 d 15.89 d 40 d 0.0179 1.2 0.038 1.2
MPO 2 4.0 d 38.72 d >60 d 0.0169 1.2 0.092
MPO 3 7.0 d 66.84 d >60d 0.0159 1.2 0.183
Latex 22 d 46.30 d 92 d 0.0393 7.06 0.353 39.5
t0 e lag time, t1 e constant release time, t2 e time when 90% nitrogen has already been released, hr e hour, d e day.
Fig. 7 e Urea mass profile in particle and core: a) constant release; b) decay release stage. There is no release during phase I.
The release occurs and reaches balance during phase II. Release rate keeps constant in phase III. And phase IV is the
decrease in release rate.
nitrogen release from polymer coated urea particle over a also presents end condition and transition between constant
wide range of sizes and release intervals. and decay release rate stages. Of interest here is the delay in
the release profile related to the mass profile within the par-
4.3. Coated urea release behaviour ticle. In phase I, the urea core quickly dissolved while the urea
mass remained constant, meaning there was no release
Figure 7a depicts the mass profile of urea within the core as observed during this period. This phenomenon can be
well as a coated particle during the stage of constant release. It explained by the movement and saturation of the urea front
Fig. 8 e Urea front movement in coating layer are presented by urea concentration in granule and water environment in 2D
plots from 100 to 15000s during constant release stage. Colour legend represents urea concentration in range of
0e0.2 mol m¡3. The movement of urea front (red) from 100 to 9000s causes a delay in the release. At time 13000s, the release
occurs and is represented by the change in environment's concentration.
b i o s y s t e m s e n g i n e e r i n g 1 3 0 ( 2 0 1 5 ) 3 4 e4 2 41
Fig. 9 e Urea concentration profile in coating layer during constant release stage: a) phase I (nitrogen movement through
coating layer and there no release), b) phase II (the release occurs and increases its rate), c) phase III (nitrogen releases at
constant rate); and decay release stage: phase IV (d, e, f). Urea concentration inside the core reduces by time (from d to f).
White arrows present urea diffusive flux, which related to nitrogen release rate. It becomes smaller from d to f. Colour
legend represents urea concentration in range of 0e1.1 £ 104 mol m¡3.
within the coating layer before reaching the particle's outer urea core and the water zone as seen in Fig. 7b. Hence the
shell as seen in Fig. 8. From time t ¼ 100e13000s, we observe diffusion flux (white arrows) decreased from day 24 to 57
the movement of the urea front from the particle's core to its during Phase IV as seen in Fig. 9. This change in diffusion flux
surface. Figure 9a also shows no release of urea into the water is the main reason for the reduction of the nitrogen release
zone. Many previous researchers did not account for this rate.
phenomenon which is an oversight that may have increased
discrepancies between modelling and experiment results.
During phase II when the urea front reached the interface 5. Conclusion
between the particle and the environment, urea began to
release to the environment and the rate of release increased The FEM and 2D geometry model as supported by Comsol
until it reached a constant rate. This transient stage happened Multiphysics 4.3 was modelled and then validated with an
very rapidly compared to the entire release period and is Agrium® urea release test. The model was able to reliably
difficult to observe experimentally. simulate the nitrogen release from polymer coated particles
During phase III, urea release gained a steady state and by the investigation of various coated urea types as per cited
maintained a constant rate. Figure 8 shows an increase of literate. This model suitable predicted nitrogen release for
accumulated urea to the water zone that is experimentally most polymer coated urea particles introduced to the market
measureable. However, this steady state is not maintained which covers a wide range of particles sizes and release pe-
indefinitely and drops to an uneven pace when the solid urea riods. In addition, the model demonstrated nitrogen release
core is completely dissolved, at which time the decay release behaviour(s) and mechanism that led to discrepancies be-
stage begins. tween prior simulations and experimental results. However,
Fig. 9d, e, f depict Phase IV of urea concentration profile challenges remain for the model's further development with
during the decay release rate at days 24, 38, 57, respectively. respect to the effect of urea concentration on effective diffu-
Urea concentration within the core constantly decreased as sivity as well as effects from particle shape and population
urea concentration in the water zone increased. This led to a size, etc. Nonetheless, the present model confidently pre-
decrease in the gradient of urea concentration between the dicted the nitrogen release behaviour of coated fertiliser.
42 b i o s y s t e m s e n g i n e e r i n g 1 3 0 ( 2 0 1 5 ) 3 4 e4 2