JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 6, DECEMBER 2003 855
Silicon Micromachined Hollow Microneedles for
Transdermal Liquid Transport
Han J. G. E. Gardeniers, Regina Luttge, Erwin J. W. Berenschot, Meint J. de Boer, Shuki Y. Yeshurun, Meir Hefetz,
Ronny van’t Oever, and Albert van den Berg
Abstract—This paper presents a novel process for the fab-
rication of out-of-plane hollow micro needles in silicon. The
fabrication method consists of a sequence of deep-reactive ion
etching (DRIE), anisotropic wet etching and conformal thin film
deposition, and allows needle shapes with different, lithography-
defined tip curvature. In this study, the length of the needles
varied between 150 and 350 micrometers. The widest dimension
of the needle at its base was 250
m
. Preliminary application
tests of the needle arrays show that they are robust and permit
skin penetration without breakage. Transdermal water loss
measurements before and after microneedle skin penetration are
reported. Drug delivery is increased approximately by a factor of
750 in microneedle patch applications with respect to diffusion
alone. The feasibility of using the microneedle array as a blood
sampler on a capillary electrophoresis chip is demonstrated.
[999]
Index Terms—Deep-reactive ion etching (DRIE), diagnostics,
drug delivery, microneedle, point-of-care, transdermal.
I. INTRODUCTION
N
OWADAYS, typical routes for drug delivery are either
through hypodermic needles, via (iontophoretic) patches
or by oral or respiratory administration, while diagnostic sam-
pling in most cases requires extraction of blood through a hy-
podermic syringe needle, followed by analysis of blood compo-
nents in a specialized laboratory environment. These methods
all have some disadvantages, depending on the particular drug
that isto be administered, or the analysis that is to be performed.
For example, in some cases the methods are too slow, not effec-
tive, or wasteful with respect to the volumes of blood or other
substances that are consumed, or receive insufficient patient’s
compliance, because of pain and skin irritation or skin damage.
The introduction of MEMS offers exciting opportunities
to advance the minimally invasive medical field. In addition,
miniaturization of analysis methods enables the development
of versatile portable equipment for “Point-of-care” monitoring
Manuscript received January 24, 2003; revised August 7, 2003. This work
was supported by NanoPass Ltd., Israel. Subject Editor K. D. Wise.
H. J. G. E. Gardeniers was with Micronit Microfluidics BV, En-
schede, The Netherlands. He is now with the MESA
+
Research
Institute, University of Twente, 7500 AE Enschede, The Netherlands
(e-mail: j.g.e.gardeniers@utwente.nl).
R. Luttge, E. J. W. Berenschot, M. J. de Boer, and A. van den Berg are with
the MESA
+
Research Institute, University of Twente, 7500 AE Enschede, The
Netherlands.
S. Y. Yeshurun and M. Hefetz are with NanoPass Technologies, Ltd., Haifa
31043, Israel.
R. van’t Oever is with Micronit Microfluidics BV, 7500 AE Enschede, The
Netherlands.
Digital Object Identifier 10.1109/JMEMS.2003.820293
Fig. 1. Two types of microneedles. Top: in-plane design. Bottom: out-of-plane
design. Drawings are cross sections along the length of the needles.
and treatment of patients. Ultimately, drug-on-demand possi-
bilities, using micro system technology to integrate monitoring
and dispensing components in an intelligent feedback system
that is so small that it can be carried on the body of the patient
without obstructing his movement, may become feasible.
In this paper we focus on the micromachining of an array
of hollow microneedles for transdermal liquid transfer. Such
hollow needles can be used for body fluid sampling (e.g., blood
or interstitial fluid) or drug delivery. Drug infusion applications
require that the drug is injected into the viable epidermis, which
forms a cell layer extending to a depth of 60–130
below the
stratum corneum (the 10–20
thick outer layer of the skin,
which forms the main barrier for transport of drugs across the
skin). For blood withdrawal, a deeper penetration is generally
required, to ca. 1 mm below the surface of the skin, into the
1–2 mm thick dermis that contains the blood vessels and the
nerve ends.
The basic idea behind the microneedle approach is that due to
the small size of the needles, tissue damage will be limited and
pain sensation can be reduced [1] or even completely avoided.
To obtain the necessary fluid flow (generally of the order of
below 1
per hour up to ca. 100 per hour) for acceptable
pumping pressures, a large number of microneedles on the same
device area can be used.
Several approaches to the micromachining of this type of de-
vice are known, and roughly these can be divided in in-plane
and out-of-plane designs, the plane in this case being the sur-
face of e.g., a silicon wafer. Fig. 1 shows schematic drawings of
both these needle types. The in-plane version is the most conve-
nient to fabricate with state-of-the-art planar technology, com-
prising surface micromachining and different techniques of (sil-
1057-7157/03$17.00 © 2003 IEEE
856 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 6, DECEMBER 2003
icon) etching, and creates a good degree of flexibility with re-
spect to different needle designs. Illustrative examples are the
sophisticated hollow neural probes of Wise et al., fabricated by
anisotropic wet etching of a silicon substrate combined with
deep diffused boron etch stops and having CMOS electronic
circuitry integrated on the chip [2], [3], the microneedles man-
ufactured by anisotropic wet etching of a silicon substrate, with
surface micromachined polysilicon-based fluidic channels on
them, by Pisano’s group [4], [5], which were later refined by
the use of silicon-on-insulator substrates and isotropic etching
[6], the hollow micromachined needle arrays by Brazzle et al.
[7], [8], and the surface micromachined polysilicon micronee-
dles with permeable polysilicon on one side which serves as a
microdialysis membrane [9], [10].
The density of needles that can be obtained with the in-plane
methodis limited, sinceonlyone rowofneedlescan be madeper
chip. Modern analytical equipment for blood analysis requires
at least 5
and more typically 50 of blood for a reliable
analysis. Therefore, for blood withdrawal, with the typical mi-
croneedle inner diameters of10–100
reported in literature, a
higher needle density may be desirable in order to achieve suf-
ficient liquid flow. Furthermore, if needle diameters are made
smaller in order to limit pain sensations, a larger number of par-
allel microneedles may give the necessary redundancy in case
some of the needles for whatever reason get blocked during op-
eration. Although it is possible to achieve high needle densi-
ties by hybrid integration of in-plane needle arrays [11], from a
technological point of view the out-of-plane versionprovidesan
easier route to obtain a large density of needles per chip, while
the fact that the needles in this case are backed with a few
area flat plate with a thickness of less than 1 mm makes this
configuration suitable for integration in a patch to be worn on a
patient’s body.
Published early examples of out-of-plane micromachined mi-
croneedles for transdermal applications [12]–[14] have as a dis-
advantage that their flat hollow tips tend to punch the skin, with
the risk that the punched material may at least partially obstruct
liquid flow through the needle. Therefore it is preferred that the
flow channel through the needle is positioned off-center from
the needle tip. Very promising results were obtained by Stoeber
and Liepmann [15], who used directional reactive ion etching
(RIE) to define a narrow flow channel through a silicon wafer
and thin film protection of this channel followed by isotropic
etching from the other side of the wafer to fabricate such a
needle. A totally different and equally promising approach was
followed by Griss and Stemme, who, using the special etching
properties of RIE combined with passivation steps in-between
the etching cycles, fabricated out-of-plane microneedles with
openings on the side, by which clogging due to punching should
be avoided [16].
In this paper we present a fabrication procedure, in which
both anisotropic wet etching and directional (deep) reactive ion
etching processes are combined, to achieve out-of-plane mi-
croneedle structures in array format [17] (the process and struc-
ture of the needles are patent pending [18]). The needle design
anticipates a good mechanical strength and adequate skin pen-
etration properties, while the needles contain flow channels po-
sitioned off-center of the needle tips. Needles with different and
Fig.2. Microneedlefabricationsequence.Thedrawings ontheleftgivea cross
section of the structure after each processing step, along the dotted line in the
top view drawings that are shown on the right. The thick black line represents
a silicon nitride coating that is used as a protection layer during KOH etching.
Refer to text for details.
well-definedtip shapeswerefabricatedand are discussed below.
The feasibility of the needles for transdermal water loss, drug
delivery and diagnostics are presented.
II. F
ABRICATION PROCESS
Fig. 2 describes the fabrication process for the out-of-plane
needle design. The method builds on the directional deep
RIE work at cryogenic temperatures that was previously
developed by us to fabricate buried microchannels in silicon
[19] and anisotropic wet etching of silicon as performed by
Albrecht et al. to fabricate AFM tips [20], and requires only
two lithographic masks. More details about the possibilities of
cryogenic RIE and the procedures that are used to achieve deep
structures in a silicon substrate, were recently published in this
journal by us [21].
Essential features of the design are that the location of the
flow channel opening does not coincide with the needle tip and
in our case can be positioned freely. This flow channel is etched
as follows: The fabrication process starts with a RIE step using
plasma chemistry in an Oxford Plasmalab 100 machine
(typical etching rate 1.5
) by which in a silicon
substrate a flow channel (see Fig. 2a) is etched. This flow
channel through the needle does not extend to the other side
of the substrate, but is etched only as deep as is necessary to
achieve the desired needle length. The length is defined by the
depth of slot Fig. 2b, which is etched in the same process as
the flow channel Fig. 2a.
Next,achannel Fig. 2cthatconnects the flowchannelthrough
the needle to the backside of the wafer is etched by RIE with
GARDENIERS et al.: SILICON MICROMACHINED HOLLOW MICRONEEDLES FOR TRANSDERMAL LIQUID TRANSPORT 857
the same parameters as in the RIE step described above. In this
step, it is possible to include a fluidic channel structure on the
backside of the substrate, e.g., for sample treatment or charac-
terization, or drug storage purposes. Furthermore, positioning
of channel Fig. 2c, which can have a much wider cross-section
than the flow channel through the needle, is not critical, and
other etching steps may be used instead of the RIE used in this
study. The requirements are that the two channels connect, and
that the channel from the backside does not connect with the
slot.
Subsequently the inner surfaces of the holes and the slot are
coated with a conformal layer that is resistant against KOH.
In our case we used low-pressure chemical vapor deposition
(LPCVD) of a low-stress (ca. 300 MPa [22]) silicon-rich sil-
icon nitride layer with a thickness of 500 nm. Conditions for
this deposition step were: gas flows 70 sccm
and 18
sccm
, temperature 850 , pressure 200 mTorr.
After removal of the protective layer at the top surface of
the wafer with a directional
-plasma RIE process [19], in
Fig. 2, step 4 anisotropic wet etching is performed, which leaves
a structure boundby a slow-etching
plane on one side, in-
dicated by Fig. 2d. A similar method was used by Albrecht et
al. to fabricate AFM tips [20]. The coated slot sidewalls pre-
vent etching from the sides. Finally, the protective silicon ni-
tride layer is stripped in a 50% Hydrogen Fluoride (HF) solu-
tion. This etchant removes the silicon nitride layer at a rate of
ca. 4 nm/min, but since the selectivity of silicon nitride etching
with respect to single-crystalline silicon is extremely high, the
exact etching time is not critical.
Fig. 3(a) shows a scanning electron microscope (SEM) pho-
tograph of the resulting structure after in Fig. 2, step 4. Fig. 3(b)
gives a schematic drawing of a similar structure as shown in
Fig. 3(a), also after Fig. 2, step 4. It can be seen that the needle
boundaries are determined by the slow-etching
plane
and the nonetching silicon nitride walls of the slot. The needle
tip shape and tip position are determined by the alignment of
the crystallographic directions in the silicon with the design of
the slot. The advantage of this method over that of Stoeber et
al., who etched the connecting hole through the wafer, starting
from the backside of the substrate up to the surface at which
the needle tip is defined later, is that the essential lithographic
definition of respective hole (flow channel through needle) and
needle tip positions is done on the same side of the wafer. There-
fore in our case back-to-front alignment is less critical, and the
shape of the flow channel at the position where it really matters,
i.e., at the needle surface, is precisely defined.
III. F
ABRICATION RESULTS
Fig. 4 shows a typical result of a 350 high microneedle
with a triangular tip shape, a base of 250
, and a maximum
hole width of 70
. The center of the elliptical flow channel
in the needle is positioned ca. 40
from the tip of the needle.
It can be seen that the vertical sidewall of the needle has a rough
surface,whichis caused by photolithographic mask edge imper-
fections in combination with the characteristics of the applied
RIE process. It is unknown whether this roughness will affect
the performance of the needles for skin penetration.
(a)
(b)
Fig. 3. (a) The slot and flow channel, bound by silicon nitride walls that are
resistant against KOH etching, after step 4 of the process sequence in Fig. 2b
Schematic drawing of slot walls, flow channel walls, andremaining needle. The
dark gray area is the
Si
f
111
g
surface, the light gray area is the side face of the
needle as it is defined by the walls of the silicon nitride coated slots.
Fig. 4. SEM picture of a 350
m
high microneedle, with a base of 250
m
(measuredinwidest direction). The ellipticalflowchannelis70
m
inits widest
direction.
The needle length as it was fabricated here is sufficient for
most drug infusion applications. The fabrication method pre-
sented here also allows needles with the appropriate length for
blood withdrawal, however in our case the length was limited
by the thickness of the 100 mm diameter silicon wafers, which
was approximately 530
.
858 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 6, DECEMBER 2003
Fig. 5. Close-up of the tip of the needle of Fig. 4. Note the steps present on
the surface.
Fig. 6. Array of needles of Fig. 4, with a pitch of 555
m
.
Fig. 5 shows a close-up of the sloped sidewall of the needle,
which is formed by a silicon surface that is oriented along the
plane. The picture shows a number of very low steps,
originating from the tip of the needle that was originally bound
by the protective layer of the slot. These steps are a typical fea-
ture of the etching process on an atomically flat crystal plane
being in contact with a nonetching material [23], and are also
observed on the pyramidal hillocks that frequently form on sil-
icon
surfaces during etching in KOH solutions [24]. Due
to these “step trains,” the slope of the needle surface is some-
what less than that of a perfectly oriented
plane. This
has less to do with the fact that the top of the needle has been
exposed longer to the etchant than the base, but is due to the
nucleation of etching steps at the lines where the silicon
surface meets the silicon nitride walls. This effect was studied
in detail and explained in a recent paper by van Veenendaal et
al. [25].
The presented fabrication method allows a high needle
density, see Fig. 6, with excellent uniformity across the wafer
surface. Other needle tip designs can easily be achieved by
changing the design of the slot, see Fig. 7 which shows a
microneedle with a round tip shape.
As was learned from discussions with several European
silicon micromachining foundries, the procedures used for the
fabrication of the needles can be developed into a large-volume
Fig. 7. SEM picture of a microneedle with a tip with large radius of curvature.
Fig.8. Measurement of transdermalwaterloss with andwithout a microneedle
chipwith a6by 6arrayof 350
m
height.A: without chip;B: chip withbackside
toskin; C: withneedlearray inserted (atthetime indicatedbythe arrow);D: with
needle chip removed.
production process. Smaller volume production runs have
already been performed (Micronit Microfluidics B.V.).
IV. P
ERFORMANCE TESTS
The effectiveness of the microneedle array for transdermal
application was tested by measuring the transdermal water loss
(TEWL) on human skin. Transdermal water loss rates can be
used to monitor changes in the barrier function of the stratum
corneum. The test was conducted using standard CORTEX Der-
maLab equipment [26] which has a probe that is placed on the
skin; the probe contains two humidity sensors that measure the
water vapor pressure gradient in the direction perpendicular to
the skin, which gradient is a measure for the TEWL [27]. En-
vironmental conditions during the tests were: relative humidity
(RH) 29.1%, temperature 22.0
.
Fig. 8 shows a typical result. The test procedure was as fol-
lows: first the TEWL (in g
) was measured without
chip (see zone A in Fig. 8) at intervals of 1 min. The 1-min in-
terval is necessary to ensure that steady-state conditions are ob-
tained inside the probe, which in most cases take 30–45 s [28].
Next, a reference test was performed, in which the microneedle
GARDENIERS et al.: SILICON MICROMACHINED HOLLOW MICRONEEDLES FOR TRANSDERMAL LIQUID TRANSPORT 859
Fig. 9. Drug plasma concentrations obtained after 6 h application of an
increasing number of microneedles (preliminary results in Sprague–Dawley
400 gr. Male rat).
chip was repetitively attached with its backside to the skin sur-
facefor 10s at thesameskinpositionas where thepreviousmea-
surements without the chip were done, and 50 s after removal
of the chip the TEWL was measured (see zone B in Fig. 8). It
canbe seen that no changeinthe TEWL occurs forthissituation.
Thiswas followedby a singletestinwhich the chip wasattached
for 10 seconds with its front side to the skin, such that the mi-
croneedles penetrated the skin, again at the same skin position,
50 s after which the TEWL was measured again (see zone C
in Fig. 8), followed by repeat of measurements alike those in
zone A (see zone D in Fig. 8). It follows from Fig. 8 that when
the needles penetrate the skin the TEWL increases with a factor
of 2.3, it is observed that after removal of the microneedles the
skin recovers and the TEWL goes back to normal within a few
minutes.
Preliminary drug infusion tests were performed by mea-
suring the diclofenac plasma levels in a test animal after treat-
ment with the microneedle device. The common administration
route of diclofenac is oral, however, this type of administration
has disadvantages, such as fast first-pass metabolism and nega-
tive side effects [29]. Alternative routes of diclofenac delivery
have been proposed, and in this regard the skin has become
increasingly important. Since diclofenac is not extensively ab-
sorbed through the skin due to its hydrophilic nature [30], it is
expected that for this particular drug a microneedle approach
will be beneficial.
The test was done with a microneedle chip mounted in a
patch-like prototype (NanoPass Ltd., Israel) that was attached
to the skin of the animal by an adhesive layer. Fig. 9 compares
the infusion process using different microneedle arrays with the
direct diffusion through the skin. It can be seen that drug infu-
sion increases with the number of microneedles on the chip.
A more comprehensive test was performed with insulin that
was transdermally delivered via an insulin pump at a rate of
1 U (U stands for the internationally accepted insulin unit of
0.045 mg of the pure crystalline insulin product) per hour to a
small number of rats with diabetes, with a similar microneedle
patch prototype as used for the diclofenac. Fig. 10 shows the
change in glucose level in percentage of the baseline for dif-
ferent treatment techniques, where “inverted chip” means that
the chip was turned with its back to the skin, resulting in trans-
port through the same effective flow area but without mechan-
ical assistance of the needles.
Fig. 10. Effect of insulin delivery on basal glucose levels in diabetic rats
(preliminary results in small numbers of Sprague–Dawley 400 gr. Male rats);
regular Insulin (Humulin R, Ely Lilly) was delivered via insulin pumps at
1 U/h through microneedle arrays, and compared to conventional subcutaneous
delivery. The longer needles are 350
m
, the shorter ones 150
m
long.
A last performance test consisted in the use of the silicon mi-
croneedle array as a diagnostic device. Since the microneedles
as fabricated in this study were not long enough for transdermal
blood withdrawal, the test was performed with blood collected
after puncturing a human finger with a Haemolance (Haemedic
AB, Sweden) disposable safety lancet, which has a puncture
depth of 1.8 mm.
Fig. 11(a) shows a schematic of the setup as used for
interfacing the microneedle array to a glass Capillary Elec-
trophoresis (CE) chip of 2 cm separation length with a channel
cross section of 60
6 with end-column integrated
electrodes for conductivity detection. The silicon needle chip
was clamped onto the CE sample inlet via a PDMS seal.
In a first run, a calibrated solution of alkali ions containing
10 mM
, , and was electrokinetically loaded
through the microneedles into the chip. A standard pinching
procedure was applied to define and inject a sample plug, which
was subsequently separated by Capillary Zone Electrophoresis
(CZE). 20 mM 2-(N-morpholino)ethanesulfonic acid (MES;
Sigma, Steinheim Germany) plus 20 mM histidine (His; Fluka,
Buchs Switzerland) was used as the background electrolyte.
(BGE) Details on the CZE procedure were described earlier
[31]. A reference experiment was performed with the same
solutions, without the microneedle chip. Fig. 11(b) shows
the resulting electropherograms. In both experiments the
concentration (peak area) and migration time of the alkali ions
are nearly the same, indicating that no dilution occurs after
transport through the needles and confirming that the needle
array does not constitute a restriction to ionic transport.
In a second experiment, ca. 30
of the collected blood was
dispensed on the microneedle chip just before the separation
experiment, which was run with an identical voltage scheme.
Fig. 11(c) compares the calibration graph for sampling through
the needles of Fig. 11(b) with a measurement of the ions in
blood. For the blood sample, two peaks were detected and
identified as potassium and sodium. The ratio of concentra-
tions of
and in whole blood, which is ca. 4:140 in
blood plasma and serum, here is found to be approximately 1:1,