PIEZORESISTIVE MEMBRANE-TYPE SURFACE STRESS SENSOR

ARRANGED IN ARRAYS FOR CANCER DIAGNOSIS THROUGH

BREATH ANALYSIS

Frédéric Loizeau1, Hans Peter Lang

2, Terunobu Akiyama

1, Sebastian Gautsch

1, Peter

Vettiger1, Andreas Tonin

2, Genki Yoshikawa

3, Christoph Gerber

2 and Nico de Rooij

1

1Ecole Polytechnique Fédérale de Lausanne, Switzerland

2University of Basel, Switzerland

3National Institute for Materials Science, Japan

ABSTRACT

We present the fabrication, characterization and

successful medical application of a membrane-type

surface stress sensor (MSS), arranged in arrays for

molecular detection in gaseous phase. Made out of

SOI substrate, a round membrane with a diameter of

500 µm and a thickness of 2.5 µm is suspended by

four sensing beams with integrated p-type

piezoresistors, composing a full Wheatstone bridge.

The membrane is coated with a thin polymer layer,

which reacts with volatile molecules and produces a

deflection of the membrane. The membranes were

functionalized with various polymers and

characterized as humidity sensors with a sensitivity of

87 mV/%RH and a time constant (Tau63%) of 1.3 s.

Finally, through breath analysis and the use of

principal component analysis (PCA), we were able, in

a double blind trial, to distinguish cancer patients and

healthy persons.

INTRODUCTION

While cancer is still one of the most lethal

diseases in the world, it is mostly highly curable if

detected early. Cancer diagnosis is usually performed

by endoscopy and taking a biopsy of suspect lesions.

In order to reduce the complexity and the risks of

such operations, non-invasive diagnosis tools are

proposed using cantilever sensors [1]. The major

sensing principle is based on analyte-induced surface

stress, which makes a cantilever bend. Volatile

organic compounds (VOCs) related to certain

diseases [2], e.g. head and neck or lung cancer, can be

detected.

In the perspective of point-of-care systems or

sensor networks, the bulky experimental setups

placed in laboratories need to be reduced in size to fit

in portable and easy-to-handle devices. Therefore, we

focus on surface stress sensors with integrated

piezoresistors rather than larger optical readout. To

overcome the sensitivity issue of cantilevers with

piezoresistive readout, we recently developed a

membrane-type surface stress sensor (MSS) with a

sensitivity increased by a factor of 20 compared to

conventional piezoresistive cantilevers [3].

Figure 1 shows a schematic of a MSS. A typical

shape of a MSS is a round membrane with a thickness

of 2.5 µm and a diameter of 500 µm. The membrane

is supported with four sensing beams, onto which

transvers and longitudinal piezoresistors are

integrated. Depending on the target analyte, a specific

polymer is coated on the top side of the membrane. Its

swelling, upon absorption of volatile molecules, will

produce the membrane deflection, which is detected

by the piezoresistors. In this paper, we present the

fabrication and characterization of the latest

generation. We finally used arrays of membranes in a

clinical trial to compare breath samples of four cancer

patients to those from four healthy persons.

Figure 1: Graphical representation of MSS. A

round membrane is suspended by four constricted

beams with integrated piezoresistors connected in a

Wheatstone bridge. The membrane is coated with a

polymer that reacts to surrounding molecules.

METHODS AND RESULTS

The fabrication of MSS is based on SOI wafers

that are structured by deep reactive ion etching

(DRIE). Figure 2 shows the main steps in the

fabrication process of the membrane. The cross

section shows one of the four constricted beams with

a partial view of the suspended membrane.

MEMS 2013, Taipei, Taiwan, January 20 – 24, 2013978-1-4673-5655-8/13/$31.00 ©2013 IEEE 621

Figure 2: Process flow of the membrane-type sensor

fabrication. The cross section represents one of the

four constricted beams with an integrated

piezoresistor. The membrane is separated from the

beam for clarity (e to f).

We start with an SOI wafer with a device layer of

3.5 µm that is thinned down to 2.5 µm by thermal

oxidation and subsequent removal of the oxide (Fig.

2a). The p-type piezoresistors are then created with

two distinct doping processes: deep and highly

conductive zones are initially created with the

deposition of boron silicate glass (BSG) and diffusion

of the boron atoms into the silicon at 1000°C for 30

minutes. Shallow layers, which are very sensitive to

surface stress, are then created by ion implantation

(Fig. 2b and 2c). The boron ions are implanted with a

dose of 2.5E+14/cm2 followed by a rapid thermal

annealing (RTA) at 1000°C for 10 s. After a

deposition of an 80 nm thick layer of Si3N4 as a

passivation layer (Fig. 2d), aluminium wires are

structured and encapsulated with a thick oxide layer

of 1 µm. The membrane and its sensing beams are

then structured into the device layer with deep

reactive ion etching (DRIE) (Fig. 2e). A 5 µm thick

layer of Parylene is deposited on the front side to

protect the membrane prior to its final backside

release by DRIE (Fig. 2f and 2g). The Parylene serves

both as a mechanical stiffening of the membrane

during the DRIE and as a protection layer during the

BHF wet etching of the buried oxide. It is removed

afterwards by oxygen plasma. Figure 3 shows SEM

pictures of the released membranes before their

functionalization.

Figure 3: SEM pictures of (a) a silicon chip

containing two arrays of membrane-type sensors and

(b) a close view on one membrane suspended by four

constricted beams.

As preliminary characterization, MSS was tested

as a humidity sensor since polymers react strongly

with water molecules. Two sensors within an array

were functionalized by inkjet spotting with cellulose

acetate butyrate (CAB). 500 droplets (10 nl each) of a

1 mg/ml solution were spotted onto each membrane,

resulting in a thickness of 700 nm after evaporation of

the solvent. The dynamic response of two

functionalized MSS to four relative humidity pulses

of 62% is displayed in Fig. 4. We measured a

response time of 1.3 s during the purging. The output

signal of a non-coated membrane is also shown on the

plot as a reference. We also characterized the static

response of an MSS within the humidity range of our

setup, i.e. 0% and 80% of humidity (Fig. 5). The

response is linear with a sensitivity of 87 mV/%RH.

622

Figure 4: Dynamic responses of three membranes to

four humidity pulses of 62% relative humidity. One of

the membranes was used as a reference without

coating, while the others were coated with a CAB

layer. The others showed a response time of 1.3 s. The

gain of the amplification stage is 500.

Figure 5: Static response of the membrane #1 to

humidity values from 0% to 80%. The measured

sensitivity is 87 mV/%RH.

Finally, an array of 2x8 MSS was functionalized

with 16 different polymers. The various absorption

properties of these polymers make them react

differently to the same analyte, creating a unique

signature (Fig. 6). Principal component analysis

(PCA) is used to differentiate complex mixtures of

gases and molecules.

Breath samples of both healthy people and

patients suffering from a head and neck cancer were

analyzed in a double blind trial. The breath samples

were transported and stored in Tedlar bags at a

temperature of 4 °C for 5 hours before analysis. Each

sample was pumped into a gas chamber containing

the functionalized MSS array. The sensors’ responses

were recorded six times and transformed by PCA to

reveal the differences between samples (Fig. 7). The

results show a good discrimination between the breath

samples of healthy persons and the ones of patients

suffering from a head and neck cancer.

Figure 6: Dynamic responses of six functionalized

MSS to ethanol pulses and nitrogen purges. The

functionalization polymers are PSS, Dextran, CMC

and PVP.

Figure 7: Principal Component Analysis case

scores for breath samples of 4 healthy persons and

4 cancer patients. Each sample has been measured

6 times (colored dots). A breath sample bag

containing saturated water vapor has been

measured as a control (blue dots). Healthy persons

can be clearly distinguished from cancer patients

(the ellipses are a guide to the eye.)

CONCLUSION

A new type of surface stress sensor consisting of

a suspended membrane and its fabrication process

have been presented in this paper. By clamping the

membrane in the four directions, the whole stress is

efficiently transduced on the constricted beams,

where the piezoresistors are located. Its sensitivity is

therefore increased compared to a standard

piezoresistive cantilever with a free-end. With the use

of different functionalization polymers, we created an

artificial nose that is sensitive enough to diagnose a

head and neck cancer through the analysis of the

breath.

623

ACKNOWLEDGEMENTS

We gratefully thank the Microsystems

Technology Division of CSEM for their help and

technical support during the microfabrication as well

as Dr. Laure Aeschimann for the fruitful technical

discussions. We also would like to acknowledge Dr.

Giorgio Mattana for his help during the inkjet

deposition of CAB. We express our appreciation to

Dr. Heinrich Rohrer for his initial investigations and

discussions on piezoresistive surface stress sensing,

which led to the invention and development of the

MSS concept and technology. We thank Jean-Paul

Rivals, Agnès Hiou and Pedro Romero from LICR at

CHUV for providing the breath samples. This project

is supported by Nano-Tera, an initiative of the Swiss

National Foundation (SNF).

REFERENCES

[1] M. Baller et al., A cantilever array-based

artificial nose, Ultramicroscopy 82, 9 (2000).

[2] D. Schmid et al., Diagnosing disease by

nanomechanical olfactory sensors – system

design and clinical validation, European Journal

of Nanomedicine 1, 44–47 (2008).

[3] G. Yoshikawa et al., Nanomechanical membrane-

type surface stress sensor, Nano Letters 11,

1044–8 (2011).

CONTACT

Frédéric Loizeau, +41-32-720-5469,

frederic.loizeau@epfl.ch

624