TYPESETTING SAMPLES
Standard TeX article style
<d id="100100621" t="none" n="621">
<issn></issn>
<cod>SEPARATION SCIENCE AND TECHNOLOGY,
35(12),pp. 1313–1828, 2000</cod>
<cau fn="ARIJIT" sn="BOSE"><af><od>DEPARTMENT
OF CHEMICAL ENGINEERING</od><org>UNIVERSITY OF
RHODE
ISLAND</org><cty>KINGSTON</cty><zip>02881</zip><st>RHODE
ISLAND</st><cny>USA</cny></af></cau>
<au fn="ALAZAR" mn="N."
sn="GHEBREMESKEL"><af><od>DEPARTMENT OF
CHEMICAL ENGINEERING</od><org>UNIVERSITY OF
RHODE
ISLAND</org><cty>KINGSTON</cty><zip>02881</zip><st>RHODE
ISLAND</st><cny>USA</cny></af></au>
<ti>A Flow-Through, Hybrid
Magnetic-Field-Gradient, Rotating Wall Device
For Magnetic Colloidal Separations</ti>
<re>August 19, 1999</re>
<red>January 2000</red>
<abs><p>A 2.0 cm internal diameter 1 m long,
axially-rotating horizontal glass tube, with
four axially located repeating hybrid magnetic
units, is used as part of a flow-through,
colloidal magnetic affinity separation device.
Each magnetic unit consists of an alternating
current solenoid surrounding the chamber
followed by four azimuthally distributed
permanent magnets that rotate with the
chamber. Experiments were carried out on a
model feed system consisting of a mixture of
1.0 μm diameter biotinylated latex beads
(targets) and 9.7 μm diameter
nonfunctionalized latex beads (nontargets) at
a 1:1 number ratio. Streptavidin labeled
magnetic particles (2.8 μm diameter
polystyrene—Dynabeads) at a number
concentration of ∼3 × 3 10<sp>6</sp>
bead/mL were used as the separation
vehicles. Two feed flow rates of 12 and 35
m/min were used until a total of 600 mL of
sample were processed for each. At the low
rate we achieved a capture efficiency of 60
± 4% and a separation factor of
18.2 ± 1.2 with 95 ± 4%
purity. For the higher flow rate the capture
efficiency was 40 ± 4% and the
separation factor was 18.6 ± 1.5 with
87 ± 4% purity. A single stage
batch process, where a total of 10 mL of feed
was processed at identical feed and magnetic
bead concentrations required ∼2 hours,
produced a capture efficiency of 42 ±
4% and a separation factor of 3.8
± 1 with 79 ± 4% purity.
Three batch stages (2 hours processing time
per stage) were required to surpass the
capture efficiency of the flow-through device
at the smaller flow rate. Thus, this
flow-through separation device can lead to a
large increase in processing volume and
reduced “down” time, compared to
current batch processes, without loss of
either efficiency or purity, potentially
opening up magnetic colloidal separations for
large scale applications.</p></abs>
<h1><ti>1. INTRODUCTION</ti>
<p>Separation schemes where colloidal
particles serve as the affinity substrates for
targets in solution offer several important
features that can be exploited for different
applications. These include the large active
surface area/volume (this quantity scales
as ∼1/<tex>$D$</tex> where
<tex>$D$</tex> is the particle diameter), the
ability of the particles to remain in
suspension by Brownian motion facilitating
good mixing, and the potential to selectively
mobilize target-bound particles to specific
locations where they can be isolated for
further processing. In colloidal magnetic
affinity separation schemes, the substrates
consist of ligand-coated superparamagnetic
particles. Highly specific linkages between
these surface-functionalized particles and
target materials (either native or also
surface-derivatized) are used to
preferentially magnetize the targets. Steady
magnetic field gradients are then employed to
immobilize and then isolate these targets. The
paramagnetic nature of the magnetic particles
means that they can be resuspended upon
removal of the external field and reused if
necessary. This simple strategy has been used
for cell separations (<lk
id="1|2|3|4|5"></lk>), other biological
macromolecules (<lk id="6|7"></lk>), and for
metals (<lk id="8|9|10"></lk>). Surfaces of
particles containing magnetic cores can be
derivatized with a large repertoire of
functional groups or specific adsorbents,
making this idea potentially feasible for many
unexplored applications.</p>
<p>However, several important limitations of
currently available technology have restricted
the applicability of colloidal magnetic
separation. These include inadequate
specificity, often caused by inadequate mixing
(for example, in diffusion-limited situations
where the nontargets are smaller than the
target macromolecules, the nontargets have
higher mobility and thus have more contact
with the active surfaces on the magnetic
particles, creating problems if there is
nonspecific binding) and the long time
necessary to achieve the required degree of
separation if target viability has to be
maintained, because agitation can produce
damaging shear forces. Since magnetic
particles have specific gravities that are
significantly larger than water or aqueous
salt solutions, they have a tendency to
sediment, and must be kept suspended by
Brownian motion. This severely restricts their
size, and, because the magnetic susceptibility
scales with particle volume, requires use of
high magnetic field gradients to mobilize them
through the surrounding liquid phase at
practically useful speeds. Such gradients can
be produced inexpensively by introducing fine
diameter soft magnetic material, such as steel
wool or nickel spheres, within the solution
and exposing the whole sample to a uniform
magnetic field (<lk id="11"></lk>). However,
the introduction of additional surfaces in the
solution increases the potential for
nonspecific binding and increases the
possibility of undesirable particle trapping.
Furthermore, most existing devices operate in
the batch mode. This limits throughput and
leads to large amounts of down time. The
economics of this procedure have made it
useful only for very high value products and
processes such as cell sorting, DNA
purification, protein capture, and
microorganism isolation (<lk
id="12|13"></lk>). Technological advances that
speed up this process without a concomitant
loss of target specificity can make a
significant impact to this burgeoning area. We
have built and completed preliminary testing
on a new flow-through, multiunit device that
reduces separation time and increases the
sample volume by two orders of magnitude
without any loss of capture efficiency, target
purity, and separation efficiency over an
equivalent batch process. The essential
features of the device are described in
Section <lk id="s2"></lk> Section <lk
id="s3"></lk> describes the experimental
procedure, while the key results are described
in Section <lk id="s4"></lk>.</p></h1>
<h1><ti>2. SEPARATION DEVICE</ti>
<p>The separation device is shown
schematically in Fig. <lk id="f1"></lk>. It
consists of a 2.0 cm internal diameter, 1.0 m
long glass tube with four repeating magnetic
units. Each unit consists of a stationary
alternating current solenoid that surrounds
the tube, followed by two pairs of 1 kGauss
Al-Ni-Co magnets positioned azimuthally on the
tube at a distance of 2.0 cm from the end of
the solenoid. Each permanent magnet pair
consists of magnets at diametrically opposite
ends, the second pair being located 1.0 cm
downstream from the first, and positioned at
90° to the first. The tube axis is
horizontal, and the chamber rotates about this
axis at ∼25 rpm. The slow rotation
simulates a low-gravity environment within the
chamber and significantly reduces
sedimentation of nonneutrally buoyant
particles without introducing centrifugal
forces, a critical feature of this device. The
permanent magnets rotate with the tube.
Separate peristaltic pumps drive the feed
mixture and the magnetic colloid suspension
through a rotary coupler (Deublin Inc.) into
one end of the chamber. A second rotary
coupler at the other end of the chamber allows
the exiting liquid to flow into a stationary
collection vessel.</p>
<fig><n>1</n><ti>Schematic representation of
the 2.0 cm internal diameter 1 m long axially
rotating horizontal chamber with four
repeating magnetic units. Each unit consists
of a stationary alternating current solenoid
surrounding the chamber followed by two pairs
of azimuthally distributed permanent magnets
that rotate with the chamber.</ti></fig>
<p>Each solenoid has 14 turns of copper wire
over a length of 2.5 cm has a diameter of 2.5
cm, and carries an alternating current of
amplitude 10 A at a frequency of 60 Hz. As the
particles and feed mixture flow through the
chamber, they are first acted upon by the
magnetic field gradient produced by the
solenoid. For a magnetic particle entering the
solenoid, the axial component of the magnetic
field gradient varies in magnitude as the
current changes but points in the same
direction as the base flow. Along the axis of
a solenoid of radius <tex>$R$</tex> and length
<tex>$L$</tex>, this field <tex>$B$</tex> is
given by
<dtex>
\begin{eqnarray}
B(x) = \mu_0NI[\{x + L/2\}\{R^2 + (x +
L/2)^2\}^{-1/2}\\
- \{x - L/2\} \{R^2 + (x - L/2)^2\}^{-1/2}]/2L
\end{eqnarray}
</dtex>
where <tex>$x$</tex> is measured from the
center of the solenoid, <tex>$N$</tex> is the
number of turns, <tex>$I$</tex> is the
current, and μ<sb>>0</sb> is the
permeability of free space. The local field
and field-gradient impart a time-varying axial
force on the paramagnetic 2.8 μm diameter
Dynabead particles, inducing particle motion
through the surrounding liquid beyond that
produced by the base flow. As an illustration
of this effect, the solenoid-current-induced
transient velocities for a magnetic bead
positioned at the solenoid entrance <tex>$x =
-L/2$</tex> and on the axis, calculated using
Stokes‘ law (particle magnetization data
provided by Dynal Corp.) and Eq. (<lk
id="fr1"></lk>) are shown in Fig. <lk
id="f2"></lk>. The maximum induced velocity is
∼60 μm/s. As the particle moves
toward the center of the solenoid, the axial
magnetic force reduces. Past the midpoint, the
direction of the time-varying axial magnetic
field gradient reverses, creating a particle
velocity opposite to the base flow. Because
the lengt/diameter ratio of the solenoid
is ∼1, fringing effects dominate and no
location within the solenoid has a uniform
axial magnetic field, as shown in Fig. <lk
id="f3"></lk>. This minimizes magnetically
“dead” regions within the
separation chamber. The radial component of
the magnetic field within the solenoid is also
nonuniform. Thus each particle also
experiences a time-varying radial force that
results in local radial motion. In addition,
the time varying magnetic field induces an
oscillating torque on each particle. The
expected consequence of the alternating
current in the solenoid is a transient motion
of the particles superimposed on a steady one
imposed by the base flow, thus enhancing the
mixing within the chamber without introducing
additional surfaces within it. (<i>Note</i>:
The base flow has a nonzero vorticity which
adds to rotary motion of the particles.)</p>
<fig><n>2</n><ti>Axial velocities induced by
the alternating current for a magnetic
particle located at the entrace of the
solenoid and on the axid of the chamber,
calculated using Stokes’ law and
particle magnetization data provided by the
manufacture. the induced velocities would
differ in direction at the solenoid exit.
These velocities, along with a radial
component not shown as well as a
torque-induced rotary oscillation, superimpose
on the particle motion created by the base
flow and result in micromixing.</ti></fig>
<fig><n>3</n><ti>Magenetic field versus
position in the solenoid. Solenoid diameter
2.5 cm, length 2.5 cm, current 10 A. No
location within the solenoid has a uniform
axial magnetic field, thus excluding
magnetically dead regions from the
chamber.</ti></fig>
<p>In the absence of any magnetic forces, the
residence time for particles in the chamber
for the flow rates used is of the order of a
few minutes. The permanent magnet strength
must be enough to permit target particles to
move a distance equal to the tube radius in a
time that is short compared to this residence
time. Using Stokes’ law, a force of
∼1.6 ×
10<sp>−3</sp><tex>$D$</tex> dynes is
needed to move a particle of diameter
<tex>$D$</tex> cm at a radial velocity of 0.1
cm/s (this would mean 10 seconds for a
particle at the axis to reach the wall)
through a liquid of waterlike viscosity. For
the 2.8 μm Dynabeads used in our
experiments, the magnetic field gradient
required to create this force is ∼0.5
kGauss/cm. Figure <lk id="f4"></lk> shows
the experimentally measured field as a
function of radial position along the diameter
connecting two facing Al-Ni-Co permanent
magnets, and demonstrates that these magnets
are strong enough to move the particles to the
wall well within the required time.</p>
<fig><n>4</n><ti>Magnetic field versus radial
position for the A1-Ni-Co permanent magnets
arranged at diametrically opposites ends of
the chamber. The field gradients is ∼ 1
kGauss/cm.</ti></fig>
<p>The trajectory taken by a magnetic particle
entering the device is affected by the
geometry of the system, including the
placement of the solenoids and permanent
magnets and the direction of the flow. Figure
<lk id="f5"></lk> is a qualitative
representation of the forces acting on a
particle as it moves through the device.
(<i>Note</i>: Gravity and centrifugal forces
are ignored because of the slow rotation of
the device about a horizontal axis.) In Region
A, the particle is far from the influence of
the solenoid or the permanent magnets. It
moves along a streamline at a constant
velocity corresponding to the base flow, so
that there are no forces acting on it. In
Region B, still far from the influence of the
permanent magnets, the particle experiences an
axial force created by the solenoid and a drag
force in the opposite direction. In Region C,
the influence of the permanent magnets becomes
important, and the effect of the solenoid is
negligible. The particle then experiences a
radial force, which is opposed by drag. It is
only in Region C that the particle trajectory
deviates from the streamline, allowing capture
at the tube wall.</p>
<fig><n>5</n><ti>The forces acting on a
magnetic particle as it moves through the
device. <i>Region A</i>: This region is
“far“ from the influence of the
solenoid and the permanent magnets. The
particle moves along a streamline at a
constant velocity set by the base flow, and no
forces ct on it. (<i>Note</i>: The <tex>$\sim
m/\mu D,$</tex> where <tex>$m$</tex> is the
mass of the particle, <tex>$D$</tex> is its
diameter, and μ the liquid viscosity. This
is of the order of a few microseconds for a
Dynal magnetis particle.) <i>Region B</i>:
Here the particle experiences an axial force
created by the solenoid, and an opposing drag
force. It is not influenced by the permanent
magnets. <i>Region C</i>: The permanent
magnets induce a radial magnetic force. An
opposing drag force is also created. Thsi
region is far from the solenoid. Particles
leave the streamlines in Region
<tex>$C$</tex>, and are captured on the tube
walls.</ti></fig></h1>
<h1><ti>3. EXPERIMENTAL PROCEDURE</ti>
<p>M-280 (2.8 μm diameter) streptavidin
coated magnetic beads were obtained from Dynal
Inc. These beads have a specific gravity of
1.3, so they would sediment rapidly if left
unperturbed in an aqueous solution. The target
particles were 1 μm diameter biotin-labeled
polystyrene beads (Sigma Chemical Company)
while the nontarget particles were 9.7 μm
diameter nonfunctionalized but
charge-stabilized polystyrene particles
(Interfacial Dynamics Corporation). Both the
target and nontarget particles are essentially
neutrally buoyant. Single distilled water was
passed through a four cartridge Millipore
“Mill Q” system until its
resistivity reached 18 MΩ-cm. This water
was used for preparing all the
suspensions.</p>
<p>The streptavidin beads were used at a
particle number concentration of ∼3
× 10<sp>6</sp> bead/mL. All particle
concentrations are measured in a hemocytometer
mounted on a Nikon optical microscope. The
feed consisted of biotinylated polystyrene
beads mixed in a 1:1 number ratio with the
nonfunctionalized beads at an overall particle
number concentration of ∼6 ×
10<sp>4</sp>/mL. 100 mL of the magnetic
beads and an equal volume of a sample
containing the target and nontarget material
were fed simultaneously at two different flow
rates specified below. The target and
nontarget particles were sufficiently
different in size so that they could be easily
distinguished using optical microscopy.</p>
<p>In a typical experiment (some control
experiments were also performed, where the
chamber was not rotated, or there was no
current in the solenoid), the chamber is first
filled with distilled water using the
peristaltic pumps. The feed and magnetic
particle flows are then initiated, along with
chamber rotation and current in the solenoids.
The liquid flowing out through the end of the
chamber, called the supernatant, is collected
continuously. When the appropriate liquid
volume is collected (usually 600 mL), the flow
of the feed and magnetic particle suspensions
is interrupted and the total volume of feed
that has entered the chamber is recorded. The
right end of the chamber is then opened, and
the remaining supernatant allowed to flow out
and is added to that already collected. The
chamber is then closed, and 500 mL of
distilled water is pumped through, flushing
out all the particles in the supernatant. This
material is also added to the already
collected liquid. The number concentrations of
target particles in the feed and the
supernatant are multiplied by the total feed
and supernatant volumes to obtain the number
of particles in the feed (<tex>$N_{\rm
{fT}}$</tex>) and in the supernatant
(<tex>$N_{\rm ST}$</tex>). Similarly, the
number of nontarget particles in the feed and
supernatant, <tex>$N_{\rm {fN}}$</tex> and
<tex>$N{\rm {SN}}$</tex> respectively, are
obtained. The particle capture
efficiency,η , is evaluated using
<dtex>
\begin{equation}
\eta = 100(N_{\rm fT} - N_{\rm ST})/N_{\rm fT}
\end{equation}
</dtex></p>
<p>The permanent magnets are then removed, and
distilled water is allowed to flow through the
chamber. The magnetic particl/target
complexes that had been immobilized at the
chamber walls are now resuspended into the
chamber, and driven out from the other end by
the bulk flow. The magnetic-particle-rich
solution collected in this way is designated
as the suspension from the pole region. The
exact counting of target particles from this
suspension proved difficult because it was
hard to distinctly identify individual ones
when they were clustered around the magnetic
beads. Thus an indirect, mass balance approach
was used, with the target particle number
percent at the pole region, <tex>$X_T$</tex>,
being given by
<dtex>
\begin{equation}
X_{\rm T} = 100(N_{\rm ft} - N_{\rm
ST})/{N_{\rm ft} - N_{\rm ST}) + (N_{\rm fN} -
N_{\rm SN}}
\end{equation}
</dtex>
The number percent of targets in the
supernatant, <tex>$Y_T$</tex>, is obtained
directly from
<dtex>
\begin{equation}
Y_{\rm T} = 100N_{\rm ST}/(N_{\rm ST} + N_{\rm
SN})
\end{equation}
</dtex>
The quantity <tex>$X_{\rm T}$</tex> is a
measure of the purity of the target material
at the poles. The separation factor, β,
is then calculated using
<dtex>
\begin{equation}
\beta = \{X_{\rm T}/(100 - X_{\rm
T})\}/\{Y_{\rm T}/(100 -Y_{\rm T})\}
\end{equation}
</dtex>
The numbers reported represent an average from
five experiments, with five samples withdrawn
from each region for each experiment. Clearly
β must be different from 1 for the
separation to be successful.</p></h1>
<h1><ti>4. RESULTS AND DISCUSSION</ti>
<p>Experiments were performed to confirm that
the rotation of the chamber and the
alternating current in the solenoid are indeed
crucial for the separation. Two overall feed
flow rates were used: 12 and 35 mL/min,
and a total of 600 mL of sample was processed.
The capture efficiencies for all of the
experimental conditions are shown in Fig. <lk
id="f6"></lk>. For the conditions probed in
these experiments, the first magnetic unit was
located 40.0 cm downstream from the entrance
of the chamber. In the experiment with no
chamber rotation, nearly all of the magnetic
particles sedimented before arriving at the
first magnetic unit, while most of the target
particles exited through the end of the
chamber, leading to the extremely low capture
efficiency. Introduction of the alternating
current in the solenoid enhances the capture
efficiencies, nearly doubling it for the lower
flow rate, and increasing it by a factor of
50% for the higher flow rate. A
dramatic increase in efficiency is observed
for both flow rates when rotation is
initiated, clearly pointing to the important
consequence of keeping the magnetic particles
suspended in solution and promoting contact
with the neutrally-buoyant targets. When both
the mixing caused by the current in the
solenoid as well as chamber rotation are
included, the average capture efficiencies
reach 40 and 60% for the higher and
lower flow rates, respectively.</p>
<fig><n>6</n><ti>Capture effeciencies for the
four-magnetic unit device at two feed flow
rates, 12 and 45 mL/min. the enhancements
produced by current flowing in the solenoids
as well chamber roation are
apparent.</ti></fig>
<p>The importance of each magnetic unit toward
the overall capture efficiency was examined by
starting with one (consisting of the solenoid
and four permanent magnets) and sequentially
adding the others. The results, for a flow
rate of 12 mL/min are shown in Fig. <lk
id="f7"></lk>. One unit gives a capture
efficiency of 22%. Each additional unit
produced a further separation of the target
molecules, up to a level of 60% when
all four are in place. These capture
efficiencies are not sensitive to the exact
positions of each of the magnetic units,
indicating that chamber rotation is effective
in keeping the magnetic particles in
suspension.</p>
<fig><n>7</n><ti>Capture efficiencies versus
total number of magnetic units at a feed flow
rate of 12 mL/min.</ti></fig>
<p>At this flow rate the separation factors
β as each of the repeating units is added
are shown in Fig. <lk id="f8"></lk>. One unit
produces a separation factor ∼3.4. As the
other units are added, the monotonic drop in
<tex>$Y_{\rm T}$</tex> causes the separation
factor to rise systematically to an average
final value of 18.2. The average separation
factor achieved at the higher flow rate was
18.6. Note that this device is equivalent to a
single-stage unit from the perspective of a
cocurrent separation scheme. This dramatically
high separation factor can clearly be
exploited in a multistage cascade, each stage
consisting of the chamber described here.
Furthermore, the judicious use of reflux can
provide significant additional benefits for
separation.</p>
<fig><n>8</n><ti>Separation factor versus
total number of magnetic units for a feed flow
rate of 12 mL/min. The drop in the
particle number concentration in the
supernatant is responsible for the rise in
separation factor as each unit is
added.</ti></fig>
<p>We obtained the purity of the target
material at the pole region (number
concentration of targets divided by the total
number concentration of target and nontarget)
for each flow rate studied. Figure <lk
id="f9"></lk> shows results for the lower flow
rate—a 95% purity for the model
system used in this study. At the higher flow
rate the purity was 81%. As discussed
later, the concentration of nontarget
particles in the pole regions is not caused by
nonspecific binding, but rather is a physical
particle trapping effect.</p>
<fig><n>9</n><ti>Target purity versus total
number of units for a feed flow rate of 12
mL/min. The purity of the target material
is not affected by the number of magnetic
units and remains at
∼95%.</ti></fig>
<p>The efficacy of this flow-through device
can be evaluated by comparing its performance
with a batch processing scheme, shown
schematically in Fig. <lk id="f10"></lk>. To
make a meaningful comparison, the processing
was done in a vial of diameter 2.0 cm, and the
same set of Al-Ni-Co permanent magnets was
used. All concentrations were similar to those
utilized for the flow-through experiments in
our device. Thus, the biotinylated polystyrene
beads were mixed in a 1:1 number ratio with
nonfunctionalized beads and ∼3 ×
10<sp>6</sp> bead/mL of streptavidin
coated magnetic beads at an overall particle
number concentration of ∼6 ×
10<sp>4</sp>/mL. The total sample volume
was 10 mL. This volume limit was dictated by
the amount of time required for the magnetic
particles to collect at the poles after
application of the field gradient. The
suspension is shaken continuously for 1 hour,
then exposed to the permanent magnets for 45
minutes. Placement of the magnet near the top
of the tube moves the magnetized target
particles, concentrating them at the poles,
while the supernatant region contains the
nontarget particles. Samples are then
withdrawn from both the supernatant as well as
the pole region, and target and nontarget
particle concentrations counted in a
hemocytometer. Figure <lk id="f11"></lk> shows
the capture efficiencies for each stage in a
three-stage scheme. A single batch stage gives
a capture efficiency of 42%, far lower
than the 60% efficiency in our device
at a flow rate of 12 m/min and comparable
to that obtained in our device for the higher
flow rate. The separation factor β for a
single batch stage is 3.4, much lower than the
factor of ~18 in our device. In addition, the
purity obtained here is 79%, also
comparable to that in our flow-through device.
Note that we needed three batch stages, each
requiring ∼2 hours of operation, to get
the capture efficiency beyond that in our
device at the lower flow rate.</p>
<fig><n>10</n><ti>The arrangement used in the
laboratory for applying the magnetic field
gradient for multistage batch processing.
Magnetic particles are added to the test tube
containing the mixture of target and hontarget
molecules, and shaken continuously for ∼ 1
hour. A pair of the A1-Ni-Co magnets are then
placed near the top of the tube, and the
magnetized target particles are allowed to
concentrate at the poles for 45 minutes. The
supernatant contains most of the nontarget
particles. Samples are then withdrawn from
both the supernatant as well as the pole
region for analysis.</ti></fig>
<fig><n>11</n><ti>Capture effeciencies versus
number of stages for the batch process. Note
that three batch stages are required to
surpass the capture efficiency of our device
at a flow rate of 12 mL/min.</ti></fig>
<p>A part of the difference between the
performance of our device and a batch process
can be understood by examining the material
that collects at the poles for each. These are
shown in Fig. <lk id="f12"></lk>. The cluster
size for the flow-through device is routinely
smaller than that in batch units. This is a
consequence of particle motion created by the
flow, as well as the magnetic field gradients.
The smaller clusters lead to a larger
available surface area per unit volume, and
lead to better capture efficiencies.</p>
<fig><n>12</n><ti>Optical micrograph showing
clusters from the pole region in (a) the batch
device and (b) the flow-through device. The
cluster sizes for the flow-through device are
smaller, allowing more surface area to be
available for target attachment.</ti></fig>
<p>In order to examine the extent of
nonspecific binding, some control batch
experiments were conducted. The nontarget
particles were incubated with the streptavidin
beads for 2 hours. The Al-Ni-Co magnets were
then used to concentrate the magnetic
particles, samples were withdrawn from this
concentrated region, gently diluted, and
examined by optical microscopy. While an
extremely small number of nontarget particles
were detected, none were attached to the
magnetic beads. Thus we see no experimental
evidence of nonspecific binding. The loss of
purity is therefore a consequence of trapping
of the nontarget particles by clusters of the
magnetic beads as they travel toward the poles
of the permanent magnets, a physical rather
than a chemical effect.</p>
<p>Our experimental results using a model
system indicate that this flow-through
multiunit separation device can lead to a
large increase in processing volume and
reduced “down” time compared to
current batch processes, without a significant
reduction in either efficiency or purity,
potentially opening up magnetic colloidal
separations for medium and large-scale
applications. The mode of operation currently
demonstrated would apply directly to a
negative selection strategy, where the
nontarget particles are the ones of interest
and are collected primarily in the
supernatant, or the target particles are
contaminants that need to be removed from a
large volume of solution.</p></h1>
<h1><ti>5. CONCLUSIONS</ti>
<p>A new flow-through, multimagnetic-unit
device, consisting of a slowly rotating
horizontal chamber, has been designed and
demonstrated for colloidal magnetic affinity
separation. Each magnetic unit consists of an
alternating-current-carrying solenoid
surrounding the chamber, and two pairs of
permanent magnets located downstream from the
solenoid, that rotate with the chamber. The
chamber rotation simulates a low gravity
environment, severely attenuating any
sedimentation of nonneutrally buoyant magnetic
particles as well as feed, thus promoting good
particleŠtarget contact throughout the chamber
volume. The oscillating magnetic field
gradient produced by the solenoid introduces
translational and rotary microparticle
oscillations, enhancing mixing, while the
permanent magnets immobilize the targets on
the chamber walls. For a model feed system
consisting of a ∼50% mixture of
biotinylated latex beads (target) and
nonfunctionalized latex beads (nontarget), we
have been able to achieve a maximum separation
capture efficiency of 60% and a
separation factor of ∼18.2 with purity as
high as 95%. The total feed volume we
processed was 600 mL at a flow rate of 12
mL/min. At a flow rate of 35 m/min the
capture efficiency was reduced to 40%,
but without a significant change in the
separation factor or purity. A single-stage
batch process using the same particle
concentrations required 2 hours, and gave a
capture efficiency of 42%, a separation
factor of 3.8, and a purity of 80%.
Three batch stages were required to surpass
the low flow rate capture efficiency of our
device. These results show the potential of
this device for magnetic colloidal separation
at medium and large scales.</p></h1>
<ack><p>Financial support for this work was
provided by the National Science Foundation
(CTS 9618635) and the University of Rhode
Island Foundation. We thank A. C. Nunes for
several useful discussions.</p></ack>
<bib>
<bic t="p"><au sn="Richards" fn="A."
mn="J."></au><au sn="Roath" fn="O."
mn="S."></au><au sn="Smith" fn="R."
mn="J.S."></au><au sn="Watson" fn="J."
mn="H.P."></au><ti></ti><jrl><i>IEEE Trans.
Magn.</i></jrl><yr>1996</yr><v><i>32</i></v><pn>460</pn></bic>
<bic t="p"><au sn="Bahaj" fn="A."
mn="S."></au><au sn="James" fn="P."
mn="A.B"></au><au sn="Moeschler" fn="F."
mn="D."></au><ti></ti><jrl><i>Ibid</i></jrl><yr>1996</yr><v><i>32</i></v>
<pn>5106</pn></bic>
<bic t="p"><au sn="Van der Wolf" fn="J."
mn="M."></au><au sn="Van Beckhoven " fn="R."
mn="C.M."></au><au sn="deVries" fn="Ph."
mn="M."></au><au sn="Van Vurrde" fn="J."
mn="W.L."></au><ti></ti><jrl><i>J. Appl.
Bacteriol.</i></jrl><yr>1994</yr><v><i>77</i></v><pn>160</pn></bic>
<bic t="p"><au sn="Thomas" fn="T."
mn="E."></au><au sn="Lansdrop" fn="P."
mn="M."></au><ti></ti><jrl><i>Prog. Clin.
Biol.
Res.</i></jrl><yr>1994</yr><v><i>389</i></v><pn>65</pn></bic>
<bic t="p"><au sn="Chalmers" fn="J."
mn="J."></au><au sn="Zborowski"
fn="M."></au><au sn="Sun" fn="L."></au><au
sn="Moore" fn="L."
mn="R."></au><ti></ti><jrl><i>Biotechnol.
Prog.</i></jrl><yr>1998</yr><v><i>14</i></v><pn>141</pn></bic>
<bic t="p"><au sn="SafaríK"
fn="I."></au><au sn="Safaríkov´"
fn="M."></au><au sn="Forsythe" fn="S."
mn="J."></au><ti></ti><jrl><i>J. Appl.
Bacteriol.</i></jrl><yr>1995</yr><v><i>78</i></v><pn>575</pn></bic>
<bic t="p"><au sn="Kapperud" fn="G."></au><au
sn="Vardunt" fn="T."></au><au sn="Skjerve"
fn="E."></au><au sn="Hornes" fn="E."></au><au
sn="Michaelsen" fn="T."
mn="E."></au><ti></ti><jrl><i>Appl. Environ.
Microbiol.</i></jrl><yr>1993</yr><v><i>59</i></v><pn>2938</pn></bic>
<bic t="p"><au sn="Ebner" fn="A."
mn="D."></au><au sn="Ritter" fn="J."
mn="A."></au><au sn="Ploehn" fn="H."
mn="J."></au><au sn="Kochen" fn="R."
mn="L."></au><au sn="Navratil" fn="J."
mn="D."></au><ti></ti><jrl><i>Sep. Sci.
Technol.</i></jrl><v><i>34</i></v><pn>1277</pn></bic>
<bic t="p"><au sn="Nuñez"
fn="L."></au><au sn="Kaminski" fn="M."
mn="D."></au><ti></ti><jrl><i>Chemtech</i></jrl><yr>1998</yr>
<v><i>28</i></v><pn>41</pn></bic>
<bic t="p"><au sn="Nuñez"
fn="L."></au><au sn="Bucholz" fn="B."
mn="A."></au><au sn="Vandergrift" fn="G."
mn="F."></au><i><jrl>Sep. Sci.
Technol.</i></jrl><yr>1995</yr><v><i>30</i></v><pn>1455</pn></bic>
<bic t="b"><au sn="Watson" fn="J."
mn="H.P."></au><sti>Solid-Liquid
Separation:</sti><ti>High Gradient Magnetic
Separation</ti><edr sn="Sarovsky"
fn="L."></edr><n>3</n><pbl><nm>Butterworth-Heinemann</nm><add><cty>Oxford</cty><st></st></add>
<yr>1990</yr></pbl><pn></pn></bic>
<bic t="p"><au sn="Haukanes" fn="B."
mn="I."></au><au
sn="Kvam"></au><ti></ti><jrl><i>Bio/Technology</i></jrl>
<yr>1993</yr><v><i>11</i></v><pn>60</pn></bic>
<bic t="p"><au sn="Olsvik" fn="O."></au><au
sn="Popovic" fn="T."></au><au sn="Skjerve"
fn="E."></au><au sn="Cudjoe" fn="K."></au><au
sn="Hornes" fn="E."></au><au sn="Ugelstad"
fn="J."></au><au sn="Uhlen"
fn="M."></au><ti></ti><jrl><i>Clin. Microbiol.
Rev.</i></jrl><yr>1994</yr><v><i>7</i></v><pn>43</pn></bic>
</bib>
</d>
Standard LaTeX article style
\documentclass[10pt,]{article}
\setlength{\textwidth}{26pc}
\font\dropletterfont=cmr10 at 26pt
\def\cornorbox#1#2#3{\setbox1=\hbox{#1}
\dp1=1pt\par \hangindent\wd3
\hangafter-#2\hskip-1.5\wd1\raise#3\box1}
\def\dropletter#1{\cornorbox{\vtop{\kern4pt\hbox{\dropletterfont#1\kern2pt}}}{2}{9pt}}
\font\abs = cmr8
\begin{document}
\begin{center}
What Do We Know About Variance in Accounting
Profitability?\\
\bigskip
Anita M. McGahan \\
{\it Boston University School of Management},
\\{\it 595
Commonwealth Avenue},\\
{\it Boston, Massachusetts 02215}\\
{\it amcgahan@bu.edu}\\
\bigskip
Michael E. Porter\\
{\it Harvard Business School},\\
{\it Soldiers Field},\\
{\it Boston, Massachusetts 02163}\\
{\it mporter@hbs.edu}\\
\bigskip
\end{center}
\hrule
\begin{quotation}
{\abs{\dropletter{I}n this paper, we analyze
the variance of
accounting profitability $~~~~$among a broad
cross-section of
firms in the American economy from 1981--1994.
The purpose of the
analysis is to identify the importance of
year, industry,
corporate-parent, and business-specific
effects on accounting
profitability among operating businesses
across sectors. The
findings indicate that industry and
corporate-parent effects are
important and related to one another. As
expected,
business-specific effects, which arise from
competitive
positioning and other factors, have a large
influence on
performance. The analysis reconciles the
results of previous
studies by exploring differences in method and
data. We also
identify the broad contributions and
limitations of the research,
and suggest avenues for further study. New
approaches are
necessary to generate significant insights
about the relationships
between industry, corporate-parent, and
business influences on
firm profitability.}}
\end{quotation}
\hrule
\bigskip
\leftline{(Performance; Sustainability;
Industry Structure;
Corporate Strategy)}
\newpage
\begin{center}
{\bf MCGAHAN AND PORTER}\\
{\it Variance in Accounting
Profitability}
\end{center}
\leftline{{\bf 1. Introduction}}
Researchers in the economics and strategy
fields have long been
interested in understanding the determinants
of firm
profitability. During the 1960s and 1970s, a
large empirical
literature in industrial organization employed
cross-sectional
regression analysis to explain firm
performance based on industry
characteristics, including seller
concentration, advertising, and
R \& D intensity. The aim was to explore the
relationship between
structural entry barriers, tacit collusion,
and industry
performance. These studies were challenged in
the 1980s because
they tended to assume that industry structure
is fixed
independently of firm performance. In a review
of the literature,
Schmalensee (1989) reinterpreted the
structure-performance
findings as descriptive of empirical
regularities rather than as
conclusive evidence of causal relationships.
Viewed in this way,
the literature of the 1960s and 1970s on firm
performance
generated important insights about the
variation in accounting
profitability.
\bigskip
Partly in response to the limits of the
early research, a new style of work emerged in
the 1980s. This new
approach, pioneered by Schmalensee (1985),
decomposed the variance
in profitability across business segments into
components
associated with year, industry, the
corporate-parent, and
business-specific effects.\footnote{During the
mid-l980s,
questions were raised about the information
contained in
accounting returns about real economic
activity. The classic
expression of concern by Fisher and McGowan
(1983) emphasized that
accounting returns do not capture the net
present value of all
returns on investment. A famous debate, which
included comments by
Horowitz (1984), Long and Ravenscraft (1984),
Martin (1984), Van
Breda (1984), and a reply by Fisher (1984),
raised questions about
whether or not accounting rates of return
reflect monopoly rents
and whether or not booked assets are fairly
depreciated. In this
study, we investigate the importance of year,
industry,
business-specific, and corporate-parent
effects on accounting
profitability, but do not address the sources
of the effects.}
Over the past dozen years, several studies in
this research stream
have explored profit variance (Rumelt 1991,
Roquebert et al. 1996,
McGahan and Porter 1997a, yielding somewhat
different conclusions.
The purpose of these studies was to describe
the importance of
industry, corporate-parent, and business
influences on
profitability. Despite the advances associated
with the new
approach, the generality of results has been
limited by breadth of
data and statistical limitations.
\bigskip
This first objective of this study is to
reconcile results from
various studies in the recent
\bigskip
The authors are grateful to two anonymous
referees, Rebecca
Henderson, Jan Rivkin, Richard Rumelt, Richard
Schmalensee,
participants in the NBER Productivity group,
and attendees at the
Academy of Management meetings for comments
and discussions
related to this paper. Special thanks to
Arthur Schleifer for
extensive discussions about statistical
methods. Thanks to Todd
Eckler, Dan Elfenbein, Lucia Marshall, Michael
Susanto, Geoff
Verter, Sarah Woolverton, and especially Jan
Rivkin for help in
compiling the data. The Division of Research
at the Harvard
Graduate School of Business Administration
provided financial
support for this project. The first author
thanks the SRC and
BUILDE at Boston University for generous
research support.
\end{document}