Accepted Manuscript

Title: Disposable MEMS optrode array integrated with single

LED for neurostimulation

Authors: Sunghyun Yoo, Hongkyun Lee, Sang Beom Jun,

Yong-Kweon Kim, Chang-Hyeon Ji

PII: S0924-4247(17)31727-2
DOI: https://doi.org/10.1016/j.sna.2018.02.039
Reference: SNA 10661

To appear in: Sensors and Actuators A

Received date: 26-9-2017

Revised date: 6-2-2018
Accepted date: 26-2-2018

Please cite this article as: Yoo S, Lee H, Jun SB, Kim Y-K, Ji C-H, Disposable MEMS
optrode array integrated with single LED for neurostimulation, Sensors and Actuators:
A Physical (2010), https://doi.org/10.1016/j.sna.2018.02.039

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Disposable MEMS Optrode Array Integrated with Single LED for Neurostimulation

Disposable MEMS Optrode Array Integrated with Single LED for

Neurostimulation
Sunghyun Yoo,a Hongkyun Lee,a Sang Beom Jun,b Yong-Kweon Kim,a and Chang-Hyeon Ji,b,*

a
Department of Electrical and Computer Engineering, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul
08826, Republic of Korea
b
Department of Electronic and Electrical Engineering, Ewha Womans University, 52 Ewhayeodae-gil, Seodaemun-

T
gu, Seoul 03760, Republic of Korea

IP
E-mail: cji@ewha.ac.kr

R
SC
Highlights:
 An LED-waveguide-based neural probe array using single commercial µLED presented

 System comprises of a disposable MEMS optrode array and a reusable unit with µLED and circuitry

 U
A 4×4 array of 6 mm-long optical fibers are assembled with thermally reflowed square-shaped glass microlenses
N
 Delivered optical power and light delivery efficiency analyzed
A
M

Abstract. We present an LED-waveguide-based neural probe array with the highest total light power delivery from a single
commercial LED. A separable configuration between the LED and the waveguide has been proposed and verified experimentally
for a disposable probe tip. The system comprises of a disposable MEMS optrode array and a reusable unit. The optrode array
D

consists of a 4×4 array of 6 mm-long optical fibers assembled with thermally reflowed square-shaped glass microlenses. The
TE

reusable unit includes a domed top LED and driving circuitry. Measured light delivery efficiency of the optrode array and the
total power efficiency of the probe were −10.6 dB and −21.6 dB, respectively.
EP

1. Introduction

Recently, optogenetics has gained increasing research interest due to the unique characteristics of cell-type
CC

selectivity and the high spatial and temporal resolution for neural stimulation and probing. In optogenetics, activities

of the neural cells can be controlled by photostimulation of genetically modified neurons with a specific wavelength
A

[1]. Researches have been carried out in mainly two representative branches, which are the development of opsin

genes to modify the neurons and the development of control interface, which includes light sources and waveguides

for light delivery to targeted neurons. Up to the present, various types of light delivery probes with enhanced

performance and extended functionality have been reported. Examples include silicon-based µLED (micro Light
Emitting Diode) [2], polymer waveguides [3, 4], glass optrode array [5], and fiber-optic probes [6, 7].

Fiber-optic probes were the principal drivers of the light delivery experiment in early stage development. A bare

optical fiber with tapered tip has been directly utilized as a light delivery probe. Direct illumination via optical fiber

inherits the advantages of conventional fiber-optic systems such as coherence with narrow bandwidth and low light

loss. Availability of various peripheral tools equipped with mature technologies is another advantage. Compared with

other type of approaches, fiber-based configuration provides superior scalability in probe length, which can be critical

T
in some applications approaching deep brain regions. In terms of biocompatibility, many researchers have already

IP
experimentally validated the safety of fiber insertion into a living body and direct contact to neural cells. Nevertheless,

direct insertion of µLED into brain still needs utmost care on both localized heat dissipation and electrical leakage.

R
Most polymers are not fully validated in terms of biocompatibility and long-term stability, which are significant in

SC
chronical applications [8, 9].

U
Despite the various advantages mentioned above, conventional fiber-based configuration still has several

fundamental limitations [6, 7]. First, it is difficult to implement an untethered and wirelessly controlled system due
N
to the need for an external light source and resulting mechanical robustness issue of the fiber. The problem becomes
A

even more complicated when a bundle of fibers is used or when the fibers are integrated along with additional
M

electrodes for read out. Difficulties with forming an array and relatively low spatial resolution of the fiber are also
D

challenging issues. Several approaches have been proposed as the solutions to these issues, while maintaining the

advantages of the fiber-based configuration [10-12]. These probes are implemented through integration of segmented
TE

optical fibers and µLEDs. Direct integration of a two-dimensional (2D) array of µLEDs and fibers have been
EP

demonstrated and the feasibility for a wirelessly controllable system have been tested [8, 9]. Wirelessly illumination

of a single ferruled fiber has been demonstrated successfully [12]. These approaches are basically based on
CC

monolithic integration of the fiber and the light source, namely the butting method [13]. It is well known that direct

butting of a fiber to a Lambertian light source maximizes the coupling efficiency when the areal ratio of the fiber
A

cross-section to the luminous surface of the LED is less than unity, as demonstrated in above devices.

For the optical neural probes used in optogenetics, delivery of sufficient amount of light is critical. As the

maximum power of light is predetermined by the performance of the LED, effective delivery of light becomes more

crucial for a wirelessly controlled system operated by limited amount of stored energy. A breakthrough technology
Disposable MEMS Optrode Array Integrated with Single LED for Neurostimulation

to overcome the limitations of conventional butting method would be required, as the method still wastes a

considerable amount of light even at the ideal maximum efficiency condition. For in vitro stimulation with a blue

light source, light intensity required to activate the light-sensitive opsin is in the range of 1-5 mW/mm2 [14, 15].

However, in practice, the experiment involves a couple of undesirable optical phenomena, which are absorption and

scattering in the brain tissue. Previous experiments showed that the intensity decreases to below 10% of the value at

the tissue surface within only a few hundreds of micrometers [16, 17]. Moreover, considering the misalignment

T
between the targeted neurons and the probe tip after insertion, in vivo neurostimulation would require a much higher

IP
optical power at the probe tip compared to the previously reported in vitro test requirements. This is indirectly

supported by the fact that the average maximum instantaneous optical power used in several animal experiments

R
reported so far exceeds 0.9 mW [3, 8, 9, 18-23], which corresponds to the light intensity of 28.6 mW/mm2 when light

SC
is delivered via a single 200 μm-diameter optical fiber.

U
Moreover, cost issue in the conventional LED-waveguide-based probes cannot be overcome with the systems

based on monolithic integration of µLEDs and waveguides because as substantial portion of the system including
N
LED and waveguides has to be disposed after biological experimentation. The issue becomes even more challenging
A

when the port count of the fibers and µLEDs increases or additional structures are integrated for advanced
M

functionalities. Although it might be possible to use the probe repeatedly for another animal experiments after
D

appropriate cleaning and sterilization, the potential risk of contamination and infection cannot be fully avoided.

In this paper, we present a new type of light delivery probe array system for optogenetic neurostimulation, where
TE

the total light power delivered at the probe tip from a single commercial LED has been substantially improved via a
EP

MEMS (Micro-Electro-Mechanical System) optrode array (MOA). Moreover, the MOA can be fully separated from

the LED circuitry. The MOA consist of a 4×4 array of square-shaped glass microlens monolithically integrated with
CC

silicon guiding structure and assembled with high numerical aperture (NA) optical fibers. By utilizing the MEMS

fabrication technology, we have fabricated disposable optrode array while improving the total light power delivery.
A

Proposed approach can potentially lead to a new alternative for the design and fabrication of wireless fiber-optic

neural probe for in vivo optogenetic applications.

2. Design
2.1. Overall design

Figure 1 shows the schematic diagram of the proposed light delivery probe in cross-sectional view. The probe

comprises two mechanically separable parts, which are the reusable LED driver unit and the disposable MOA. Both

parts are packaged with customized polycarbonate housings and a set of latch structures is utilized for non-permanent

assembly of two parts. During animal experiments, probes are inserted into the brain and the housing of the disposable

MOA is fixed to the skull. Upper part of the housing, which contains LED and control circuitry, can be separated and

T
reused after the experiment. In addition to the reusability for minimization of wasted part, the device can be equipped

IP
with added functionalities, which were impossible to implement in previous configurations. By replacing the reusable

unit, light source with different wavelength can be used without having to replace the inserted probes.

R
SC
U
N
A
M
D
TE

Fig. 1. Schematic diagram of the proposed light delivery probe comprised of disposable MEMS optrode array and
EP

reusable LED unit.

CC

The reusable top unit consists of a domed top LED and its driver circuitry. The disposable bottom unit with a MOA

consists of a 4×4 array of high NA optical fibers and matching array of square-shaped glass microlens. The microlens
A

is formed on a silicon substrate and the optical path is secured via hole formed through the substrate. The via hole

works as a fiber guide structure which provides both lateral and vertical alignment of the fiber to the microlens and

substrate. Although only 4×4 array has been demonstrated in this work, proposed fabrication approach can be

advantages in terms of the scalability of the probe array size, as the fibers are passively assembled with silicon
Disposable MEMS Optrode Array Integrated with Single LED for Neurostimulation

substrate with integrated glass microlens array. Probe array with large port count can enhance the total output power

and potentially be advantageous in researches on large area neural systems or applications targeted for larger

mammals such as non-human primates [24, 25].

Maximum achievable efficiency in coupling of an optical fiber with light source having Lambertian emission

pattern is known to be proportional to the square of the fiber NA and the areal ratio of the fiber cross-section to the

LED luminous surface, while unaffected by additional collimating optics when the areal ratio is less than unity [13].

T
In order to improve the light coupling efficiency above this limitation, we have utilized a LED whose luminous area

IP
is smaller than the total light receiving area of the fiber array. The method is analogous to the well-known approaches

of enhancing light coupling efficiency by either coupling the bulb-ended fiber to planar LED [26] or deploying a

R
spherical microlens between the fiber and the domed top LED [27]. Proposed method enlarges the light receiving

SC
area rather than reducing the size of the LED to less than a single fiber cross-section by adopting microlens array.

U
This method has the advantage of simple fabrication process by utilizing commercially available LEDs with

necessary requirements.
N
For the proof-of-concept, light delivery probe consisting of sixteen high NA optical fibers (FP200URT, Thorlabs,
A

Inc.) with core diameter of 200 µm and NA of 0.5 has been proposed. The fibers are arranged in a square 4×4 array
M

configuration with fiber pitch and a gap of 320 µm and 95 µm in both directions, respectively. Although it is true
D

that relatively large diameter of the optical fiber and narrow gap between the optical fibers may cause damage to the

tissue during the deep insertion, all the geometric parameters can be modified if required, at the cost of modification
TE

of optics design.
EP

In contrast to bare LEDs, domed top LEDs generally have narrower viewing angle down to a few degrees and

higher luminous intensity in the region of interest due to integrated converging microlens. Domed top LED
CC

(XZFBB78W, SunLED Electronic Co., Ltd.) with p-n junction size of approximately 200×200 µm2 has been selected

considering the total light receiving area of the fiber array which is 0.50 mm2. In reference to [13], calculated
A

achievable maximum efficiency is over unity which indicates that ideal collimating optics could theoretically achieve

light delivery from LED to fiber distal end without any loss.

The diameter of the dome and viewing angle are 1.8 mm and ±10˚, respectively. As the selected LED has relatively

low viewing angle, emission from the source has been assumed to be a plane wave for design simplification. In this
case, areal ratio of the light receiving region of the microlens array to the light transmitting region of the LED

indicates the coupling efficiency between the LED and the MOA. Microlenses are arranged in order to maximize the

light delivery efficiency of the probe while satisfying the given geometrical and fabrication constraints. Footprint of

single microlens and the gap between the microlenses are 300×300 µm2 and 20 µm, respectively. Resulting coupling

efficiency between the LED and the MOA is calculated to be −2.47 dB.

Surface profile of the microlens is determined by computational modeling described in the following section.

T
However, both LED specification and geometrical dimension of the microlens can be modified depending on the

IP
type of the optical fiber used and configuration of the array required by application.

R
2.2. Computational modeling of thermally reflowed square-shaped microlens

SC
Square-shaped glass microlens array fabricated with thermal reflow process has been proposed. During the in vivo

U
experiment of neural cells, MOA contacts living cells on one end and the other end faces the LED in close proximity.

In addition to the high optical transparency, glass microlens can potentially be advantageous in terms of thermal
N
stability compared to polymer counterparts.
A

Despite the originally undesirable aspects, such as asymmetric radius of curvature and resulting light dispersion at
M

the focal plane, square-shaped microlens can be more efficient in terms of light delivery in proposed microlens and
D

probe array design. In contrast to that of a spherical microlens array, maximum fill-factor can be achieved with

square-shaped microlens array. Due to fabrication constraints, minimum loss in the gap between the nearby
TE

microlenses cannot be avoided as illustrated in Fig. 2. Compared to fill-factor of square-shaped microlens in lattice
EP

configuration, spherical microlens have relatively lower value of 69.0% and 79.7% in lattice and hexagonal

configurations, respectively.
CC
A
Disposable MEMS Optrode Array Integrated with Single LED for Neurostimulation

Fig. 2. Comparison of the effective light receiving area of the microlens arranged in array.

FEA (finite element analysis) simulation using computational fluid dynamics (CFD) and geometrical optics have

been sequentially carried out in order to obtain the optimized geometry of the microlens which can deliver maximum

light power from the LED to the fiber tip. Modeling and microfabrication process of the spherical glass microlens

using thermal reflow process have been validated in the previous work [28]. Despite the well-organized reports on

T
numerical modeling of thermally reflowed square-shaped microlens [29, 30], not many work has been done on the

IP
prediction of microlens profile, which requires modeling based on fluid dynamics.

As depicted in Fig. 3(a-b), square-shaped microlens is formed by annealing a square column at temperature higher

R
than the softening point of the lens material. In the simulation, two-phase level-set method has been utilized to find

SC
the stabilized geometry of the microlens as a function of the initial column height T ranging from 15 to 55 µm.

U
Dynamic viscosity of the molten borosilicate glass (BSG) was set to 1.82 MPa·s. The reliability of the result has been

validated through volume consistency before and after the thermal reflow process. Effective annealing time of 40
N
minutes has been determined by evaluating the saturation point of the maximum fluid velocity.
A

The light delivery efficiency has been evaluated by comparison between the initial power emitted to the microlens
M

and power at the fiber distal end determined by ray tracing simulation (Fig. 3(c)). In order to take the loss caused by
D

less than unity fill-factor into account, the silicon gap surrounding each microlens has been included in the geometry.

The light rays are assumed to be 470 nm-wavelength plane wave with uniform intensity distribution and normal
TE

incidence. Refractive index of the air, BSG, core and cladding of the optical fiber were set to 1.000, 1.472, 1.464,
EP

and 1.376, respectively. Microlens and the optical fiber were assumed to be in direct contact and Fresnel reflection

has been neglected in the simulation.

CC

Simulation results are summarized in Fig. 3(d). The results shows that the light delivery efficiency of the square-

shaped microlens is higher than that of the spherical microlens in the initial glass column height range of 30 µm to
A

40 µm, considering that the fill-factor of the spherical microlens (0.797) is the achievable maximum in the given

geometrical condition. From the results, T was set to 35 µm, at which normalized light delivery has been estimated

to be 0.83. Considering the marginal loss due to the silicon lattice surrounding the microlens, over 94% of the light

received by the microlens is transmitted to the distal end of the optical fiber, resulting in a coupling loss of −0.25 dB.
In addition to the near maximum light delivery efficiency, variation of efficiency is relatively less sensitive to the

possible initial glass column height variation due to fabrication imperfections. Loss of light due to Fresnel reflection

will occur mainly at the air/microlens and core/air interfaces, where the losses are −0.16 dB and −0.15 dB,

respectively. Consequently, considering the coupling efficiency discussed above, estimated total light delivery

efficiency of the MOA can be as high as −3.03 dB.

T
R IP
SC
U
Fig. 3. Simulation results of (a-b) the reflow modeling of square-shaped microlens, (c) ray tracing through the
N
reflowed microlens, and (d) normalized light delivery as a function of glass column height prior to the thermal reflow
A
process.
M

3. Microfabrication
D

3.1. Square-shaped glass microlens array with fiber structure

TE

Microfabrication procedure is illustrated in Fig. 4. The process starts with a 510 μm-depth cavity patterning on a

1 mm-thick 4 inch silicon wafer using metal mask and deep reactive ion etching (DRIE) (Fig. 4(a)). The cavity is
EP

filled with molten BSG (Borofloat 33, Schott AG) via anodic bonding and subsequent annealing at 850 ˚C for 8 hours

(Fig. 4(b)). After the first thermal reflow process, glass remaining on top of the cavity is removed by chemical
CC

mechanical polishing followed by 40 μm-depth leveling down of glass layer using hydrofluoric acid solution (Fig.

4(c)). Plasma treatment is carried out on the surface of the BSG wafer prior to the bonding in order to suppress the
A

surface crystallization on the surface between the BSG and silicon [31]. Preventing the surface crystallization on both

BSG/air and BSG/silicon interface is critical in maintaining the optical transparency of the reflowed glass.

Crystallized BSG, cristobalite, has different physical and chemical characteristics, including altered coefficient of

thermal expansion (CTE), compared to amorphous BSG. As a result, cristobalites on the BSG/silicon interface cause
Disposable MEMS Optrode Array Integrated with Single LED for Neurostimulation

irregular glass geometry after the leveling process in Fig. 4(c), which directly affects the property of the microlenses.

Moreover, mismatch in CTE causes micro-cracks [31] on the microlens surface during the microlens formation step

(Fig. 4(e)), which significantly degrades the optical transmittance of the microlens. As depicted in Fig. 4(d), BSG

columns are formed using metal mask and DRIE. The second thermal reflow process is carried out at the same

temperature as the first reflow process for 40 minutes (Fig. 4(e)). Concurrent formation of the optical path and fiber

guide structure is carried out by backside DRIE. The diameter and depth of the silicon guide are 250 µm and 500 µm,

T
respectively. Finally, the microlens array is diced into 5×5 mm2 dies.

IP
Measurement results of the fabricated microlens array are summarized in Fig. 5. The scanning electron microscope

(SEM) image of the microlens array verified the successful fabrication of the microlens array. Fig. 5(b) shows the

R
microscopic image at focal plane of the microlens array when backlight is provided on the convex side of the

SC
microlens. The result is consistent with the SEM measurement validating the uniform optical characteristics among

microlenses. Surface profile of the microlens is measured with 3D surface profiler (μsurf, Nanofocus AG). Measured
U
center cross-section and the diagonal cross-section of the microlens array are shown in Fig. 5(c) and Fig. 5(d),
N
respectively. In both figures, the black line indicates the simulation result while the hollow blue dots indicate the
A

measured profile. The surface profiles are in good agreement with the simulation result. Measured average surface
M

roughness was 47 nm. Silicon grass formed in the gap region between the microlenses is due to imperfect etching of
D

silicon during the glass column formation step (Fig. 4(d)). Alternative etch mask materials for DRIE, such as silicon

oxide which is known to generate less silicon grass, can be utilized for better etch results. Despite the silicon grass
TE

formation in some regions, computational modeling and microfabrication results of the thermally reflowed square-
EP

shaped glass microlens array have been verified successfully.

CC
A
 Fig. 4. Microfabrication procedure of the square-shaped glass microlens array with fiber guide structure.

T
IP
Fig. 5. Measurement results of the fabricated square-shaped glass microlens array including (a) SEM image of the

R
microlens array, (b) microscopic image at the focal plane when light illuminated on the microlens, surface profile of

SC
(c) the center cross-section, and (d) the diagonal cross-section of the microlens array.

3.2. Fiber assembly procedure U

N
Optical fibers are prepared through several procedures before the assembly. First, the Tefzel coating of the fiber is
A

removed using general-purpose fiber stripper (T12S21, Thorlabs, Inc.). The fiber pieces which are approximately 16
M

mm-long are placed on a silicon V-groove substrate which is fabricated beforehand using silicon wet etch process.
D

The fibers are fixed to the substrate by spin coating a photoresist over the fibers. Fibers attached to the substrate are

uniformly diced into 6 mm-long pieces using dicing saw. Segmented fibers are released from the silicon V-groove
TE

by removing the photoresist in acetone solution and are cleaned with isopropyl alcohol solution in ultrasonic bath.
EP

Optical fibers and microlens array are assembled using additional silicon substrate and an aluminum aligner, as

shown in Fig. 6(a). The substrate has been fabricated by forming through-holes inside a silicon wafer with the same
CC

mask used in the formation of optical path in the microlens array (Fig. 4(f)). Substrate with through-holes has been

diced into 5×5 mm2 pieces, whose size was the same as that of the die with microlens array. The microlens array and
A

the silicon substrate are inserted into top and bottom socket of the aluminum aligner. The optical fibers are inserted

into silicon guide structure with index matching epoxy at the tip and are cured in the oven. Optical image of the

fabricated MOA is shown in Fig. 6(b). Optical fibers with consistent length are successfully assembled with the

microlens array. However, dicing process for the optical fiber should be optimized or additional polishing procedure
Disposable MEMS Optrode Array Integrated with Single LED for Neurostimulation

should be carried out in order to fully remove undesirable residual cladding near the facet of the optical fiber.

Fig. 6. Optical images of (a) the assembly apparatus, and (b) the fabricated MEMS optrode array. The scale bar

T
indicates 5 mm in both images.

R IP
4. Experimental results and discussion

SC
4.1. Light delivery efficiency

Light delivery efficiency of the MOA is analyzed through comparison of measured light output power with four

U
separate samples (A, B, C, D), as illustrated in Fig. 7. Sample A is the LED only condition, which is a reference point
N
for evaluating the light delivery efficiency of the MOA. Sample B is an arrangement in which the LED and a
A
separately fabricated aperture are combined to measure the amount of light reaching the microlens array. The aperture
M

is fabricated by patterning metal on a glass substrate with the same mask used in the silicon cavity formation (Fig

4(a)). Sample C comprises of the LED and the MOA, which corresponds to the proposed neural probe. Finally,
D

sample D consists of 4×4 optical fibers assembled within a silicon substrate with through-holes to evaluate the role
TE

of glass microlenses.

The power is measured with photodiode power sensor (S121C, Thorlabs, Inc.) and the power meter (PM100D,
EP

Thorlabs, Inc.). In order to evaluate the efficiency of the probe while excluding the loss caused by inaccurate
CC

alignment between MOA, LED, and the power sensor, all components are individually clamped to 6-DOF stages

controlled by micromanipulators. A thorough alignment of each component has been carried out until the light output
A

power has been maximized.Measurement result is summarized in Fig. 8. While the current applied to LED was

changed from 5 to 25 mA in DC, the light delivery efficiency of the MOA (C−A), remained constant at a level of

−10.6 dB. By assuming a uniform light power distribution among the fibers, the light delivery efficiency of a single

set of microlens and optical fiber is expected to be −22.6 dB. The result is comparable to previously reported results
which range between −25.9 to −16.3 dB [10, 11]. Note that the compared values are recalculated using the measured

light intensity results and specification of the optical fiber and LED mentioned in the work.

However, the superiority of the proposed probe has been sufficiently validated in terms of the total light power

delivery. In contrast to conventional configurations, proposed design is advantageous in constructing array of fibers,

considering that the numerical extension of the fibers directly increases the amount of light delivered with no

additional arrangement of LEDs. Total power efficiency, which is the ratio of the total light output power to the

T
applied electrical power, is crucial in the wirelessly controlled applications since the living animal experiment

IP
requires much higher light output power, compared to the in vitro experiment, with low power consumption. Total

light power efficiency of the proposed probe, which utilizes single LED, is measured to be −21.6 dB, when the

R
applied voltage and current are 3.3 V and 20 mA, respectively. This result is significant when compared with the

SC
power efficiency of conventional LED-waveguide-based devices, which ranges from −32.3 dB to −24.6 dB [9-11].

U
In addition, by adjusting the property and arrangement of the microlenses, for example enlarging the interval between

the fibers and size of the microlenses, the light delivery efficiency for single set of microlens and optical fiber can be
N
improved due to increased amount of converged light.
A
M
D
TE
EP
CC
A

Fig. 7. Illustration of four separate measurement condition for evaluation of light delivery efficiency of the

MOA.
Disposable MEMS Optrode Array Integrated with Single LED for Neurostimulation

T
IP
Fig. 8. Comparison of the measured light output power between the samples A to D.

R
SC
Estimated light intensity at the fiber tip is 1.06 mW/mm2, which is calculated by dividing the output power by total

U
area of the fiber core cross-section, when the applied current of the LED is 25 mA. Achieved intensity satisfies the

commonly required light level for microbial opsin activation [14, 32]. Maximum instantaneous intensity can be
N
improved either by pulse modulation of the LED or by utilizing a different type of LED, taking into account that the
A

maximum direct current rating of the present LED is 30 mA.

M

The effectiveness of the glass microlens array has been verified by comparing the measured light power of sample
D

C and D. Improvement of light delivery efficiency by 3.13 dB is well matched with the simulated expectation of 3.5

dB shown in Fig. 3(d). However, the amount of light actually transmitted from the LED to the microlens array,
TE

estimated by comparing the measurement results of sample A and B, was measured to be −6.04 dB, which is lower
EP

than the expected value of −2.47 dB by 3.57 dB. The degradation can be caused by the light emission pattern of the

utilized LED. Although a plane wave has been assumed in the simulation, domed top LED had relatively large
CC

dispersive light distribution and light leakage over the given viewing angle.

When the 3.57 dB of light loss is taken into account, additional 4.04 dB of light loss needs to be accounted for
A

considering the measured and predicted total light delivery efficiency of -10.6dB and −3.03 dB. One of the dominant

cause of loss can be attributed to the optical coupling loss at the rough end facet of the diced optical fiber. Figure 9

shows the optical image and measured surface roughness of the optical fiber facet. Average surface roughness

measured with 3D surface profiler after thermal deposition of aluminum on the facet was 1.272 μm. To our
knowledge, a precise analysis on the optical coupling loss at the end facet of an optical fiber with average roughness

in the order of micrometers has not been reported. However, additional loss of −4.04 dB is comparable with the

previously reported coupling loss of −3.88 dB at the end facet of diced polymer waveguide with surface roughness

of approximately 0.9 μm, measured with center wavelength of 850 nm [33].

T
R IP
Fig. 9. Optical image (a) and measured surface average roughness (b) of the end facet of diced optical fiber.

SC
4.2. Disposable configuration
U
N
The MOA and LED (Fig. 10(a)) are assembled with lower and upper housings (Fig. 10(b)). The silicon side of the

MOA facing the nerve cell is sealed with opaque epoxy to prevent possible light leakage and to avoid in-use
A

contamination of the optical components. The total weight of the disposable MOA and the reusable unit is 6.9 g.
M

Measured light output power is summarized in Fig. 10(c). In order to verify the stable operation during repetitive
D

assembly and separation, the output power has been measured for 5 times while changing the relative direction of

the top and bottom housings. Measured average light output power is 465.5 μW, which is 87.5% of the power
TE

measured with the optimized alignment via 6-DOF stages. Standard deviation of the output power was less than 2%.
EP
CC
A

Fig. 10. Optical images of (a) domed top LED with LED driver, (b) the overall device including the MEMS optrode
Disposable MEMS Optrode Array Integrated with Single LED for Neurostimulation

array and the reusable unit, and (c) measured light output power

5. Conclusion

In the present work, a light delivery probe for optogenetic neurostimulation has been proposed with a disposable

MEMS optrode array. By utilizing a microlens array and a domed top LED, the total light power delivery is improved

to a practical level for in vivo optogenetics experiment on living animal. The MEMS optrode array and the domed

top LED have been packaged in separate housings to realize a reusable light source and a disposable optrode array.

T
Proposed design successfully verified the feasibility and provided a solution to minimizing the cost for a large scale

IP
animal experiment for neuroscience studies. Moreover, in contrast to conventional configurations with a direct butt

coupling between the fiber and the LED, a collimating optics was adopted to enhance the light delivery efficiency of

R
the fiber-LED-based neural probe. To our knowledge, this is the first attempt to utilize microlenses in optical neural

SC
probe applications. For further improvement of the device performance, smoothing the end facet of the fiber should

U
be carried out in order to prevent significant degradation of light delivery efficiency.
N
Acknowledgement
A
This work was supported by the Center for Integrated Smart Sensors funded by the Ministry of Science, ICT &
M

Future Planning as Global Frontier Project (CISS-2012M3A6A6054204) and by Basic Science Research Program

through the National Research Foundation of Korea (NRF) funded by the Ministry of Education
D

(2015R1D1A1A09058627).
TE

References
EP

[1] Deisseroth, G. P. Feng, A. K. Majewska, G. Miesenbock, A. Ting, and M. J. Schnitzer, Next-generation

optical technologies for illuminating genetically targeted brain circuits, J Neurosci 26 (2006) 10380-10386.
CC

[2] A. N. Zorzos, J. Scholvin, E. S. Boyden, and C. G. Fonstad, Three-dimensional multiwaveguide probe array

for light delivery to distributed brain circuits, Opt Lett 37 (2012) 4841-4843.
A

[3] F. Wu, E. Stark, M. Im, I. J. Cho, E. S. Yoon, G. Buzsaki, K. D. Wise, and E. Yoon, An implantable neural

probe with monolithically integrated dielectric waveguide and recording electrodes for optogenetics

applications, J Neural Eng 10 (2013).

 [4] K. Y. Kwon, A. Weber, and W. Li, Varying-length polymer microneedle arrays fabricated by droplet

backside exposure, J Microelectromech S 23 (2014) 1272-1280.

[5] T. V. Abaya, S. Blair, P. Tathireddy, L. Rieth, and F. Solzbacher, A 3D glass optrode array for optical neural

stimulation, Biomed Opt Express 3 (2012) 3087-3104.

[6] E. Stark, T. Koos, and G. Buzsaki, Diode probes for spatiotemporal optical control of multiple neurons in

freely moving animals, J Neurophysiol 108 (2012) 349-363.

T
[7] S. Y. Chen, W. H. Pei, Q. Gui, Y. F. Chen, S. S. Zhao, H. Wang, and H. D. Chen, A fiber-based implantable

IP
multi-optrode array with contiguous optical and electrical sites, J Neural Eng 10 (2013).

[8] B. Fan, K. Y. Kwon, A. Weber, and W. Li, An implantable, miniaturized SU-8 optical probe for

R
optogenetics-based deep brain stimulation, in Engineering in Medicine and Biology Society (EMBC), 2014

SC
36th Annual International Conference of the IEEE(2014).

[9]
U
K. Y. Kwon, H. M. Lee, M. Ghovanloo, A. Weber, and W. Li, Design, fabrication, and packaging of an

integrated, wirelessly-powered optrode array for optogenetics application, Frontiers in systems neuroscience
N
9 (2015).
A

[10] M. Schwaerzle, P. Elmlinger, O. Paul, and P. Ruther, Miniaturized 3×3 optical fiber array for optogenetics
M

with integrated 460 nm light sources and flexible electrical interconnection, in Micro Electro Mechanical
D

Systems (MEMS), 2015 28th IEEE International Conference on (Estoril, 2015), pp. 162-165.

[11] J. G. Bernstein, B. D. Allen, A. A. Guerra, and E. S. Boyden, Processes for design, construction and
TE

utilisation of arrays of light-emitting diodes and light-emitting diode-coupled optical fibres for multi-site
EP

brain light delivery, The Journal of Engineering 1 (2015).

[12] S. T. Lee, P. A. Williams, C. E. Braine, D. T. Lin, S. W. M. John, and P. P. Irazoqui, A Miniature, Fiber-
CC

coupled, wireless, deep-brain optogenetic Stimulator, Ieee T Neur Sys Reh 23 (2015) 655-664.

[13] M. C. Hudson, Calculation of maximum optical coupling efficiency into multimode optical-waveguides,
A

Appl Optics 13 (1974) 1029-1033.

[14] E. S. Boyden, F. Zhang, E. Bamberg, G. Nagel, and K. Deisseroth, Millisecond-timescale, genetically

targeted optical control of neural activity, Nat Neurosci 8 (2005) 1263-1268.

Disposable MEMS Optrode Array Integrated with Single LED for Neurostimulation

[15] F. Zhang, L. P. Wang, M. Brauner, J. F. Liewald, K. Kay, N. Watzke, P. G. Wood, E. Bamberg, G. Nagel,

A. Gottschalk, and K. Deisseroth, Multimodal fast optical interrogation of neural circuitry, Nature 446 (2007)

633-U634.

[16] A. R. Adamantidis, F. Zhang, A. M. Aravanis, K. Deisseroth, and L. De Lecea, Neural substrates of

awakening probed with optogenetic control of hypocretin neurons, Nature 450 (2007) 420-U429.

[17] O. Yizhar, L. E. Fenno, T. J. Davidson, M. Mogri, and K. Deisseroth, Optogenetics in Neural Systems,

T
Neuron 71 (2011) 9-34.

IP
[18] K. M. Tye, J. J. Mirzabekov, M. R. Warden, E. A. Ferenczi, H. C. Tsai, J. Finkelstein, S. Y. Kim, A. Adhikari,

K. R. Thompson, A. S. Andalman, L. A. Gunaydin, I. B. Witten, and K. Deisseroth, Dopamine neurons

R
modulate neural encoding and expression of depression-related behaviour, Nature 493, 537-+ (2013).

SC
[19] B. Rubehn, S. B. E. Wolff, P. Tovote, A. Luthi, and T. Stieglitz, A polymer-based neural microimplant for

[20] U
optogenetic applications: design and first in vivo study, Lab Chip 13 (2013) 579-588.

B. Fan, K. Y. Kwon, R. Rechenberg, M. F. Becker, A. J. Weber, and W. Li, A hybrid neural interface
N
optrode with a polycrystalline diamond heat spreader for optogenetics, Technology 4 (2016) 15-22.
A

[21] T. I. Kim, J. G. McCall, Y. H. Jung, X. Huang, E. R. Siuda, Y. H. Li, J. Z. Song, Y. M. Song, H. A. Pao, R.
M

H. Kim, C. F. Lu, S. D. Lee, I. S. Song, G. Shin, R. Al-Hasani, S. Kim, M. P. Tan, Y. G. Huang, F. G.

D

Omenetto, J. A. Rogers, and M. R. Bruchas, Injectable, cellular-scale optoelectronics with applications for

wireless optogenetics, Science 340 (2013) 211-216.

TE

[22] H. Cao, L. Gu, S. K. Mohanty, and J. C. Chiao, An integrated mu LED optrode for optogenetic stimulation
EP

and electrical recording, Ieee T Bio-Med Eng 60 (2013) 225-229.

[23] K. Y. Kwon, B. Sirowatka, A. Weber, and W. Li, Opto-mu ECoG array: A hybrid neural interface with
CC

transparent mu ECoG electrode array and integrated LEDs for optogenetics, Ieee T Biomed Circ S 7 (2013)

593-600.
A

[24] T. V. F. Abaya, M. Diwekar, S. Blair, P. Tathireddy, L. Rieth, and F. Solzbacher, Deep-tissue light delivery

via optrode arrays, J Biomed Opt 19 (2014).

[25] M. R. Warden, J. A. Cardin, and K. Deisseroth, Optical neural interfaces, Annu Rev Biomed Eng 16 (2014)

103-129.
[26] B. S. Kawasaki, and D. C. Johnson, Bulb-ended fiber coupling to LED sources, Opt Quant Electron 7 (1975)

281-288.

[27] D. R. Gabardi, and D. L. Shealy, Coupling of domed Light-Emitting-Diodes with a multimode step-index

optical fiber, Appl Optics 25 (1986) 3435-3442.

[28] S. Yoo, J. Y. Jin, J. G. Ha, C. H. Ji, and Y. K. Kim, Two-dimensional optical scanner with monolithically

integrated glass microlens, Journal of Micromechanics and Microengineering 24 (2014).

T
[29] M. He, X. C. Yuan, K. J. Moh, J. Bu, and X. J. Yi, Monolithically integrated refractive microlens array to

IP
improve imaging quality of an infrared focal plane array, Opt Eng 43 (2004) 2589-2594.

[30] S. M. Kim, M. Choi, H. Kim, J. Lim, and S. Kang, Development of method to model surface of reflowed

R
square microlens for image sensor, Electronics Letters 44 (2008) 492-U410.

SC
[31] S. Yoo, C. H. Ji, J. Y. Jin, and Y. K. Kim, Suppression of surface crystallization on borosilicate glass using

[32]
RF plasma treatment, Appl Surf Sci 316 (2014) 484-490.
U
B. Y. Chow, X. Han, A. S. Dobry, X. F. Qian, A. S. Chuong, M. J. Li, M. A. Henninger, G. M. Belfort, Y.
N
X. Lin, P. E. Monahan, and E. S. Boyden, High-performance genetically targetable optical neural silencing
A

by light-driven proton pumps, Nature 463 (2010) 98-102.

M

[33] H. Baghsiahi, K. Wang, W. Kandulski, R. C. A. Pitwon, and D. R. Selviah, Optical waveguide end facet
D

roughness and optical coupling loss, J Lightwave Technol 31 (2013) 2959-2968.

TE
EP
CC
A
Disposable MEMS Optrode Array Integrated with Single LED for Neurostimulation

Author Biographies:

Sunghyun Yoo received the B.S., M.S., and Ph.D. degrees in electrical engineering and computer science from Seoul National
University, Seoul, Korea, in 2010, 2012, and 2017, respectively. His research interests include micro-optical devices,
endoscopic OCT (Optical coherence tomography) imaging, microlens and microlens array, and glass thermal reflow process.

Hongkyun Lee received the B.S. degree in the school of electronics engineering from Kyungpook National University, Daegu,
Korea, in 2016. He is currently with Micro Sensors and Actuators (MiSA) Laboratory in the department of electrical and
computer engineering, Seoul National University, Seoul, Korea, where he is currently working toward his M.S. degree.

T
Sang Beom Jun received B.S. and M.S. degrees in electrical engineering from Seoul National University (SNU), Seoul,

IP
Korea, in 1999 and 2001, respectively. He received the Ph.D. degree in electrical engineering and computer science from SNU
in 2007. From 2001 to 2008, he worked for Nurobiosys Corp. developing cochlear implant and deep brain stimulation systems.

R
From 2007 to 2008, he was a visiting postdoctoral fellow at Wadsworth Center, New York State Department of Health. He

SC
continued to be trained as a postdoctoral fellow until 2010 at laboratory for integrative neuroscience in National Institute on
Alcohol Abuse and Alcoholism (NIAAA/NIH) Bethesda, MD. Currently, Dr. Jun is an associate professor in Department of
Electronics and an adjunct professor in Department of Brain & Cognitive Sciences, Ewha Womans University, Seoul, Korea.

U
His research is focused on the development of neural interface for probing and modulating neural activity by electrical and
optical means. Primary research goal is to develop novel neural interfaces combining knowledge in neurophysiology and
N
electronic engineering and to apply them for neuroscience studies as well as development of neural prosthetic systems.
A
Yong-Kweon Kim received the B.S. and M.S. degrees in electrical engineering from Seoul National University in 1983 and
M

1985 respectively, and the Dr. Eng. degree from the University of Tokyo in1990. His doctoral dissertation was on modeling,
design, fabrication and testing of micro linear actuators in magnetic levitation using high critical temperature superconductors.
D

In 1990, he joined the Central Research Laboratory of Hitachi Ltd. in Tokyo as a researcher and worked on actuators of hard
disk drives. In 1992, he joined Seoul National University, where he is currently a Professor in the Department of Electrical and
TE

Computer Engineering. His current research interests are modeling, design, fabrication and testing of electric machines,
especially micro electro-mechanical systems, micro sensors and actuators.
EP

Chang-Hyeon Ji received the B.S. and M.S. degrees in electrical engineering and the Ph.D. degree in electrical engineering
and computer science from Seoul National University, Seoul, Korea, in 1995, 1997, and 2001, respectively. His doctoral
CC

dissertation concerned the design, fabrication, and testing of electromagnetic micromirrors for microphotonic applications. He
was with the LG Electronics Institute of Technology, Seoul, from 2001 to 2006 as a Senior and Chief Research Engineer,
where he developed microactuators for various types of applications, including optical communication and raster scanning
A

laser display system. From 2006 to 2011, he was a Postdoctoral Fellow at the Georgia Institute of Technology, Atlanta, where
he researched micro power generators and energy harvesters, through-wafer interconnection technology for integrated power
electronics, and microfabricated components for wireless power transfer and energy storage. In 2011, he joined the faculty of
the Department of Electronics Engineering, Ewha Womans University, Seoul, where he is currently working as a Professor.
His current research interests include power microelectromechanical systems (MEMS), bio-MEMS, optical MEMS and
nanofabrication technology.