Content uploaded by Michel Demierre
Author content
All content in this area was uploaded by Michel Demierre on Mar 26, 2019
Content may be subject to copyright.
A vertical Hall device in CMOS high-voltage technology
Enrico Schurig
a,*
, Michel Demierre
a
, Christian Schott
b
, Radivoje S. Popovic
a
a
Swiss Federal Institute of Technology, Institute of Microsystems, CH-1015 Lausanne, Switzerland
b
SENTRON AG, CH-6300 Zug, Switzerland
Received 13 June 2001; received in revised form 17 August 2001; accepted 26 September 2001
Abstract
In this paper we demonstrate for the ®rst time, how vertical Hall devices manufactured in CMOS technology attain sensitivities comparable
to those of conventional silicon plate-shape devices without modifying the standard process or adding any post-processing steps. This was
achieved by taking advantage of the low-doped deep wells provided by a high-voltage technology and by applying additionally an
unconventional doping reduction method. It is demonstrated that deliberate violation of design rules can increase sensitivity without negative
in¯uences on the devices. The current-related sensitivity of the presented devices varies from 18 V/AT up to 127 V/AT for different sensor
geometry and doping concentrations. The linearity error is less than 0.04% for magnetic ®elds up to 2 T. #2002 Elsevier Science B.V. All
rights reserved.
Keywords: Vertical Hall; CMOS
1. Introduction
In-plane sensitive magnetic sensors are preferred in
important applications, such as angular position sensing
[1]. The vertical Hall device (VHD) [2] is a sensor naturally
adapted to these requirements. The principle of such a
device, realized in a special vertical Hall (VH) technology
is shown in Fig. 1 [3]. However, production costs for high
sensitivity VHDs (current-related sensitivities up to
SI400 V/AT) are high, since to obtain low-doped and
deep structures, a special technology is needed. Highly
sensitive VHDs in CMOS technology not only bear a cost
advantage, they also open the ®eld to co-integration with
electronics on the same chip. This way an integrated sensor
with compensated offset (by using the spinning current
method), low temperature dependence and ampli®ed sensor
output comes into grasp. Previous attempts to manufacture
them resulted in devices with poor sensor performances
(SI20 V/AT) or they required additional non-standard
processing steps in order to increase sensitivity [4±6].
A comparison between VHDs in VH and CMOS technol-
ogy can be seen in Fig. 2. The VH technology uses an n-type
wafer (constant doping concentration of 2 1014 cm
3
).
The sensor is open downwards and allows a deep current
¯ow (e.g. 30 mm). In the CMOS technology (substrate p),
the current ¯ow in the sensor is limited to 7 mm. The doping
pro®le of the diffusion layer is a Gaussian one with a surface
concentration of about 2 1016 cm
3
. The current ¯ow is,
therefore, concentrated near the surface (ca. 0±3 mm).
There are two major challenges to resolve for CMOS
technology. The VH devices, which we will discuss in this
paper, have to overcome the small n-layer depth and the high
doping concentrations with a Gaussian diffusion pro®le.
2. Sensor layout
2.1. General layout
Our sensor layout (Figs. 2c and 3) is based on the original
geometry given by Popovic [2]. The device consists of ®ve
contacts arranged in a line on top of a low-doped, active n-
diffusion region. A p-diffusion layer surrounds the active
area laterally. Since a deep and low-doped active area is
essential for high magnetic sensitivity, we decided to use a
high-voltage CMOS process, which provides n-diffusion
layers up to 7 mm deep.
2.2. Initial sensor and geometry variations
A ®rst sensor (VHS) was realized in a deep n-diffusion
layer (DNTUB, depth 7 mm) on a p-substrate. The plate
thickness is t9mm, the contact width w1:5mm and the
distance between two contacts is d10 mm. Starting from
this basic design, several variations were implemented to
Sensors and Actuators A 97±98 (2002) 47±53
*
Corresponding author. Tel.: 41-21-693-6732; fax: 41-21-693-6670.
E-mail address: enrico.schurig@epfl.ch (E. Schurig).
0924-4247/02/$ ± see front matter #2002 Elsevier Science B.V. All rights reserved.
PII: S0924-4247(01)00859-7
increase sensor sensitivity as well as to ®nd out the limits of
the technology for our application. For standard electronics
like MOS transistors, these limits are given by the design
rules in order to guarantee the function and reliability.
However, for our Hall elements we may try to go beyond
them for performance improvement. The most important
limit of CMOS technology for our purpose is the small depth
of the active n-region. One could assume that smaller VHDs
are less in¯uenced by the small depth than bigger ones,
because of the shallower current ¯ow. So our aim became to
scale down the sensor dimensions. In order to increase the
sensor performance one should consider the sensitivity
dependence on several physical and geometrical parameters.
For the current-related sensitivity S
I
the following relation
is given
SI/1
NDt;(1)
Fig. 1. Principle layout of a VH sensor. The Hall voltage is measured between the two probe contacts while a magnetic field perpendicular to the plane is
applied. The image shows the current flow as well as equipotential lines [3].
Fig. 2. (a)±(d) Comparison of VH sensors in different technologies. The VH technology uses an n-type wafer and the sensor is open downward and allows a
deep current flow (e.g. 30 mm) (a). The corresponding doping profile is constant in the depth with a concentration of 2 1014 cm
3
. In the CMOS
technology (c) (substrate p), the current flow in the sensor is limited to 7 mm. The doping profile of the diffusion layer is a Gaussian one with a surface
concentration of about 2 1016 cm
3
. The current flow will, therefore, be concentrated near the surface (ca. 0±3 mm).
48 E. Schurig et al. / Sensors and Actuators A 97±98 (2002) 47±53
while N
D
is the doping concentration and tthe thickness of
the Hall plate. Thus, the sensitivity S
I
can be increased either
by a decrease of the sensor thickness or by a reduction of the
doping concentration.
For the voltage-related sensitivity S
V
of a Hall plate, one
can ®nd
SVmH
w
lG;(2)
where m
H
is the Hall mobility of the majority carriers, w=l
the width-to-length ratio of the equivalent rectangle, and G
the geometrical correction factor. In order to increase w=lwe
must either decrease the contact distance or increase the
input contact width. We did not pursue the second possibility
since we wanted to have a structure with ®ve contacts of the
same size and the output contacts should be small for a high
G. In our ®rst optimization step we reduced the sensor
thickness to 6 mm (VHT) and 3 mm (VHVT), which is even
less than allowed by the technology design rules (5.4 mm).
The VHT structure has also been further modi®ed to reduce
the overlap of DNTUB and PTUB (Fig. 2c) (VHT2) in order
to see, if these regions have a positive or negative in¯uence
on the device sensitivity.
In another design (VHSH), the sensor length was reduced
by placing the contacts closer together, which should
increase S
V
according to formula (2). Instead of the
10 mm, we used d5mm while all the other geometrical
parameters were the same as for VHS.
2.3. Reduction of the doping concentration
2.3.1. Partial doping
According to formula (1), an increase in S
I
can be
achieved by reducing the doping concentration. A very
Fig. 3. Photograph of the realized VH sensor on-chip with magnified image of the sensor.
Fig. 4. Transformation of the conventional implantation mask (a) into one
that uses partial implantation (b), which means that one part of the active
area is covered during implantation. The covering stripes have to be small
(1 ...2mm) in order to achieve a continuous implanted layer after high
temperature diffusion.
E. Schurig et al. / Sensors and Actuators A 97±98 (2002) 47±53 49
ef®cient, patented method to do so, is to structure the
implantation mask for the n-well in stripes with a distance
smaller than tolerated by the design rules [7], instead of
making an opening over the entire active area. Fig. 4 shows
the transformation of the conventional implantation mask
into one that uses partial implantation. After high tempera-
ture diffusion, the originally separate zones join to form a
continuous layer with a lower overall carrier concentration
of Nd51015 cm
3
instead of 2 1016 cm
3
near the
surface. This partial implantation is illustrated by process
simulation results (Fig. 5). Applying this method to the VHS
design we got a more resistive sensor (VHL).
2.3.2. Shallow n-layer
A second way to achieve a lower doping level is to use a
shallow n-layer (SNTUB) with a depth of 5.5 mm (VHSN).
Even though this layer has a reduced depth compared to the
DNTUB, the four times higher sheet resistance should
increase S
I
considerably.
3. Experimental results
3.1. Sensor resistance and sensitivity
The reference sensor (VHS) yielded a current-related
sensitivity of SI18 V/AT and a voltage-related sensitivity
S
V
of 0.015 V/VT. This result is similar to the one achieved
by other authors [3].
In the next step, we monitored the effect of a thinner
device. Reducing the sensor thickness to 6 mm and ®nally
3mm yielded an increase in S
I
to 29 and 39 V/AT, respec-
tively. The voltage-related sensitivity is almost the same,
0.015 and 0.012 V/VT, respectively. However, for VHVT
the diffusion of the PTUB layer into the active area has
obviously a negative effect on the sensitivity, and therefore
the thickness should not be made smaller than about 4 mm.
If we compare the results for the VHT2 (SV0:023 V/
VT and SI58 V/AT) with those of the sensor VHS
(SV0:015 T
1
), we see that the DNTUB/PTUB overlap
regions (Fig. 2c) decrease the sensitivity and should be
avoided.
The reduction of the distance between contacts resulted
for VHSH in an increase of S
V
from 0.015 to 0.022 T
1
according to formula (2). S
I
did slightly decrease (16 V/AT)
because of the lower input resistance.
The partial doping technique resulted in an about four
times higher sensitivity S
I
(77 V/AT) because of the higher
resistance and a slightly higher S
V
due to the higher Hall
mobility. Using the high resistive SNTUB layer even higher
sensitivities were achieved (127 V/AT and 0.019 T
1
). The
presented results are summarized in Table 1.
Fig. 5. Illustration of the partial implantation technique. (a)±(c) Show the formation of the continuous layer during a high temperature diffusion process: (a) after
implantation; (b) after short diffusion at 1150 8C; (c) after complete diffusion at 1150 8C, the additional p-tub layer shapes the dimensions of the n-tub region.
Table 1
Basic sensor characteristics
Sensor
name
I
bias
±relative
sensitivity, S
I
(V/AT)
V
bias
±relative
sensitivity, S
V
(V/VT)
Input resistance,
R
in
(kO)
VHS 18 0.015 1.2
VHT 28 0.015 1.9
VHVT 39 0.012 3.6
VHT2 58 0.023 2.5
VHSH 16 0.022 0.7
VHL 77 0.016 4.8
VHSN 127 0.019 6.6
50 E. Schurig et al. / Sensors and Actuators A 97±98 (2002) 47±53
3.2. Non-linearity and temperature behavior
Non-linearity at high magnetic ®elds up to 2 T is for
all sensors smaller than 0.04% (Fig. 6). There is also a
certain sensitivity drift depending on the supply current
(Fig. 7). It is due to the junction ®eld effect. Since the
depletion layer of the p±n junction limiting the active sensor
area becomes larger with higher bias voltages, the input
resistance and the current-related sensitivity of the sensor
are increasing.
The sensors have a temperature coef®cient of about
4104K
1
(Fig. 8), which is of the same order of
magnitude as conventional silicon plate-shape devices.
The offset voltage of the sensors is up to several millivolts
corresponding to about 30 mT.
3.3. B-field cross-sensitivity
For the use as a 2-axes sensor, it is interesting to see how
the sensor responds to a magnetic ®eld parallel to a line
through its contacts. To measure this effect the sensor was
rotated in a constant magnetic ®eld of different values
around an axis perpendicular to the surface with an angular
position accuracy better than 0.0058.
The cross-sensitivity in¯uence is shown in Fig. 9. It
indicates the sensor output for a parasitic ®eld from 0.2
to 2 T. The calculated cross-sensitivity is about 0.01 V/AT at
B2 T and only 1/8000 of the Hall sensitivity of the sensor.
Since it is about quadratic with the B-®eld and since most
angle sensor applications work with ®elds in the millitesla
range, the effect of cross-sensitivity can be neglected.
Fig. 6. Current-related magnetic response of the Hall (VHL) device. The
signal deviation due to non-linearity is calculated to be less than 0.04%
between 0.3 and 2 T.
Fig. 7. Dependence of the supply current-related sensitivity on the bias
current (VHL). The sensitivity increase of about 7%/mA is due to the
junction field effect.
Fig. 8. The dependence of sensor sensitivity and input resistance on the
temperature (VHL). The calculated temperature coefficients are
cT;S4:1104and cT;R5:6103K1. The sensor resistance is
about 15 times more sensitive to temperature changes than the Hall
sensitivity.
Fig. 9. Cross-sensitivity of the sensor VHL. The field B
cross
is
perpendicular to the sensitive direction of the sensor. The sensor is biased
with 1 mA and the sensitivity between 0.8 and 2 T is about 0.01 V/AT. The
sensor output signal for B
cross
is about 0.0013% of the one at B
norm
.
E. Schurig et al. / Sensors and Actuators A 97±98 (2002) 47±53 51
3.4. Noise
To obtain an idea about the in¯uence of noise on sensor
detectivity, we may concentrate on thermal noise, since low-
frequency noise can be compensated applying the spinning
current method.
The voltage spectral density S
NV
of the thermal noise
because of the output resistances is given as
SNV4kTR;(3)
with Rthe output resistance, Tthe temperature and kthe
Boltzmann constant.
It can be related to S
I
and Iby the equivalent magnetic
noise density b
N
bN
4kTR
p
SII:(4)
The results for b
N
are presented in Table 2 for a constant bias
current at about 5 V and 298 K. The noise equivalent ®eld is
about 50 nT/
Hz
p, giving a limit for the minimum detectable
®eld.
3.5. Simulations
Process and current ¯ow simulations (2D), using the
software package ``ISE-TCAD'', show the doping concen-
tration pro®le of the sensor and the distribution of the current
density (Fig. 10).
They were executed for various sensor dimensions pre-
dicting resistances in good agreement with the real values
(errors less than 30%). The information regarding the cur-
rent ¯ow is very useful for sensitivity optimization of the
sensors. It helps also to see the precise diffusion behavior
and understand the monitored results. For a better analysis of
the results further 2D simulations (even better 3D) are
necessary and will be performed in the near future.
3.6. Optimized sensor
Although we already achieved a sensor with good S
I
,a
widely optimized sensor can be designed from the obtained
data. It should have a small thickness of about t4mm,
contact length of d5mm or even smaller, small overlap
region DNTUB/PTUB and the SNTUB layer as active area.
If necessary, S
I
could be further increased by the partial
implantation method. Sensitivities of about SI400 V/VT
as in the VH technology seem realizable. However, up to
now it was not found how the voltage-related sensitivity
might become larger than 0.05 T
1
.
4. Conclusions
For the ®rst time a VH sensor with a good sensor
performance (SI130 V/AT) has been realized in CMOS
technology without any post-processing step. Different
aspects regarding the geometry and the doping level of
the active area have been presented.
Combining the advantages of low doping (VHSN), partial
implantation (VHL) and narrow geometry (VHVT) CMOS
VH sensors can be further optimized for even higher current-
related sensitivity and low power consumption.
Certain design rules can be broken for further minimiza-
tion of the devices or decreasing the doping level without
diminishing the sensor function or its reliability. These
violations are important for the production of highly sensi-
tive devices.
The basic recipe how to implement high quality VH
sensors in CMOS technology is given by this work with
the potential to make a special technology obsolete and an
fully integrated sensor with high performance possible.
Acknowledgements
The authors would like to thank the ``Swiss Committee
for Technology and Innovation'' for funding this CTI project
Table 2
Output resistance R
out
and equivalent magnetic noise density b
N
for
thermal noise of various VH sensors, for the sensor working at 5 V at
298 K
Sensor name R
out
(kO)b
N
(nT/
Hz
p)
VHS 1.4 32
VHT 2.4 37
VHVT 4.5 64
VHT2 3.1 28
VHL 5.5 56
VHSN 8.2 55
Fig. 10. Simulation of the technology process (left half) and the current flow (right half) of the VH sensor. The contacts to bias the sensor are under the
arrows I
in
and I
out
. The Hall voltage is measured between the contacts S
1
and S
2
. A positive N-value stands for n-doping and the negative for p.
52 E. Schurig et al. / Sensors and Actuators A 97±98 (2002) 47±53
as well as the industrial partners, ``Austria Microsystems
(AMS)'' and ``SENTRON AG'' (Switzerland) for their
cooperation and support.
References
[1] F. Burger, P.A. Besse, R.S. Popovic, New fully integrated 3D silicon
Hall sensor for precise angular position measurements, Sensors
Actuators A 67 (1998) 72±76.
[2] R.S. Popovic, The vertical Hall-effect device, IEEE Electron Dev.
Lett. EDL-5 (9) (1984) 357±358.
[3] A.M.J. Huiser, H.P. Baltes, Numerical modeling of vertical Hall-effect
devices, IEEE Electron Dev. Lett. EDL-5 (11) (1984) 482±484.
[4] M. Paranjape, L. Ristic, W. Allegretto, A vertical Hall magnetic field
sensor using CMOS-compatible micromachining techniques, in:
Proceedings of the IEEE Solid-State Sensor and Actuator Workshop,
1992, pp. 32±34.
[5] R. Steiner, F. Kroener, T. Olbrich, R. Baresch, H. Baltes, In-plane
sensitive vertical trench-Hall device, in: Proceedings of the Interna-
tional Electron Devices Meeting, 1998, pp. 479±482.
[6] M. Kahrizi, M. Paranjape, L.M. Landsberger, Complementary metal-
oxide±semiconductor-compatible micromachined two-dimensional
vertical Hall magnetic field sensor: a modified design, J. Vac. Sci.
Technol. A 16 (2) (1998) 873±875.
[7] Procedure for the production of Hall elements. European Patent No.
1045 461 A2.
Biographies
Enrico Schurig was born in 1972 in Elsterwerda, Germany. He obtained
the university diploma degree in Physics in 1997 at the Martin-Luther
University, Halle-Wittenberg. His first field of interest was solid-state
physics for optical applications. After his graduation, he became interested
in taking a research occupation abroad. That is why he joined the Institute
for Microsystems at the Swiss Federal Institute of Technology, Lausanne
(EPFL) in1998, where he work in the field of magnetic sensor design
during his PhD studies.
Michel Demierre was born in Lausanne, Switzerland, in 1974. He received
his MSc in Micro-engineering from the Swiss Federal Institute of
Technology, Lausanne (EPFL), in March 1996. He is now a research
assistant in the Institute of Microsystems (EPFL), where he is engaged in
magnetic microsystems. His research interests include sensors and
systems.
Christian Schott was born in 1965 in Hildesheim, Germany. He graduated
from the Technische Hochschule Karlsruhe in 1992, where he got a
university diploma degree in electrical engineering. After his work on
several projects in technical development, he started in 1995 as a PhD
student at the Swiss Federal Institute of Technology, Lausanne, where he
had a special interest for magnetic field measurements with Hall sensors.
At the end of 2000, he joined a Swiss company (SENTRON AG) that is
specialized in magnetic Hall sensors.
Radivoje S. Popovic was born in Yugoslavia (Serbia) in 1945. He obtained
the Dipl. Ing. degree in applied physics from the University of Beograd,
Yugoslavia, in 1969, and the MSc and DrSc degrees in electronics from the
University of Nis, Yugoslavia, in 1974 and 1978, respectively. From 1969
to 1981 he was with Elektronska Industrija Corp., Nis, Yugoslavia, where
he worked on research and development of semiconductor devices and
later became head of the company's CMOS department. From 1982 to
1993 he worked for Landis & Gyr Corp., Central R&D, Zug, Switzerland,
in the field of semiconductor sensors, interface electronic and micro-
systems. There he was responsible for research in semiconductor device
physics (1983±1985), for microtechnology R&D (1986±1990) and was
appointed Vice President (Central R&D) in 1991. In 1994, he joined the
Swiss Federal Institute of Technology at Lausanne (EPFL) as Professor for
microtechnology systems. He teaches conceptual products and system
design and microelectronics at the Department of Microengineering of the
EPFL. His current research interests include sensors for magnetic, optical,
and mechanical signals, the corresponding microsystems, physics of
submicron devices, and noise phenomena.
E. Schurig et al. / Sensors and Actuators A 97±98 (2002) 47±53 53