of 6
Techniques of cryogenic reactive ion etching in silicon for fabrication
of sensors
M. David Henry
a

Department of Applied Physics, California Institute of Technology, Pasadena, California 91125
Colin Welch
b

Oxford Instruments Plasma Technology, Yatton, Bristol BS49 4AP, United Kingdom
Axel Scherer
Department of Applied Physics, California Institute of Technology, Pasadena, California 91125

Received 25 February 2009; accepted 12 July 2009; published 25 August 2009

Cryogenic etching of silicon, using an inductively coupled plasma reactive ion etcher

ICP-RIE

,
has extraordinary properties which can lead to unique structures difficult to achieve using other
etching methods. In this work, the authors demonstrate the application of ICP-RIE techniques which
capitalize on the cryogenic properties to create different sensors geometries: optical, electrical,
magnetic, and mechanical. The three techniques demonstrated are

1

single step deep etches with
controllable sidewall profiles. Demonstrating this, silicon pillars with over 70

m depth and less
than 250 nm sidewall roughness were etched using only 1.6

m of photoresist for use as solar cells.

2

Using the cryogenic etch for thick metallization and liftoff with a thin photoresist mask.
Demonstrating this second technique, a magnetic shim was created by deposition of 6.5

m of iron
into 20

m deep etched trenches, using the remaining 1.5

m photoresist etch mask as the liftoff
mask. Using the same technique, 15

m of copper was lifted off leaving a 20

m deep plasma
enhanced chemical vapor deposition silicon oxide coated, silicon channel with copper.

3

Use of a
two step cryogenic etch for deep etching with reduced sidewall undercutting. This was demonstrated
by fabrication of deep and anisotropic microelectromechanical systems structures; a mechanical
resonator was etched 183

m deep into silicon with less than 3

m of undercutting. This work also
describes the etch parameters and etch controls for each of these sensors.
© 2009 American Vacuum
Society.

DOI: 10.1116/1.3196790

I. INTRODUCTION
Cryogenic inductively coupled plasma reactive ion etch-
ing

ICP-RIE

of silicon has significant advantages over that
of other silicon etches such as chopping Bosch or Cl
2
chem-
istries. In cryogenic etching, the silicon sample is cooled
down to subzero temperatures, typically −100 to −140 °C.
1
These low temperatures not only enable profile control by
virtue of condensation of a SiO
x
F
y
passivation layer but also
provide extremely high etch selectivities of the etch mask
over silicon. It is understood that the silicon etch product
from the SF
6
etch chemistry, SiF
x
, condenses on the side-
walls and combines with the reactive oxygen to create the
SiO
x
F
y
passivation layer.
2
4
This condensate offers sidewall
protection from the SF
6
chemical attack, similar to the role
C
4
F
8
plays in Bosch etching. Anisotropic silicon etching can
thus be realized, resulting in geometries in which the verti-
cality of the sidewalls is tightly controlled by the passivation
rate. Although the etch rate and the passivation rate equa-
tions are coupled, various mechanisms for controlling the
passivation rate are also separable from the etching rate
equation, such as silicon surface temperature and oxygen
concentration in the etching plasma. Since the passivation is
occurring simultaneously with the etching, sidewall rough-
ness introduced by scalloping, usually seen with chopping
etch-passivation chemistries such as Bosch, can be elimi-
nated yielding very smooth sidewalls with sub-250 nm
roughness. The very low temperature ensures preferential
etching of silicon to that of the photoresist or silicon dioxide
etch mask due to the low Arrhenius activation energy for
silicon. Selectivity ratios of better than 70:1 for photoresist
and 150:1 for silicon dioxide have been achieved. These two
advantages allow for unique structures to be fabricated in
silicon.
Here, we demonstrate devices of micron sized structures
using cryogenic etching recipes developed within an induc-
tively coupled plasma system. We also detail exact etching
conditions along with general trends of pattern transfer pa-
rameters for quickly reproducing this work. The device ap-
plications of the pillars described below are microfabricated
radial pillar solar cells. The applications of the exceptionally
thick liftoff layers of copper and iron coils embedded in sili-
con enable thick metal liftoff with thin photoresist layers
enable the definition magnetic shims and electromagnetic
coils. Finally, very deep anisotropic silicon etches described
here have been optimized for the creation of microelectro-
mechanical system

MEM

silicon mechanical resonators.
a

Electronic mail: mdhenry@caltech.edu
b

Electronic mail: colin.welch@oxinst.com
1211
1211
J. Vac. Sci. Technol. A 27
5
...
, Sep/Oct 2009 0734-2101/2009/27
5
...
/1211/6/$25.00 ©2009 American Vacuum Society
II. SMOOTH ETCH SIDEWALLS AND ANGLE
CONTROL: PILLAR ARRAYS
The first example devices presented here are silicon pil-
lars used for characterizing silicon solar cells.
5
,
6
The pillars
were lithographically defined in hexagonal arrays with pillar
separations equal to their diameter. Sets of 5, 10, 20, and
50

m diameter pillars were all etched in 1 mm square ar-
rays on the same silicon sample allowing aspect ratio depen-
dencies to become apparent. The fabrication sequence begins
with cleaning a
P
-type, 1–10

cm,

100

silicon piece,
spinning AZ5214e photoresist, exposing and developing the
pattern into the

1.6

m thick photoresist. The silicon
sample, roughly 1 in.
2
, is then mounted to a 150 mm carrier
silicon wafer using a thin film of Fomblin oil. This results in
excellent thermal conductivity between the sample piece and
the carrier wafer, with the oil easily removed later using
isopropyl alcohol. Before etching, the chamber was cleaned
using a high pressure SF
6
plasma, then conditioned on a
blank wafer using the etch recipe for 30 min. We find that the
conditioning step is exceedingly important for obtaining re-
producible results.
The sample was then cryogenically etched in an Oxford
Instruments PlasmaLab System 100 ICP-RIE 380. The
samples were etched under the following typical conditions

Table
I

.
This process, running for 40 min, resulted in 5

m diam-
eter pillars approximately 35

m tall. Increasing the etch
time to 80 min for the 5

m diameter pillars results in an
etch height of 76

m, Fig.
1
. Surface defects from the etch
were typically less than 250 nm with occasional 600 nm de-
fects spaced greater than 5

m apart. This is significantly
smoother than the regular pattern of chopping marks, ap-
proximately 2

m, created using Bosch etching. Aspect ra-
tios of over 17.5 were achieved, defined here as the ratio of
etched depth to width between pillars. Since the different
sized pillar arrays were created on the same substrate, it al-
lowed for characterization of aspect ratio etching rate depen-
dence, Fig.
2
. The divergence of the etch depths between the
different diameter pillars, in Fig.
2
, clearly indicates that as-
pect ratio dependent etching is occurring.
More than four contributors to aspect ratio dependent
etching have been suggested.
7
The most likely candidates are
Knudsen transport of neutrals, ion shadowing, neutral shad-
owing, and differential insulating charging. Keil and
Anderson
8
correctly assessed that differentiation and mea-
surement between the contributors in standard etchers are
difficult and as a substitute rate equations may be used to
describe their effects. To characterize this effect here, a dif-
ferential equation for the etch rate was established similar to
their Eq. 4,
dD
dt
=
E
b

D
W
.

1

For Eq.

1

,
D
is the etch depth,
E
is the etch rate for an
aspect ratio of zero,
b
is a coefficient to be fitted describing
the etch rate reduction due to aspect ratio, and
W
is the width
of the minimum etched spacing between the pillars. This
model is intended as an approximation for etching rates and
should closely approximate the rates seen for etching
trenches of similar aspect ratios. The assumption made for
this model is that the aspect ratio dependent etch rate scales
linearly, the simplest possible model describing the aspect
ratio effects. The etch depth verse time data points are plot-
ted with the graphical solutions to the differential equation
for the different pillar heights, presented in Fig.
2
.
E
was
T
ABLE
I. Etch parameters for 5, 10, 20, and 50

m diameter pillars in
silicon.
Etch parameter
Units
Value
Etch
parameter
Units
Value
SF
6

SCCM

70
ICP

Watts

900
O
2

SCCM

4.5
RIE

Watts

5
Temperature

Celsius

−120 Pressure

milliTorr

10
E
o


m
/
min

1.15
Helium

Torr

10
F
IG
. 1. Cross sectional SEM of 76

m high, 5

m diameter silicon pillars
cryogenically etched with 4.5 SCCM of O
2
. 800 nm of photoresist remained
after the etch for a selectivity of 89:1 and an etch rate of 0.96

m
/
min.
F
IG
. 2. Etch depth dependence on time for cryogenically etched silicon
pillars. The curves generated are from the solutions to the generalized etch-
ing rate equation with the
E
coefficient set to 1.15

m
/
min and
b
coefficient
set to 0.040 99.
1212
Henry, Welch, and Scherer: Techniques of cryogenic reactive ion etching in silicon for fabrication of sensors
1212
J. Vac. Sci. Technol. A, Vol. 27, No. 5, Sep/Oct 2009
determined to be 1.15

m
/
min and
b
was 0.040 99

m
/
min
by curve fitting the width normalized data,
D
/
w
and
t
/
w
,
resulting in an
R
-square fit of 0.9972,
D
=
E

w
b


1−
e
b

t
/
w

.

2

The etch profile is easily controlled by either changing the
temperature of the table or by changing the oxygen flow
rate.
1
Altering either one of these conditions affects only the
passivation rate with relatively good decoupling from the
etching rate. However, we noted that temperature further
controls notching at the top of pillars with high aspect ratios.
Warmer temperatures produced notching which was round
and smooth but pitted the surface of the pillar, whereas the
colder temperatures produced notching which was squared
and jagged. Optimization between these two conditions for
the cryogenic etch was found to be −120 °C and profile con-
trol was then tuned using O
2
flow rates. By setting the O
2
to
3.5 SCCM

SCCM denotes cubic centimeter per minute at
STP

, the angle of the pillar sidewall was made slightly re-
entrant, and by increasing the O
2
to 6.5 SCCM the pillar
sidewall was made to taper positive, Fig.
3
. A flow rate of
4.5 SCCM was determined to give approximately vertical
sidewalls. Black silicon began to appear at 7.5 SCCM in the
trenches between the pillars while still increasing the posi-
tive taper. The arrival of black silicon prevents further tuning
for a positive taper, Fig.
4
. For our system, the mass flow
controller for the oxygen becomes unstable at flow rates of
approximately 2 SCCM which prevents tuning the sidewall
angle more in the re-entrant direction. These limitations
combine to give about 7° of tunability across the different
pillar widths. Similar pillars to those just described were then
doped and used to study radial
p
-
n
junction solar cells.
5
III. METAL LIFTOFF WITH PHOTORESIST:
MAGNETIC SHIMS AND PLANAR MICROCOILS
Another positive advantage of the cryogenic silicon etch
is the high selectivity of the etch rate of photoresist over
silicon. This advantage can be utilized for improving metal-
lization lift-off on silicon. Typically when lifting off a met-
allization layer using photoresist, care must be taken in ob-
taining resist sidewalls that are vertical or even slightly re-
entrant, and the photoresist has to be substantially thicker
than the deposited metal layer.
9
The approach which we de-
scribe here transfers the difficult lift-off profile requirements
from the photoresist to the cryogenic silicon etch. As we
demonstrated with the pillars, the sidewall profile is very
easy to control and reproducible by optimizing etch
parameters.
4
,
7
,
10
The high selectivity improves the relative
height between the top of the photoresist to the silicon sur-
face being metallized, thereby permitting thicker metal lay-
ers to be deposited. This enables creation of passive mag-
netic shims and electromagnetic coils by deposition of thick
layers of iron and copper into silicon, Fig.
5
. To the authors
F
IG
. 3. Profile control of cryogenically etched silicon pillars. Error bars were
placed to indicate sidewall angle variations due to temperature fluctuations
of the substrate. Lines are linear fits to data intended to guide the eye.
F
IG
. 4. Cross sectional SEM of 36

m high, 5

m diameter silicon pillars
cryogenically etched with 7.5 SCCM of O
2
for an angle of 3.6°. Note the
onset of black silicon between the pillars.
F
IG
. 5. Cross sectional SEM of a 6.5

m thick evaporated iron ring in a
18.5

m deep silicon trench. The 1.6

m thick photoresist, used as the
cryogenic etch mask, was also used for metal liftoff. This transferred the
sidewall requirements for liftoff from the resist to the more controllable
etch. Note that the pattern roughness along the edge of the trench was
transferred from the mask to the silicon by the cryogenic etch.
1213
Henry, Welch, and Scherer: Techniques of cryogenic reactive ion etching in silicon for fabrication of sensors
1213
JVST A - Vacuum, Surfaces, and Films
knowledge, this is the first time using the etch mask as also
the metallization lift-off mask for deposition of thick metal
or oxide layers on silicon.
With sample preparation similar to the previously men-
tioned approach, a 2 in.
p
-type silicon wafer was cryogeni-
cally etched to 18.5

m depth to define patterns with various
radii and widths, steps a and b in Fig.
6
. Optimization for
vertical sidewalls led to the following conditions

Table
II

.
Profilometer measurements taken before and after etching
indicated photoresist losses only 0.2

m, an approximate se-
lectivity of 90:1. With over 1

m of photoresist remaining,
iron was thermally evaporated at a rate of 8 A
/
s to a final
thickness of 6.5

m, step d in Fig.
6
. Lift-off was then per-
formed using acetone, step e in Fig.
6
. This created thick iron
rings embedded in silicon, which was then placed on a 2 in.
diameter NdFeB permanent magnet for passively shaping the
magnetic field in nuclear magnetic resonance experiments.
This technique was also used in creating planar microcoil
detectors in silicon,
11
Fig.
7
. The previous etch recipe was
applied to a 3 in.,
p
-type, 1 m

cm silicon wafer to etch a
four turn 750

m radius microcoil. 1.5

m of SiO
2
was then
deposited in a plasma enhanced chemical vapor deposition
reactor at 135 °C to provide electrical insulation between the
copper wires being created and the silicon substrate, step c in
Fig.
6
. The insulating SiO
2
layer was followed by a 40 nm
amorphous silicon deposition to improve metal adhesion.
Prior to deposition, the center contact is mechanically frac-
tured away to allow for copper to silicon Ohmic contacting
and removing the need for a bridging contact, Fig.
8
. Next,
10

m of copper was thermally evaporated into the 20

m
deep trenches. Lift-off was again performed with acetone
and a cotton swab. Due to the vertical profile of the etch and
brittle nature of silicon dioxide precision shearing of the ox-
ide can be seen at the top edge of the trench.
IV. DEEP SILICON ETCHING USING TWO STEP
PROCESSES MECHANICAL RESONATOR
Micromachining technology using silicon, MEMs, typi-
cally requires significant etch depths to be achieved, over
200

m for the device described here. A widely used silicon
etch to achieve this is the “Bosch” etch process which alter-
nates an etching step with a passivation step.
12
,
13
While this
etch allows for etch depths of several hundred microns to be
achieved, it leaves “chopping” marks along the sidewalls
F
IG
. 6. Fabrication sequences for metallization in silicon:

a

1.6

mof
photoresist is patterned on a silicon substrate,

b

silicon is cryogenically
etched to depths of approximately 20

m,

c

SiO
2
is deposited

microcoil
sequence only

,

d

thermal evaporation of thick layers of metal, and

e

lift-off using acetone.
F
IG
. 7. Cross sectional SEM of 10

m of evaporated copper in a 20

m
deep silicon trench insulated by 1.5

mofSiO
2
. The brightest material is
the copper insulated by silicon dioxide, dark layer between the copper, and
substrate.
F
IG
.8.SEMofa750

m radius planar microcoil fabricated in silicon using
a single photolithography mask. The copper turns are imbedded into silicon
cryogenically etched and metal lift-off is performed using the 1.5

m thick
etch mask. No bridging contact to center is needed as the current is sourced
into the silicon substrate as an Ohmic contact at the center of the coil.
T
ABLE
II. Etch parameters for etching trenches for iron and copper coils in
silicon.
Etch parameter
Units
Value
Etch
parameter
Units
Value
SF
6

SCCM

90
ICP

Watts

1000
O
2

SCCM

6RIE

Watts

3
Temperature

Celsius

−120 Pressure

milliTorr

10
Etch rate


m
/
min

2.3
Helium

Torr

10
1214
Henry, Welch, and Scherer: Techniques of cryogenic reactive ion etching in silicon for fabrication of sensors
1214
J. Vac. Sci. Technol. A, Vol. 27, No. 5, Sep/Oct 2009
which are approximately several microns in size.
14
This
marking can be detrimental when the device applications
rely on interactions with the sidewalls such as the mechani-
cal resonator fabricated here. Using the cryogenic etch for
this application offers the elimination of the chopping marks.
However, maximum etch depth for the cryogenic etch is ul-
timately mask erosion limited or aspect ratio limited, the
later limitation becomes significant after several tens of mi-
crometer depths as previously described.
15
To improve the mask erosion quality either the forward
power is reduced or a thicker mask may be employed. Hin-
dering fabrication with thicker silicon dioxide masks is the
slow growth or deposition rates required with the added dif-
ficulty of etching a pattern into silicon dioxide while hinder-
ing resists masks thicker than 1.5

m are the problems of
thermal gradients on the mask causing cracking. Increasing
the RIE power imparts more momentum to the ions allowing
for the gases to achieve greater depths, improving the aspect
ratio limitation. However, since the milling action of the ions
has increased, the selectivity of the mask is typically reduced
and maximum depth attainable is subsequently reduced. In-
creasing the RIE power also requires an increase in passiva-
tion gas, oxygen, moving the etch into the black silicon re-
gime, as demonstrated with the pillars.
Combining two etches then enables optimization for the
required depth. The first etch, using lower RIE power to etch
as deep as possible, while minimizing damage to the etch
mask, is followed by a second etch with higher RIE power.
Thus deeper cryogenic etches can be achieved. This tech-
nique is used for creation of MEM mechanical resonators.
The diameter of such a resonator is 7 mm, defined with a
5.5

m thick silicon dioxide etch mask. Structures of the
resonator rings are 20

m wide ridges separated by 80

m.
The 80

m separations are the etch trenches. This two step
cryogenic etch achieves over 200

m of depth, 2

m of ox-
ide mask loss, and with less than 3

m of lateral etching or
undercutting; a remarkable 1.5% ratio, Figs.
9
and
10
.Asa
comparison, the second etch step was performed for 140 min
using the same thickness oxide mask. The mask had eroded
and began etching the silicon after only 160

m of etch
depth, Fig.
11
.
To achieve the depth required for the resonator, the fol-
lowing two-step etch process is performed

Tables
III
and
IV

.
V. CONCLUSIONS
This work has demonstrated how the cryogenic ICP-RIE
etch can be utilized to perform deep silicon etches for appli-
cations ranging from solar cell construction to the definition
of liftoff-fabricated electromagnetic structures. With rela-
tively little photoresist, 100

m tall pillars can be con-
structed with smoother sidewalls than possible by using
chopping etches such as the Bosch etch. We also describe the
T
ABLE
III. Etch parameters for first step of MEM resonator etch in silicon.
Etch parameter
Units
Value
Etch
parameter
Units
Value
SF
6

SCCM

90
ICP

Watts

900
O
2

SCCM

6RIE

Watts

3
Temperature

Celsius

−140 Pressure

milliTorr

10
Etch rate


m
/
min

1.3
Helium

Torr

10
F
IG
. 9. SEM of a cross section of the MEM resonator etched using a two
step cryogenic etch and a 5.5

m thick silicon dioxide etch mask to a depth
of 200

m.
F
IG
. 10. SEM of a cross section of the MEM resonator’s sidewalls etched
using a two step cryogenic etch to a depth of 200

m. Note that 3.5

mof
etch mask still remains.
F
IG
. 11. SEM of a cross section imaged at an angle of 45° of the MEM
resonator’s sidewalls etched using only the second step of the two step etch
in Table
III
. Note that the silicon dioxide mask has been eroded through
after only achieving about 160

m etch depth.
1215
Henry, Welch, and Scherer: Techniques of cryogenic reactive ion etching in silicon for fabrication of sensors
1215
JVST A - Vacuum, Surfaces, and Films
limits of the fabrication depths of cryoetched trenches and
the control over the sidewalls of lift-off metallization. In this
liftoff procedure, the fabrication requirements replace chal-
lenging photoresist chemistry to control sidewalls of thick
polymer layers with the geometric control over the sidewalls
and etch depth available from cryogenic silicon etching pro-
cedures. Cryogenic etching can also be utilized for deep sili-
con etching of MEMs mechanical resonator structures using
a two step process. This allows for the etch to minimize the
etch mask thickness requirements needed for deep reactive
ion etching. Etch recipes to reproduce all these etches were
also provided.
ACKNOWLEDGMENTS
M.D.H. gratefully thanks the Hertz Foundation for gener-
ous support. He also thanks Mike Shearn and Andrew
Homyk for helpful conversations. The authors also thank
Craig Ward of Oxford Instruments for technical equipment
support. The authors also acknowledge funding from
DARPA under HR0011-04-1-0054 and NSF under ECS-
0622228 and the CIAN ERC center as well as the Boeing
SRDMA program.
1
M. J. de Boer, J. G. E. Gardeniers, H. V. Jansen, E. Smulders, M. J. Gilde,
G. Roelofs, J. N. Sasserath, and M. Elwenspoek, J. Micromech. Mi-
croeng.
11
, 385

2002

.
2
X. Melhaoui, R. Dussart, T. Tillocher, P. Lefaucheux, P. Ranson, M.
Boufnichel, and L. J. Overzet, J. Appl. Phys.
98
, 104901

2005

.
3
A. F. Isakovic, K. Evans-Lutterodt, D. Elliot, A. Stein, and J. B. Warren,
J. Vac. Sci. Technol. A
26
, 1182

2008

.
4
T. Tillocher, R. Dussart, L. J. Overzet, X. Mellhaoul, P. Lefaucheux, M.
Boufnichel, and P. Ranson, J. Electrochem. Soc.
155
, D187

2008

.
5
B. M. Kayes
et al.
, 33rd IEEE Photovoltaic Specialist Conference, 2008.
6
L. Sainiemi, H. Keskinen, M. Aromaa, L. Luosujarvi, K. Grigoras, T.
Kotiaho, J. M. Makela, and S. Franssila, Nanotechnology
18
, 505303

2007

.
7
R. A. Gottscho, C. W. Jurgensen, and D. J. Vitkavage, J. Vac. Sci. Tech-
nol. B
10
, 2133

1992

.
8
D. Keil and E. Anderson, J. Vac. Sci. Technol. B
19
,2082

2001

.
9
S. A. Campbell,
The Science and Engineering of Microelectronic Fabri-
cation
, 2nd ed.

Oxford University Press, Oxford, 2001

, pp. 283 and 295.
10
M. W. Pruessner, W. S. Rabinovich, T. H. Stievater, D. Park, and J. W.
Baldwin, J. Vac. Sci. Technol. B
25
,21

2007

.
11
M. D. Henry and A. Scherer

unpublished

.
12
F. Laermer and A. Schilp, U.S. Patent No. 5,501,893

26 March 1996

.
13
J. Min, J. Lee, S. H. Moon, and C. Kim, J. Vac. Sci. Technol. B
23
, 1405

2005

.
14
J. Yeom, Y. Wu, J. C. Selby, and M. A. Shannon, J. Vac. Sci. Technol. B
23
, 2319

2005

.
15
C. C. Welch, A. L. Goodyear, T. Wahlbrink, M. C. Lemme, and T. Mol-
lenhauer, Microelectron. Eng.
83
,1170

2006

.
T
ABLE
IV. Etch parameters for second step of MEM resonator etch in
silicon.
Etch parameter
Units
Value
Etch
parameter
Units
Value
SF
6

SCCM

90
ICP

Watts

900
O
2

SCCM

6RIE

Watts

15
Temperature

Celsius

−140 Pressure

milliTorr

10
Etch rate


m
/
min

1.3
Helium

Torr

10
1216
Henry, Welch, and Scherer: Techniques of cryogenic reactive ion etching in silicon for fabrication of sensors
1216
J. Vac. Sci. Technol. A, Vol. 27, No. 5, Sep/Oct 2009