Journal of Micromechanics and Microengineering

Related content
- Design and fabrication of a highly
A robust MEMS probe card with vertical guide for manufacturable MEMS probe card for high
speed testing
a fine pitch test - Design and Fabrication of
Microelectromechanical Systems Probe
Card with Vertical Trench Guide for Fine
To cite this article: Bong-Hwan Kim et al 2007 J. Micromech. Microeng. 17 1350 Pitch Probing

- Cantilever-Type Microelectromechanical
Systems Probe Card with Through-Wafer
Interconnects for Fine Pitch and High-
Speed Testing
View the article online for updates and enhancements.

Recent citations
- Yong-Soo Lee et al

- Jaouad Marzouk et al

- MEMS probes for on-wafer RF microwave

characterization of future microelectronics:
design, fabrication and characterization
Jaouad Marzouk et al

This content was downloaded from IP address 140.115.78.89 on 03/11/2020 at 04:17

IOP PUBLISHING JOURNAL OF MICROMECHANICS AND MICROENGINEERING
J. Micromech. Microeng. 17 (2007) 1350–1359 doi:10.1088/0960-1317/17/7/018

A robust MEMS probe card with vertical

guide for a fine pitch test
Bong-Hwan Kim1,2, Hyeon-Cheol Kim2, Soon-Don Choi3,
Kukjin Chun2, Jong-Bok Kim1 and Jong-Hyun Kim1
1
UniTest Inc., Gyeonggi-Do Yongin-City, 446-901, Korea
2
Seoul National University, Seoul, Korea
3
Yeungnam University, Gyeongsangbuk-do, Korea
E-mail: [email protected]

Received 18 March 2007, in final form 10 May 2007

Published 11 June 2007
Online at stacks.iop.org/JMM/17/1350

Abstract
A vertically guided MEMS probe card was designed to deflect 50 µm at a
force of 1.5 g and achieve less than 50 µm of pad pitch. Based on our
experimental results, the measured average contact resistance of a device
under test (DUT) was approximately 0.2
at 1.44 g of force and the
leakage current between two tips in the distance of one pitch was about
10 pA. In addition, tip planarity was about ±6 µm with x–y alignment errors
within ±8 µm. A reliability test showed that the average contact resistance
was 0.34
and the probe tip wear was less than 1 µm after the 10 000 timed
touchdowns. To be capable of fine pitch probing, a cantilever beam was
constructed by dry etching using a positive photoresist. After the cantilever
beam was formed by silicon etching using a deep RIE etcher inside a
deep-recessed trench, a vertically guided structure was created from the
cantilever beam. Furthermore, to make a horizontally stopped structure, the
cantilever beam was designed to have a pyramid tip with a width bigger than
that of the beam itself. This kind of structure is mechanically stable when
the tip is applied with an oblique force. Because the probe card can be
guided vertically and horizontally, it can be neither broken nor deformed by
any directional force. In that respect, this newly proposed probe card is
suitable for wafer-level testing and fine pitch device testing.
(Some figures in this article are in colour only in the electronic version)

1. Introduction be solved by decreasing the force applied to individual pads.

A force at each pad has decreased from 12 g force (gf) to
In testing of a DUT, the probe card comes in contact with 1 gf. Furthermore, the pitch, which can be used to estimate
the device’s metal pads. As the integrated circuit becomes the technological limit of the probe cards, has decreased by
faster and more complex, the number of input/output pads 80 µm in memory devices and by 40 µm in liquid crystal
increases drastically and it becomes more difficult to fabricate displays (LCD) [1].
the probe cards with conventional technology. As a result, the Nowadays, the probe card has to be batch processed
probe card, made of tungsten needles, suffers greatly during and fine pitch manufactured as well as has a desired low
testing. The conventional epoxy-ring-type probe card has contact force and high wear resistance of tip at a touchdown
some limitations in controlling the electrical interface between of 100 000 times or more. For batch processes, probe cards
the DUT and card, tends to easily damage the pad, and have have been developed by using MEMS technology [2–9] and
difficulties in fine pitch probing. With the advent of high- for fine pitch manufacturing, the through-hole interconnection
density circuits with integration higher than 4 G bytes and has been used [10–18].
more number of pads, the total contact force has increased Generally, the MEMS probe card, which is capable
relatively on the DUT, damaging the pads. This problem can of batch processes and fine pitch probing, can be easily

0960-1317/07/071350+10$30.00 © 2007 IOP Publishing Ltd Printed in the UK 1350

 A robust MEMS probe card with vertical guide for a fine pitch test

(1a) (2a)

(1b) (2b)

(1c) (2c)

(1d )
(2d )

Figure 1. Concept of a vertically guided probe beam.

(1e)
(2e)
processed by electroplating with materials such as copper (Cu),
nickel (Ni) and gold (Au). Plating adoption rates have been
increasing in the fields of semiconductors and MEMSs because (2 f )
of high deposition rates, relatively low cost and easiness in
satisfying high aspect ratio features. In addition, pulse current
plating can fill the trench completely without any defects such
Glass wafer PR Ti/Cu Nickel
as void or seam, offer fine pitch, and prepare the probe card for
high speed testing. Plating can be divided into electroplating, Figure 2. Process flow of an electroplated nickel test sample.
which requires a seed layer, and electrolessplating, which does
not. In order to process a high aspect ratio of though-hole the fabrication process flow. The samples were fabricated
to wafer, the probe card is filled by plating. In addition, the by two types of electroplating method: one is bottom-up
adhesive strength of plating with a through-hole and the trench electroplating method, which is the normally used method,
interface wall should be high to prevent plating from being while the other is the trench electroplating method, used in
separated from the wall. In this regard, the electroplating actual probe card fabrication. The detailed processes are as
method is more suitable than electrolessplating since it is follows. For the bottom-up electroplating sample, a titanium
simpler and provides good adhesion. As mentioned [10–18], and copper seed layer was deposited on a 4 inch Corning glass
the through-hole interconnection technology generally uses 7740 wafer to a thickness of 200/1500 angstroms (Å) as shown
the plating method [19]. in figure 2(1a). Then, the thick photoresist (PMER-LA900)
In this paper, a vertically and horizontally guided MEMS was exposed and developed after spinning to a thickness
probe card was proposed in order to make low contact forced, of 30 µm (figure 2(1b)). Nickel and nickel–cobalt were
fine pitched and easy processed probe card as shown in
electroplated at 50 ◦ C in a nickel sulfate-chloride nickel plating
figure 1.
bath (figure 2(1c)). After etching the seed layer (figure 2(1d)),
the glass substrate was isotropic etched in liquid hydrofluoric
2. Mechanical and electrical issues in the MEMS acid (HF) to obtain the free-standing nickel and nickel–cobalt
probe card structure (figure 2(1e)). The beam widths varied from 10 to
20 µm and the lengths varied from 400 to 1000 µm. For
2.1. Material properties of nickel the trench electroplating sample, the AZ 4620 photoresist was
The mechanical properties of the materials used in spun on the 4 inch p-type silicon (1 0 0) wafer as shown in
microelectronics and MEMS applications have become figure 2(2a). The sample was etched by a deep reactive ion
critically important. The material properties often deviate etching (RIE) system (figure 2(2b)). Then, after electroplating
from bulk properties as the probe card becomes smaller, and on titanium and copper seed layers (figure 2(2c)), nickel
they are highly influenced by fabrication process conditions. and nickel–cobalt were electroplated (figure 2(2c)) and then
In the MEMS probe card, the probe structure should have high planarized by a chemical and mechanical polishing (CMP)
hardness, yield strength, wear resistance and low resistivity. machine (figure 2(2e)). Finally, after etching silicon with
In addition, the metal should be easily depositable and a deep RIE etcher, the sample was released by potassium
electroplatable, up to hundreds of microns on silicon wafer. hydroxide (KOH) etchant (figure 2(2f )). Figure 3 shows the
The best material to fulfill these requirements is nickel or fabricated devices. Using an UV adhesive, the two samples
nickel alloy. were attached to the microtensile system [20], which measured
To obtain more accurate data of the metal, Young’s mod- their material properties. For nickel electroplating, the average
ulus, hardness, tensile stress and yield strength of metal-free value of the tensile strength, the yield strength which is
standing films were estimated by a microtensile test method normally 0.2% offset of the strain curve, the Young’s modulus,
[20–25]. A surface micromachining process, that includes which is obtained from the slope of the partial unloading curve,
nickel and nickel–cobalt electroplating, fabricated test samples and the hardness, which is obtained from nanoindentation,
for nanoindentation and microtensile testing. Figure 2 shows were 460.68 MPa, 282.03 MPa, 188.7 GPa and 570.23 MPa

1351
B-H Kim et al

(a) (b)

(c) (d)

Figure 3. Fabricated results.

Table 1. Measured material properties of electroplated nickel.

Tensile strength (MPa) Yield strength (MPa) Young’s modulus (GPa) Hardness (MPa)
Beam width 10 µm 20 µm 10 µm 20 µm
Beam thickness 30 µm 40 µm 30 µm 40 µm Both Both
Plating 1a 334.64 236.86 220.5 151.3 187.4 542.412
Plating 2b 460.68 404.06 282.03 191.3 188.7 570.228
a
Plating 1 is the bottom-up method.
b
Plating 2 is the trench filling method.

(Hv, Vickers Hardness), respectively (see table 1). For nickel– The wear phenomenon can be defined in terms of wear
cobalt electroplating, yield strength varied from 537.2 MPa rate and wear intensity [26]. First, the wear rate is defined as
to 616.7 MPa. These material properties of the electroplated the volume of material that is worn from a unit area of surface
nickel and nickel–cobalt samples were used in the MEMS per unit time. The probe tip wear is the sum of adhesive,
probe card design. abrasive, fretting and fatigue wears. The wear rate can be
expressed as follows:
2.2. Wear dw∗ dw∗
=V (1)
dt dl
Wear of contacting surfaces is another process that occurs

dw∗

fV fV fV
during contact interactions. Wear literally means the = Kad + Kab + Kf t f V + Kf r . (2)
progressive loss of material from surfaces from their fracture dt
total H H H
due to friction interaction, gradually changing the dimensions Here, the wear coefficient K means modeling of wear
and shape of the contacting bodies [26–33]. The aluminum differences according to the level of material complexity in
pad and probe tip continuously make contact in probe cards. quality, shape of structure and temperature. Wear is extremely
Usually, a probe tip must survive a recursive test, in which influenced by adhesive, abrasive and fatigue wear. However,
the tip is forced to make more than 100 000 touchdowns. The in the case of adhesive and abrasive wear, the wear coefficient
probe wear influences the contact resistance and changes the is relatively small since most wear occurs in soft materials
contact area as well as the planarity of the probe tip. [31]. The adhesive wear coefficient changes from 10−7 to

1352
 A robust MEMS probe card with vertical guide for a fine pitch test

in the z direction and sliding distance increase. Hence, wear

decreases as the applied force decreases.

2.3. Contact resistance

A probe card must ensure a predictable, consistently low
contact resistance. Electrical contact means a releasable
junction between two conductors that can carry an electric
current [34]. An ideal electrical contact is established with
a good conductor when closed and a good insulator when
open. Contact resistance is defined as the difference between
the actual resistance of contact and what would be expected
Figure 4. Contact movement of the probe tip.
solely from the bulk resistivity of the metals comprising the
−2 contact.
10 . The abrasive wear coefficient is normally calculated by
2 × 10−3 even though it changes from 10−4 to 10−1, according Contact resistance can be divided into two components.
to the measurement environment. Fatigue wear, however, Since the surface of the contacts is microscopically rough, the
changes from 10−7 to 10−2 [32]. Therefore, the probe tip real contact area is smaller than the apparent contact area. For
wear can be calculated primarily based on fatigue wear [33], the nickel probe tip and aluminum pad, the contact resistance
and accordingly, the probe tip wear can be expressed as is calculated from


dw∗

ρ1 + ρ2 πH σf H
≈ Kf t f V . (3) RC = + . (6)
dt
total 4 F F
The wear intensity is defined as The tunneling resistivity, σ f, is 1 × 10−7
m2. When ρ 1 is


dw∗

Kf t f V 6.8 × 10−8
m for nickel, ρ 2 is 2.7 × 10−8
m for aluminum
≈ = Kf t f (4)
dl
total V and hardness is 180 MPa for soft metal (aluminum), the
calculated contact resistance at 1.5 g force is approximately
and the probe tip wear can be given by 0.13
.
W∗ = Kf t f L. (5)
Figure 4 shows the wear model when the probe tip makes 3. MEMS probe card design
contact with an aluminum pad and slides a distance of L in
the opposite direction of y. Here, the wear of the probe card Most cantilever-type MEMS probe cards have a structure
is directly proportional to fatigue wear coefficient, load and where the width of the cantilever beam is larger than the
sliding distance. The wear increases as both the applied force thickness of the beam as shown in figure 5(a). This causes

F F
F F F
x F

Stable : width > thickness Unstable : width < thickness Stable : width < thickness
Generalized cantilever Generalized cantilever Proposed cantilever

(a) (b) (c)

Figure 5. Cantilever moving concept.

1353
B-H Kim et al

the first vibration mode to move in a top–down direction

1
(z direction). If the width of the beam is relatively smaller NODAL SOLUTION
SEP 3 2004
STEP=1
than the thickness, the first vibration mode shape of the beam SUB =1
17:51:49
TIME=1
changes its direction to the right–left (x direction, figure 5(b)). SEQV (AVG)
DMX =50.367
This consequently incurs problems that probe tips hit and SMN =.001657
SMX =498.997

damage other tips in the fine pitch probe card.

In contrast with the conventional MEMS probe card
structure, the proposed MEMS probe card has a beam width Y MXX

that is smaller than its thickness. In figure 5(b), when force is Z

not applied in a vertical direction to the probe but obliquely, MN

the beam will bend. This means the structure is still unstable.
In order to overcome this instability, the proposed probe beam
was formed inside a deeply recessed trench to guide the beam
vertically as shown in figure 5(c). Therefore, the probe beam
slipped vertically. In addition, to construct a horizontally
stopped structure, the probe beam must possess a pyramid tip .001657 110.889 221.777 332.665 443.553
55.446 166.333 277.221 388.109 498.997
whose width is larger than that of the beam. This, thereby,
(a)
protects the probe tip by stopping and preventing further
damage causing motion. Thus, even if the probe card is 1
DISPLACEMENT
guided vertically or horizontally, it does not break or become STEP=1
SEP 3 2004
18:01:27
deformed by an applied force, regardless of direction. SUB =1
FREQ=7818
DMX =5846
When the probe cards are tested, sufficient contact force
must be applied. The structure of the probe card should not be
damaged to ensure contact repeatability. Accordingly, the
design parameters must be considered. The deflection of
the beam is inversely proportional to the width of the beam.
Y
The beam width is determined according to the limitation Z X
parameter, known as pad pitch, which can be controlled by
two-dimensional mask work. The length of the beam is also
determined according to a limitation condition, taking into
account the chip pad arrangement. Because the length of the
beam is inversely proportional to the triple ratio of the beam
deflection for an applied load, it is regarded as an important
design variable for obtaining the mechanical characteristics of
the beam. Similar to the effect of the length of beam, that of the (b)
thickness of beam is inversely proportional to the triple ratio
of the beam deflection. First of all, the easy adjustability of 1
DISPLACEMENT
the beam thickness should be taken into account in obtaining STEP=1
SEP 3 2004
18:02:12
SUB =2
accurate force-deflection characteristics for the beam. The FREQ=15831
DMX =5839
thickness of the beam, if accurately controlled, can be a very
useful parameter in probe card design.
The beam material should have low structure attenuation
and high elasticity, because probe card specifications require Y X
more than 100 000 touchdowns. Silicon is commonly used Z

as a structural material in the MEMS field because it has

very low structure attenuation, high elasticity and mechanical
hardness. If the silicon structure exceeds the elastic limit, it
breaks upon impact. If the manufacturing process is judged in
terms of process ease and degree of freedom in manufacturing
as well as the structure attenuation, elasticity and mechanical
hardness of the beam material, the suitable beam materials
would be nickel, tungsten, nickel alloy, copper alloy, or such
materials which can facilitate electroplating. (c)
The probe structure was designed with a 50 µm overdrive
(this means deflection) at 1.5 gf contact force by the MEMS Figure 6. Simulation of cantilever when L is 1620 µm, w is 35 µm,
and t is 90 µm.
process. When employing conventional needle types, a contact
resistance of less than 3
at 50 µm overdrive, ±9 µm planarity the length of the beam. The thickness of the beam was 90 µm
and ±12 µm x–y alignment are required in ordinary cases. because the beam was inside the deep trench. To implement
As previously mentioned, among the probe card’s design less than 50 µm pitch and apply an actual commercialized
parameters, the thickness of the beam is easier to change than device, the width of the beam was set to 35 µm. Figure 6

1354
 A robust MEMS probe card with vertical guide for a fine pitch test

Figure 7. Concept of the MEMS probe card.

(1a) (1g) (1m) (2a)

(1b) (1h) (2b)

(1n)

(1c) (1i)
(1o) (2c)

(1d ) (1j)
(1p) (2d )

(1e) (1k)

(1f ) (1l) (1q) (2e)

Figure 8. Process flow of the probe card.

shows the ANSYS simulation results of the proposed probe our experiment, the cantilever-type probe card was fabricated
structure with the results of table 1. When the length was through the following procedure, as shown in figure 8. First,
1620 µm, width was 35 µm, thickness was 90 µm, and contact the starting material was a 4 inch, p-type (1 0 0), and double
force was 1.5 gf, the probe card deflected 50 µm. Naturally, the side polished silicon wafer with a thickness of 400 µm. After
first vibration mode (figure 6(b)) was to the right–left direction the probe beam was immersed in an oxidizing solution, which
(x direction). was composed of H2SO4:H2O2 = 4:1 at 120 ◦ C for 10 min,
it was patterned by using AZ 1512 photoresist (figure 8(1a)).
Then, the probe beam was defined inside the silicon trench
4. Fabrication by using a deep RIE etcher (figure 8(1b)). In addition, after
double side alignment of the probe beam by using a MA6
The MEMS probe card consists of a probe structure, a ceramic aligner (Karl Suss, Germany), the through-hole was patterned
PCB and a main PCB as shown in figure 7. A probe structure and etched (figure 8(1c)). Subsequently, a thermally grown
that is fabricated on a silicon wafer and ceramic PCB are silicon dioxide possessing a thickness of 3000 Å and a low
connected by soldering. The ceramic PCB and the main pressure chemical vapor deposited (LPCVD) silicon nitride
PCB are connected with a pogo pin. Fabrication can be with 3000 Å were formed to insulate the structure electrically
divided into trench etching, beam formation, through-hole between the holes and wafer. Then, LPCVD undoped poly-
filling, planarization, probe tip formation and assembly. In silicon with thickness of 3000 Å was deposited. A 3 µm

1355
B-H Kim et al

Table 2. Nickel bath and pulse electroplating condition. Table 4. Nickel–tungsten bath and pulse condition.
Basic bath Basic bath
NiSO4 7H2O 110 g L−1 Nickel sulfate (NiSO4 ·6H2O) 0.06–0.18 mol L−1
NiCl2 6H2O 6 g L−1 Sodium tungstate (Na2WO4·2H2O) 0.14–0.42 mol L−1
Boric acid 30 g L−1 Citric acid (C8H7O6·H2O) 0.5 mol L−1
Additives Ammonium chloride (NH4Cl) 0.5 mol L−1
Electroplating conditions Electroplating conditions
Cathode Silicon wafer Cathode Silicon wafer
Anode Nickel ball Anode Nickel ball
Temperature 50 ◦ C Temperature 70 ◦ C
pH 4 pH 8.0
Current density 3 mA cm−2 Current density 100–300 mA cm−2
Current on-time 10 µs Current on-time 100 ms
Current off-time 90 µs Current off-time 100–500 ms
Duty cycle 10% Duty cycle 20–50%
Electroplating time 1 h 20 min Electroplating time 2h

Table 3. Nickel–cobalt bath and pulse electroplating condition.

electroplate when the thickness is greater than 10 µm because
Basic bath it is fragile and its tensile stress is much greater than that of
Ni(NH2SO3)2 5H 110 g L−1 nickel–cobalt. The probe tip and beam were bonded with
H3BO3 30 g L−1 electroplated tin (Sn) (figure 8(1p)). Finally, the probe beam
Nickel chloride 6 g L−1 was released by a hydrofluoric acid solution (figure 8(1q)),
Cobalt powder 4 g L−1 and the probe structure (figure 9) was assembled with ceramic
Additives (Brighterner) 10 ml L−1 PCB and main PCB. A blade type probe card was fabricated
Electroplating conditions in order to compare with the proposed probe card as shown in
Cathode Silicon wafer figure 8(2a–2e) [35–37].
Anode Nickel ball
Temperature 55 ◦ C
pH 3.5 5. Results and discussions
Current density 3.2 mA cm−2
Current on-time 20 µs The fabricated and assembled MEMS probe card was
Current off-time 80 µs
Duty cycle 20% measured by HP 4155 and Precision Point VX2 Extension
Electroplating time 1 h 10 min to obtain the mechanical characteristics such as planarity,
contact force, x–y alignment and electrical characteristics such
as leakage current and contact resistance. First, the planarity
thick tetraethylorthosilicate (TEOS) oxide was deposited at the base level was measured by scanning the probe at an
on the front side of the structure (figure 8(1d)). Then, increment of 0.25 µm from the check plate. The contact force
the trench was filled by nickel–cobalt–nickel electroplating was measured by using the electronic scale as the equipment
(figure 8(1e)). After planarizing with extra nickel alloy by probe tip pressed down the fabricated probe tip. In addition,
using CMP (see tables 2 and 3), nickel–cobalt–nickel structure after inputting the base coordinate plane, the x–y alignment
was formed (figure 8(1f )). Nickel–cobalt was composed was measured by using two cameras. Also, the leakage current
of 52% nickel and 48% cobalt, as measured by the energy is the current between the tips when the probe tip that is to be
dispersive spectrometer (EDS, or EDAX). As the content of measured is applied with 5 V and all others are grounded.
cobalt increased, the yield strength, the hardness, the fragility Finally, making to contact with a gold plate after a current of
and stress still increased only moderately. 20 mA was passed through the probe tips, the contact resistance
In addition, after the through-hole was solvent cleansed was measured.
and sputtered with titanium/copper it was filled with The average planarity and the x–y alignment were ±6 µm
electroplated copper on the backside (figure 8(1g), (1h)). The and ±8 µm, respectively. They were influenced by the flexure
electroplated copper was planarized by the CMP process and of wafer, plating and CMP. For the contact force as a function
then gold was sputtered on the copper film to prevent oxidation of overdrive, the measured value and simulation value were
of the copper (figure 8(1i), (1j )). Then, the seed layer was found to be almost identical. When the overdrive was 50 µm,
deposited by using Ti/Cu, and nickel was deposited after bump the contact force was measured by 1.44 gf for 1.5 gf of the
patterning (figure 8(1k)). Using a stripper, the photoresist was design value. The leakage current was about 10 pA. This time,
then removed, and the backside silicon etched (figure 8(1l)). the resistance was about 1T
. The average total resistance
Probe tip was fabricated on another silicon wafer. After silicon was approximately 4.1
between the two tips. Therefore, the
nitride patterning, the pyramid tip was formed by KOH wet resistance of one probe can be 2.05
. Finally, the measured
etching (figure 8(1m), (1n)). Then, nickel–tungsten/nickel contact resistance as a function of contact force on aluminum
electroplating formed the tip base (see table 4 and and gold film varied from 0.3
to 0.45
. The results were
figure 8(1o)). Nickel–tungsten is composed of 83% nickel different from the results obtained with equation (6). They
and 17% tungsten and has an average grain size of about differed by approximately 0.3
for aluminum and 0.27
for
11 nm. However, nickel–tungsten is much more difficult to gold. This suggests that there is another source of resistance

1356
 A robust MEMS probe card with vertical guide for a fine pitch test

(a)

(b)

(c)

Figure 9. Fabricated and assembled probe card.

making up the contact resistance, besides constriction and film Table 5. Measured tunneling resistivity and trace resistance.
resistance. This resistance, the so-called trace resistance, is Tunneling resistivity Trace
due to the contamination of the test equipment and test wafer. Contact metal (10−11
cm2) resistance (
)
When the resistance in the previous contact resistance model
Au 1 0.27
was added to the trace resistance, the model and experiment
Al 3 0.3
values were the same. Figure 10 shows the change in the
contact resistance by contact force. On the basis of the contact
resistance model and experiment results, the contact resistance was measured by using a computer-assisted controllable
can be given by equipment, which can repeat up–down movement for a 4 s
 cycle. First, the test wafer was brought up to stage after
ρ1 + ρ2 πH σf H attaching the probe tip, and the stage was moved upward in 2 s
RC = + + Rtrace , (7)
4 F F and then moved downward in 2 s. Consequently, the stage
where Rtrace is the trace resistance caused by contamination. In could give 50 µm overdrive to the probe tip. The probe tip
this experiment, the tunneling resistivities and trace resistance was worn by about 0.3 µm at a touchdown of 20 000 times as
were measured as shown in table 5. shown in figure 11.
To evaluate the reliability of the probe card, wear and Contact resistance was measured by using the same test
contact resistance by touchdowns were measured. Wear method used in wear measurement, where aluminum and

1357
B-H Kim et al

the aluminum pad wafer was approximately 0.34
during

1.0
20 000 touchdowns, while the gold pad wafer was 0.31
as
Al wafer shown in figure 12.
Au wafer
0.8
Contact Resistance (ohm)

ρ1 + ρ 2 ηπ H σfH 6. Conclusion
RC = + + Rtrace
0.6 4 F F
In this paper, a MEMS probe card was developed. This new
cantilever type probe card can implement a minimum pitch
0.4 of 30 µm, unachievable through conventional probe cards.
The probe card has a vertically slid and horizontally stopped
cantilever structure. For implementing a fine pitch of less than
0.2
50 µm, a cantilever beam should have a width smaller than
its thickness. This leads to unstable conditions. To form a
0.0 stable structure, a deeply recessed trench type cantilever was
0.0 0.2
.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
used, and to give more than 50 µm of overdrive, nickel and
Contact Force (gf)
nickel alloys were used as the tip and structure materials. The
Figure 10. Contact resistance of the MEMS probe card. planarity and x–y alignment were within ±8 µm. The total
resistance was 4.1
between the shortened tips. The average
leakage current was about 10 pA between insulated tips as
0.6
measured by HP 4155 equipment. The contact resistance at
50 µm overdrive was about 0.4
, and this value corresponded
Measurement
Calculation
well with the calculation. In conclusion, this probe card is
0.4 suitable for wafer-level and function testing.
Wear (µm)

Acknowledgments
0.2 W *=KftfL
The authors would like to thank Tae-Un Chun for measurement
Kft=13 and Soo Yun for help. The authors would also like to thank
Inter-University Semiconductor Research Center (ISRC) and
0.0 its staff.

0 5000 10000 15000 20000 References

Number of Touchdowns
[1] Schwartz R and Kosecki D 2003 Advances in probe card
Figure 11. Wear by touchdowns. analyzer: test and maintenance of very high pin count cards
Southwest Test Workshop (Long Beach, CA, June 2003)
[2] Bates D 1997 Int. Test Conf. (Washington, DC) p 533
1.0 [3] Beiley M, Leung J and Wong S S 1995 IEEE Trans. Compon.
Packag. Manuf. Technol. B 18 184
0.9
[4] Zhang Y, Zhang Y and Marcus R B 1999 IEEE J.
Microelectromech. Syst. 8 43
Contact Resistance (ohm)

0.8
[5] Kim Y, Sim J, Nam J and Lee J 1988 Microprocesses and
0.7 Nanotechnology Conf. (Gyeongju, Korea) pp 182–3
0.6
Al wafer [6] Pruitt B, Choi D, Florando J, Martens R, Wenzel S,
Au wafer
Reynolds C, Nix W and Kenny T 2001 11th Int. Conf.
0.5 Solid-State Sensors and Actuators (TRANSDUCERS’01,
EUROSENSORS XV) (Munich, Germany) p 1032
0.4
[7] Itoh T, Kataoka K and Suga T 2001 11th Int. Conf. Solid-State
0.3 Sensors and Actuators (TRANSDUCERS’01,
EUROSENSORS XV) (Munich, Germany) p 222
0.2
[8] Park S, Kim B and Cho D 2001 Proc. SEMI Tech. Symp.
0.1 (STS): Innovation in Semiconductor Test, Assembly &
Packaging, SEMI WEST 2001 (San Jose, CA, USA) p 81
0.0 [9] Park S, Kim B, Kim J, Paik S, Choi B, Jung I, Chun K and
0 5000 10000 15000 20000
Cho D 2002 J. Micromech. Microeng. 12 650
Number of Touchdowns
[10] Khoury S L, Burkhard D J, Galloway D P and Scharr T A
Figure 12. Contact resistance as a function of touchdowns. 1990 Proc. 40th Electronic Components and Technology
Conf. p 768
[11] Sawada Y, Yamaguchi A, Oka S and Fujioka H 2002 IEEE
gold pad layers were used on the silicon substrate. For Trans. Compon. Packag. Manuf. Technol. 25 73
[12] Lu J, Busch S C and Baldwin D F 2001 IEEE Trans. Electron.
the aluminum pad, even though a soft brush cleaned the
Packag. Manuf. 24 154
tips every 1000 touchdowns in order to prevent aluminum [13] Fujita Y, Kawamura Y and Mizuishi K 1993 Proc. 43rd
sticking, contact resistances fluctuated. Contact resistance on Electronic Components and Technology Conf. p 1081

1358
 A robust MEMS probe card with vertical guide for a fine pitch test

[14] Linder S, Baltes H, Gnaedinger F and Doering E 1994 IEEE [26] Goryacheva I G 1998 Contact Mechanics in Tribology
Workshop on Micro Electro Mechanical Systems (MEMS (Dordrecht: Kluwer Academic Publishers)
’94) (Oiso, Japan) p 349 [27] Archard J F 1953 Contact and rubbing of flat surfaces J. Appl.
[15] Soh H T, Yue C P, McCarthy A, Ryu C, Lee T H, Wong S S Phys. 24 981–8
and Quate C F 1999 Japan J. Appl. Phys. 38 2393 [28] Kragelsky I V 1965 Friction and Wear (London: Butterworths)
[16] Wu J H, Alamo J A D and Jenkins K A 2000 Tech. Dig. of Int. [29] Rabinowicz E 1965 Friction and Wear of Materials (New
Electron Devices Meeting (IEDM 2000) p 477 York: Wiley)
[17] Oh S and Kondo K 2002 Proc. 3rd Japan–Korea Joint [30] Harricks P L 1976 The mechanism of fretting and the
Workshop on Advanced Semiconductor Process and influence of temperature Indust. Lubrication Technol.
Equipments (Hakone, Japan) p 190 (January/February) 9–17
[18] Lee D, Ono T, Abe T and Esashi M 2001 14th IEEE Int. Conf. [31] Bhushan B 2001 Modern Tribology Handbook (Boca Raton,
on Micro Electro Mechanical Systems (MEMS ’01) FL: CRC Press LLC) chapter 7
(Interlaken, Switzerland) p 204 [32] Holm R 1967 Electric Contacts, Theory and Application
[19] Kim B, Park S, Lee B, Lee J, Cho D and Chun K 2002 15th 4th edn (Berlin: Springer-Verlag) p 234
IEEE Int. Conf. on Micro Electro Mechanical Systems [33] Ling F F and Pan C H T 1986 Approaches to Modeling of
(MEMS’2002) (Las Vegas, NV) p 368 Friction and Wear (Berlin: Springer) p 15
[20] Son D I 2004 Dissertation Seoul National University [34] Holm R 1967 Electric Contacts, Theory and Application
[21] Weihs T P, Hong S, Bravman J C and Nix W D 1988 J. Mater. 4th edn (Berlin: Springer-Verlag) p 1
Res. 3 931 [35] Kim J-H, Kim H C and Chun K 2005 A small size broadband
[22] Beams J W 1959 Proc. Int. Conf. Sponsored by Air Force MEMS antenna for 5 GHz WLAN applications
Office of Scientific Research, Air Research and J. Semicond. Technol. Sci. 5 204–9
Development Command and the General Electric Research [36] Kim I, Seok S, Kim H C, Kang M K and Chun K 2005
Laboratory (Bolton Landing) pp 182–92 Wafer level vacuum packaged out-of-plane and in-plane
[23] Petersen K E and Guarnieri C R 1979 J. Appl. Phys. 50 6761 differential resonant silicon accelerometers for
[24] Oliver W C and Pharr G M 1992 An improved technique for navigational applications J. Semicond. Technol. Sci.
determining hardness and electric modulus using load and 5 58–66
displacement sensing indentation experiments J. Mater. Res. [37] Kim B-H, Park S-J, Chun K, Cho D-I (D), Park W-Ki, Jun T-U
7 1564–83 and Yun S 2004 A fine pitch MEMS probe unit for flat
[25] Kapels H, Aigner R and Binder J 2000 J. IEEE Trans. panel display as manufacturing MEMS application Sensors
Electron. Devices 47 1522 Actuators A 115 46–52

1359