Biomaterials 23 (2002) 42214231

Why degradable polymers undergo surface erosion or bulk erosion

a,
.
Friederike von Burkersrodaa, Luise Schedlb, Achim Gopferich
*
a

Department of Pharmaceutical Technology, University of Regensburg, 93040 Regensburg, Germany

b
Department of Pharmaceutical Technology, University of Erlangen, Germany
Received 18 October 2001; received in revised form 22 March 2002; accepted 30 April 2002

Abstract
A theoretical model was developed that allows to predict the erosion mechanism of water insoluble biodegradable polymer
matrices. The model shows that all degradable polymers can undergo surface erosion or bulk erosion. Which way a polymer matrix
erodes after all depends on the diffusivity of water inside the matrix, the degradation rate of the polymers functional groups and the
matrix dimensions. From these parameters the model allows to calculate for an individual polymer matrix a dimensionless erosion
number e: The value of e indicates the mode of erosion. Based on e; a critical device dimension Lcritical can be calculated. If a matrix
is larger than Lcritical it will undergo surface erosion, if not it will be bulk eroding. Lcritical values for polymers were estimated
based on literature data. Polyanhydrides were found to be surface eroding down to a size of approximately Lcritical 104 m while
poly(a-hydroxy esters) matrices need to be larger than Lcritical 101 m to lose their bulk erosion properties. To support our
theoretical ndings it was shown experimentally that poly(a-hydroxy ester) matrices, which are considered classical bulk eroding
materials, can also undergo surface erosion. r 2002 Elsevier Science Ltd. All rights reserved.
Keywords: PLA; PLGA; Degradation; Surface erosion; Bulk erosion; Modeling

1. Introduction
Thirty years ago, polyesters made of glycolic acid
were among the rst synthetic biodegradable polymers
applied in medicine. Based on their ability to degrade
in an aqueous environment they were used for the
manufacture of resorbable sutures [1]. Since then
the number of applications and the variety of biodegradable polymers that are not water soluble increased
substantially. Polymers such as polyanhydrides [1],
poly(a-hydroxy esters) [2], poly(b-hydroxy esters) [3],
poly(ortho esters) [4] and many more were used for
numerous medical and pharmaceutical applications [5]
such as matrices for drug delivery [6], scaffold materials
for tissue engineering [7], and degradable screws in
orthopedic surgery to mention just a few [8,2].
A closer look at biodegradable polymers even
those that were commercialized as adjuvants in pharmaceutical preparations such as polyanhydrides and
*Corresponding author. Tel.: +49-941-943-4843; fax: +49-941-9434807.
E-mail address: achim.goepferich@chemie.uni-regensburg.de (A.
.
Gopferich).

poly(a-hydroxy esters) reveals, however, that there are

still many unsolved problems that hinder us to take full
advantage of these materials. One example is the lack of
understanding polymer erosion, which is the loss of
mass from degradable polymer matrices [9]. Despite the
importance of erosion which controls essential processes
like the release of drugs from polymer implants or
mechanical stability, the mechanism of erosion is only
insufciently investigated for many polymers. Our
current knowledge seems to reveal only the tip of the
iceberg. All we know about erosion mechanisms dates
back to the early 1980s when degradable polymers were
classied into surface eroding and bulk eroding ones
[10]. The only hint how this may be related to the
polymer structure was the observation that materials
built from functional groups with short hydrolysis half
lives tend to be surface eroding (Fig. 1). In the case of
surface eroding polymers, erosion proceeds at constant
velocity at any time during erosion [11]. For bulk
eroding polymers such as poly(d,l-lactic acid) (PLA)
and poly(d,l-lactic-co-glycolic acid) (PLGA), however,
things are more complicated as they have no constant
erosion velocity [12]. Usually, they do not erode for
long periods of time after which erosion sets in

0142-9612/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 2 - 9 6 1 2 ( 0 2 ) 0 0 1 7 0 - 9

4222

F.V. Burkersroda et al. / Biomaterials 23 (2002) 42214231

polymers and nd an answer to the question why they

undergo bulk erosion.
Our effort aimed at providing the basis for a better
understanding of PLGA and PLA erosion, which served
as role models for the erosion of bulk eroding polymers
in general. Once a better understanding of erosion is
provided, one will be able to use these polymers more
efciently for the design of drug delivery systems in the
future.

Fig. 1. Schematic illustration of the changes a polymer matrix

undergoes during surface erosion and bulk erosion.

2. Materials and methods

2.1. Materials

spontaneously. It is, however, still not clear why

polymers erode one way or the other.
In contrast to polymer erosion, there is plenty of
literature on polymer degradation which is the process
of chain cleavage [9]. In recent years a number of
parameters have been identied that affect the degradation velocity. Among them are the copolymer composition [13], autocatalysis by acidic degradation products
inside a matrix [14,15] and the presence of a protolytic
drugs [16] or other excipients [17]. However, the impact
of these parameters that increase or decrease the
degradation velocity on the erosion mechanism are not
exactly clear. It is so far not possible to predict how the
erosion process changes when the degradation rate is
altered.
To resolve the problems outlined above, we intended to
develop a simple theoretical model that allows to predict if
a polymer matrix undergoes surface or bulk erosion.
Concomitantly, we planned erosion studies to investigate
the impact of degradation conditions on the erosion
mechanism experimentally. As model polymers we chose
PLA and PLGA. One reason was their exceptional
importance due to their biocompatibility [18] and their
use in numerous approved medical devices and pharmaceutical preparations for parenteral use in humans.
Another reason was that this class of polymer has been
extensively investigated so that a decent amount of
literature is available. By changing parameters that have
an effect on the erosion mechanism according to the model
we tried to prove that predictions based on our theory are
right. Thus we performed erosion studies with geometrically well-dened PLA and PLGA matrices (12.5 mm
diameter 1.4 mm height) and systematically changed the
pH of the erosion medium. To alter the degradation rate
of the poly(a-hydroxy esters) we decided on using acid
catalysis (pH 2) as well as alkaline conditions (pH>13).
Degradation and erosion were monitored by following
mass loss, polymer swelling, changes of molecular weight
and morphology. The results of erosion under catalysis
were compared to those obtained at pH 7.4. It was our
overall goal in this study to investigate bulk eroding

PLGA molecular weight 8,000 (PLA25GA508h),

PLGA molecular weight 14,000 (PLA25GA5014), PLGA
molecular weight 47,000 (PLA25GA5047h), poly (d,llacitc acid) (PLA) molecular weight 2000 (PLA502),
PLA molecular weight 11,000 (PLA5011h) and PLA
molecular weight 17,000 (PLA5017) were obtained from
Boehringer Ingelheim, Ingelheim, Germany. The molar
ratio of lactic to glycolic acid in PLGA as well as the
molar ratio of the optical isomers d- and l-lactic acid in
PLA and PLGA was 1:1 in all polymers. PLA502,
PLA5017 and PLA25GA5014 were end-capped with
lactic-acid-ethyl-ester while the COOH-terminal of the
other polymers was non-end-capped. Acronyms reect
the molar ratios of d- to l-lactic acid units and glycolic
acid units according to [13], the molecular weights and
the state of the COOH-terminal (non-end-capped
materials are marked h). Chloroform, sulfuric acid
and potassium iodide were obtained in analytical grade
from Fluka, Buchs, Switzerland. Iodine, 1 m hydrochloric acid solution, 1 m sodium hydroxide solution, sodium
hydroxide, sodium dihydroxide-phosphate-dihydrate,
disodium hydrogen phosphate-dihydrate, trimethylsilane and sodium azide were purchased also in analytical
grade from Merck, Darmstadt, Germany. Deuterated
chloroform (CDCl3) was purchased from Deutero
GmbH, Kastellaun, Germany.

3. Methods
3.1. Manufacture of polymer matrices
Cylindric polymer matrices with a diameter of
12.5 mm and a height of 1.4 mm were produced from
polymer powder by compression. First, the polymer was
lled into a compression tool that allowed for evacua.
tion (evacuable mini-press, Perkin-Elmer, Uberlingen,
Germany). With the punches in place, the dyce was
evacuated for 2 min with an RV5 two stage oil pump
from Edwards, Crawley, Sussex, UK. Prior to compression, the compression tool was kept for 3 min at 701C

F.V. Burkersroda et al. / Biomaterials 23 (2002) 42214231

between the heated plates of the hydraulic press (PW 10,

Weber Maschinen-Apparatebau GmbH, Remshalden,
Germany). Finally, the powder was compressed to
matrices by applying a force of 5 kN for 20 s. The
average matrix weight was approximately 220 mg.
3.2. Erosion of polymer matrices
Erosion studies were performed at 371C with 39
matrices of each polymer. To examine the effect of
degradation velocity on the erosion mechanism, matrices were either eroded in a volume of 10 ml 0.1 m
phosphate buffer pH 7.4 containing 0.02% sodium
azide or in 0.01 m hydrochloric acid (pH 2) or in a 10%
aqueous sodium hydroxide solution (pH>13). The
erosion media were changed regularly so that the pH
was maintained and polymer matrix samples were taken
after time intervals that allowed to follow erosion until a
mass loss of approximately 75% was reached. The
weight of eroded matrices prior to drying was taken as a
measure for polymer swelling. The samples were dried in
vacuum using an RV 5 two stage oil pump (Edwards)
until a nal pressure of 0.5 Pa was reached. The weight
of dried matrices was taken as a measure for erosion.
Height and diameter of wet and dried samples of
matrices eroded at pH>13 were determined with a
caliper rule with a resolution of 50 mm.
3.3. Determination of molecular weight by gel permeation
chromatography (GPC)
The molecular weight of polymer matrices was
determined by GPC using a HPLC-setup from Shimadzu, Kyoto, Japan which consisted of an LC-10 AT VP
pump, a SIL-10 AD autosampler, a SCL-10A interface,
a CTO-10AC column oven and an LC-10AT refractive
index detector. Three columns in series served as
stationary phase: a 5 mm 50  7.8 mm Phenogelt guard
column, a 5 mm 300  7.8 mm2 mixed bed Phenogelt
column (MW range 175,000) and a 5 mm 300  7.8 mm2
( (MW range
Phenogelt column with a pore size of 500 A
50015,000) all from Phenomenex, Torrance, CA.
Samples were analyzed at 351C oven- and detector
temperature using a ow rate of 0.9 ml chloroform/min.
Molecular weights were calculated from calibration
curves obtained for polystyrene standards (molecular
weight range 0.279 kDa, Phenomenex) with narrow
molecular weight distribution, using the Class VP 4.2
size exclusion chromatography software package from
Shimadzu.
3.4. Scanning electron microscopy (SEM)
To investigate the morphology of dried samples, SEM
pictures were taken using a JEOL JSM-840 scanning
electron microscopy (Jeol Ltd., Tokyo, Japan). Dried

4223

polymer matrix samples were brittle enough so that they

could be broken in the middle. They were glued to
25 mm diameter aluminum sample holders using conductive carbon paint (Leit-C, Neubauer Chemikalien,
Munster,
.
Germany) and gold sputtered for 4 min under
argon atmosphere using a Polaron E 5200 sputter coater
from Polaron Equipment Ltd. (Watford, UK).

4. Results and discussion

4.1. Development of an erosion model
The intention that we had with the development of
an erosion model was to identify parameters that
determine why polymers such as PLA are bulk eroding
while others such as polyanhydrides are surface eroding.
The model should furthermore allow to forecast if a
polymer is undergoing surface or bulk erosion. The
following assumptions were made:
*

the polymers for which the model is valid are water

insoluble,
degradation triggers erosion and is a matter of
hydrolysis only,
other processes than degradation have no impact on
erosion.

From the literature it was known that bulk eroding

PLA and PLGA matrix cylinders similar to the ones
that we used in our study degrade right from the
beginning of an erosion experiment. Assuming that the
polymer degradation velocity is uniform inside a
polymer matrix, one can hypothesize that two major
processes have an impact on the erosion kinetics [5]:
1. the diffusion of water into the polymer bulk,
2. the degradation rate of the polymer backbone.
If the diffusion of water into the polymer is faster than
the degradation of polymer bonds, the polymer will
undergo bulk erosion, because degradation is not
conned to the polymer surface. If, however, the
degradation of the polymer bonds is faster than the
diffusion of water, it will be consumed by the hydrolysis
of bonds on the polymer surface and will thus be
prevented from diffusion into the bulk. Degradation
processes are then strictly conned to the matrix surface
and we have in an ideal case, i.e. when the degradation
products are reasonably water soluble, a surface eroding
polymer.
The strategy that we pursued was to nd a measure
for the velocity of both processes: diffusion of water in
the polymer matrix and degradation of polymer bonds.
The ratio of the velocity of both processes, would then
be a convenient parameter that indicates which way a
degradable polymer will erode.

F.V. Burkersroda et al. / Biomaterials 23 (2002) 42214231

4224

The velocity of water inside degradable polymers can

most conveniently be described by applying diffusion
theory. The time, tdiff ; water needs to travel a mean
distance /xS can be estimated by applying random
theory to the motion of water in the polymer [19]:
tdiff

/xS2 p

;
4Deff

where Deff is the effective diffusion coefcient [20] of

water inside a polymer. Eq. (1) provides a measure for
the velocity of water inside a polymer matrix that we
were looking for.
To describe the degradation velocity, we assumed that
the scission of a bond in the backbone of a degradable
polymer can be regarded a random event that follows
Poisson kinetics [21]. The time, t; for a bond to degrade
is called zero-order waiting time and is distributed
according to a zero-order Erlang distribution [22]:
f t lelt ;

where f t is the likelihood that a bond is cleaved at time

t: l is a rate constant that accounts for the differences in
the reactivity of polymer functional groups. l is
equivalent to a rst-order rate constant like they are
used in reaction kinetics and, therefore, directly related
to the half-life of a polymer bond.
To degrade over an average distance /xS (the same
distance that we used to clock the velocity of diffusion)
from the surface into the polymer bulk of a matrix, n
polymer bonds have to be hydrolyzed along that way. n
can be estimated, by dividing /xS through the cubic
root of the polymer volume containing one degradable
bond:
/xS
n q;
3
Mn =NA N  1r

of degradation:
0
2s31
3
1@
Mn
5A:
ln/xS  ln4
Etn
l
NA N  1r

After having found a measure for the velocity of water

diffusion, tdiff ; and polymer degradation, Etn ; one can
compare the time frames of these processes which are
given in Eqs. (1) and (6). For this purpose it was most
convenient to divide tdiff by Etn ; whereby a dimensionless number, e; was obtained that we named erosion
number
e

/xS2 lp

q;
3
Mn =NA N  1r
4Deff ln/xS  ln

where e is dimensionless and can be considered as

Deborah number [23,24]. The values of e fall into three
categories: For polymer matrices with lb1; the reaction
between water and polymer is fast compared to water
diffusion. Such systems undergo surface erosion. For
polymer matrices with l51; the diffusion of water is
faster than the erosion process. These devices are
invaded by water prior to signicant degradation and
undergo bulk erosion. For polymer matrices with l 1;
the erosion mechanism cannot be predicted unequivocally.
Fig. 2 shows a plot of the erosion number e as a
function of the matrix dimension, L; of a polymer
device, the diffusivity of water, Deff ; in the matrix and
the degradation rate constant of the polymer bonds l: L
can be considered the half-thickness of a slab and
replaces the mean distance /xS in Eq. (7) that the
processes of diffusion and degradation travel inside a

where Mn is the number average molecular weight, N

the average degree of polymerization which is the
number of monomers per polymer chain, NA is
Avogadros number and r the density of the polymer.
As the degradation of each single bond is a Poisson
process, the degradation of n bonds is also a Poisson
process with a value of expectation Etn [22]:
Etn

n
1X
1
;
l i1 i

where Etn is the time that is on averagePneeded to

degrade n polymer bonds. Approximating ni1 1=i by
lnn one obtains
1
Etn lnn:
l

Upon substitution of n in Eq. (5) by the expression in

Eq. (3), one nally obtains the measure for the velocity

Fig. 2. Dependence of the erosion number, e, on the diffusivity of

water inside the polymer, Deff, the dimensions of a polymer matrix, L,
and the polymer bond reactivity, l, calculated from equation 7. The
white plane represents the area of surface erosion, the gray one the
area of bulk erosion.

F.V. Burkersroda et al. / Biomaterials 23 (2002) 42214231

4225

Table 1
Estimated values of e and Lcritical for selected degradable polymers
Chemical structure

Polymer

l (s1)

ea

Lcritical b

Poly(anhydrides)

1.9  103 Ref. [30]

11,515

75 mm

Poly(ketal)

6.4  105 Ref. [30]

387

0.4 mm

Poly(ortho esters)

4.8  105 Ref. [30]

291

0.6 mm

Poly(acetal)

2.7  108 Ref. [30]

0.16

2.4 cm

Poly(e-caprolactone)

9.7  108 Ref. [31]

0.1

1.3 cm

Poly(a-hydroxy-esters)

6.6   109 Ref. [30]

4.0  102

7.4 cm

Poly(amides)

2.6  1013 Ref. [30]

1.5  106

13.4 m

q
3
For a 1 cm thick device, D 108 cm2 s1(estimated from Ref. [32]) and ln
Mn =NA N  1r 16:5:
q
3
b
D 108 cm2 s1 (estimated from Ref. [32]) and ln
Mn =NA N  1r 16:5:
a

matrix. The response function shows that erosion

numbers fall into two areas. The white plane indicates
e > 1 which means that the matrix will undergo surface
erosion, while the gray area represents values of eo1
indicating that bulk erosion will dominate. The gure
also indicates, however, that changes in l=Deff or L can
lead to changes in the erosion mechanism. Thus, a
polymer matrix that is undergoing bulk erosion may
become surface eroding when its dimension L is
increased or when the ratio l=Deff is increased. This
means that extremely large matrices are surface eroding
irrespective of the type of polymer they are made of.
Small matrices in contrast are only undergoing surface
erosion when the diffusivity of water in the polymer
matrix is low or the reactivity of polymer bonds with
respect to hydrolysis is high.
Although it is convenient to know that the erosion
mechanism is prone to changes, it would be useful to
know when a surface eroding polymer becomes a bulk
eroding polymer and vice versa. This break point can
be calculated from Eq. (7). Substituting L for /xS;
the device dimension, and assuming that for e 1 the
erosion behavior of a polymer changes, Eq. (7) can
be solved for L: Under these conditions L becomes the

critical dimension Lcritical at which the erosion mechanism is not unequivocally dened. If the dimension of a
polymer matrix is larger than Lcritical the matrix is
undergoing surface erosion. When matrices have dimensions smaller than Lcritical they are undergoing bulk
erosion. Lcritical is an individual characteristic of a
polymer and can accordingly be used to rank polymers.
Table 1 shows the values for e and Lcritical that were
estimated for some classes of degradable polymers based
on literature data. It is obvious that these data can be
subject to dramatic changes depending on the variability
that exists within a polymer class based on the chemical
variability. Nevertheless, the values can give a good
impression of how tremendous the differences in erosion
properties of the different polymer classes can be. Fig. 3
shows the values Lcritical for degradable polymers based
on literature data for Deff and l: It is obvious, that fast
eroding polymers, such as polyanhydrides, are already
surface eroding on a micrometer scale while slow
eroding polymers might need a thickness of meters to
show surface erosion behavior. Fig. 3 makes also clear
that even for fast degrading polymers it is impossible to
undergo surface erosion, when it is processed to polymer
devices that are smaller than Lcritical irrespective of how

4226

F.V. Burkersroda et al. / Biomaterials 23 (2002) 42214231

Fig. 3. Critical thickness, Lcritical, that a polymer device has to exceed

to undergo surface erosion (calculated from Eq. (7), data shown in
Table 1).

reactive its polymer bonds might be. It has to be stressed

at this point that according to this result there is no such
material as a surface eroding polymer or a bulk eroding
polymer. All degradable polymers can rather erode both
ways.
To support our theory, we decided to show experimentally that a polymer that is considered bulk eroding
according to the literature can be made surface eroding.
To this end, we manufactured PLA and PLGA matrix
discs and eroded them in vitro. It is a well-accepted fact
that these polymers undergo bulk erosion under the said
conditions. According to our theory, there are two
parameters that can be used to change the erosion
mechanism: the ratio of l=Deff and the dimension L of
the matrices. Of all three parameters, it was most
convenient to increase l because the diffusivity inside
the polymer is hard to change over several orders of
magnitude and an increase of L would have been
accompanied by experimental difculties such as large
devices and long experimental erosion times. We tried to
increase l simply by catalysis as it is well known that the
hydrolysis of esters is affected tremendously by pH
variations [25].
4.2. PLA and PLGA degradation and erosion at pH 7.4
Prior to erosion under catalytic conditions, we
investigated the erosion of PLA (PLA5017 and
PLA5011h) and PLGA (PLA25GA5014, PLA25GA5047h
and PLA25GA508h) matrix discs at pH 7.4 (Figs. 4a
and b). All matrices show the same principal erosion
behavior: after a period of no erosion, the matrices start
to erode at a dened time. However, the time point at
which erosion sets in, the erosion onset, is quite
different. Overall, the PLGA polymers eroded faster
than the PLA polymers. A good example are the nonend-capped polymers PLA5011h, PLA25GA508h and
PLA25GA5047h. While the PLA polymer had an erosion
onset of approximately 41 days, the two PLGA

Fig. 4. Erosion proles of poly(a-hydroxy esters) at pH 7.4: (a)

PLA5011h (J) and PLA5017 (&), (b) PLA25GA508h (E). PLA25GA5014 (K) and PLA25GA5047h ().

polymers were substantially less stable. PLA25GA508h

started to erode after 5 days and PLA25GA5047h after
15 days. This shows that the copolymer composition is
of vital importance for the stability of the polymers [26].
The PLGA polymers also show that an increase of
molecular weight enhances the erosion resistance
signicantly. This is generally true for end-capped as
well as non-end capped PLA and PLGA. End-capping
has also a pronounced effect on the erosion kinetics as
the example of PLA25GA5047h and PLA25GA5014
shows. Despite its substantially higher molecular weight,
PLA25GA5047h starts to erode earlier (day 15) than
PLA25GA5014 (day 21).
However, despite its impact on the duration of the
erosion process, the copolymer composition, the molecular weight and end-capping do not change the general
prole of bulk erosion. Processes that have so far been
identied to be responsible for this odd erosion behavior
are autocatalysis [27] and percolation phenomena [11].
Autocatalysis arises from the uptake of water by the
polymers which are shown in Figs. 5a and b and from
the increasing amount of carboxylic acid end groups
that are created due to the degradation of the polymer
[14]. Figs. 5a and b show that the polymers take up a
signicant amount of water even before the onset of

F.V. Burkersroda et al. / Biomaterials 23 (2002) 42214231

Fig. 5. Water uptake of poly(a-hydroxy esters) at pH 7.4: (a)

PLA5011h (J) and PLA5017 (&), (b) PLA25GA508h (E), PLAGA5014 (K) and PLA25GA5047h ().

erosion is reached. As no swelling fronts are observed,

one can assume that the matrices take up water
homogeneously and cause the polymer chains to
degrade all over the matrix cross sections. This creates
free carboxylic acid groups which lead to a decrease of
pH inside the swollen polymer matrix and according to
what is known from ester hydrolysis to an acceleration
of degradation inside the polymer bulk [25]. As the
surface of the polymer is kept at neutral pH, a pH
gradient develops that slows down the degradation of
the polymer matrix surface compared to the center. The
surface layer breaks at some point when a critical
osmotic pressure builds up inside the matrix due to the
accumulation of degradation products. Percolation
phenomena [11] are based on the fact that degradation
products can apparently not leave the matrices prior to
the erosion onset. Only after a critical degree of
degradation is reached, the polymers form a network
of pores that allows for the release of monomers and
oligomers [12].
4.3. PLA and PLGA degradation and erosion at pH 2
In a rst attempt to prove our erosion theory right, we
tried to increase the degradation rate of the polymer via
acid catalysis. We, therefore, degraded PLA and PLGA
matrix discs at pH 2, a value that is assumed to be close

4227

Fig. 6. Erosion of poly(a-hydroxy esters) at pH 2: (a) PLA502 (B),

PLA5011h (J) and PLA5017 (&), (b) PLA25GA508h (E), PLA25GA5014 (K) and PLA25GA5047h ().

to the pH that is reached inside the polymer during

degradation at pH 7.4 under the impact of autocatalysis
[28].
Figs. 6a and b show that the erosion proles for the
PLA and the PLGA polymers did not change signicantly. Except for PLA5011h, the erosion onset is at later
times. This is surprising as one would assume that
accelerated degradation should somehow have an
impact on the erosion kinetics. We expected that either
the phenomenon of bulk erosion with its typical delay
period disappears or that at least erosion is accelerated.
That both did not happen suggests that the degradation
rate did not change at pH 2 because even the
degradation at neutral pH is already acid catalyzed
under the impact of autocatalysis. The pH gradient that
develops across the matrix during erosion at pH 7.4 may
well be not linear so that the majority of polymer chains
are degraded at pH 2. Low pH of the degradation
medium, therefore, cannot spur degradation and erosion
any more. Furthermore, the result shows that autocatalysis alone cannot be held responsible for the bulk
erosion proles of PLA and PLGA because degradation
at pH 2 does not allow for the buildup of pH gradients.
Cross sections of the matrices also reveal that the
matrices appear homogenous in contrast to matrices

4228

F.V. Burkersroda et al. / Biomaterials 23 (2002) 42214231

Fig. 7. Degradation of poly(a-hydroxy esters) at pH 2: (a) PLA502

(B), PLA5011h (J) and PLA5017 (&), (b) PLA25 GA508h (E),
PLA25 GA5014 (K) and PLA25GA5047h ().

degraded at neutral pH where a soft and almost liquid

core forms under the impact of autocatalysis [28].
Figs. 7a and b conrm that the degradation kinetics
cannot account for the biphasic bulk erosion proles.
The degradation proles of PLA (Fig. 7a) and of PLGA
(Fig. 7b) reveal a continuous decrease of molecular
weight right from the beginning of the experiment.
The mass of the eroded matrices prior to drying which
is again a measure for polymer swelling is shown in
Figs. 8a and b. The matrices take up again substantial
amounts of water during the erosion experiment. The
progressive degradation of the polymers seems to lead to
an increase of degradation products with high osmotic
pressure which attracts even more water. The maximum
weight of wet matrices, therefore, coincides with the
onset of erosion which is the time point at which the
matrix contains a maximum amount of degradation
products. The results conrm our assumption, that bulk
eroding polymers are completely intruded by water prior
to the start of erosion.
The conclusion that we may draw from PLA and
PLGA erosion at pH 2 is that autocatalysis seems to be
the motor for PLA and PLGA bulk erosion but that
other processes such as the release kinetics of degradation products in conjunction with percolation phenomena may contribute to this type of degradation behavior
as well. This does not exclude that other mechanisms
which we may have not discovered yet contribute to the

Fig. 8. Water uptake of poly(a-hydroxy esters) at pH 2: (a) PLA502

(B), PLA50 11h (J) and PLA50 17 (&), (b) PLA25GA508h (E),
PLA25GA5014 (K) and PLA25GA5047h ().

complicated erosion behavior as well. Acid catalysis

does not allow to prove our hypothesis that an increase
of l leads to a change of the PLA and PLGA erosion
mechanism. Therefore, we eroded selected PLA and
PLGA matrices at alkaline pH.

4.4. PLA and PLGA degradation and erosion at pH>13

The erosion kinetics of PLA and PLGA change
tremendously under base catalysis. The erosion proles,
obtained at pH>13 for PLA5017, PLA25GA5047h and
PLA25GA5014 matrix discs are shown in Fig. 9. Mass
loss starts without any delay right from the beginning of
each erosion experiment. All three polymers showed
linear erosion proles which are typical of surface
eroding polymers [29]. The time frame of erosion
changed signicantly. While PLA and PLGA erosion
was a matter of days and weeks at pH 7.4 and 2, it is
complete after a couple of hours at pH>13. The striking
difference in the duration of erosion is illustrated in
Fig. 10 where the time until a 50% mass loss is
compared for erosion at pH 2 and pH>13. The data
shows a reduction by a factor of 40 for PLA5017, a
factor of 102 for PLA25GA5047h and factor of 150 for
PLA25GA5014. The effect of copolymerization might

F.V. Burkersroda et al. / Biomaterials 23 (2002) 42214231

explain why the reduction is higher for the two PLGA

polymers compared to PLA5017.
The data for water uptake of matrices at pH>13 is
shown in Fig. 11. In contrast to erosion at pH 2 and 7.4,
the matrices do not take up signicant amounts of
water. The slight difference between wet and dry weight

Fig. 9. Erosion of poly(a-hydroxy esters) at pH>13: PLA25GA5014

(K) and PLA5017 (&), PLA25GA5047h ().

4229

at pH>13 might be due to the water adhered or

adsorbed to the surface. Whereas polymers eroded at
pH 2 and 7.4 changed their original geometric shape
within a few days to an irregular form, matrices eroded
at pH>13 maintained their geometry until the end of
the erosion study. Height and diameter, wet and dry,
developed almost parallel to erosion (data not shown).
Both parameters decreased linearly from the beginning
of the erosion study. In summary, one can conclude that
the mechanism of erosion has changed from bulk to
surface erosion. To scrutinize this, eroded matrices were
investigated by GPC and SEM.
The degradation prole (Fig. 12) of PLA and PLGA
polymers t well to the ndings of the erosion study.
During the erosion of PLA5017, PLA25GA5014 and
PLA25GA5047h, the molecular weight of the matrices
remained constant. Apparently, the chain cleavage
proceeds at pH>13 much faster than polymer swelling.
As swelling cannot be observed we hypothesize that
water which enters the polymer is mainly trapped by
degradation. This connes the processes of degradation
and erosion to the surface. This can explain why the
molecular weights of the matrices remained unchanged.
To investigate if corresponding morphological changes
can be observed, cross section of polymers were
investigated by SEM.

Fig. 10. Time until 50% of mass are lost by erosion at pH 2 (white
bars) and pH>13 (black bars).

Fig. 11. Water uptake of poly(a-hydroxy esters) at pH>13: PLA25GA5014 (K) and PLA5017 (&), PLA25GA5047h ().

Fig. 12. Degradation of poly(a-hydroxy esters) at pH>13: (a)

PLA5017 (&), (b) PLA25GA5014 (K) and PLA25GA5047h ().

4230

F.V. Burkersroda et al. / Biomaterials 23 (2002) 42214231

Fig. 13. SEM pictures of PLA5017 during erosion at pH>13: (a) t

0; (b) t 108 h.

Figs. 13a and b show cross sections through PLA5017

matrices prior to and during erosion at pH>13. Initially
the cross sections were homogenous without any aws.
After 108 h the shape of the matrices is unchanged.
However, the matrix diameter is reduced and on the
surface a porous zone has developed. Such erosion zones
are typical of polymers that undergo surface erosion
[29].
4.5. Summary and conclusions
We developed an erosion model that predicts the
erosion mechanism of water-insoluble biodegradable
polymers. The mechanism depends on the diffusivity of
water inside the matrix, the degradation rate of the
polymer and the matrix dimensions. A sufcient
increase of the ratio l=D or an increase of L can lead

to a change of the erosion mechanism from surface

erosion to bulk erosion. To support our model we tried
to change the erosion mechanism of bulk eroding PLA
and PLGA to surface erosion by acid and base catalysis.
Erosion of PLA and PLGA at pH 2 did not change the
erosion proles that were obtained during erosion at pH
7.4. This indicates, that autocatalysis is indeed an
important factor for the erosion of these polymers but
not the only one responsible for the odd bulk erosion
proles of PLA and PLGA polymer matrix discs. At
pH>13 in contrast, the erosion kinetics changed
tremendously. Linear erosion proles were obtained
similar to the ones of polymers that are considered to be
surface eroding such as polyanhydrides or polyorthoesters. The investigation of polymer molecular weight by
GPC revealed constant molecular weights throughout
erosion which is also indicative of surface erosion. SEM
pictures of PLA matrices eroded at pH>13 showed
erosion fronts similar to the ones that were observed in
polyanhydrides before.
Although the model allows for the calculation of
erosion numbers e; simplifying assumptions were necessary for developing it. In some cases this may introduce a
substantial variability. Thus tdiff is estimated based on
diffusion theory Eq. (1). The variation of tdiff will increase
for large times which affects the accuracy of e: Other
simplications are due to neglecting structural features of
the polymers such as crystallinity which was shown to
lead to different half lives for crystalline and amorphous
polymer areas [11]. Most important is polymer molecular
weight that does not only translate to longer degradation
times as estimated by the model (cf. Eq. (3)) but also to a
change of polymer properties that may have an effect on
water diffusivity or polymer degradation. The results of
Table 1 and Fig. 3 should, therefore, be interpreted with
the necessary circumspection.
Despite the simplications that were necessary to
develop the model, it allows to draw a number of
conclusions. Thus it shows that degradable polymers can
undergo surface erosion and bulk erosion depending on
the erosion conditions and the geometry. Below a critical
dimension Lcritical ; a polymer matrix will always undergo
bulk erosion while above Lcritical ; it will be a surface
eroding material. Despite the identication of these
parameters, the presented model provides still ample
room for improvement as important parameters such as
polymer swelling, the impact of molecular weight and
other that have not been taken into account so far.
However, the presented model may help us on our way
towards a better understanding of polymer erosion.

Acknowledgements
Thanks are due to Boehringer Ingelheim for providing PLA and PLGA. This work was funded by

F.V. Burkersroda et al. / Biomaterials 23 (2002) 42214231

the Deutsche Forschungsgemeinschaft with Grant GO

565/3-2.

References
[1] Benicewicz BC, Hopper PK. Polymers for absorbable surgical
sutures. J Bioact Compat Polym 1991;6(1):6494.
[2] Leenslang JW, Pennings AJ, Ruud RM, Rozema FR, Boering G.
Resorbable materials of poly(l-lactide), VI. Plates and screws for
internal fracture xation. Biomaterials 1987;8:703.
[3] Holland SJ, Jolly AM, Yasin M, Tighe BJ. Polymers for
biodegradable medical devices II. Hydroxybutyratehydroxyvalerate copolymers: hydrolytic degradation studies. Biomaterials
1987;8:28995.
[4] Heller J. Controlled drug release from poly(ortho esters). A
surface eroding polymer. J Controlled Release 1985;2:16777.
[5] Peppas NA, Langer R. New challenges in biomaterials. Science
1994;264(5162):10657.
[6] Holland SJ, Tighe BJ, Gould PL. Polymers for biodegradable
medical devices. 1. The potential of polyesters as controlled
macromolecular release systems. J Controlled Release 1986;4(3):
15580.
[7] Freed LE, Marquis JC, Nohria A, Emmanual J, Mikos AG,
Langer R. Neocartilage formation in vitro and in vivo using cells
cultured on synthetic biodegradable polymers. J Biomed Mater
Res 1993;27(1):1123.
[8] Reed AM, Gilding DK. Biodegradable polymers for use in
surgeryPGA/PLA homo and copolymers. Polymer 1981;22(4):
4948.
.
[9] Gopferich
A. Mechanism of polymer degradation and elimination. In: Domb AJ, Kost J, Wiseman DM, editors. Handbook of
biodegradable polymers. Amsterdam: Harwood Academic Publishers, 1997. p. 45171.
[10] Tamada J, Langer R. Erosion kinetics of hydrolytically degradable polymers. Proc Natl Acad Sci 1993;90(2):5526.
.
[11] Gopferich
A, Langer R. Modeling of polymer erosion. Macromolecules 1993;26:410512.
.
[12] Gopferich
A. Polymer bulk erosion. Macromolecules 1997;30(9):
2598604.
[13] Vert M, Christel P, Chabot F, Leray J. In: Hastings GW,
Ducheyne P, editors. Macromolecular materials. Boca Raton, FL:
CRC Press, 1984. p. 119 (Chapter 4).
[14] Li SM, Garreau H, Vert M. Structureproperty relationships in
the case of the degradation of massive aliphatic poly-(a-hydroxy
acids) in aqueous media, Part 1. J Mater Sci: Mater Med
1990;1:12330.

4231

[15] Li SM, Garreau H, Vert M. Structureproperty relationships in

the case of the degradation of massive aliphatic poly-(a-hydroxy
acids) in aqueous media, Part 3. J Mater Sci: Mater Med 1990;
1:198206.
[16] Li S, Girod-Holland S, Vert M. Hydrolytic degradation of
poly(d,l-lactic acid) in the presence of caffeine base. J Controlled
Release 1996;40:4153.
[17] Heller J. Control of polymer surface erosion by the use of
excipients. Polym Sci Technol 1986;34:35768.
[18] Vert M, Li SM, Garreau H. New insights on the degradation of
bioresorbable polymeric devices based on lactic and glycolic acids.
Clin Mater 1992;10:38.
[19] Atkins PW. Physical chemistry, NewYork: Freeman, 1990,
p. 770.
[20] Langer R, Peppas N. Chemical and physical structure of
polymers as carriers for controlled release of bioactive agents: a
review. J Macromol Sci Rev Macromol Chem Phys 1983;C23:
61126.
.
[21] Gopferich
A. Mechanisms of polymer degradation and erosion.
Biomaterials 1996;17:10314.
[22] Drake AW. Fundamentals of applied probability theory. New
York: McGraw Hill, 1988.
[23] Reiner M. The Deborah number. Phys Today 1964;1:62.
[24] Metzner AB, White JL, Denn MM. Constitutive equations for
viscoelastic uids for short deformation periods and for rapidly
changing ows: signicance of the Deborah number. AIChE J
1967;12:8636.
[25] Sykes PM. A guidebook to mechanism in organic chemistry, 4th
ed. London: Longman, 1975.
[26] Park TG. Degradation of poly(lactic-co-glycolic acid) microspheres: effect of copolymer composition. Biomaterials 1995;15:
112330.
[27] Li SM, Vert M. Hydrolytic degradation of the coral/poly(d,llactic acid) bioresorbable material. J Biomater Sci Polym Ed
1996;7(9):81727.
[28] Herrlinger M.PhD thesis, Heidelberg, 1994.
.
[29] Gopferich
A, Langer R. The inuence of microstructure and
monomer properties on the erosion mechanism of a class of
polyanhydrides. J Polym Sci Part A: Polym Chem 1993;31:
244558.
[30] Park K, Shalaby WSW, Park H. Biodegradable hydrogels for
drug delivery. Lancaster: Technomic Publications, 1993.
[31] Pitt CG, Gratzel MM, Kimmel JL, Surles J, Schindler A.
Aliphatic polyesters II. The degradation of poly(d,l-lactide),
poly(e-caprolactone), and their copolymers in vivo. Biomaterials
1982;2:21520.
[32] Crank J, Park GS. Diffusion in polymers. London: Academic
Press, 1968.