Bending instability of electrically charged liquid jets of polymer solutions in

electrospinning
Darrell H. Reneker, Alexander L. Yarin, Hao Fong, and Sureeporn Koombhongse

Citation: J. Appl. Phys. 87, 4531 (2000); doi: 10.1063/1.373532

View online: http://dx.doi.org/10.1063/1.373532
View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v87/i9
Published by the American Institute of Physics.

Related Articles
Kinetic energy budgets near the turbulent/nonturbulent interface in jets
Phys. Fluids 25, 015114 (2013)
Preferential states of rotating turbulent flows in a square container with a step topography
Phys. Fluids 25, 015109 (2013)
Large-eddy simulation of highly underexpanded transient gas jets
Phys. Fluids 25, 016101 (2013)
The generation of zonal jets by large-scale mixing
Phys. Fluids 24, 126601 (2012)
Effects of passive control rings positioned in the shear layer and potential core of a turbulent round jet
Phys. Fluids 24, 115103 (2012)

Additional information on J. Appl. Phys.

Journal Homepage: http://jap.aip.org/
Journal Information: http://jap.aip.org/about/about_the_journal
Top downloads: http://jap.aip.org/features/most_downloaded
Information for Authors: http://jap.aip.org/authors

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
JOURNAL OF APPLIED PHYSICS VOLUME 87, NUMBER 9 1 MAY 2000

Bending instability of electrically charged liquid jets of polymer solutions

in electrospinning
Darrell H. Renekera)
Department of Polymer Science, The University of Akron, Akron, Ohio 44325-3909
Alexander L. Yarin
Technion-Israel Institute of Technology, Haifa 32000, Israel
Hao Fong and Sureeporn Koombhongse
Department of Polymer Science, The University of Akron, Akron, Ohio 44325-3909
共Received 20 September 1999; accepted for publication 25 January 2000兲
Nanofibers of polymers were electrospun by creating an electrically charged jet of polymer solution
at a pendent droplet. After the jet flowed away from the droplet in a nearly straight line, it bent into
a complex path and other changes in shape occurred, during which electrical forces stretched and
thinned it by very large ratios. After the solvent evaporated, birefringent nanofibers were left. In this
article the reasons for the instability are analyzed and explained using a mathematical model. The
rheological complexity of the polymer solution is included, which allows consideration of
viscoelastic jets. It is shown that the longitudinal stress caused by the external electric field acting
on the charge carried by the jet stabilized the straight jet for some distance. Then a lateral
perturbation grew in response to the repulsive forces between adjacent elements of charge carried by
the jet. The motion of segments of the jet grew rapidly into an electrically driven bending instability.
The three-dimensional paths of continuous jets were calculated, both in the nearly straight region
where the instability grew slowly and in the region where the bending dominated the path of the jet.
The mathematical model provides a reasonable representation of the experimental data, particularly
of the jet paths determined from high speed videographic observations. © 2000 American Institute
of Physics. 关S0021-8979共00兲03609-4兴

I. INTRODUCTION stability in electrospinning is discussed. The three-

dimensional equations describing the dynamics of the bend-
Electrospinning is a straightforward and cost effective ing of electrospun jets are derived and the calculated
method to produce novel fibers with diameters in the range behavior is compared with experimental observations of jets.
of from less than 3 nm to over 1 ␮m, which overlaps con-
temporary textile fiber technology. Polymer nanofibers are A. History and patents
being used, or finding uses, in filtration, protective clothing, Interest in the behavior of thin liquid jets in electric
biomedical applications including wound dressings, drug de- fields dates back to the work of Rayleigh.1 Later Zeleny2
livery systems, the design of solar sails, light sails, and mir- studied liquid jets in strong electric fields. Theoretical and
rors for use in space, the application of pesticides to plants, experimental activities in this area in the last 30 years were
as structural elements in artificial organs, and in reinforced revitalized by several important contributions by Taylor.3–6
composites. Ceramic or carbon nanofibers made from poly-
He produced useful experimental evidence, and calculated
meric precursors make it possible to expand the list of pos-
the conical shape of the protrusion from which a jet some-
sible uses for nanofibers. The electrospinning process easily
times leaves the surface of a liquid. Taylor also made the first
incorporates particles of materials such as pigments, carbon
black particles, and many others into the nanofibers that are attempt to calculate the growth rates of both varicose and
produced. Flexible fibers are needed on a scale commensu- bending perturbations of electrified liquid jets in the linear
rate with micro- or nanoelectrical, mechanical and optical approximation when the perturbations are small. In his the-
systems. The use of Coulomb forces to fabricate polymer oretical analysis, only inviscid fluids were considered.
objects may lead to ways to make such fibers in situ for such The bending instability of thin, highly viscous jets mov-
devices. ing in air, with no electrical forces, is a kindred phenomenon
Observations and a theoretical model of the electrospin- to that treated in the present work. The theory of the bending
ning process are presented in this article. The theory ac- instability of uncharged jets was developed and described by
counts for the nonlinear effects that are characteristic of fi- Yarin and co-workers.7–10
nite perturbations, as well as for the rheological behavior of Electrified jets of polymer solutions were investigated as
viscoelastic liquids. The general reason for the bending in- routes to the manufacture of polymer nanofibers.11–15 Since
1934, when a U.S. patent on electrospinning was issued to
a兲
Electronic mail: [email protected] Formhals,16 over 30 U.S. patents have been issued.

0021-8979/2000/87(9)/4531/17/$17.00 4531 © 2000 American Institute of Physics

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
4532 J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al.

B. Formation of bending instability TABLE I. Symbols employed and their definitions.

During the electrospinning of an aqueous solution of Unit

high molecular weight polyethylene oxide, a straight jet was Symbol Definition 共cgs兲
formed, as a consequence of electrical forces, from a conical a Cross-section radius cm
protrusion, often called a Taylor cone,3 on the surface of a a0 Initial cross-section radius cm
pendent drop of solution. The electrically charged jet trav- e Charge 共g1/2 cm3/2兲/s
eled for a few centimeters in a straight line. At the end of this fa Air friction force per unit length g/s2
fg Gravity force per unit length g/s2
straight segment, a diaphanous shape, also conical, with its
G Elastic modulus g/共cm s2兲
vertex at the end of the straight segment, was seen when h Distance from pendent drop to grounded collector cm
proper illumination was provided. This cone is the envelope, L Length scale, L⫽(e 2 / ␲ a 20 G) 1/2 cm
in space, of the complicated set of paths taken by a jet during Lz Length of the straight segment cm
the observation time. Images obtained with short exposure l Length of the ideal rectilinear jet cm
times by Baumgarten11 and by Warner et al.17 indicated that m Mass g
t Time s
the jet was continuously bending for as far as it could be v Velocity cm/s
followed after it entered the envelope cone. V0 Voltage g1/2 cm1/2/s
W Absolute value of velocity cm/s
␣ Surface tension g/s2
C. Electrical charge ␨ Initial segment length cm
␭ Perturbation wavelength cm
The electrical charges referred to in this article are the ␮ Viscosity g/共cm s兲
excess charges, whose electrical fields at long distances are ␯ Kinematic viscosity cm2/s
␴ Stress g/共cm s2兲
not canceled by nearby counterions. The common assump-
␳ Density g/cm3
tion that charge moves instantaneously through a metal is not ␳a Air density g/cm3
appropriate for ionic conductivity in a moving fluid. In an ␪ Relaxation time(⫽ ␮ /G) s
uncharged ionic solution, there are the same numbers of ␻ Frequency of the perturbation s⫺1
positive and negative ions in each volume element of the
solution and no external field is created. When an external
electric field is applied to the solution, the positive and nega-
E. Tables of symbols and dimensionless groups
tive ions in the polymer fluid tend to move in opposite di-
rections. Negative ions are forced toward the positive elec- The international system of units 关Système International
trode, and positive ions are forced toward the negative 共SI兲兴 was used to report the values of experimental measure-
electrode. The difference in the number of positive and nega- ments. Gaussian units that are customary in fluid mechanics
tive ions in a particular region is often called the excess and electrostatics have been also used, as well as dimension-
charge or, simply, the charge. For example, a volume ele- less combinations of parameters to provide concise coverage
ment of the fluid near the negative electrode will then con- of the multidimensional parameter space. The numerical re-
tain more positive ions than negative ions. The excess charge sults from the calculations were converted to SI units for
establishes an electrical field that extends for large distances. comparison with the experimental observations. Table I of
Adding a soluble salt, which dissociates into equal numbers symbols and Table II of dimensionless groups of parameters
of positive and negative ions, increases the electrical conduc- are provided.
tivity of the solution by increasing the number of ions per
unit volume, but cannot increase the excess charge. The F. Outline of the article
higher conductivity may, however, shorten the time required
for the excess charge, in the form of ions, to move to a In Sec. II the experimental setup is described. Section III
particular region in response to changes in the electrical contains the results of the experimental observations. In Sec.
field, or in the shape of a segment of the jet. IV the mathematical model of the phenomenon is given, and
the reasons for the jet bending are explained. In Sec. V the
theoretical results are presented and compared with the ex-
D. Ion mobility perimental data, and a discussion is also presented. Summary
and conclusions of the work are presented in Sec. VI.
The mobility of ions18,19 through the polymer solution is
around 10⫺6 m2/V s. The value of the electric field, deter-
II. EXPERIMENT
mined by dividing the applied potential by the distance be-
tween the surface of the pendent drop and the collector plate, Figure 1 is a sketch of the experimental apparatus. In
was typically 100 000 V/m. The drift velocity of ions inside this article, words such as up, down, top and bottom are used
the jet is then estimated to be 0.1 m/s. The velocity of the to simplify the description of the experimental arrangements
segment at the end of the straight segment of the jet was and the observations. The jet flowed downward from the
observed to be about 1 m/s in our experiments and about 5 surface of a pendent drop of fluid toward a collector at a
m/s in the work of Warner et al.17 In many cases it is useful distance h below the droplet. An electrical potential differ-
to simplify this by assuming that the ionic charge is fixed in ence, which was around 20 kV, was established between the
the fluid and moves with the jet. surface of the liquid drop and the collector. The distance, h,

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al. 4533

TABLE II. Dimensionless groups and parameters employed and their definitions.

Dimensionless Dimensionless
Symbol group parameter Definition

A Surface tension ( ␣ ␲ a 20 ␮ 2 )/(mL 2 G 2 )

H Distance from pendent h/L
drop to grounded collector
Ks Perturbation frequency ␻ ␮ /G
Kt Perturbation wave number 2 ␲ L/␭
Q Charge (e 2 ␮ 2 )/(L 3 mG 2 )
V Voltage (eV 0 ␮ 2 )/(hLmG 2 )
F ve Elastic modulus ( ␲ a 20 ␮ 2 )/(mLG)
t̄ Time t/( ␮ /G)
W̄ Absolute value of W/(LG/ ␮ )
velocity
l¯ Length of the rectilinear l /L
part of the jet
v̄ Velocity v /(LG/ ␮ )
¯␴ Stress ␴ /G

was around 0.2 m. The nanofibers formed a mat on the col- on an insulator, although a way to neutralize the charge car-
lector. The coordinates used in the mathematical description ried by nanofibers must be provided in order to collect many
are also shown. A magnified segment of the jet near the top layers of nanofibers. Airborne ions from a corona discharge
of the envelope cone shows the electrical forces that cause provide an effective way to neutralize the charge on the jets
the growth of the bending instability. These forces are de- and on the nanofibers. Nanofibers may also be collected on
scribed in detail in Sec. IV E and Fig. 12. the surface of a liquid.
In general, the pendent drop may be replaced by other Experiments on electrospinning15,20 typically use setups
fluid surfaces such as films on a solid or shapes generated by similar to that sketched in Fig. 1. All experiments were per-
surface tension and flow. The collector is usually a good formed at room temperature, which was about 20 °C. Poly-
electric conductor. The charged nanofibers may be collected ethylene oxide with a molecular weight of 400 000, at a
weight concentration of 6%, was dissolved in a mixture of
around 60% water and 40% ethanol. Fresher solutions pro-
duced jets that traveled further before the first bending insta-
bility appeared. The solution was held in a glass pipette with
an internal diameter of about 1 mm. At the beginning of the
experiment, a pendent droplet of polymer solution was sup-
ported at the tip of the capillary. The liquid jet formed on the
surface of the pendent drop of solution. When the electrical
potential difference 共measured in volts, and often referred to
as the applied voltage兲 between the capillary and the
grounded collector is increased, the surface of the liquid be-
comes charged by the electrical field induced migration of
ions through the liquid. Instability of the droplet set in when
the potential difference was high enough that electrical
forces overcame the forces associated with surface tension.3
Above this threshold, a stable liquid jet emerged. The jet
carried away excess ions that migrated to the surface when
the potential was applied. A higher potential difference cre-
ated a higher charge on the jet. For low conductivity solu-
tions, a significant time may be required for the charge to
reach a saturation value after the applied potential changes,
since charge transport within the fluid is limited by the finite
mobility of the ions.
A region about 5 mm across near the vertex of the en-
velope cone was imaged with a lens that had a focal length
of 86 mm and an f number of 1.0. The lens was placed about
20 cm from the jet to avoid disturbing the electrical field near
the jet. The image produced by this lens was observed using
FIG. 1. Schematic drawing of the electrospinning process, showing the jet
path, reference axes, relative arrangement of parts of the apparatus at dif- a 12.5–75 mm, f 1.8 zoom lens on an electronic camera that
ferent scales, and the region where the bending instability grew rapidly. recorded up to 2000 frames per second with exposure times

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
4534 J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al.

FIG. 2. Stereographic images of an electrically driven bending instability.

The exposure time was 0.25 ms. The arrow marks a maximum lateral ex-
cursion of a loop. FIG. 3. Image of the end of the straight segment of the jet. The exposure
time was 0.25 ms.

as short as 0.0125 ms, although the exposure times used in

this work were longer. with growing amplitude, the jet followed a bending, winding,
The light source was a 50 W halogen lamp with a fac- spiraling and looping path in three dimensions. The jet in
eted parabolic reflector. A Fresnel condenser lens was used each loop grew longer and thinner as the loop diameter and
to project an image of the halogen lamp and its reflector onto circumference increased. Some jets, which are shown in
the region occupied by the cone. The Fresnel lens had a focal Figs. 3–5, drifted downwards at a velocity much slower than
length of 19 cm and a diameter of 30 cm. The central 15 cm the downward velocity of the smaller loops close to the ver-
diameter part of the Fresnel lens was covered so that the tex of the envelope cone. After some time, segments of a
camera received the light scattered from the jet superimposed loop suddenly developed a new bending instability, similar
upon the dark background produced by the covered part of to, but at a smaller scale than, the first. Each cycle of bending
the Fresnel lens. instability can be described in three steps.
Images for stereographic viewing were obtained by re- Step 共1兲 A smooth segment that was straight or slightly
moving the 86 mm lens, which reduced the magnification so curved suddenly developed an array of bends.
that a region about 1 cm wide is shown in each image in Fig. Step 共2兲 The segment of the jet in each bend elongated
2. A pair of wedge prisms that were 40 mm high and 55 mm and the array of bends became a series of spiraling loops
wide were placed about 20 cm in front of the jet. Each prism with growing diameters.
deflected the light beam that passed through it by 5°. The Step 共3兲 As the perimeter of the loops increased, the
zoom lens on the electronic camera, viewing the jet through cross-sectional diameter of the jet forming the loop grew
the two prisms, produced side by side images of the jet from smaller; the conditions for step 共1兲 were established on a
two directions that were 10° apart on each frame that was smaller scale, and the next cycle of bending instability be-
recorded. These paired images were viewed stereoscopically gan.
during playback to produce a slowed down, three- This cycle of instability was observed to repeat at an
dimensional image of the moving jet. Image processing and even smaller scale. It was inferred that more cycles occur,
analysis was done with Adobe Photoshop, Corell Photopaint reducing the jet diameter even more and creating nanofibers.
and the software supplied with the electronic camera. After the second cycle, the axis of a particular segment may
point in any direction. The fluid jet solidified as it dried and
III. OBSERVATIONS electrospun nanofibers were collected some distance below
the envelope cone.
A. Jet paths
The vector sum of forces from the externally applied
The region near the vertex of the envelope cone was field, the charge momentarily held in space by the jet, and air
imaged at 2000 frames per second. These images showed the drag caused the charged segments to drift towards the col-
time evolution of the shape of the jet clearly and in detail. lector. Except for the creation of the pendent droplet, the
Stereographic images such as those in Fig. 2 showed the electrospinning process discussed in this article depends only
shape in three dimensions. The expanding spiral in Fig. 2 is slightly on the gravity of the Earth.
a simple example of the kinds of paths that were observed. Figure 3 shows the jet entering in the upper left corner,
After a short sequence of unstable bending back and forth, near the end of the straight segment of a jet, and the vertex of

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al. 4535

FIG. 4. Evolution of electrical bending instability. The exposure times were FIG. 5. Images of secondary and tertiary cycles of bending instabilities. The
0.25 ms. The width of each image was 5 mm. exposure time was 0.25 ms. The width of each image is 5 mm.

the envelope cone, where the first bending instability grew. These dots are not evidence of the familiar varicose instabil-
Several segments of the jet are shown, including segments ity that may cause a liquid jet to become a series of droplets.
from slowly moving loops that formed earlier. All these seg- No varicose instability was observed in this experiment.
ments are connected by segments that are not shown. Two Using a set of image files created by the electronic cam-
smooth segments cross each other in this image, as they run era, it was often possible to follow the evolution of the shape
nearly horizontally across the bottom of the image. These of spiraling segments, such as those shown in Fig. 3, back to
two segments are noticeably thinner than the jet entering the the straight segment that entered the upper left corner of the
image because the jet elongated as time evolved. These image. In Figs. 4 and 5, the light ellipse in the first image
slowly moving segments were part of large loops and were marks a segment that evolved in an interesting way. The
affected both by air drag and by the disturbance of the ap- selected segment of the jet was followed forward in time,
plied electrical field caused by the presence of both charged from the moment it entered the region contained in the im-
segments of the jet and charged nanofibers below the region ages until it elongated, looped, became unstable, bent, en-
being observed. Such slowly moving segments remained in tered the next cycle, and ultimately became too thin to form
view for many frames. an image.
Two thinner segments that formed even earlier are also Figure 4 starts with a bend near the end of the straight
included in Fig. 3. One runs across the top half of the image, segment of a jet entering the image at the upper left. The
and the other runs across the bottom half. In the lower of onset of the electrically driven bending instability occurred
these segments the successive bends 共step 1 of the second just before the jet entered the image. The straight segment of
cycle兲 were apparent. In the upper segment, the bends had the jet extended upward, and is not shown. The segment of
already developed into spiraling loops 共step 2 of the second the jet that is highlighted by the white ellipse was followed
cycle兲. The pattern of dots visible in the lower left corners of for 27.5 ms in a series of images that were recorded at 0.5 ms
Figs. 3–5 was caused by the pattern of facets on the reflector intervals. The thinner segments of the jet were emphasized
of the halogen lamp used to illuminate this experiment. by using the Photopaint 6 software to reproduce them. Places

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
4536 J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al.

where the faint image of the jet was ambiguous are indicated
by dots, seen, for example, in the image at 22.5 ms.
Eleven images were selected from this series of 55 im-
ages to show the evolution of the highlighted segment. The
time intervals between the images that are shown vary. Many
images that show only a gradual evolution of the path were
omitted to simplify Fig. 4. The time at which the first image
was captured is taken as time zero. The elapsed time at
which each of the following images was recorded is given in
Fig. 4.
The looping segment being observed at zero time elon-
gated for 10 ms in Fig. 4. Its further elongation was not
followed, because the loop had extended entirely across the
image. The rate of increase in the length of the highlighted
segment was around 120 mm/s. After 22 ms the visible part
of the highlighted segment still appeared in Fig. 4 as a
smooth, slightly curved line. In the short time interval be-
tween 22.0 and 22.5 ms, this long, slightly curved, smooth
segment suddenly became unstable. A linear array of bends
appeared, marking the beginning of the second cycle. The
lateral amplitude of the bends grew to about 1 mm, and the
spatial period of the bends along the segment was also about
1 mm.
These smaller bent segments of the jet continued to elon-
gate, but the images of the trajectories grew fainter and soon
were ambiguous. The elongation and the associated thinning
presumably continued as long as the charge on the jet sup-
plied enough force. Meanwhile, the elongational viscosity
increased as the jet dried. Eventually the jet solidified and the
FIG. 6. A jet splits off the primary jet and splays to a different direction.
elongation stopped. The evolution of the solidification pro-
cess remains to be investigated.
The first image in Fig. 5 shows a selected segment that evidence of a process that splayed the primary jet into many
was tracked back to the highlighted area near the bottom of smaller jets. The smaller jets were supposed to emerge from
the straight segment. This loop grew in diameter as the jet the region just below the apex of the envelope cone. Figure
elongated and became thinner. After 18 ms, an array of 7共a兲 shows an image from a video frame with an exposure of
bends that had a relatively long wavelength developed. 16.7 ms. The envelope cone was illuminated with a single
These bends evolved gradually to the path shown at 30.5 ms. bright halogen lamp that projected a narrow beam, through
Then a tertiary array of bends developed on the highlighted the envelope cone, toward, but not directly into, the lens, so
segment during the next 0.5 ms, and quickly evolved to the that most of the light that entered the video camera was
path shown at 31.5 ms. The growth of the tertiary excursions scattered from the jets.
was followed until 38.5 ms after the first image, at which Figure 7共b兲 shows a jet similar to that shown in Fig. 7共a兲
point the jet was so thin that its image could no longer be that was illuminated with light from two halogen lamps and
followed.

B. Jet splaying
The circled region in Fig. 6 shows a jet that split into
two jets that splayed apart, with the axis of the thinner
branch generally perpendicular to the axis of the primary jet.
The thinner jet disappeared in a few milliseconds, in some
cases because it rapidly became even thinner, and in other
cases because its path left the field of view. No bending
instability was observed on the thinner segment, probably
because it was not observed long enough for an instability to
develop. Only a few such events were observed in the thou-
sands of images of polyethylene oxide solution examined.
Before the high frame rate, short exposure time images
of Figs. 4 and 5 were available, visual observations and FIG. 7. Images of electrospinning jet with longer camera exposure times: 共a兲
video images of electrically driven jets were interpreted as 16.7 and 共b兲 1.0 ms.

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al. 4537

FIG. 8. Scanning electron micrograph of coiled and looped nanofibers on FIG. 9. Viscoelastic dumbbell representing a segment of the rectilinear part
the surface of an aluminum collector. of the jet.

photographed with a video camera. The two lamps were camera. The abundance and single coil of the coiled loops
above and behind the jet. One was to the left and the other to depended on the distance below the vertex at which they
right. This provided a broader source of illumination than were collected.
that used for Fig. 7共a兲, but not as uniform as the Fresnel lens The well-known tendency9 of a straight liquid jet mov-
arrangement shown in Fig. 1. An exposure time of 1 ms was ing in its axial direction to coil when it impacts a hard, sta-
used. The part of the straight jet with small bending ampli- tionary surface and buckles could account for some of the
tude is visible as are the loops containing segments, which observed coils. This mechanical effect is easily observed
had turned so that the axis of the segment formed a high when a gravity driven jet of honey falls onto a hard surface.
angle with the axis of the straight segment. The parts of the The occurrence of mechanical buckling during impact is
jet nearer the vertex of the envelope cone appeared only as likely to be infrequent because most of the long segments of
short, unconnected lines. Specular reflections of the beam of the jet were moving in a sidewise direction as they encoun-
light, called glints, from one or the other of the two halogen tered the collector. It is interesting to hypothesize that in
lamps off nearly horizontal segments of downward moving these experiments the coils and loops solidified before col-
loops were shown to be the cause of these bright spots. Simi- lection. Then, the collected coils and loops provide informa-
lar bright spots moved downwards during the longer expo- tion about the smallest bending instabilities that occurred.
sure of Fig. 7共a兲, and created the lines that are prominent in
Fig. 7共a兲.
IV. MATHEMATICAL DESCRIPTION OF A JET
The video frame rate of 30 frames per second was not
fast enough to follow the smooth development of the jet A. Viscoelastic model of a rectilinear electrified liquid
path. At this frame rate, for any particular frame, the preced- jet
ing and the following frames showed loops and spirals in Consider first a rectilinear electrified liquid jet in an
completely different positions. Only after the illumination electric field parallel to its axis. We model a segment of the
was improved, described in Sec. II, and the high frame rate jet by a viscoelastic dumbbell as shown in Fig. 9. In the
electronic camera used was it obvious that the envelope cone mathematical description, we use the Gaussian electrostatic
was occupied by one long, flowing, continuous, and ever system of units. Corresponding SI units are given when pa-
thinner jet. The repeated cycles of ever smaller electrically rameters are evaluated. Table I lists the symbols and their
driven bending instabilities created a complex path in which units.
the directions of the axes of the connected segments were Each of the beads, A and B, possesses a charge e and
often different and changing, sometimes by large angles. mass m. Let the position of bead A be fixed by non-Coulomb
forces. The Coulomb repulsive force acting on bead B is
C. Coiled and looped jets captured on a hard surface ⫺e 2 / l 2 . The force applied to B due to the external field is
⫺eV 0 /h. The dumbbell, AB, models a viscoelastic Max-
Nanofibers were sometimes collected by moving a glass wellian liquid jet. Therefore the stress, ␴, pulling B back to
microscope slide, a metal screen, or other solid surfaces A is given by21
through the conical envelope. Figure 8 shows that coiled and
looped nanofibers collected in this way were similar in shape d␴ dl G
⫽G ⫺ ␴, 共1兲
to the bending instabilities photographed with the high speed dt l dt ␮

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
4538 J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al.

where t is time, G and ␮ are the elastic modulus and viscos-

ity, respectively, and l is the filament length. It should be
emphasized that according to Refs. 9 and 22, the phenom-
enological Maxwell model adequately describes rheological
behavior of concentrated polymeric systems in strong
uniaxial elongation, which is the case in the present work.
The momentum balance for bead B is
dv e 2 eV 0
m ⫽⫺ 2 ⫺ ⫹ ␲ a 2␴ , 共2兲
dt l h
where a is the cross-sectional radius of the filament, and v is
the velocity of bead B which satisfies the kinematics equa-
tion
dl FIG. 10. Longitudinal stress ¯␴ in the rectilinear part of the jet, and the
⫽⫺ v . 共3兲 longitudinal force F v e ¯␴ / l¯ . Q⫽12, V⫽2, F v e ⫽12.
dt
B. Introduction of dimensionless parameters
specific expression for the evaporation rate. Evaporation is
We adopt dimensionless descriptions, as is customary in
not expected to introduce qualitative changes in jet dynamics
fluid mechanics23 共see Table II兲. We define the length scale
in the main part of the jet path. However, the effect of sol-
L⫽(e 2 / ␲ a 20 G) 1/2, where a 0 is the initial cross-sectional ra- vent evaporation on the values of the rheological parameters
dius at t⫽0, and render l dimensionless by L, and assume L of the polymer solution, which are not fully known at
to also be an initial filament length which is not restrictive. present, ultimately leads to the solidification of the jet into a
To make them dimensionless, we divide t by the relaxation polymer fiber.
time ␮ /G, stress ␴ by G, velocity v by LG/ ␮ , and radius a
by a 0 . Denoting W⫽⫺ v and applying the condition that the
volume of the jet is conserved, C. Stress and force along the jet axis

␲ a 2 l ⫽ ␲ a 20 L, 共4兲 Numerical solutions of the system, Eqs. 共5兲, may be

found using the following initial conditions t̄ ⫽0:
we obtain Eqs. 共1兲–共3兲 in the following dimensionless forms:
l¯ ⫽1, 共7a兲
d l¯
⫽W̄, 共5a兲 W̄⫽0, 共7b兲
d t̄
¯␴ ⫽0. 共7c兲
dW̄ ¯␴ Q
⫽V⫺F v e ⫹ , 共5b兲 Rheological and electrical parameters of the polymer so-
d t̄ l¯ l¯ 2 lution are at present not fully known from experiments.
Therefore here and hereinafter the calculations were done
d ¯␴ W̄ with the best values available for the dimensionless groups.
⫽ ⫺ ¯␴ , 共5c兲 In certain cases the values were chosen as close as possible
d t̄ l¯ to plausible estimates of the physical parameters involved. In
where the dimensionless parameters are denoted by bars, and these cases we list the values of the physical parameters
the dimensionless groups are given by along with the values of the dimensionless groups based on
them. The calculated results in Fig. 10 show that the longi-
e 2␮ 2 tudinal stress ¯␴ first increases over time as the filament
Q⫽ , 共6a兲
L 3 mG 2 stretches, passes a maximum, and then begins to decrease,
since the relaxation effects always reduce the stress at long
eV 0 ␮ 2 times. The dimensionless longitudinal force in the filament,
V⫽ , 共6b兲
hLmG 2 F v e ¯␴ / l¯ , passes its maximum before ¯␴ does. At the condi-
␲ a 20 ␮ 2 tions corresponding to the maximum of ¯␴ the value of the
F ve⫽ . 共6c兲 force is already comparatively small and decreasing rapidly.
mLG We will see below that this small value of the longitudinal
It is emphasized that, in this momentum balance, we force allows the onset of an electrically driven bending in-
temporarily neglect the secondary effects of the surface ten- stability. Therefore, we identify the filament length, l¯ * , at
sion, gravity and the air drag force. Note also that using the the condition when ¯␴ passes the maximum and the longitu-
definition of L, we obtain from Eqs. 共6a兲 and 共6c兲 that Q dinal force is already small, as the length of the rectilinear
⬅F v e . It should also be mentioned that, here in Eq. 共4兲 and segment of the electrospun jet at which the bending instabil-
hereinafter in this model, we neglect mass losses due to ity begins to grow rapidly. The relationship of this theoreti-
evaporation. In principle, they can be accounted for using a cally defined segment to the observed length of the straight

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al. 4539

FIG. 11. Length of the rectilinear part of the jet l¯ * as a function of the
dimensionless voltage V. Q⫽12 and F v e ⫽12. FIG. 12. Illustration of the Earnshaw instability, leading to bending of an
electrified jet.

segment is not yet determined. The length, l¯ * , increases

with the applied voltage as is seen in Fig. 11. Near the pen-
dent drop, the longitudinal force is also small, but the jet ⫻10⫺4 kg/s2. The compressive stress along the jet axis of
does not bend, since its radius there is large, and the corre- the air drag is negligibly small in comparison with the
sponding bending stiffness is large. stretching due to gravity, and is much smaller than the
Thin rectilinear liquid jets are unstable to capillary 共vari- stretching due to the electrical forces. Buckling of the elec-
cose兲 perturbations driven by surface tension. Longitudinal trospun jet due to the compressive force from air drag does
stretching can stabilize the jet in the presence of these not occur, since the electrical forces that tend to elongate the
perturbations.24 In electrospinning, jets are stretched along jet are larger and dominate any perturbation that might lead
their axis by the external electric field and are elongated to buckling.
further by the repulsive force between charges on adjacent
segments. The resulting tensile forces prevent development
of capillary instability in the experiments reported here. E. Bending instability of electrified jets
The reason for the observed bending instability may be
D. Effects from aerodynamics and gravity are small understood in the following way. In the coordinates that
For an uncharged jet moving in air at high speed, an move with a rectilinear electrified jet, the electrical charges
aerodynamically driven bending instability may set in if can be regarded as a static system of charges interacting
␳ a v 2 ⬎ ␣ /a, where ␳ a is the air density, v the jet velocity, mainly by Coulomb’s law 共without the external field兲. Such
and ␣ the surface tension coefficient.7–9 Taking, for example, systems are known to be unstable according to the Earn-
␳ a ⫽1.21 kg/m3, v ⬃0.5 m/s, ␣ ⬃0.1 kg/s2, and a⬃10⫺4 m, shaw’s theorem.26 To illustrate the instability mechanism
we estimate ␳ a v 2 ⬃0.3 kg/m s2, which is much smaller than that is relevant in the electrospinning context, we consider
␣ /a⬃103 kg/m s2. Therefore, under the conditions character- three point-like charges, each with a value e, and originally
istic of experiments on electrospinning, the aerodynamically in a straight line at A, B, and C as shown in Fig. 12. Two
driven bending instability does not occur. Coulomb forces having magnitudes F⫽e 2 /r 2 push against
The air drag force per unit jet length, which tends to charge B from opposite directions. If a perturbation causes
compress the jet along its axis, is given by25 point B to move off the line by a distance ␦ to B ⬘ , a net force

冉 冊 ⫺0.81
F 1 ⫽2F cos ␪⫽(2e2/r3)•␦ acts on charge B in a direction per-
2va pendicular to the line, and tends to cause B to move further
f a ⫽ ␲ a ␳ a v 2 0.65 , 共8兲
␯a in the direction of the perturbation away from the line be-
where ␳ a and ␯ a are the air density and kinematic viscosity, tween fixed charges, A and C. The growth of the small bend-
respectively. The gravity force per unit length pulling the jet ing perturbation that is characterized by ␦ is governed in the
downward in the experimental geometry shown in Fig. 1 is linear approximation by the equation

f g⫽ ␳ g ␲ a 2, 共9兲 d 2 ␦ 2e 2
m ⫽ 3 ␦, 共10兲
dt 2 l1
where ␳ is the liquid density and g is the acceleration due to
gravity. where m is the mass.
In the momentum balance, Eq. 共2兲 or 共5b兲, we neglected The growing solution of this equation, ␦
f g as a secondary effect. The air drag force f a is even smaller ⫽ ␦ 0 exp关(2e2/ml 31 ) 1/2t 兴 , shows that small perturbations in-
than f g . Taking ␳ a ⫽1.21 kg/m3, ␯ a ⫽0.15⫻10⫺4 m2/s, ␳ crease exponentially. The increase is sustained because elec-
⫽1000 kg/m3, v ⫽0.5 m/s, and a⫽150 ␮ m we obtain from trostatic potential energy of the system depicted in Fig. 12
Eqs. 共8兲 and 共9兲 f a ⫽1.4⫻10⫺5 kg/s2 and f g ⫽6.9 decreases as e 2 /r when the perturbations, characterized by ␦

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
4540 J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al.

and r, grow. We believe that this mechanism is responsible

for the observed bending instability of jets in electrospin-
ning.
If charges A, B, and C are attached to a liquid jet, forces
associated with the liquid tend to counteract the instability
caused by the Coulomb forces. For very thin liquid jets, the
influence of the shearing force related to the bending stiff-
ness can be neglected in comparison with the stabilizing ef-
fect of the longitudinal forces since the shearing forces are of
the order of O(a 4 ), which is much smaller than the longitu-
dinal forces,9 which are of the order of O(a 2 ). The longitu-
dinal force, at the moment when the bending instability sets
in, was calculated above for the stretching of a rectilinear
filament. Its value is given by f l ⫽ ␲ a 2 ␴ * 共or in dimension-
less form by F v e ¯␴ * / l¯ * 兲. The values of ␴ and l¯ at the mo-
ment when ␴ 共or ¯␴ 兲 passes its maximum are denoted by
asterisks. The forces f l are directed along BC or BA in Fig.
12, and are opposite to the local Coulomb force F. If F is
larger than the viscoelastic resistance f l , the bending per-
turbation continues to grow, but at a rate decelerated by f l .
It might be thought that bending perturbations of very
short lengths can always overcome the viscoelastic resistance FIG. 13. Bending electrospun jet modeled by a system of beads connected
f l , since the Coulomb force increases when the wavelength by viscoelastic elements.
of the perturbation decreases. In fact, the surface tension al-
ways counteracts the bending instability because bending al-
ways leads to an increase of the area of the jet surface.9
Surface tension resists the development of too large a curva- d ␴ di 1 d l di G
⫽G ⫺ ␴ di . 共12b兲
ture by the perturbation ABC in Fig. 12, and therefore limits dt l di dt ␮
the smallest possible perturbation wavelengths. All these fac-
The total number of beads, N, increases over time as
tors are accounted for in the description of the three-
new electrically charged beads are inserted at the top of Fig.
dimensional bending instability of electrospun jets in Sec.
13 to represent the flow of solution into the jet. The net
IV F.
Coulomb force acting on the ith bead from all the other
beads is given by
F. Three-dimensional equations of the dynamics of
the electrospun jets fC ⫽ 兺 2
e2
j⫽1,N R i j
j⫽i
冋 i
x i ⫺x j
Rij
⫹j
y i ⫺y j
Rij
⫹k
z i ⫺z j
Rij
, 册 共13兲
We represent the electrospun jets by a model system of
beads possessing charge e and mass m connected by vis- where i, j, and k are the unit vectors along the x, y, and z
coelastic elements as shown in Fig. 13, which generalizes the axes, respectively, and
models of Figs. 9 and 12. It needs to be mentioned that these R i j ⫽ 关共 x i ⫺x j 兲 2 ⫹ 共 y i ⫺y j 兲 2 ⫹ 共 z i ⫺z j 兲 2 兴 1/2. 共14兲
imaginary beads are not the same as the physical beads20
resulting from the varicose instability. The parameters corre- The electric force imposed on the ith bead by the electric
sponding to the element connecting bead i with bead (i field created by the potential difference between the pendent
⫹1) are denoted by subscript u 共up兲, those for the element drop and the collector is
connecting bead i with (i⫺1) by subscript d 共down兲. The V0
lengths l ui and l di of these elements are given by f0 ⫽⫺e k. 共15兲
h
l ui ⫽ 关共 x i⫹1 ⫺x i 兲 2 ⫹ 共 y i⫹1 ⫺y i 兲 2 ⫹ 共 z i⫹1 ⫺z i 兲 2 兴 1/2,
It is clear that the gravitational force, already shown to
共11a兲
be small, may be included in f0 , as is done later in this
l di ⫽ 关共 x i ⫺x i⫺1 兲 2 ⫹ 共 y i ⫺y i⫺1 兲 2 ⫹ 共 z i ⫺z i⫺1 兲 2 兴 1/2, article to create a hypothetical uncharged jet.
共11b兲 The net viscoelastic force acting on the ith bead of the
respectively, where x i ,y i ,z i ,..., are the Cartesian coordi- jet is
nates of the beads.
The rates of strain of the elements are given by
(d l ui /dt)/ l ui and (d l di /dt)/ l di . The viscoelastic forces
fv e ⫽ ␲ a 2ui ␴ ui i冋 x i⫹1 ⫺x i
l ui
⫹j
y i⫹1 ⫺y i
l ui
⫹k
z i⫹1 ⫺z i
l ui
册
acting along the elements are similar to Eq. 共1兲,
d ␴ ui 1 d l ui G
⫺ ␲ a 2di ␴ di i 冋 x i ⫺x i⫺1
l di
⫹j
y i ⫺y i⫺1
l di
⫹k
z i ⫺z i⫺1
l di
册 ,
⫽G ⫺ ␴ ui , 共12a兲
dt l ui dt ␮ 共16兲

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al. 4541

where, when mass is conserved and evaporation neglected, and the charge on each element of the jet is removed as it
the filament radii a ui and a di are given by arrives at the collector. Such a calculation mimics the devel-
opment of the electrically driven bending instability. The cal-
␲ a 2ui l ui ⫽ ␲ a 0 L,
2
共17a兲
culation begins with only two beads, N⫽2. As the jet flows,
␲ a 2di l di ⫽ ␲ a 0 L,
2
共17b兲 the number of beads in the jet, N, increases.
It is also possible to consider the temporal instability of
which is similar to Eq. 共4兲. an established jet. In this case the calculation began from a
The surface tension force acting on the ith bead, and long rectilinear filament 0⭐z⭐h containing a fixed number
tending to restore the rectilinear shape of the bending part of of beads. The filament was perturbed by moving it laterally,
the jet, is given by at t⫽0, everywhere along its axis by the function

冉 冊
␣ ␲共 a 2 兲avk i
fcap⫽⫺ 关 i兩 x i 兩 sign共 x i 兲 ⫹j兩 y i 兩 sign共 y i 兲兴 , 共18兲 2 ␲ h⫺z
共 x 2i ⫹y 2i 兲 1/2 x⫽10⫺3 L cos z , 共22a兲
␭ h
where ␣ is the surface tension coefficient, k i is the jet curva-
ture calculated using the coordinates of beads (i⫺1), i and
(i⫹1), and (a 2 ) a v ⫽(a ui ⫹a di ) 2 /4. The meaning of ‘‘sign’’
y⫽10⫺3 L sin 冉 冊2 ␲ h⫺z
␭
z
h
, 共22b兲
is as follows:
where ␭ is the wavelength of the perturbation. Then the tem-
sign共 x 兲 ⫽1, if x⬎0, poral evolution of the path was calculated.
sign共 x 兲 ⫽⫺1, if x⬍0, 共19兲 In all cases, the system of Eqs. 共12兲 and 共20兲 was solved
numerically, assuming that the stresses ␴ ui and ␴ di and the
sign共 x 兲 ⫽0, if x⫽0. radial velocity dri /dt were zero at t⫽0. The equations were
Setting the forces described in Eqs. 共13兲, 共15兲, 共16兲, and made dimensionless by the same scale factors as those in
共18兲 equal to mass times acceleration, according to Newton’s Sec. IV B. Since here it is necessary to account for the sur-
second law, we obtain the equation governing the radius vec- face tension and for the perturbing displacements, three new
tor of the position of the ith bead ri ⫽ix i ⫹jy i ⫹kz i in the dimensionless groups emerge in addition to those of Eqs. 共6兲,
following form: ␣ ␲ a 20 ␮ 2
A⫽ , 共23a兲
d 2 ri e2 V0 ␲ a 2ui ␴ ui mL 2 G 2
m 2⫽
dt 兺
3 共 ri ⫺r j 兲 ⫺e
j⫽1,N R i j h
k⫹
l ui
共 ri⫹1 ⫺ri 兲
j⫽i K s ⫽ ␻ ␮ /G, 共23b兲
␲ a di
2
␴ di ␣ ␲共 a 2 兲avk i 2␲L
⫺ 共 ri ⫺ri⫺1 兲 ⫺ K t⫽ . 共23c兲
l di 共 x 2i ⫹y 2i 兲 1/2 ␭
⫻ 关 i兩 x i 兩 sign共 x i 兲 ⫹j兩 y i 兩 sign共 y i 兲兴 . 共20兲 The last dimensionless group needed is formed by dividing
For the first bead, i⫽1, and N, the total number of the distance h, from the collector to the pendent droplet, by
beads, is also 1. As more beads are added, N becomes larger L,
and the first bead i⫽1 remains at the bottom end of the
h
growing jet. For this bead, all the parameters with subscript H⫽ . 共24兲
d should be set equal to zero since there are no beads below L
i⫽1.
V. RESULTS AND DISCUSSION
G. Spatial and temporal perturbation of the rectilinear A. Jet path calculated from the electrically driven
segment bending instability
Both space and time dependent perturbations lead to the We begin with the calculation of the development of the
development of the electrically driven bending instability. To temporal perturbation into the bending instability. Figure 14
model the way a spatial perturbation develops, we denote the shows that the small perturbation, Eqs. 共22兲, increased dra-
last bead pulled out of the pendent drop and added at the matically as the Earnshaw-like instability grew. The path of
upper end of the jet by i⫽N. When the distance l d,N be- the electrified jet at t̄ ⫽0.89 is represented by a helix of
tween this bead and the pendent drop becomes long enough, increasing radial dimension and a pitch commensurate with
say, h/25 000, a new bead i⫽N⫹1 is inserted at a small
the radial dimension. It is emphasized that the jet at t̄ ⫽0
distance, say, h/50 000, from the previous one. At the same
would appear on the scale of Fig. 14 as a straight line, since
time a small perturbation is added to its x and y coordinates,
the initial helix radius is very small in comparison to the
x i ⫽10⫺3 L sin共 ␻ t 兲 , 共21a兲 other dimensions. As the calculation progresses, the beads
move further and further apart. Since, in this model, the
y i ⫽10⫺3 L cos共 ␻ t 兲 . 共21b兲
beads are connected by straight lines, the graph becomes
Here ␻ is the perturbation frequency. The condition that quite irregular when the separation between the beads is
the collector at z⫽0 is impenetrable is enforced numerically, larger than the radius of the spiral path.

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
4542 J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al.

shown at the same scale in the inset at the upper right of each
part of Fig. 15. An ellipse in each inset encloses the part of
the jet path shown in the corresponding coordinate box. The
pendent drop was always at x⫽0, y⫽0, and z⫽h.
The experimental evidence shows a self-similar, fractal-
like process of development of the electrically driven bend-
ing instabilities. The diameter of the first generation of bend-
ing loops becomes larger and the jet becomes thinner. Then
much smaller bending perturbations set in on these loops and
begin to grow also. This self-similar process continues at
smaller and smaller scales until viscoelastic force, surface
tension or solidification of the jet arrest further bending. The
numerical results in Fig. 15 describe only the emergence and
growth of the first cycle of the loops. This is a consequence
of the fact that the distances between the beads increase
enormously in the simulation of the development of the first
cycle. No new beads were added except at the top of the
FIG. 14. Temporal growth of the bending instability along the straight seg- rectilinear segment. Therefore the capability of the present
ment of a charged jet subject to a small perturbation that is initially periodic computer code to elucidate smaller details of the path de-
in space. The growth of the lateral excursions is larger than that usually
observed to show details of the model. The following values of the param-
creases as the jet elongates enormously.
eters were used: Q⫽50, V⫽40, F v e ⫽50, A⫽0, K t ⫽1, and H⫽100; t̄ Figure 16 shows the path of a charged jet calculated
⫽0.89. The number of beads N⫽100. from a realistic but different set of dimensionless parameters
and perturbations than was used in Fig. 15. The path displays
a bending instability generally similar to that shown in
Consider now the development of perturbations into a Fig. 15.
bending instability in a realistic jet 共spatial instability兲. We To show that the bending instability is driven by the
estimate the charge carried by the jet to be 1 C/ l, which is of Coulomb interaction, the charge, e, on the beads is taken to
the same order as the values measured.20 We also estimate be zero so that Q⫽0. The electrical driving force for the
that the relaxation time ␪ is 10 ms, a 0 is 150 ␮m, ␳ is bending instability is then zero, but the other parameters are
103 kg/m3, h is 2 m, V 0 is 10 kV, ␣ is 0.7 kg/s2, and ␮ is exactly the same as those in Fig. 16. If a jet were then pulled
103 kg/共m s兲. The value of ␮ is taken to be much larger than downward by gravity, which can supply a downward com-
the zero-shear viscosity ␮ 0 reported,20 since the strong lon- ponent of force that acts on the segment in the same way as
gitudinal flows we are dealing with in the present work lead the downward component of the electrical force from the
to an increase, by several orders of magnitude, of the elon- electrical field, one would expect the uncharged jet to be
gational viscosity from ␮ 0 . 9,22,27,28 The dimensionless pa- almost straight in spite of the small perturbations applied to
rameters are as follows: Q⫽F v e ⫽78359.6, V⫽156.7, A it, since the perturbations would not develop into a bending
⫽17.19, and H⫽626.9. The length scale is L⫽3.19 mm.
instability. The calculated result with the same parameters as
The charge on the bead e⫽8.48 共g1/2 cm3/2兲/s
⫺9 those in Fig. 16, but Q⫽0, is in fact a straight jet growing
⫽2.83⫻10 C. The mass of each bead is m⫽0.283
⫻10⫺8 kg. The value of K s is taken as 100. Since ␪ ⫽ ␮ /G downward, even at a later time ( t̄ ⫽8.99). Increasing the
⫽10 ms, this value corresponds to ␻ ⫽104 s⫺1, which is in ratio of the surface tension to the Coulomb force also stabi-
the frequency range of typical noise in the laboratory. lizes a charged jet. If A is increased to 9, by increasing the
Figures 15共a兲–15共e兲 illustrate the development of a typi- surface tension while all the other parameters are kept the
cal jet path. The time periodic perturbation, Eqs. 共21兲, that same as those in Fig. 16 practically no bending occurs. The
grows along the jet is similar to the case shown in Fig. 14. results for the gravity driven jet and for the high surface
The jet flows continuously from the pendent drop in response tension jet are not shown because the calculated jet path
to the electric field established by the externally applied po- cannot be distinguished from a straight line at the scale in
tential between the droplet and the collector. This electric Fig. 16.
field also causes the jet to be charged as it leaves the pendent
drop. At t̄ ⫽0.99 in Fig. 15共e兲 the instantaneous path of the
jet is similar to the patterns recorded in experiments using a
high speed video camera such as those shown in Fig. 2. It is
B. Measurements of the trajectory and velocity of a
emphasized that the stresses ␴ ui and ␴ di are positive along
particular segment of a jet
the entire jet in Figs. 15共a兲–15共e兲, which means that the
whole jet is stretched continuously. Figure 17 is derived from measurements of a sequence
In Fig. 15, a long segment near the vertex of the enve- of stereographic images of the part of a jet just below the
lope cone is plotted in the x, y, and z coordinates at various apex of the envelope cone. The polyethylene oxide solution
times and scales to show details of the jet path. The entire in water and ethanol that was used is the same as that used
length of both the straight segment and the spiral part is for Figs. 4 and 5. The exposure time was 0.25 ms. The po-

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al. 4543

FIG. 15. Perturbations develop into a bending instability. The dimensionless groups have the following values: Q⫽F v e ⫽78359.6, V⫽156.7, A⫽17.19,
K s ⫽100, H⫽626.9. 共a兲 t̄ ⫽0.19, 共b兲 0.39, 共c兲 0.59, 共d兲 0.79, and 共e兲 0.99.

sition projected into the plane of the image of the maximum Figure 17 shows the downward displacement of the
lateral excursion of a loop, an example of which is marked maximum lateral excursion from its initial position. The
with an arrow in Fig. 2, was followed as a function of time downward motion of the jet predicted by the model and the
for 8 ms. value measured from the experiment are of the same order of

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
4544 J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al.

FIG. 18. Downward velocity of a maximum lateral displacement in a bend-

ing instability as a function of time.

FIG. 16. Charged jet with values of the dimensionless parameters that are
ment of the jet would be projected into the image plane. The
realistic but different from those used in Fig. 15. Q⫽F v e ⫽12, V⫽2, A
assumption that the velocity and displacement of the maxi-
⫽0.9, K s ⫽100, H⫽100, t̄ ⫽4.99.
mum lateral excursion represent the motion of a particular
segment of the jet is essentially correct if the curve is not
rotating, but only elongating along the axis of the segment,
magnitude. Figure 18 shows the downward velocity, which
and moving radially outwards and downwards.
decreased from about 1.1 to about 0.8 m/s. The increase in
The motion of a splay point such as that shown in Fig. 6,
the radial displacement of the maximum is shown in Fig. 19.
which is attached to a particular point on the jet, was ob-
The measured radial velocity was nearly constant at a value
served in a series of stereo pictures. The motion of the splay
of 0.5 m/s, as shown in Fig. 20.
point was only downward and outward in a plane that con-
The displacements of the maximum lateral excursion,
tained the y axis and a single radial direction, indicating that
given in Figs. 17 and 19, were fitted by the following expres-
rotation of the path was small.
sion, where z is measured from the pendent drop downward:
x⫽0.0257t 2 cos t, C. Bending instability leads to large area reduction
ratios
2
y⫽0.0257t sin t, 共25兲
The ‘‘area reduction ratio,’’ which is defined as the ratio
z⫽20 exp共 ⫺0.0595t 兲 . of the cross-sectional area of the upper end of a segment to
Figure 21 is a three-dimensional plot of this equation. the cross-sectional area at the lower end of the same seg-
The position of ‘‘maximum lateral excursion of a loop’’ that ment, is equal to the draw ratio if the volume of material in
was used as a reference to determine the velocities does not the segment is conserved. Since in this experiment the sol-
necessarily move in the same way as a marker embedded in vent evaporated during the process, the area reduction ratio
the jet and carried with the jet. If a path with an elliptical was related to the draw ratio by the time varying concentra-
loop were to rotate about its vertical axis, a changing seg- tion of the solution. Here we assume that the fluid jet was

FIG. 19. Lateral movement of a maximum lateral displacement in a bending

FIG. 17. Downward movement of a maximum lateral displacement in a instability as a function of time. The inset shows the experimental results on
bending instability as a function of time. an expanded distance scale.

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al. 4545

FIG. 20. Lateral velocity of a maximum lateral displacement in a bending

instability as a function of time. The inset shows the experimental velocity
on an expanded velocity scale. FIG. 21. Three-dimensional reconstruction of the bending jet.

drawn to a particular draw ratio and then the solvent was instability occurs repeatedly at smaller and smaller scales,
evaporated to produce a dry nanofiber with the observed di- and the process becomes more and more fractal like, which
ameter. is practically impossible to follow with numerical calcula-
The examples that follow show that both the total area tions of the sort used here. As a result, the total draw ratio
reduction ratio and the total draw ratio were large for a seg- and the total decrease in jet radius are both underestimated in
ment that contained the entire jet, starting at the surface of the calculations.
the droplet and ending in the dry nanofiber that was collected
at the end of the process. For a jet that had a concentration of
D. Longitudinal strain rate and molecular orientation
6% polymer and a starting diameter of 100 ␮m, that pro-
duced a nanofiber with a diameter of 100 nm, the draw ratio The high value of the area reduction ratio and the asso-
not accounting for evaporation would be 106 . The draw ratio ciated high longitudinal strain rate imply that the macromol-
of the jet accounting for evaporation is, however, 0.06 ecules in the nanofibers should be stretched and axially ori-
⫻106 ⫽60 000. ented. Most electrospun nanofibers, even those made from a
A draw ratio of 60 000 was routinely achieved in these styrene-butadiene-styrene triblock copolymer, are
experiments. Segments of the jet were drawn simultaneously birefringent.29
in many expanding loops. If the jet were drawn in a straight The longitudinal strain rate was different at different
line with a segment at the bottom end of the straight segment places along the jet. The longitudinal strain rate for three
moving at a velocity of around 1.0 m/s, and then being elon- different parts of the jet was determined.
gated to a draw ratio of 60 000, the velocity achieved at the 共1兲 The jet velocity in the downward direction was de-
nanofiber end of the jet 共about 60 000 m/s兲 would be 176 termined from the sequential images by determining the ve-
times faster than the speed of sound in air. The path actually locity of a particular maxima of the growing bending insta-
observed achieves very high elongation without such an un- bility and the length of the straight segment. In this work, the
reasonably high velocity by enlarging the diameters of many length of the straight segment (L z ) was 5 cm and the velocity
loops during the same time interval. was 1 m/s. Therefore the longitudinal strain rate was
The calculation from the bending instability model sup- ( ␦ L z /L z ␦ t),
ports the conclusion that, in the loops of the bending insta-
␦ L z /L z ␦ t⫽V ␦ t/L z ␦ t⫽V/L z ⫽ 共 1 m/s兲 • 共 1/0.05 m兲
bility, the total draw ratio is very high. The initial distance
between any two beads used in the calculation was h/50 000, ⫽20 s⫺1, 共27兲
whereas the final distance between adjacent beads in the cal- which is the rate of strain in the straight segment of the jet.
culated loop, for example, in Fig. 15共e兲 is about 1.97h. As- Warner et al.17 measured the velocities along the straight
suming that the polymer concentration of the jet was 6%, the segment with a laser Doppler velocimeter, and reported a
final cross-sectional radius of a dry fiber a f , when all the velocity that increased from 1 to 15 m/s. Doppler velocim-
solvent evaporates, is estimated as follows: etry is most useful for monitoring the longitudinal strain rate,
h particularly where the jet diameter is large. The Doppler ve-
␲ a 2f •1.97h⫽ ␲ a 20 •0.06. 共26兲 locimetry data become more difficult to interpret when the
50 000
bending instabilities are encountered.
This yields a f ⯝0.78⫻10⫺3 a 0 . For a 0 ⫽150 ␮ m, we 共2兲 The observation of expanding loops provided a sec-
obtain a f ⯝0.117 ␮ m. This value is of the same order as ond measure of the longitudinal strain rate for the segment
those measured experimentally in Ref. 20, as well as those that formed the loop. A typical loop grew from a diameter of
obtained in the experiment with h⫽0.2 m. The correspond- 1 to 8 mm in 7 ms. The resulting longitudinal strain rate in
ing draw ratio is 98 619. Also in the experiment, the bending such a loop was 1000 s⫺1.

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
4546 J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al.

共3兲 The overall longitudinal strain rate can be estimated area of the jet was achieved in this manner and the corre-
using the data in Sec. V C. The time that a typical segment of sponding large increase in the jet length occurred in a region
the electrospun jet is in flight ( ␦ t) can be estimated as the that was only a few centimeters across. The associated high
distance between the pendent droplet and the collector 共20 longitudinal strain rate implies that, in the electrospun jet, the
cm兲 divided by the average downward velocity of the jet 共1 macromolecular coils are stretched along the axis of the jet.
m/s兲. The resulting ␦ t is 0.2 s. The longitudinal strain rate is Bending instabilities in electrospun jets were modeled
␦ ␨ /( ␦ t• ␨ ), where ␨ is the initial segment length, and ␦␨ is by a system of connected viscoelastic dumbbells. Beads in
the growth in length. Since ␦␨ is much greater than ␨, ␦␨ is the dumbbells possess appropriate mass and charge to repre-
approximately the final segment length. The ratio ␦␨/␨ was sent the observed jets. The beads interacted with each other
around 105 , and therefore the longitudinal strain rate was according to Coulomb’s law. The beads were also subject to
around 0.5⫻106 s⫺1. Using the estimate, Eq. 共27兲, we find the electrical forces from the electrical field created between
that in the straight segment the length of a liquid element has the pendent droplet and the collector by the imposed poten-
been approximately doubled, and the cross-sectional radius tial difference. Springs and dashpots, connecting the beads in
decreased by a factor of 4. Then the longitudinal strain rate the dumbbells, mimicked the Maxwellian viscoelastic resis-
in the loops becomes of the order of 105 s⫺1. The actual tance to elongation of the jet. Surface tension effects were
value will be lower due to the effects of evaporation and also included. Estimates showed that effects of the Earth’s
solidification. gravity, as well as aerodynamic forces 共the drag and distrib-
Theory suggests that the transformation from a random uted lift force兲, were negligibly small in the electrospinning
coil to a stretched macromolecule occurs when the strain rate experiments.
multiplied by the conformational relaxation time of the mol- Equations of motion of the beads with all the forces
ecule is greater than 0.5.27,28 Since the relaxation time of this acting in combination were used to follow, numerically, the
polymer solution is about 0.01 s, then ␦ ␨ /( ␦ t• ␨ ) multiplied evolution of the path of the electrically charged polymer jets
by the relaxation time was equal to 10– 103 , which is much in the presence of spatial and temporal perturbations. The
greater than 0.5. Therefore the longitudinal flow in the elec- results are in reasonable agreement with the experimental
trospun jet is strong, and the macromolecules are likely to be evidence. In spite of the fact that some material parameters
stretched in the direction of the jet axis. can only be estimated at present, order of magnitude, or bet-
ter, agreement between the theory and experimental observa-
tions was achieved. The theory showed that the viscoelastic
VI. SUMMARY AND CONCLUSIONS force along the jet and the surface tension both tend to sta-
The initial straight segment of a typical electrically bilize the charged jet.
charged jet of polyethylene oxide aqueous solution was cre- The entire electrospinning process and the electrically
ated by an electrical potential applied between the pendent driven bending instabilities of an electrospun fluid can each
drop and the collector. Then an electrically driven bending be viewed as particular examples of the very general Earn-
instability, triggered by perturbations of the lateral position shaw theorem in electrostatics. This theorem leads to the
and lateral velocity of the jet, grew. The repulsive forces conclusion that it is impossible to create a stable structure in
between the charges carried with the jet caused every seg- which the elements of the structure interact only by Cou-
ment of the jet to lengthen continuously along a changing lomb’s law. Charges on or embedded in a polymer fluid
path until the jet solidified. The geometrically simple idea move the fluid in quite complicated ways to reduce their
that the jet lengthened in a straight line along its axis leads to Coulomb interaction energy. Electrospinning, and perhaps
an implausibly high velocity at the thin, leading end of such other useful processes, utilize this behavior to produce inter-
a straight jet. Instead, the jet bent and developed a series of esting and useful polymer objects.
lateral excursions that grew into spiraling loops. Each of For additional information on this topic, see Ref. 30.
these loops grew larger in diameter as the jet grew longer
and became thinner.
About 20 ms after a loop had first formed, occurring ACKNOWLEDGMENTS
within a time interval of less than a millisecond, a new set of
electrically driven bending instabilities appeared on the now Support for this work was provided by the National Sci-
thinner, smoothly curved loop. The new bending instabilities ence Foundation under Grant Nos. DMI-9813098 and CTS-
grew when the jet in the loop was thin enough and when the 9900949, by the U.S. Army Research Office, by the U.S.
viscoelastic stress along its axis had relaxed enough. These Army Soldier and Biological Systems Command, and by the
new bending instabilities developed into a smaller set of spi- Nonmetallic Materials Division of the U.S. Air Force Re-
rals that looped around the path of the first loop. search Laboratory, Wright Patterson Air Force Base. The
The envelope cone observed at the end of the straight high-speed camera was acquired with support from the
segment defined the region inside which this complex path of Hayes Investment Fund of the Ohio Board of Regents. A gift
the jet developed. The cycles of bending instability repeated from the DUPONT Company made it possible for Dr. Alex
in a self-similar manner until the solvent evaporated, and the Yarin to make a productive visit to the University of Akron.
remaining polymer fiber resisted further elongation by the The help of Dale Ertley and Dr. Dan Galehouse in the con-
Coulomb forces of the charge that was still present on the jet. struction of many specialized pieces of equipment is greatly
A very high reduction 共as much as 105 兲 in the cross-sectional appreciated.

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 87, No. 9, 1 May 2000 Reneker et al. 4547

1 19
X. Lord Rayleigh, London, Edinburgh, Dublin Philos. Mag. 44, 184 J. S. Chang, A. J. Kelly, and J. M. Crowley, Handbook of Electrostatic
共1882兲. Processes 共Dekker, New York, 1995兲, p. 248.
2
J. Zeleny, Phys. Rev. 10, 1 共1917兲. 20
H. Fong, I. Chung, and D. H. Reneker, Polymer 40, 4585 共1999兲.
3
G. I. Taylor, Proc. R. Soc. London, Ser. A 280, 383 共1964兲. 21
R. B. Bird, R. C. Armstrong, and O. Hassager, Dynamics of Polymeric
4
G. I. Taylor, J. Fluid Mech. 22, 1 共1965兲. Liquids, 2nd ed. 共Wiley, New York, 1987兲, Vol. 1.
5
G. I. Taylor, Proc. R. Soc. London, Ser. A 291, 145 共1966兲. 22
A. L. Yarin, J. Non-Newtonian Fluid Mech. 37, 113 共1990兲.
6
G. I. Taylor, Proc. R. Soc. London, Ser. A 313, 453 共1969兲. 23
R. W. Fox and A. T. McDonald, Introduction to Fluid Mechanics, 3rd ed.
7
A. L. Yarin, J. Appl. Mech. Tech. Phys. 23, 39 共1982兲.
8
A. L. Yarin, J. Appl. Mech. Tech. Phys. 27, 828 共1986兲. 共Wiley, New York, 1985兲.
24
9
A. L. Yarin, Free Liquid Jets and Films: Hydrodynamics and Rheology D. V. Khakhar and J. M. Ottino, Int. J. Multiphase Flow 13, 71
共Longman, Harlow, and Wiley, New York, 1993兲. 共1987兲.
10
V. M. Entov and A. L. Yarin, J. Fluid Mech. 140, 91 共1984兲. 25
High-Speed Fiber Spinning, edited by A. Ziabicki and H. Kawai 共Wiley,
11
P. K. Baumgarten, J. Colloid Interface Sci. 36, 71 共1971兲. New York, 1985兲.
12
L. Larrondo, and R. St. J. Manley, J. Polym. Sci., Part B: Polym. Phys. 19, 26
J. Jeans, The Mathematical Theory of Electricity and Magnetism 共Cam-
909 共1981兲. bridge University Press, Cambridge, 1958兲.
13
L. Larrondo and R. St. J. Manley, J. Polym. Sci., Part B: Polym. Phys. 19, 27
P. G. de Dennes, J. Chem. Phys. 60, 5030 共1974兲.
921 共1981兲. 28
H. Chang and A. S. Lodge, Rheol. Acta 11, 127 共1972兲.
14
L. Larrondo and R. St. J. Manley J. Polym. Sci., Part B: Polym. Phys. 19, 29
H. Fong and D. H. Reneker, J. Polym. Sci., Part B: Polym. Phys. 37, 3488
933 共1981兲.
15
D. H. Reneker and I. Chun, Nanotechnology 7, 216 共1996兲. 共1999兲.
30
16
A. Formhals, U.S. Patent No. 1,975,504 共filed 1934兲. See E-PAPS Document No. E-JAPIAU-87-O36009 for a film on bending
17
S. B. Warner, A. Buer, S. C. Ugbolue, B. C. Rutledge, and M. Y. Shin, instability 共Bending-movie.avi兲. This document may be retreived via the
National Textile Center Annual Report No. 83-90 共1998兲. E-PAPS homepage 共http://www.aip.org/pubservs/epaps.html兲 or from ft-
18
A. G. Bailey, Electrostatic Spraying of Liquid 共Wiley, New York, 1998兲, p.aip.org in the directory/epaps/. See the E-PAPS homepage for more
p. 21. information.

Downloaded 18 Feb 2013 to 142.150.190.39. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions