Journal of Controlled Release 73 (2001) 59–74

www.elsevier.com / locate / jconrel

Fabrication of PLG microspheres with precisely controlled and

monodisperse size distributions
a b a,
Cory Berkland , Kyekyoon (Kevin) Kim , Daniel W. Pack *
a
Department of Chemical Engineering, University of Illinois, Urbana, IL 61801, USA
b
Department of Electrical and Computer Engineering, University of Illinois, Urbana, IL 61801, USA

Received 26 December 2000; accepted 28 February 2001

Abstract

The size distribution of biodegradable polymer microspheres critically impacts the allowable routes of administration,
biodistribution, and release rate of encapsulated compounds. We have developed a method for producing microspheres of
precisely controlled and / or monodisperse size distributions. Our apparatus comprises spraying a polymer-containing solution
through a nozzle with (i) acoustic excitation to produce uniform droplets, and (ii) an annular, non-solvent carrier stream
allowing further control of the droplet size. We used this apparatus to fabricate poly( D,L-lactide-co-glycolide) (PLG) spheres.
The acoustic excitation method, by itself, produced uniform microspheres as small as 30 mm in diameter in which $95% of
the spheres were within 1.0–1.5 mm of the average. The carrier stream method alone allowed production of spheres as small
as |1–2 mm in diameter from a 100-mm diameter nozzle, but generated broader size distributions. By combining the two
devices, we fabricated very uniform spheres with average diameters from |5 to .500 mm. Furthermore, by discretely or
continuously varying the experimental parameters, we fabricated microsphere populations with predefined size distributions.
Finally, we demonstrate encapsulation and in vitro release of a model drug compound, rhodamine B. In summary, our
apparatus provides unprecedented control of microsphere size and may allow development of advanced controlled-release
delivery systems.  2001 Published by Elsevier Science B.V.

Keywords: Controlled release; Uniform microspheres; Poly(lactide-co-glycolide); Drug delivery

1. Introduction [1]. Unlike traditional small-molecule drugs, how-

Rapid advances in biotechnology have led to the
discovery of numerous protein and peptide thera-
peutics, many of which have recently reached the
marketplace or are currently under regulatory review
by the United States Food and Drug Administration
*Corresponding author. Tel.: 11-217-244-2816; fax: 11-217-
333-5052.
E-mail address: dpack@uiuc.edu (D.W. Pack).
ever, proteins and peptides generally cannot be
administered orally; injection or infusion is most
often required. Further, because of their fragility and
short in vivo half-lives, encapsulation of proteins in
biodegradable polymeric devices, from which the
drug can be delivered, locally or systemically, for a
prolonged period of time, has been a promising and
intensely studied solution to these problems [2–4].
Biodegradable microspheres comprising a variety of
polymers have been the most studied devices due to
relatively simple fabrication and facile administration

0168-3659 / 01 / $ – see front matter  2001 Published by Elsevier Science B.V.

PII: S0168-3659( 01 )00289-9
60 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74
to a variety of locations in vivo through a syringe
needle.
Several methodologies for microsphere fabrication
have been described in the literature including
precipitation [5], spraying [6,7], phase separation,
and / or emulsion techniques [8–10]. The emulsion
and spraying approaches have been commonly used
both at the bench and industrial scales [11,12].
Sphere size and size distribution are reproducible but
often poorly controllable. Standard deviations equal
to 25–50% of the mean diameter are not uncommon.
Control of sphere size and size distribution has
several important implications for controlled-release
drug delivery. For example, there typically is an
ideal sphere size that provides a desired release rate
and route of administration. Spheres that are ‘too
small’ exhibit poor encapsulation efficiency, may
migrate from the site of injection, and may exhibit
undesirably rapid release of their payload. Spheres
that are ‘too large’ may not easily pass through a
syringe needle. Thus, the typically polydisperse
spheres generated by conventional fabrication tech-
niques must be filtered or sieved to isolate particles
within the desired size range, and the polymer and
drug composing spheres outside that range are
wasted. Further, precise size control may lead to
advanced delivery systems not possible with polydis-
perse spheres. Uniform microspheres approximately
1–5 mm in diameter would be ideal for passive
targeting of professional antigen-presenting cells
(APCs) such as macrophages and dendritic cells
[13,14]. Similarly, microspheres 10–20 mm in diam-
eter could be used to target the tortuous capillary bed
of tumor tissues by chemo-embolization [15]. A
system capable of precise microsphere fabrication,
therefore, is needed and could allow the optimal size
for such applications to be identified and provide an
efficient route to commercial manufacture and clini-
cal implementation.
A long-sought goal for controlled-release drug
delivery technologies is the ability to precisely
control the release rate of encapsulated compounds,
and microsphere size is a major determinant of
release kinetics. Larger spheres generally release
encapsulated compounds more slowly and over
longer time periods, other properties (polymer mo-
lecular weight, initial porosity, drug distribution
within the sphere, etc.) being equal [16]. A constant
(i.e. zero-order) release rate is often preferred
[17,18], while variable drug release rates can be
beneficial for many important indications [19]. For
example, intermittent high doses of antibiotics may
alleviate evolution of resistance in bacteria, and
discontinuous administration of vaccines often en-
hances the immune response [20,21].
Methods to control drug release rate include (i)
choice of polymer chemistry (anhydrides, esters, etc.)
and comonomer ratios, (ii) conjugating the drug to
the polymer [22], (iii) varying the microsphere
formulation parameters, and thus the physical
characteristics of the resulting particles [23–27], and
(iv) manipulating the sphere size and distribution
[16,28,29]. The success of the latter studies was
limited by the relatively broad microsphere size
distributions.
In recent years, there have been several reports of
the fabrication of biodegradable polymer micro-
spheres with controlled, uniform size [16,23,29–31].
However, none of these methods was successful in
generating particles in a size range appropriate for
drug delivery (|1–100 mm) while maintaining nar-
row size distributions. In addition, these previous
methods appear to be difficult to scale-up for com-
mercial applications.
We have developed a methodology, based on
spraying a polymer solution through a small orifice,
for fabricating polymeric microspheres. Here we
report the capabilities of two techniques, individually
and in combination, for generating monodisperse
microspheres with precisely controlled sizes from |1
to .500 mm in diameter. We have further demon-
strated the capability of the apparatus to fabricate
predefined particle size distributions via continuous
variation of the process parameters.
2. Materials and methods
2.1. Materials
Poly( D,L-lactide-co-glycolide) polymers (50:50
lactic acid / glycolic acid; [i.v.]50.15–0.63 dl / g cor-
responding to Mw 6,000–52,000) were obtained from
Birmingham Polymers. Poly(vinyl alcohol) (PVA;
88% hydrolyzed) was obtained from Polysciences,
Inc. Rhodamine B chloride was obtained from
 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74
Sigma. HPLC grade dichloromethane (DCM) and
dimethylsulfoxide were purchased from Fisher Sci-
entific.
2.2. Preparation of microspheres
Five- to 20-ml quantities of 5% w / v PLG dis-
solved in DCM were prepared. In some cases,
rhodamine B (1, 3 or 5% w / w rhodamine / PLG) was
co-dissolved in the PLG solution for encapsulation
and release studies. The PLG solution was pumped
through a small-gauge needle or circular orifice (30–
1001 mm opening) at flow rates from 0.001 to 5.0
ml / min while an ultrasonic transducer (Valpey
Fisher) controlled by a frequency generator (Hewlett-
Packard model 3325A) disrupted the stream into
uniform droplets. In some cases, a carrier stream
(1% w / v PVA in distilled water) flowed around the
emerging PLG stream. The streams flowed into a
beaker containing approximately 500 ml of 1% PVA.
Microspheres were analyzed in free fall and during
the hardening process using a video camera and a
VCR (Panasonic), and still images were generated
using a video printer (Panasonic). The spheres were
left overnight to harden by extraction and evapora-
tion of the DCM.
Three times the hardened microspheres were
centrifuged at 500–2,000 rpm (depending on size)
for 5 min and resuspended in purified water. The
microspheres were lyophilized (Labconco benchtop
model) for a minimum of 48 h. Finally, microspheres
were stored at 2208C in the presence of a desiccant.
2.3. Rhodamine B release study
The initial loading of rhodamine B was deter-
mined as follows. A known mass (|2 mg) of
microspheres was dissolved in 50 ml dimethylsulfox-
ide. PBS (500 ml) was added and precipitated
polymer was removed by centrifugation at 12 000
rpm for 10 min. Rhodamine B concentration in the
supernatant was determined by measuring the ab-
sorbance at 550 nm in a multi-well plate spec-
trophotometer (Molecular Devices Spectra Max
340PC).
To measure release kinetics, a known mass of
microspheres encapsulating rhodamine B was re-
suspended in 2 ml of phosphate-buffered saline
61
(PBS, pH 7.4). The suspended microspheres were
continuously agitated by inversion (at |12 rpm) in a
378C incubator. At regular intervals over 10 days, the
samples were centrifuged, the supernatant was re-
moved, and the spheres were resuspended in fresh
PBS. Concentration of rhodamine B in the superna-
tant was determined using the spectrophotometer as
described above.
2.4. Light and electron microscopy
Hardened microspheres were typically visualized
by light microscopy. Drops of microsphere suspen-
sion were placed on glass coverslips and the spheres
were allowed to settle onto the substrate. Fluores-
cence microscopy (Olympus BX60) was used to
observe the loading and distribution of rhodamine B
in the spheres. Rhodamine B was excited via a
mercury lamp (550-nm bandpass filter) and the
emitted light was observed at 670 nm. Microsphere
surface structure and porosity were investigated by
scanning electron microscopy (Hitachi S-4700).
Samples were prepared by placing a droplet of an
aqueous microsphere suspension onto a silicon stub.
The samples were dried overnight and were sputter
coated with gold prior to imaging at 2–10 eV.
2.5. Particle size distribution
A Coulter Multisizer (Beckman Coulter Inc.)
equipped with a 200-mm aperture was used to
determine the size distribution of the various sphere
preparations. The lyophilized particles were resus-
pended in Isoton and a type I-A dispersant was used
to prevent microsphere aggregation. A minimum of
5,000 microspheres was analyzed for each sample.
3. Results
3.1. Description and theory of the microsphere
generating apparatus
The main apparatus (Fig. 1A), which provides
fabrication of monodisperse microparticles, is based
on passing a solution containing the sphere material,
and any drug to be encapsulated, through a small
nozzle or other orifice (20 mm to a few millimeters
62 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74
Fig. 1. (A) Schematic of the microsphere generator portraying acoustic excitation with carrier stream for microsphere production. (B)
Schematic indicating the variables used for acoustic excitation theory development.
in diameter) to form a smooth, cylindrical jet. To
controllably break the jet into droplets, the nozzle is
vibrated by a piezoelectric transducer driven by a
wave generator at a frequency tuned to match the
flow rate and the desired drop size (vide infra). The
mechanical excitation launches a wave of acoustic
energy along the liquid jet generating periodic
instabilities that, in turn, break the stream into a train
of uniform droplets (Fig. 1B). With this apparatus
alone, the minimum particle size achievable is
slightly larger than the nozzle opening. This ap-
proach represents an improvement over conventional
ultrasonic nozzles as the acoustic wave intensity is
lower and we can tightly control the match between
the frequency and solution flow rate.
We can further control sphere size, and in fact
form droplets smaller than the nozzle opening, by
employing an annular flow of a non-solvent phase
around the polymer jet. The annular stream is
pumped at a linear velocity greater than that of the
polymer stream. Thus, frictional contact between the
two streams generates an additional downward force
that effectively ‘pulls’ the polymer solution away
from the tip of the nozzle. The polymer stream is
accelerated by this force and, therefore, thinned to a
degree depending on the difference in linear veloci-
ties of the two streams.
It is necessary to understand the theory of droplet
formation as applied to this system in order to
predict the process parameters (flow rates and acous-
tic frequency) required to generate particles of a
desired size or size distribution. Lord Rayleigh first
derived the jet instability equations for a cylindrical
inviscid jet subject to disturbance [32]. Normally, a
liquid jet under the influence of surface tension alone
tends to break naturally into randomly sized drops.
Lord Rayleigh found that the most unstable wave-
length ( lmax ) of a disturbance imposed on a jet
surface, which will give rise to maximum growth
rate and consequently result in the break-up of the jet
into uniform droplets (Fig. 1B), is:
lmax 5 9.016r j (1)
where r j is the radius of the undisturbed jet (approxi-
mately equal to, but typically slightly larger than, the
diameter of the nozzle orifice). The theoretical range
of wavelengths that still results in the production of
uniform droplets was also derived by Lord Raleigh
to be:
7r j , l , ` (2)
However, above a certain wavelength, the instability
growth is so small that noise near the wavelength of
the applied acoustic wave causes random break-up of
the jet. The practical range of acoustic wavelengths
that give rise to the break-up of a liquid jet into
uniform droplets was experimentally determined to
be [33]:
 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74
7r j , l , 36r j (3)
Both upper and lower wavelength limits may vary
somewhat depending on the noise level of the system
and the amplitude of the acoustic wave.
The microsphere generator used here allows for
control of the acoustic wave frequency and am-
plitude. The wavelength produced by a set frequency
( f ) is given by:
f 5 vj /l (4)
where vj is the linear velocity of the liquid jet.
Knowing that the volume of the resulting sphere
should be equal to the volume of a cylindrical
element of the jet, the length of which is defined by
the acoustic wavelength, we find that the drop radius,
r d , is predicted to be:
r d 5 (3r 2j vj / 4f )1 / 3 (5)
At the optimum wavelength, r d,max 51.891r j . Thus,
by imposing a uniform, high-amplitude oscillation on
the nozzle, which will dominate the random, natural
instability, we can control the break-up of the stream
into droplets and predict the orifice size (|r j ),
solution flow rate (vj ), and acoustic frequency ( f )
needed to generate a desired sphere size or size
distribution.
3.2. Acoustic excitation technique for microsphere
fabrication
We first demonstrated use of the acoustic excita-
tion technique to fabricate uniform PLG micro-
spheres. PLG solutions were pumped at varying rates
(2.08–2.92 ml / min) through a 60-mm orifice, and
the streams were broken into uniform microspheres
by acoustic excitation at frequencies from 19 to 70
kHz. Generally, the sphere size increased with
increasing PLG flow rate and decreased with increas-
ing acoustic frequency (Fig. 2). Eq. (5) predicts that
the droplet radius will be proportional to (1 /f )1 / 3 and
(vj )1 / 3 . Indeed, plots of ln(r d ) versus ln(1 /f ) are
linear with a slope of 0.33–0.35 (r 2 50.99) as
expected (Fig. 3). Because our current apparatus
does not allow simultaneous measurement of the
drop size and jet diameter, we were not able to
quantitatively verify the dependence of drop size on
63
jet velocity. However, the drop size does quali-
tatively follow the expected trend.
PLG microsphere size can be controlled within a
wide range by varying the orifice diameter, the PLG
flow rate and the acoustic frequency. As shown in
Fig. 4, we have produced homogenous microspheres
from 500 to 45 mm in diameter. The reproducibility
of the sphere sizes was very precise; microspheres
generated in two separate experiments under identi-
cal flow rates and acoustic frequencies were in-
distinguishable. We have fabricated uniform spheres
as small as 25 mm using this apparatus (data not
shown). The microspheres tended to self-assemble
into hexagonal arrays on the glass slides, which is
indicative of spheres having a uniform size dis-
tribution.
While the homogeneity of sphere size is apparent
in the micrographs, we have also measured the size
distributions of large samples of PLG microspheres
(at least 5,000 spheres counted). The distributions
show 95% of the spheres are within 1.0–1.5 mm of
the targeted average diameter. In fact, the uniformity
of the microspheres is nearly indistinguishable from
commercial size standards (compare Fig. 5A and B).
Similarly narrow distributions were observed for all
sphere sizes. In fact, the absolute width of the
distributions narrowed with decreasing sphere sizes.
However, in terms of the fraction of the average
sphere diameter, the distributions become wider with
decreasing sphere diameter (vide infra).
Because sphere diameter is governed by the
polymer solution flow rate and acoustic frequency,
we can control the sphere size distribution by
varying these two parameters over the duration of the
experiment. For example, we have generated micro-
spheres exhibiting a bimodal size distribution by
maintaining a flow rate of 0.33 ml / min while
switching the acoustic frequency from 1.0 to 5.4 kHz
half-way through the experiment (Fig. 5C). Further-
more, we generated a continuously decreasing dis-
tribution (Fig. 5D) by manually increasing the
acoustic frequency from 1.60 to 3.20 kHz in 10-Hz
increments every 3 s at a constant flow rate of 0.33
ml / min. The distribution obtained is not entirely
smooth due to manual operation of the apparatus,
and we expect size distributions to be improved
through computer control of the fabrication parame-
ters.
64 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74

Fig. 2. Nascent PLG / ethyl acetate microspheres. (A) flow rate52.5 ml / min; frequency519, 35, 55 and 70 kHz. (B) frequency529 kHz;
flow rates52.08, 2.20, 2.50 and 2.92 ml / min. Scale bar5100 mm.

Although we have made PLG microspheres as
small as 25 mm using acoustic excitation, the pro-
duction of microspheres less than |30 mm becomes
increasingly difficult. The minimum size obtainable
is approximately equal to the diameter of the orifice
through which the polymer solution is extruded.
Thus, to make 20–30 mm spheres, an |25-mm
orifice is required. Such small nozzles create several
difficulties. Fabrication of small nozzles is difficult,
especially if one considers the need to employ many
nozzles simultaneously in order to scale up the
process. The risk of clogging the nozzle increases
with decreasing size, and carefully filtering the
polymer solution becomes necessary. In addition, the
pressure drop and shear forces at the orifice increase
with decreasing orifice diameter, necessitating a
high-pressure pump and potentially damaging fragile
protein or nucleic acid therapeutics.
3.3. Non-solvent carrier stream for uniform
microsphere fabrication
We have developed a second method for precise
fabrication of PLG microspheres using a parallel,
co-current, non-solvent stream (Fig. 1A) that can
generate spheres smaller than the nozzle opening.
This carrier stream surrounds the polymer solution
stream exiting the nozzle and travels with it toward
the aqueous bath where the droplets are hardened.
When the velocity of the carrier stream is higher than
 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74 65

the polymer stream, the carrier ‘pulls’ on the poly-

Fig. 3. Dependence of nascent PLG / DCM droplet size on acous-
tic frequency ( f ) at several constant polymer solution flow rates.
mer, focusing it into a jet of reduced diameter. Thus,
the jet, and therefore droplet, diameters are de-
termined by the size of the PLG nozzle orifice and
the relative flow rates of the PLG solution and the
carrier stream. We have used this method to produce
uniform microspheres from 500 to 40 mm in diam-
eter (Fig. 6), as well as more polydisperse spheres as
small as |1 mm from a 100-mm nozzle (data not
shown).
The sphere size increased with increasing PLG
solution flow rate and decreased with increasing
carrier stream flow rate. Again, we analyzed the size
distribution of a large sample of microspheres (Fig.
7). The widths of the sphere size distributions, with
average diameters greater than 30 mm, are compar-
able to those obtained with the acoustic excitation

Fig. 4. Light micrographs of uniform PLG microspheres demonstrating a portion of the size range achievable using purely acoustic
excitation for microsphere production. Sizes are approximately (A) 540 mm, (B) 215 mm, (C) 60 mm and (D) 45 mm.
66 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74

Fig. 5. (A) Size distribution of 45 mm microspheres. (B) Commercial size standard for Coulter Multisizer. (C) Bimodal distribution
obtained by changing acoustic frequency from 1.0 to 5.4 kHz during collection. (D) ‘Continuous’ distribution obtained by manually
increasing the acoustic frequency from 1.6 to 3.2 kHz by 10-Hz increments every 3 s.

method. However, using higher carrier-stream flow
rates to generate spheres smaller than |30 mm
caused the resulting size distribution to widen. As
with the acoustic excitation method, size reproduci-
bility was excellent.
Unique microsphere size distributions can be
achieved by varying the flow rates during a pro-
duction run. For example, we have generated multi-
ple particle sizes continuously by varying the carrier
stream flow rate from 1.67 to 0.67 ml / min (flow rate
was decreased by 0.33 ml / min every minute) while
maintaining a PLG solution flow rate of 0.033 ml /
min (Fig. 7C). However, when the carrier stream
pump is adjusted up or down, there is a lag time for
stabilization of the flow rate. Thus, the individual
peaks are broadened and tend to blend together. But,
this phenomenon can be used to create improved
smooth distributions resembling Fig. 5D (data not
shown).
The carrier-stream method allows production of
microspheres with diameters as small as 1% of the
nozzle orifice diameter (|1-mm spheres from a 100-
mm nozzle). The carrier-stream technique also re-
duced the formation of ‘satellite droplets’ and other
undesired spheres; the size distributions are very
‘clean’ with no very large or very small spheres
(compare Fig. 5B and Figs. 7A and B). However, the
microsphere polydispersity increased when produc-
ing microspheres less than |30 mm, in contrast to
the acoustic method. Distributions were 3–4 mm
wide when fabricating the smallest (|1 mm) micro-
spheres (this is actually similar to the distribution
 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74 67

Fig. 6. Light micrographs of uniform PLG microspheres produced from a 100-mm orifice demonstrating a portion of the size range
achievable using a carrier stream for microsphere production. Sizes are approximately (A) 875 mm, (B) 95 mm, (C) 70 mm, and (D) 45 mm.

Fig. 7. (A) and (B) Uniform microspheres produced by flow control (note the absence of undesired microsphere sizes). (C) Distribution
obtained by manually decreasing the carrier stream flow rate from 1.67 to 0.67 ml / min by 0.33 ml / min every minute while maintaining a
0.033 ml / min PLG solution flow rate during the experiment.
68 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74

widths of larger spheres, but 3–4 mm is a much
larger percentage of the diameter for 1-mm compared
to 100-mm spheres). The acoustic and carrier-stream
techniques are in fact complementary, however,
providing an opportunity for a combined and im-
proved method for microsphere production.
3.4. Combined acoustic and carrier stream
techniques for microsphere fabrication
By coupling the size reducing effects of the carrier
stream with the size control capabilities of the
acoustic excitation technique, we have fabricated
uniform microspheres as small as 5 mm in diameter
(Fig. 8). Microspheres prepared in independent
experiments utilizing the same conditions were in-
distinguishable. To determine the size distributions
for the smallest microspheres, we measured 250
randomly sampled microspheres on digitized mi-
crographs by comparing to an image of a micro-
scale. The smallest microspheres (produced using
nozzles as large as 100 mm) exhibit the narrowest
size distributions (Fig. 9A and B). Again, however,
the width of the distribution for the smallest micro-
spheres is a larger percentage of their diameter. In
other words, a 2- to 3-mm wide distribution con-
taining 95% of 100 mm microspheres is relatively
insignificant (2–3% of the targeted diameter), but a
1.0- to1.5-mm wide distribution containing 95% of 5
mm microspheres is very significant (10–20% of the

Fig. 8. Light and fluorescence micrographs of uniform PLG microspheres demonstrating the lower limit of the size range achievable using
acoustic excitation with a carrier stream for microsphere production. Sizes are approximately (A) and (B) 24 mm, (C) 11 mm, and (D)
(composite picture) 6 mm. Agitation was removed during microsphere hardening for the collection of the above microspheres resulting in
non-uniform rhodamine B loading.
 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74 69

Fig. 9. (A) 11 mm and (B) 6 mm size distributions obtained by measuring 250 randomly selected microspheres (Fig. 8C and D). (C) Coulter
multisizer distribution exhibiting the distinct and smooth regions of size distribution obtainable. (D) Distribution obtained by manually
changing the acoustic frequency from 1.6 to 3.2 kHz by 100-Hz increments for various periods of time.

targeted diameter). We expect that size distributions
can be narrowed by implementing a more precise
flow control system.
Complex size profiles comprising microspheres
less than 30 mm in diameter can be fabricated using
the precise and instantaneous control afforded by the
acoustic excitation method with the size reduction
capability of the carrier stream. Profiles can be
produced containing both discrete and smoothly
varying sizes (Fig. 9C). Also, pre-defined profiles
can be generated over a wide range of microsphere
sizes. For example, we created microspheres ex-
hibiting distinct sizes from 30 to 80 mm in diameter
in variable amounts by adjusting the duration of time
between 100 Hz changes in the acoustic frequency
while maintaining constant PLG and carrier stream
flow rates (Fig. 9D).
Problems of the individual techniques are over-
come by combining acoustic excitation with a carrier
stream (Table 1). Acoustic methods allow for precise
control of sphere sizes and instant changes in size
over a defined range. Implementing a carrier stream
limits the formation of undesired microsphere sizes,
effectively reduces the sphere size obtainable and
broadens the range of microsphere sizes achievable
from a given orifice by controlling the diameter of
the PLG solution stream. Finally, by orchestrating
the variables of acoustic frequency and intensity,
70 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74

Table 1
Comparison of microsphere production techniques
Minimum uniform Size range Advantages Disadvantages
diameter (mm) achievable (mm)
Ultrasonic |25 |25 to • Precise and instantaneous • Production of unwanted
excitation .500 control of microsphere sphere sizes (satellite drops).
diameter. • Minimum sphere size limited
by shear forces at the orifice.
• Increased clogging as nozzle
size decreases.
• Narrow size range from
a particular nozzle.
Carrier stream |30 |1 to • Nozzle clogging and • Increased microsphere
.500 shear forces avoided. polydispersity as microsphere
• Unwanted microsphere diameter decreases.
sizes eliminated. • Difficult to produce sharp changes
• Broad size range when developing pre-defined
from one nozzle. size distributions.
Combined |5 |1 to • Nozzle clogging and • Increasing polydispersity
techniques .500 shear forces avoided. for microspheres #1 mm.
• Unwanted microsphere • More complex apparatus and
sizes eliminated. control systems are required.
• Broad size range
from one nozzle.
• Precise and instantaneous
control of microsphere
diameter.

flow rate, and nozzle size, we have the ability to
produce pre-defined and monodisperse size distribu-
tions.
3.5. Encapsulation and release of a model drug
Microspheres encapsulating various compounds
can be generated by co-dissolving the drug in the
polymer solution, suspending solid drug particles in
the polymer solution, or forming an emulsion of a
drug-containing aqueous phase in the polymer phase.
We have encapsulated rhodamine B as a model drug
by co-dissolving the free-base form in methylene
chloride along with PLG. We fabricated micro-
spheres containing 1, 3, and 5% rhodamine B with
diameters of 35, 50 and 65 mm. Rhodamine B-loaded
microspheres fabricated using our technique resem-
ble those fabricated using emulsion methods in
surface structure and drug loading (Fig. 10A and B).
Also, the release of Rhodamine B closely resembles
previously reported results for the release of simple
molecules from PLG microspheres [34,35]. The
release rate increased with decreasing sphere size
confirming that microsphere size affects drug deliv-
ery rates even in the case of small molecules (Fig.
10C).
3.6. Size change upon solvent extraction
Although Eq. (5) predicts the size of the initial
PLG solution droplets, the change in droplet size
during solvent extraction must be taken into consid-
eration in order to predict the final microsphere size.
We measured microsphere size from digitized mi-
crographs taken at 15-min intervals during the
solvent extraction and evaporation process (Fig. 11).
The volume loss of each microsphere can be attribu-
ted to the extraction of methylene chloride out of the
microsphere into the aqueous solution and the
simultaneous counter diffusion of a small amount of
water into the microsphere [36]. The droplets re-
quired 30–45 min to reach their minimum diameter,
which was approximately 30% of the initial diam-
eter. Because the droplets consist of only 5 wt%
 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74
Fig. 10. (A) Scanning electron micrograph of PLG microspheres.
(B) Fluorescence micrograph of PLG microspheres loaded with
rhodamine B base. (C) The amount of rhodamine B released is
normalized for each sphere size for the comparison of release
rates.
71
Fig. 11. Plot of microsphere diameter decrease over time as
microspheres harden by phase inversion.
PLG, solvent extraction accounts for 95% of the
mass loss corresponding to 37% volume change (at
constant density). The additional volume loss is
accounted for by the fact that the polymer-rich phase
is more dense than the polymer-lean solution. We
have not measured residual solvent in the hardened
microspheres, and a further decrease in size upon
complete drying can not be excluded. However, any
such changes are likely to be minor.
4. Discussion
The sizes of biodegradable polymer microspheres
have several critical implications for controlled-re-
lease drug delivery. For example, sphere size impacts
allowable routes of administration and the final
disposition of the spheres in the body. Importantly,
microsphere diameter is a determining factor of drug
release rates and controlled manipulation of the size
may provide a means to tailor release rate profiles.
Finally, sphere size can be used to passively target
the delivery vehicles for uptake by specific types of
cells, such as professional antigen-presenting cells,
or to target specific tissues such as by embolization
in the tortuous capillary bed of cancerous tumors.
For all of these reasons, control of sphere size during
fabrication should be a critical aspect of any fabrica-
tion process as spheres outside the desired size
range, and the encapsulated drug, are often wasted.
Several approaches to fabricating polymer micro-
spheres with controlled size have been reported
72 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74
[16,23,29–31]. For example, Shiga et al. [29] have
successfully developed a membrane emulsification
method to fabricate poly( D,L-lactide) and PLG micro-
spheres with narrow size distributions and average
diameters of 1–3 mm. The sphere size was controll-
able, however, only by extruding the polymer solu-
tion through membranes of different pore sizes. This
requirement obviously reduces the flexibility of the
method, and fabrication of pre-defined size distribu-
tions is not allowed. Further, Amsden and Goosen
[23] reported an electrostatic spraying method for
control of microsphere size. The electrostatic method
was found to be incapable of producing uniform
polymer particles less than 500 mm in diameter, and
the authors concluded their method can not produce
smaller microspheres having a narrow size distribu-
tion. Amsden [30] described a hydrodynamic method
with several similarities to the methodology reported
here. This method successfully produces micro-
spheres down to 68 mm in diameter, but for all
spheres less than 100 mm the standard deviation of
the size distribution was greater than 15 mm. Thus,
our methodology surpasses each of these previous
reports in terms of both the range of sphere diame-
ters obtainable and the uniformity of the size dis-
tributions.
The droplet generation techniques described here
are rather general and are expected to be applicable
to any polymer that can be processed as a melt or
solution. In previous studies, one of us (KK) has
used very similar technologies to fabricate monodis-
perse particles of a variety of materials including jet
fuel oil, silicates, metal oxides, etc. While our
experiments with drug delivery systems have been
restricted to PLG, the technique is expected to be
applicable to other polymers including other polyes-
ters, polyanhydrides, poly(ethylene-vinyl acetate),
poly(caprolactones), poly(organophosphazenes),
polyphosphoesters, etc. We have used PLG of vary-
ing molecular weights form |6,000 to 52,000 (i.v.
0.15–0.63 dl / g). Further, we have sprayed PLG
solutions from 10 to 100 mg / ml dissolved in methyl-
ene chloride (as reported here), ethyl acetate, chloro-
form and acetone. Changes in viscosity of the
polymer solution will, of course, alter the experimen-
tal parameters required to produce spheres of a
particular size, but we have not as yet encountered a
limitation due to solution viscosity.
The uniqueness of our approach, compared with
other commonly used methods, is limited to the
mechanism of the droplet formation. In other re-
spects, the approach is similar to traditional PLG
microsphere fabrication methods. For example, we
have demonstrated here that a hydrophobic ‘drug’
molecule can be encapsulated by co-dissolution with
the polymer. In addition, preliminary experiments
have shown that water-in-oil emulsions (with use of
a surfactant to stabilize the emulsion) or suspensions
of solid drug particles can be encapsulated (data not
shown; these results will be reported in a future
manuscript focusing on drug encapsulation and
release kinetics). The microspheres reported here
were hardened by solvent extraction in an aqueous
non-solvent bath in a manner analogous to typical
emulsion-extraction procedures [8–10]. Other hard-
ening / drying methods are also applicable. Finally,
the microsphere morphology, appearance of the
surface, and drug encapsulation and release kinetics
are similar to those of traditionally fabricated micro-
spheres. The similarities to the commonly used
microsphere fabrication methods are a major advan-
tage of our approach, as they will allow easy
integration with existing processes.
5. Conclusions
We have developed a microsphere fabrication
process capable of reproducibly generating monodis-
perse particles as well as populations exhibiting pre-
defined size distributions. Using acoustic excitation,
we can precisely control microsphere diameter with-
in a range from |1- to 10-times the orifice diameter
(Table 1). By further employing an annular, non-
solvent carrier stream, the sphere-forming liquid can
be ‘focused’ into a fine jet, reducing particle size by
two orders of magnitude. The technique has been
well characterized resulting in unprecedented control
of PLG microsphere production. The experiments
reported here demonstrate the precision and accuracy
with which PLG microspheres can be produced. The
data are especially encouraging since the apparatus is
still rather crude and many improvements in the
control system are currently under investigation. The
methodology is generally applicable to fabrication of
spheres comprising virtually any polymer and can be
 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74
integrated with any of the several methods for drug
encapsulation and particle hardening. In addition,
multi-nozzle arrays for higher throughput sphere
fabrication are under development. Particle size is a
critical parameter for several aspects of controlled-
release drug delivery including control of drug-re-
lease kinetics, passive targeting to specific cell or
tissue types, biodistribution upon administration, and
available routes of administration. Thus, we believe
the ability to precisely control biodegradable poly-
mer microsphere size will not only improve current
controlled-release applications, but will hasten de-
velopment of advanced drug delivery systems, as
well.
Acknowledgements
We wish to thank Karen Gibbs of Applied En-
gineering Materials for her assistance with the
particle sizing and Dave Garland of Valpey Fisher
for donating piezoelectric transducers. This work
was funded in part by the University of Illinois
Campus Research Board.
References
[1] M.M. Struck, Biopharmaceutical R&D success rates and
development times. A new analysis provides benchmarks for
the future, Bio / Technology 12 (1994) 674–677.
[2] S.D. Putney, P.A. Burke, Improving protein therapeutics with
sustained-release formulations, Nat. Biotechnol. 16 (1998)
153–157.
[3] R. Langer, J. Folkman, Polymers for the sustained release of
proteins and other macromolecules, Nature 263 (1976) 797–
800.
[4] S.P. Baldwin, W.M. Saltzman, Materials for protein delivery
in tissue engineering, Adv. Drug Deliv. Rev. 33 (1998)
71–86.
[5] T.J. Young, K.P. Johnston, K. Mishima, H. Tanaka, Encapsu-
lation of lysozyme in a biodegradable polymer by precipi-
tation with a vapor-over-liquid antisolvent, J. Pharm. Sci. 88
(6) (1999) 640–650.
[6] F.X. Lacasse, P. Hildgen, J. Perodin, E. Escher, N.C.
Phillips, J.N. Mcmullen, Improved activity of a new an-
giotensin receptor antagonist by an injectable spray-dried
polymer microsphere preparation, Pharm. Res. 14 (7) (1997)
887–891.
[7] M. Luck, K.F. Pistel, Y.X. Li, T. Blunk, R.H. Muller, T.
Kissel, Plasma protein adsorption on biodegradable micro-
73
spheres consisting of poly( D,L-lactide-co-glycolide), poly( L-
lactide) or ABA triblock copolymers containing poly(oxy-
ethylene). Influence of production method and polymer
composition, J. Control. Release 55 (2-3) (1998) 107–120.
[8] Y. Kawashima, H. Yamamoto, H. Takeuchi, T. Hino, T.
Niwa, Properties of a peptide containing DL-lactide / glycolide
copolymer nanospheres prepared by novel emulsion solvent
diffusion methods, Eur. J. Pharm. Biopharm. 45 (1) (1998)
41–48.
[9] C. Sturesson, J. Carlfors, Incorporation of protein in PLG-
microspheres with retention of bioactivity, J. Control. Re-
lease 67 (2-3) (2000) 171–178.
[10] A.M. Tinsley-Bown, R. Fretwell, A.B. Dowsett, S.L. Davis,
G.H. Farrar, Formulation of poly( D,L-lactic-co-glycolic acid)
microparticles for rapid plasmid DNA delivery, J. Control.
Release 66 (2-3) (2000) 229–241.
[11] Y.F. Maa, C. Hsu, Microencapsulation reactor scale-up by
dimensional analysis, J. Microencapsul. 13 (1) (1996) 53–
66.
[12] M.A. Tracy, Development and scale-up of a microsphere
protein delivery system, Biotechnol. Prog. 14 (1998) 108–
115.
[13] C. Evora, I. Soriano, R.A. Rogers, K.M. Shakesheff, J.
Hanes, R. Langer, Relating the phagocytosis of microparti-
cles by alveolar macrophages to surface chemistry — the
effect of 1,2-dipalmitoylphosphatidylcholine, J. Control.
Release 5 (1998) 143–152.
[14] S. Akhtar, K.J. Lewis, Antisense oligonucleotide delivery to
cultured macrophages is improved by incorporation into
sustained-release biodegradable polymer microspheres, Int. J.
Pharm. 151 (1997) 57–67.
[15] C.R. Dass, M.A. Burton, Microsphere-mediated targeted
gene therapy of solid tumors, Drug Deliv. 6 (1999) 243–252.
[16] P. Sansdrap, A.J. Moes, Influence of manufacturing parame-
ters on the size characteristics and the release profiles of
nifedipine from poly( DL-lactide-co-glycolide) microspheres,
Int. J. Pharm. 98 (1993) 157–164.
[17] S. Stithit, W. Chen, J.C. Price, Development and characteri-
zation of buoyant theophylline microspheres with near zero
order release kinetics, J. Microencapsul. 15 (6) (1998) 725–
737.
[18] W. Chen, D.R. Lu, Carboplatin-loaded PLGA microspheres
for intracerebral injection: formulation and characterization,
J. Microencapsul. 16 (5) (1999) 551–563.
[19] B. Lemmer, Why Are So Many Biological Systems Periodic,
Wissenschaftliche Verlagsgesellschaft mbH, 1992, pp. 11–
24.
[20] D.L. Wise, D.J. Trantolo, R.T. Marino, J.P. Kitchell, Oppor-
tunities and challenges in the design of implantable bio-
degradable polymeric systems for the delivery of anti-micro-
bial agents and vaccines, Adv. Drug Deliv. Rev. 1 (1987)
19–39.
[21] J.L. Cleland, A. Lim, L. Barron, E.T. Duenas, M.F. Powell,
Development of a single-shot subunit vaccine for HIV-1: Part
4. Optimizing microencapsulation and pulsatile release of
MN rgp120 from biodegradable microspheres, J. Control.
Release 47 (1997) 135–150.
[22] Y. Capan, B.H. Woo, S. Gebrekidan, S. Ahmed, P.P. DeLuca,
74 C. Berkland et al. / Journal of Controlled Release 73 (2001) 59 – 74
Preparation and characterization of poly( D,L-lactide-co-gly-
colide) microspheres for controlled release of poly( L-lysine)
complexed plasmid DNA, Pharm. Res. 16 (4) (1999) 509–
513.
[23] B.G. Amsden, M. Goosen, An examination of the factors
affecting the size, distribution, and release characteristics of
polymer microbeads made using electrostatics, J. Control.
Release 43 (1997) 183–196.
[24] S. Cohen, T. Yoshika, M. Lucarelli, L.H. Hwang, R. Langer,
Controlled delivery systems for proteins based on poly(lac-
tic / glycolic acid) microspheres, Pharm. Res. 8 (6) (1991)
713–720.
[25] W. Dang, W.M. Saltzman, Controlled release of macro-
molecules from a degradable polyanhydride matrix, J.
Biomater. Sci., Polym. Ed. 6 (3) (1994) 297–311.
[26] K.W. Leong, B.C. Brott, R. Langer, Bioerodible poly-
anhydrides as drug-carrier matrices. I: characterization,
degradation, and release characteristics, J. Biomed. Mater.
Res. 19 (1985) 941–955.
[27] H. Rafati, A.G.A. Coombes, J. Adler, J. Holland, S.S. Davis,
Protein-loaded poly( DL-lactide-co-glycolide) microparticles
for oral administration: formulation, structural and release
characteristics, J. Control. Release 43 (1997) 89–102.
[28] C.T. Ying Men, H.P. Merkle, B. Gander, G. Corradin, A
single administration of tetanus toxoid in biodegradable
microspheres elicits T cell and antibody responses similar or
superior to those obtained with aluminum hydroxide, Vaccine
13 (7) (1995) 683–689.
[29] K. Shiga, N. Muramatsu, T. Kondo, Preparation of poly( D,L-
lactide) and copoly(lactide-glycolide) microspheres of uni-
form size, J. Pharm. Pharmacol. 48 (1996) 891–895.
[30] B. Amsden, The production of uniformly sized polymer
microspheres, Pharm. Res. 16 (1999) 1140–1143.
[31] N. Leelarasamee, S.A. Howard, C.J. Malanga, J.K.H. Ma, A
method for the preparation of polylactic acid microcapsules
of controlled particle size and drug loading, J. Microencap-
sul. 5 (1988) 147–157.
[32] L. Rayleigh, Proc. Lond. Math. Soc. 10 (1879) 4.
[33] L. Rayleigh, Philos. Mag. S.G. 14 (1882) 184.
[34] F.B. Landry, D.V. Bazile, G. Spenlehauer, M. Veillard, J.
Kreuter, Release of the fluorescent marker Prodan from
poly( D,L-lactic acid) nanoparticles coated with albumin or
polyvinyl alcohol in model digestive fluids (USP XXII), J.
Control. Release 44 (1997) 227–236.
[35] L. Lewis, R.L. Boni, C.M. Adeyeye, The physical and
chemical stability of suspensions of sustained-release di-
clofenac microspheres, J. Microencapsul. 15 (5) (1998) 555–
567.
[36] P.D. Graham, K.J. Brodbeck, A.J. McHugh, Phase inversion
dynamics of PLGA solution related to drug delivery, J.
Control. Release 58 (1999) 233–245.