united states patent us 7,968,054 b1 li - nasa · mu uuuu ui iiui iiui mu uui lull uui imi iuu mui...

mu uuuu ui iiui iiui mu uui lull uui imi iuu mui mi uii mi

(12) United States Patent (lo) Patent No.: US 7,968,054 B1Li (45) Date of Patent: Jun. 28, 2011

(54) NANOSTRUCTURE SENSING ANDTRANSMISSION OF GAS DATA

(75) Inventor: Jing Li, San Jose, CA (US)

(73) Assignee: The United States of America asrepresented by the Administrator ofthe National Aeronautics and SpaceAdministration (NASA), Washington,DC (US)

(*) Notice: Subject to any disclaimer, the term of thispatent is extended or adjusted under 35U.S.C. 154(b) by 941 days.

(21) Appl. No.: 11/715,785

(22) Filed: Mar. 7, 2007

(51) Int. Cl.GOIN7/00 (2006.01)GOIN33148 (2006.01)GOIN27100 (2006.01)GOIN31100 (2006.01)GOIN19100 (2006.01)

(52) U.S. Cl . ............. 422/83; 422/68.1; 422/98; 702/22;702/23; 702/27; 702/30; 977/953; 977/957;

977/742;977/842(58) Field of Classification Search .................... 702/27;

700/266See application file for complete search history.

(56) References Cited

U.S. PATENT DOCUMENTS6,289,328 B2 * 9/2001 Shaffer ........................... 706/206,433,702 B1 8/2002 Favreau7,312,095 B1 * 12/2007 Gabriel et al . .................. 438/497,318,908 B1 * 1/2008 Dai .............................. 422/68.17,477,993 B2 * 1/2009 Sunshine et al ................. 702/227,623,972 B1 * 11/2009 Li et al ............................ 702/27

2003/0175161 Al * 9/2003 Gabriel et al . .................. 422/902005/0233325 Al * 10/2005 Kureshy et al .................... 435/62007/0202012 Al * 8/2007 Steichen et al . ................ 422/98

OTHER PUBLICATIONSJanata, Electrochemical Sensors, Principles of Chemical Sensors,1989, 81-239, Plenum Press, New York.Kong, et al., Nanotube Molecular Wires as Chemical Sensors, Sci-ence, Jan. 28, 2000, 622-625, 287, AAAS.Li, Chemical and Physical Sensors, Carbon Nanotubes: Science andApplications, 2004, 213-233, Editor: M. Meyyappan, CRC Press,Boca Raton, FL.Li, et al., Carbon Nanotube Sensors for Gas and Organic VaporDetection, NanoLetters, 2003, 929-933, 3-7, American ChemicalSociety.Liao, et al., Telemetric Electrochemical Sensor, BiosensorsBioelectromes, 2004, 482-490, 20, Elsevier B.V.DWL-AB650, a commercial product, D-Link Systems Inc., http://www.dlinkshop.com/.MICAz, a commercial product, Crossbow Technology Inc., http://www.xbow.com/.Nanotechnology Innovation for Chemical, Biological, Radiological,and Explosive (CBRE) Detection and Protection, Workshop Report,Nanoscale .... Nov. 2002, http://www.nano.gov.

* cited by examiner

Primary Examiner In Suk BullockAssistant Examiner Jennifer Wecker(74) Attorney, Agent, orFirm John F. Schipper; RobertM.Padilla

(57) ABSTRACT

A system for receiving, analyzing and communicating resultsof sensing chemical and/or physical parameter values, usingwireless transmission of the data. Presence or absence of oneor more of a group of selected chemicals in a gas or vapor isdetermined, using suitably functionalized carbon nanostruc-tures that are exposed to the gas. One or more physical param-eter values, such as temperature, vapor pressure, relativehumidity and distance from a reference location, are alsosensed for the gas, using nanostructures and/or microstruc-tures. All parameter values are transmitted wirelessly to a dataprocessing site or to a control site, using an interleavingpattern for data received from different sensor groups, usingI.E.E.E. 802.11 or 802.15 protocol, for example. Methods forestimating chemical concentration are discussed.

10 Claims, 13 Drawing Sheets

Pmvke napae caelflGente o for N;

M slrpW_a W. p

onussaM NcoelFpe (n=1, ...,

m . 1 ..., M; N>1, M>2) for on SWCW ne6

Add known Inc emend of one or more of the (suspected)Md oonvdWmb W sae G b Provide aupmeni aultl G'

Provide response values Vp (G; mesa) end Vn (G', rtnea)for u—on SV%Z no"V2 meulda Gand G

Provide —Pans. values V (G; meal) and Vc (G'. mess)foreoatinp no. n on.N4 network forlWida G end G'

W

FP_rovidekn

an error funceon 'I"'

q, q). dependent dunown mnaentretlons c1, q , q in eultls G end G'

WOehimine wlutlona (c1, q , q) fmm minimleetion of 96nvrkndbn a wllh resPecl b ohokxa acf, q, q

https://ntrs.nasa.gov/search.jsp?R=20110013149 2020-01-13T05:09:44+00:00Z

U.S. Patent

Jun. 28, 2011 Sheet 1 of 13 US 7,968,054 B1

30

ReceiverAnalyzer

Chemical Sensor(s)

S1 '

S2M ux

SK

11^

^^ 21

MOXA RS-23222 W2150

WirelessServer

^10

12

13

0.4

14

ConstantCurrent Source

—15

16

Temperature

17

PIC VaporPressure

18

Flow

19

RelativeHumidity

20

Distance

FIG. 1

U.S. Patent

Jun. 28, 2011 Sheet 2 of 13 US 7,968,054 B1

0.00118

Q 0.00116C

0.00114CNi 0.00112V

0.0011CD

p 0.00108CLWCD 0.00106LN 0.00104CDcn 0.00102

0.001

Gas (Benzene)

1 52 103 154 205 256 307 358 409 460 511 562 613 664 715 766 817 868

Time (s)

FIG. 2A

aO 0.25

tY

0.2

C

(D 0,15

a)> 0.1

0,05

U.S. Patent Jun. 28,2011 Sheet 3 of 13 US 7,968,054 BI

0.35

0.3

analyte detection

Sensor Sensor 12

Sensor

-0.050 5 10 15 20 2!

G

TIME ( hours)

FIG. 2B

U.S. Patent Jun. 28, 2011 Sheet 4 of 13 US 7,968,054 B1

250 5 10 15 20

TIME (Hours)

FIG. 3

1

1

xX

tD

0WU<;7Q 0

d

0

0


-0.10 5 10 15 20 25

TIME (Hours)

FIG. 4

0.7'

0

W_

U

W0c^

0

Qwfyw00

a 0

0


NA

a 0.6;

Z 0.5,W_

0.4HWO 0.3-,UZO 0.2'

w 0.10y0`O 0'0

-0.1 1

18 18.5 19 19.5 20 20.5

TIME (Hours)21 21.5 22

FIG. 5

—5.

—1.(

0

0

cu

94 —4-(

U.S. Patent

Jun. 28, 2011 Sheet 7 of 13 US 7,968,054 B1

u-0— nn 1

Sensor No.

3 5 7 9 11 13 15 17 19 21 23 25 27 29 31

FIG. 6

U.S. Patent

Jun. 28, 2011 Sheet 8 of 13 US 7,968,054 B1

Sensor No.

3 5 7 9 11 13 15 17 19 21 23 25 27 29 310.00E+00 1

5.00E-02w

-1.00E-01

Q -1.50E-01(D

L -2.00E-010

(D -2.50E-01m

-3.00E-01aD

0^ -3.50E-01

-4.50E-01

-4.50E-01

FIG. 7


Acetone

C12

F3

F1 Benzene

NO2 •F2

FIG. 8


V

Provide response coefficients Qn , m for M singleconstituent gases and N coatings (n = 1, ... , N;m = 1, ..., M; N? 1, M?2) for an SWCNT network

V

Add known increment of one or more of the (suspected)fluid constituents to gas G to provide augmented fluid G'

VProvide response values Vp (G; mess) and Vp (G', meas)for uncoated SWCNT network for fluids G and G' ,

V

Provide response values Vn (G; meas) and Vn (G', mess)94for coating no. n on SWCNT network for fluids G and G' ^^

V

Provide an error function e(c1, c2, c3), dependenton unknown concentrations c1, c2, c3 in fluids G and G'

V

Determine solutions (c1, c2, c3) from minimization oferror function c with respect to choices of c1, c2, c3

FIG. 9

05


Add (distinct) increments A i c(M O ) and A2c(M O) of selectedmolecule M O to fluid G to provide augmented fluids G 1 and G 2with unknown concentration co of M 0 in G

101

102

Measure or provide response value differencesAV, (M O ) = V(G j ; meas) - V(G O ; meas) (i = 1, 2)

103

Estimate coefficient V i in an approximation for response value differencesAV; (MO ) z Vp+ Vi {c 0(MO ) + A i c(M O )} (i = 1, 2)

FIG. 10


Provide measured response values V(EPp; mess)(p = 0, 1, ..., P; P? 1) for known environmental parameter 111values EPp of a selected environmental parameter EP

112

Provide a sequence of reference values V(EPp; ref; h)(h = 0, 1, ..., H; H >1) for each of H candidate fluidcomponents and each of P reference values of the parameter EP

Determine an error valueP q

2 E (h) =L w p V(EPp ; meas) - V(EPp ; ref; h) (h = 1, ..., H)P=1

for selected non-negative weights wp and a selected positive number q

Determine E(min) = min E (h = 1), ..., E(h = H)}= E (h0

115

No

YesIs E(min <E(min, thr)

117V V 116

No candidate fluid component

Candidate fluid component(s)is likely present in the gas

h = ho is/are likely present in the gas

FIG. 11

113

114


EPV(C)

Ctr C

FIG. 12

US 7,968,054 B11

2NANOSTRUCTURE SENSING AND

is used as the channel material and the conductivity was

TRANSMISSION OF GAS DATA shown to change upon exposure to NO z and NH, (J. Kong etal, "Nanotube Molecular Wires as Chemical Sensors," Sci-

ORIGIN OF THE INVENTION ence, vol. 287, pp 622-625, January 2000). The potential for

5 using CNTs in chemical sensors was noted in the Kong et alWork related to this invention was performed under a joint

article. However, it still is a challenge to make practical sen-research agreement between the National Aeronautics and

sors, due to difficulty in fabrication complexity, low sensorSpace administration and the University of California. yield and poor reproducibility.

A different configuration of carbon nanotube-based chemi-FIELD OF THE INVENTION

10 cal sensors with much easier fabrication process was intro-

duced by J. Li et al in "Chemical and Physical Sensors,This invention relates to wireless transmission of data pro- Carbon Nanotubes: Science and Applications", M. Meyyap-

vided by nano structure-based chemical and physical sensors.

pan, ed., CRC Press, Boca Raton, Fla., 2004. First, an inter-digitated electrode ("IDE") configuration is fabricated using

BACKGROUND OF THE INVENTION

15 conventional photolithographic methods with a nominal fin-

ger width of 10 µm and gap size of 8 µm. The electrode fingersChemical sensors have been developed for decades to are made of thermally evaporated Ti and Au (20 mu and 40 mu

detect various concentration levels of gases and vapors for thickness, respectively) on a layer of SiOz, thermally growndeployment in a wide range of applications in industry, space on top of a silicon wafer. Second, a thin layer of carbonmission, environment monitoring, medical, military, and oth- 20 nanotubes forming a network is laid on the fingers using aers. The detection usually centers on change in a particular solution casting process. The conductivity of the CNT net-property or status of the sensing material (such as thermal, work changes upon exposure of the fingers to different gaseselectrical, optical, mechanical, etc) upon exposure to the spe- or vapors. Such a process is significantly simpler and pro-cies of interest. The sensing material may be any of several

duces consistent sensors based on statistical properties of the

elements from the periodic table, plus inorganic, semicon- 25 CNT network with high yield (-100 percent). The IDE con-ducting and organic compounds, in bulk or thin film form. figuration facilitates effective electric contact betweenOne of the most investigated classes of chemical sensors in SWCNTs andthe electrodes overlarge areas, while providingthe past is the high-temperature metal oxide sensors due to good accessibility for analytes in the form of gas/vaporthese sensors' high sensitivity at low ppm to ppb concentra- adsorption or contaminants adsorption/extraction in/fromtion levels, with tin oxide thin films as an example. Polymer 30 liquid to all SWCNTs including semiconducting tubes.sensors have been studied in recent years because they can be Although carbon nanotube-based chemical sensors usingoperated at room temperature with low power consumption the IDE configuration have been prototyped in laboratories,and are easily fabricated. Although commercial sensors based use of an IDE configuration for wireless sensing has not beenon these materials are available, continued research is in reported in published research. Use of an IDE configurationprogress with sensing technologies using new sensing mate- 35 requires frequent transmission of data measured at each of arials and new transducer platforms. New sensing technolo- large number of IDE fingers. A straightforward approach maygies, such as nanotechnology-based sensors, are being devel- consume substantial power and may require use of a relativelyoped to overcome the large power consumption and poor large footprint for the sensor system. This may be inconsistentselectivity of metal oxides sensors, and to improve the poor with automated use of a remote system in a confined space.sensitivity and narrow detection spectrum of polymer sen- 40 What is needed is a wireless transmission system that (1)sors. has a small footprint and whose size is compatible with a

Typical figures of merit expected from a chemical sensor nanosensor system, which may have a diameter as small asinclude sensitivity (even down to a few molecules, selectivity, 5-15 cm, (2) consumes relatively little power (e.g., about 50low power consumption, rapid response time and rapid sensor µWatt-60 mWatt, or smaller with voltage regulation not acti-recovery time. Sensors based on the emerging nanotechnol- 45 vated), (3) is reliable, with a mean time to failure exceeding 8ogy may provide improved performance on all of the above hours (representative battery life) and (4) permits frequentaspects compared to current micro and macro sensors. Nano- data transmission from each of a large number of datamaterials exhibit small size, light weight, very high surface- sources.to-volume ratio, and increased chemical reactivity comparedto bulk materials; all these properties are ideal for developing 50 SUMMARY OF THE INVENTIONextremely sensitive detectors. The potential of nanomaterialsfor detection and for protection and remediation has been

These needs are met by the invention, which provides a

outlined in a recent report by the National Science and Tech- system with one or more sensors for selected chemicals (allnology Council ("Nanotechnology Innovation for Chemical, using nanostructure sensors with small physical sizes), one orBiological, Radiological and Explosive Detection and Pro- 55 more sensors forphysical parameters, a multiplexerto receivetection," November 2002). and interleave the measured data stream values from the

One promising nanomaterial is the carbon nanotube sensors according to a selected interleaving pattern, and a("CNT"), which exhibits extraordinary mechanical, electrical

wireless transmission module to transmit the measured val-

and optical properties. These interesting properties have ues to a receiver and data analyzer.prompted wide range investigations for applications in nano- 60 The overall sensor system consists of a chemical sensorelectronics, high strength composites, field emitting devices, module, a microcontroller-based data acquisition module, acatalysts, etc. A single-walled carbon nanotube ("SWCNT")

multiplexer and constant current source module, and a wire-

has all the atoms on the surface and therefore would be

less communication module. The chemical sensor module isexposed maximally to the environment, allowing a change in

based on use of an interdigitated electrode ("IDE") configu-

its properties sensitively. The first demonstration of an 65 ration. A layer of single-walled carbon nanotubesSWCNT-based sensor was for a chemical field effect transis- ("SWCNTs") is laid on the IDE fingers. Chemical sensing istor ("CHEMFET"), where a single semiconducting SWCNT

based on detection of changes in the conductivity of the

US 7,968,054 B1

3SWCNT network. In one embodiment, the system has 32channels of chemical sensing elements. It is self-containedand portable, and wirelessly transmits measurement data to aPC, using an I.E.E.E. 802.1 la, 802.1 lb or 802.15 wirelessLAN protocol. The footprint of the invention has a diameteras small as a few cm.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 schematically illustrates a sensor array and a wire-less transmitter according to an embodiment of the invention.

FIGS. 2A and 2B graphically illustrate sensor responsesfor identical exposures.

FIG. 3 shows the variance of measurement data from thesensor in FIGS. 2A and 213, computed with a moving windowof 1000 samples of the data.

FIGS. 4 and 5 illustrate correlation of two sensors within amoving window.

FIGS. 6 and 7 graphically illustrate relative chemical sen-sor response for exposure to NO z and for Cl z, respectively.

FIG. 8 graphically illustrates separation of the gases NOz,Clz, benzene and acetone according to principal componentanalysis.

FIGS. 9,10 and 11 are flow charts of algorithm proceduresfor estimating concentrations of initially unknown constitu-ents according to the invention.

FIG. 12 graphically illustrates transition from linear tologarithmic dependence upon a variable, such as concentra-tion.

DESCRIPTION OF BEST MODE OF THEINVENTION

FIG. 1 schematically illustrates components of a system10, constructed according to the invention, including a chemi-cal and/or physical nanosensor module 11, whose outputsignals are received by a multiplexer ("MUX") 12. An outputsignal of the MUX 12 is augmented by a constant currentsource 13 and is passed through a buffer amplifier 14 (op-tional) and is received and processed by a microcontroller 15(e.g., a PIC16F688, available from Microchip Technologies).In one embodiment, wherein an array of carbon nanostruc-tures is applied to environmental sensing, the microcontroller15 also receives a signal from at least one of a temperaturemicrosensor 16, a vapor pressure microsensor 17, a flowmicrosensor (direction, velocity) 18, a relative humiditymicrosensor 19 and an acoustic or electromagnetic distancesensor 20. The microcontroller output signals (e.g., in RS-232format) are received by a wireless communication module 21,which includes a wireless server 22 (e.g., a MOXA W2150),which transmits these signals, using 802.1 lb protocol,through an antenna 23 to a local area network (LAN). Thetransmitted signals are received by a receiver (not shown) thatis spaced apart from the transmission system 11, for displayand/or further signal processing.

Each of the individual sensors, such as 16, 17, 18 and 19,

has its own data reporting cycle, and it is assumed herein thatthese reporting cycles are numerically compatible. In a firstapproach, each of the reporting cycles has the same length At,and the four sensors report to a multiplexer in a consecutiveinterleave pattern as tm n={16, 17, 18, 19, 16, 17, 18, 19, 16,17.... } (n=1, 2, 3, 4; m=1, 2.... ) In a second version, at leastone of the sensors (16) has a reporting cycle length that is Ntimes as long as the other cycle lengths (e.g., N3), and thefour sensors report in a second interleave pattern as t m n { 16,17,18,19,17,18,19,17,18,19,16,17,18,19,17,18,...}.In a third version, the reporting cycle lengths are rational,

4non-integer, multiples of each other so that the reportingpattern appears random, for example, where At(16)=(7/4)At(17)=(3/5)At(18)=(7/4)At(19), a consistent third interleave pat-tern is: tm,,,={16, 18, 16, 17, 18, 18, 19, 16, 18, 18, 16, 17, 18,

5 19.... 1. For any interleave pattern {t. ,,,}. ,„ for data report-ing, the multiplexer transmits the received data in anothertime sequence It',n „ I , ,,, where t'm n is determined with ref-erence to t. ,, (e.g., t'm,,,=tn,,,+At, where At is a substantiallyconstant time delay value).

10 In one embodiment, the chemical sensor module contains32 sensing elements, arranged in an IDE configuration, hav-ing pure SWCNTs, having polymer coated SWCNTs, and/orhaving metal nanoclusters or doped SWCNTs. At the centerof the data acquisition system is a microcontroller

15 (PIC16F688) from Microchip Technologies Inc. that sampleseach sensor element through a set of four Texas InstrumentsCD4051 multiplexers. Each MUX 12 reads signals from agroup of eight chemical sensing elements. The LM234 con-stant current source from National Semiconductor is used to

20 provide a constant current (100 µAmp) to each sensing ele-ment. Four of these devices are used to excite each group ofeight chemical sensing elements. Conductivity or resistanceis measured by supplying a constant current and measuringthe corresponding voltage difference across the sensor. Also

25 included in the data system is an AD22100K temperaturesensor from Analog Devices. The microcontroller reads all 32chemical sensor and temperature values and generates a serialdata output that can be connected directly (1) to a wirelessserial device server, for wireless data transmission, or (2) to

3o an RS-232 serial data output fora PC, for data logging. In thisdesign, a W2150 serial server from Moxa Technologies, Inc.is used to wirelessly transmit data to a laptop or to a desktopPC, using an I.E.E.E. 802.1 lb wireless LAN protocol.

An alternative server is a Crossbow MICAz, using an35 I.E.E.E. 802.15.4 protocol at 2.4 GHz, can also be used for

wireless transmission. This system has a data transport rate of250 kbits/sec, and each node can also serve as a router. Thisallows a subset of one or more nodes to serve as a collectionpoint for collecting data that were measured elsewhere and

4o routing the data to a data processing module and/or to acontrol site that can respond to any problem encountered inthe data measurements or data processing. This system usesdirect sequence spread spectrum radio transmission, which isuseful where two or more data collection nodes are transmit-

45 ting substantially simultaneously to a data processing moduleand/or to a control site. Transmission range is up to 100 Moutdoors and up to 30 M indoors. Device dimensions are58x32x7 mm (diameter =68 mm). Analog inputs as well asdigital inputs are accepted by the MICAz system.

50 Another alternative server is an Air-Pro D-Link dual bandwireless LAN, operating at 5 GHz and 2.4 GHz, using801.11a/b protocol with 64/128 and 152-bit, shared key dataencryption for authentication and enhanced data security.Each node can also provide data collection and forwarding to

55 a data processing module and/or to a control site. Data trans-port rates are 1, 2, 5.5, 11 Mbits/sec (802.1 lb) and 6, 9,12,18,24, 36, 48 and 54 Mbits/sec (802.11 a). Transmission range isup to 400 M outdoors and up to 100 M indoors. Devicedimensions are I l8x54xl3 mm (diameter =133 mm).

60 FIG.1 schematically illustrates components of the system10, which includes a chemical nanosensor module 11, whosesensor output signals are received by the MUX 12. The outputsignal from the MUX 12 is augmented by a constant currentsource 13, and this signal is passed through a buffer amplifier

65 14 (optional) and received by a microcontroller 15 (e.g., aPIC16F688, available from Microchip Technologies). Themicrocontroller 15 also receives a physical sensor signal from

NV ^2)

Y, xk =Ci30 k =1

Here, k designates the sensor, i identifies the gas, and NV isthe total number of sensors (32 in one embodiment). Theo-

35 retically, c, can be any positive number, such as 1.In addition to analyzing the relative responses, the data are

autoscaled to unit variance and zero mean, by mean-centeringfollowed by dividing by the standard deviation:

40

x;k — xk (3)xik =

Sk

where x',k is the autoscaled response, x,k is the relative sensor45 response, xk is the mean value of normalized response for that

specific sensor, and sk is the standard deviation:

NP 1/2

50 1 (xik — xk)2^)1Sk = NP — 1 LI

where NP is the number of independent responses. Autoscal-55 ing removes any inadvertent weighting that arises due to use

of arbitrary physical units. In the absence of any informationthat would preclude its use, use of autoscaling is preferablefor most applications.

The processed data set contains the sensor responses from60 all 32 channels. Pattern recognition algorithms are very pow-

erful tools to deal with a large set of data. For example,principal component analysis (PCA) was applied to the dataset that was collected from this carbon nanotube based sensorarray. The PCA is to express the main information in the

65 variables X={xklk=l, 2 ... K)} using a smaller number ofvariables

1=—{i1, i, .. i,] (A-K), (5)

(4)

US 7,968,054 B15

at least one of a temperature microsensor 16, a vapor pressuremicrosensor 17, a flow microsensor (direction, velocity)18, arelative humidity microsensor 19 and an acoustic or electro-magnetic distance sensor 20. The microcontroller output sig-nals (presented, e.g., in RS-232 format) are received by awireless server 21 (e.g., a MOXA W2150), which transmitsthese signals, using an I.E.E.E. 802.1 lb protocol, through anantenna 22 and are received by a receiver and analyzer 30,spaced apart from the transmission system 10, for displayand/or further signal processing.Sensor Response

FIG. 2A shows typical sensor responses with the calcula-tion. FIG. 2B shows the sensor responses from three identicalsensors that were observed over a period of 25 hours fromthree sensing elements. The graph illustrates change of sensorresistance changes in response to the introduction of an ana-lyte, in this case ammonia, at approximately t=19 hours andt=20.5 hours. Data from the sensors were collected at onesample per second. For this experiment, a cotton swab soakedin ammonia hydroxide solution was placed approximately 2cm above the sensor surface, but the concentration of theanalyte was not quantified. The sensor resistance for sensorsx, y, and z increases approximately 10 percent from the sensorresistance measured just prior to the time(s) of detection ofthe analyte. Statistical properties of the measurement datawere investigated to extract the information that presents thedetection of analytes. Such information can be drawn frommeasurement data of a single sensor or from multiple sensors.

FIG. 3 shows the variance of the measurement data fromsensor "x" computed over a moving window of 1000 samplesof the data. This window represents 16.7 minutes of measure-ment data. Small variance indicates small changes in mea-sured data values over this period of time. A large varianceindicates possible presence of analytes. In FIG. 3, this vari-ance is substantially larger at two instants of time (t=19 and20.5 hours) when the analyte is introduced.

In addition to computing variance of individual sensor datavalues, statistical correlation of two sensors can be computedto confirm the presence of analytes. FIG. 4 shows the corre-lation coefficient for two sensors y and z, computed over thesame moving window. FIG. 4 indicates, from an agreementbetween two sensors, that both sensors were exposed to thepresence of analytes. An enlarged view of FIG. 4 around thetwo times of analyte introduction is shown in FIG. 5. Similarresults were also observed for sensor pairs (x,y) and (x,z).FIGS. 6 and 7 graphically illustrate relative sensor responsefor exposure of the chemical sensor array to NOz gas and toClz gas, respectively. The functionalized sensor-to-sensorresponse differences can be used to determine or estimate thegas that is present.Chemical Discrimination

The nanomaterials used in this array were individuallystudied for different gases and can be classified as: 1) pristinesingle walled carbon nanotubes, 2) carbon nanotubes coatedwith different polymers (here, chlorosulfonated polyethyleneand hydroxypropyl cellulose), and 3) carbon nanotubes

loaded with Pd nanoparticles and monolayer protected clus-ters of gold nanoparticles.

When exposed to a vapor-phase analyte, each sensing ele-ment in the array responds uniquely. A reproducible combi-nation of resistances, or "smellprint", for each vapor/gas ismanifest. The sensor response is measured as a bulk relativeresistance change (alternatively, current change or voltagechange): AR/R0, as in FIG. 3A. To calculate the relativesensor response (SR), the initial sensor resistance R O (or cur-rent/voltage), and the resistance R (or current/voltage) duringthe exposure are measured, and a normalized change,

6

SR = R, — Ro(1)Ro

5is computed.

By comparing the results of several gases, it is possible todetermine which sensors will respond in which manner (i.e.large or small amplitude, positive or negative). Looking at the

10 histogram of sensor responses from 32 sensors to differentgases/vapors, patterns for each gas/vapor can be establishedfor discrimination between tested gases/vapors. When the

sensor responses of an unknown subject gas are comparedwith the smellprints of several known substances, the

15 unknown subject gas canbe identified by matching its patternto, or minimizing a pattern difference relative to, one of theknown substances in the library.

However, the sensor responses often vary with concentra-tion of the gas, temperature, and other external factors. To

20 account for this, the data gathered by the sensors is normal-ized so that the relative responses can be compared, in orderto make the sensor information more accurate. In normaliza-tion, the length of all data vectors becomes the same bymaking the sum of the squares of vector components equal to

25 a constant so that the vectors become fixed length vectors.The data were normalized using:

US 7,968,054 B17

referred to as principal components of the variable X. In thisinstance, the X is a matrix of 32 sensor responses to thesegases and vapors from the sensor array. Here, three principalcomponents are labeled as F1, F2, and F3 that correspond totl , tz , andt3 in Eq. (5). FIG. 8 illustrates separation of the gasesNOz, Clz, benzene and acetone in terms of these three prin-cipal components.

These gases and vapors are completely separated in prin-cipal component space, which indicates that this sensor arrayprovides a high discrimination power to these gases andvapors. Because these gases were tested by this sensor arrayat different concentration levels in the range of 5-45 ppm, thissensor array can be used (uniquely) for high sensitive gas andvapor detection and discrimination. This sensor array dis-criminates the gases and vapors by their chemical naturerather than their concentrations. Also, the array responses candiscriminate between gases that have some similarity in theirchemical nature.

Further analysis is often required to reliably estimate con-centration of one or more target chemicals c0. After exposureof a nanostructure to a gas, a measured electrical parametervalue EPV (e.g., impedance, conductivity, capacitance,inductance, etc.) changes with time in a predictable manner, ifa selected chemical precursor is present, and will approach anasymptotic value promptly after exposure to the target chemi-cal. The measured EPVs are compared with one or moresequences of reference EPVs for one or more known targetprecursor molecules, and a most probable concentrationvalue is estimated for each of one, two or more target mol-ecules. An error value is computed, based on differences forthe measured and reference EPVs using the most probableconcentration values. Where the error value is less than anerror value threshold, the system concludes that the targetmolecule is likely. Presence of one, two or more target mol-ecules in the gas can be sensed from a single set of measure-ments.

For relatively large concentrations, it is assumed that themeasured EPV for a given target molecule will vary linearlywith the concentration C,

EPV(C)-a+b-C,tm (6)

where the parameters a and b are characteristic of the particu-lar target molecule for which the measurements are made.Where EPV(C l 1) and EPV(C l z) are values for a pure sub-stance measured at the different concentration values C,,1 andC1,2, respectively, the parameter values a and b can be esti-mated by

a-[EPV(C11)C1,2 EPV(C 1,2)C11 14C 1 ,2 C11 1, (7-1)

8which CNS sensor sub-arrays are more sensitive to presenceof a particular target molecule. In a first embodiment, thearray is exposed to a selected target molecule, such as H2021at a selected sequence {Cq}q (q=1, ... , Q with Q=N) known

5 (not necessarily distinct) concentration values (e.g., 500 ppm,14,000 ppm, 65 ppm, 1200 ppm, 10 ppm, etc.), and a refer-ence measurement value EPV(n;q;k;ref) (n=1, 2, ... , N) isrecorded for each sensor sub-array n, each concentration C.of a reference gas containing a selected target molecule (k).

10 Measurements for one or more concentration values for theselected target molecule gas can be repeated, if desired, toobtain an NxN square matrix of values. With q fixed, an NxNmatrix {EPV(n;q;k;ref)/EPV(n;k;norm)}=E(n;q;k;ref) is

15 formed, where, EPV(n;k;norm) is a normalization factor forthe selected target molecule gas no. k and the sub-array no. n,which may be chosen as

20 1^P (10)

EPV(n; k; norm) _ I N

^ ug EPV(n; q; k; ref) l,

q=1

where {uq}q is a selected set of non-negative weight numbers25 whose sum is a selected positive number (e.g., 1 or N) and p

is a selected positive number. The normalization factor E(n;k;norm) may be a single EPV (e.g., u q ,-1 anduq_0 for q;-ql),or may be a weighted sum of two or more EPVs. This nor-malization (optional) is intended to compensate for the con-

30 centration dependence of the particular target molecule.For each sub-array n (fixed), the mean and standard devia-

tion for each of K reference gases, numbered k=1, ... , K(K?2) are computed, as

35

N (11A)

µ(n; k) _ Y' E(n; q; k; ref) / N,q=1

40 K(11B)

µ(n) _ Y' µ(n; k) / K,k=1

N 1/2 (12A)

o-(n; k) _ JE(n; q; k; ref) -µ(n; k)}2/N^q=1

45

No-(n) _ Y. o-(n; k) / K.

q=1

(12B)

b-(EPV(Cl,l)-EPV(C l,2}1{C1,2- Cl,l }, (7-2) 50 One now forms K autoscaled NxN matrices, defined byfor each target molecule. For smaller concentrations (-1-50ppm), it may be preferable to use a logarithmic approxima-tion,

S(n; q; k)-(E(n; q; k,-ref)-µ(n)}/a(n), (13-k)

and analyzes K eigenvalue equationsEPV(C)--a'+b 1 1 og,C

( 8) 55

where a'andb'are selected parameters. The parameter valuesa' and b' can be determined in a manner similar to that of Eqs.(7-1) and (13-2), by replacing the variables C1,1 and Cl,z byloge{Cl , I and loge {Cl 2}, respectively. FIG. 12 graphicallyillustrates the two concentration regimes for EPV(C), corre-sponding to Eqs. (6) and (8), which join together at a concen-tration transition value C, for which

a'+b 1 1og,C a+b-C,. (9)

Given an array of N sub-arrays of CNS sensors and a set ofK target molecules (k=1, ... , K;K?2), the EPV data for thesensors can be pre-processed in order to identify more clearly

S(n;q;k) V(k;k(k))-k(k) V(k;k(k)), (14-k)

where k (=1, ... , K) is fixed and V(k;k(k)) is a normalizedNx I vector that will usually depend upon the reference gas(k). If, as is likely, the N eigenvalues k(k) for a fixed reference

60 gas k are distinct, the corresponding eigenvectors V(k;k,(k))are mutually orthogonal (non-degeneracy). In the unusualevent (degeneracy) that two or more of the N eigenvaluesk(k)are equal, different non-zero linear combinations of thecorresponding eigenvectors V(k;k(k)) can be constructed that

65 are orthogonal to each other, within a sub-space spanned bythe reduced set of eigenvectors corresponding to the identicaleigenvalues.

US 7,968,054 B19

Each matrix equation (14-k) has a sequence of N (eigen-value;eigenvector) pairs, {(X,,(k);V„(k;X,(k))},,, for a fixedreference gas k, and it is assumed here that the eigenvalues arearranged so that

11(k)I >_1^'2(k)>_ ... >_11r,(k)i (15-k)

and so that the highest magnitude eigenvalue in each setsatisfies

T 1 (k1)J_1 1(k2)I>_... Ik 1(kM 1 , (16)

where {kl, k2, ... , kN} includes each of the integers 11,2, ... , N} precisely once. The eigenvector V(kl;X,(kl)) isidentified as a first basis vector V'(1):

v'(k1;k 1(k1))—V(k1 ;k1(k1 )) (17-1)

A second modified vector

v'(k2 ;k 1(k2))—V(k2 ;k 1(k2)) — ( V(k2 ;k 1(k2)), v'(k1;kl

(k1))} v'(k1 ;k1(k1 )) 17-2)

is computed, where {V(k2;X2(k2)) 1 V'(1)} is the scalar prod-uct (also referred to as the inner product) of the vectors

V(k2;X2(k2)) and V'(kl ). More generally, a pth modified vec-tor

V ' (kP; A1(kP)) _ (17- P)

P-1

V (kP; A1(kP)) — Y, {V (kP; A l (kP)), V' (kr; A l (kr)) 1 V' (kr ; Al (kr))r-1

is computed for p=2, ... , K. The set of vectors {V'(kr;X, (kr))}k(r=1, ... , K) is mutually orthogonal, in the sense that thescalar products satisfy

{V' (kr; Al (kr)), V' (ks; A l (ks))I = 6,,,. (18)

> 0(r = s)

= 0(r:# s).

Each of the set of vectors {V'(kr;XJkrA r(r=1 1 .... K) ismaximally independent of each of the other vectors in the set,in the sense of mutual orthonormality (Eq. (18)). Each vectorV'(kr;X,(kr)) will have relatively large (primary) contribu-tions from some of the sensor sub-arrays and will havesmaller (secondary) contributions from the remainder of theN sub-arrays. The vectors V'(kr;X,(kr)) identify a maximallyindependent set of linear combinations of EPV responsesfrom the N sub-arrays that can be used to distinguish presenceof one reference gas (target molecule kr) from presence ofanother reference gas (target molecule ks). For example, if theset of reference gases are H2O2, H2O and CH3OH, N=3 andthree matrix eigenvalue equations are to be solved in Eqs.(9-k) (kA, 2, 3). More generally, presence or absence of anyof K target molecules (K?2) may be estimated.

The linear combinations LC(kp) of EPV measurements forthe different sensor sub-arrays correspond to modified prin-cipal components for the particular reference gases chosen.Choice of another set of another set of reference gases willresult in a different set of modified principal components,although change of one or more concentration values withina reference gas may have little or no effect on the modifiedprincipal components.

The preceding analysis concerning EPV measurementsand estimates of concentration can be done wholly at the dataprocessing site, partly at each of the data processing site and

10at the chemical sensor module site, or wholly at the chemicalsensor module site. The system provides an integrated chemi-cal sensor system based on use of SWCNTs as sensing ele-ments, including some preliminary testing results. The asso-

5 ciated data transmission system is wireless, small, portable,and readily deployable in the field.

A nanostructure, or assembly of such structures, can begrown, for example, by a procedure discussed in connectionwith FIG.1 in "Controlled Patterning And Growth Of Single

io Wall And Multi-wall Carbon Nanotubes," issued to Delzeitand Meyyappan in U.S. Pat. No. 6,858,197, incorporated byreference herein.

What is claimed is:1. A system for receiving, analyzing and communicating

15 results of sensing chemical and/or physical parameter values,the system comprising:

N sensors numbered n=1, ... , N (N?2), of parameters,where each sensor senses an electrical parameter valueEPV(tm,,,;n) of a selected chemical or a physical param-

20 eter at a selected sequence of times {t tm }m(m=1,2.... ), where at least one sensor senses a data set thatincludes at least one of a local temperature, a local vaporpressure, a local relative humidity, at least one of anacoustic signal and an electromagnetic signal for esti-

25 mating a distance between the sensor and a referencesurface, and presence of a selected chemical in a test gas,and where each sensor comprises a nanostructure (NS),having a sensing element having a diameter no greaterthan about 20 mu;

30 wherein at least one NS is loaded with a selected sensitiz-ing substance drawn from a group of sensitizing sub-stances comprising An particles, located between firstand second ends of the at least one NS, and the first andsecond ends are connected to first and second terminals,

35 respectively, of at least one of a voltage source and acurrent source, and to an electrical parameter value mea-surement mechanism that measures a change in at leastone of electrical current, voltage difference, electricalresistance, electrical conductance and capacitance

40 between the first and second ends of the at least one NS;a multiplexer, having at least N input terminals and at least

one output terminal, that receives at least one of a sensedvalue EPV(tm,,,;n) and a sensed value change AEPV(tm ;n) from sensor no. n, at a sequence of times,

45 {t t'm where the time t'_ ,,, is determined with refer-ence to a corresponding time tm ,,, and where each of afirst sequence of times {tm n1 }m is interleaved with asecond sequence of times {tm, a2l m for at least two inte-gers, nl and n2, satisfying 1 -nl<n2-N;

50 a wireless transmission module, connected to the multi-plexer output terminal, to receive and transmit at leastone of the sensed value EPV(tm,,,;n) and the sensed valuechange AEPV(tm,,,;n), received from the multiplexer, forat least one sensor no. n; and

55 a computer that is programmed to receive, from the wire-less transmission module, and to compare at least one ofthe second value EPV(tm,,,;n) and the measured sensedvalue change AEPV(tm,,,) in an electrical parametervalue with a corresponding reference value and to esti-

60 mate and indicate, based upon the comparison, presenceor absence of at least one chemical in a group of K targetchemicals, numbered k1, ... , K (K?1),

where the N sensors, the multiplexer and the wireless trans-mission module are contained in a volume having a

65 diameter no greater than about 14 cm.2. The system of claim 1, wherein said multiplexer is

located adjacent to at least one of said sensors.

US 7,968,054 B1

5

10

15

20

25

30

35

40

45where w„ are selected non-negative weight coefficients and pis a selected positive number;

to provide at least K equations relating the calibrationcoefficients ao,k and a,, ,, for k=1, ... , K, by partiallydifferentiating the error function EQCk})withrespectto

50each of the concentration values Ck, and setting each ofthe partially differentiated expressions of EQCk}) equalto 0; and

to obtain solutions Ck for the at least K equations, fork=1, ... , K, and to interpret these solutions as optimal

55solutions for the concentration values Ck.

4. The method of claim 3, wherein said computer is furtherprogrammed to choose said positive number p to be p=2P,where P is a positive integer.

125. The system of claim 3, wherein said computer is further

programmed:to receive a test measurement value for each of said N

sensors for a new test gas, where said concentrationvalue of at least one of said K target molecules in the newtest gas is not yet known;

to provide said error function EQCk}), where said calibra-tion coefficients, a o,k and a,, ,,, are determined as in claim3 and said concentration values Ck (k=1, ... , K) aretreated as initially unknown;

to minimize said error function EQCk}) with respect to achoice of value of at least one selected concentrationvalueC(k=k1) (1-k1-K); and

to interpret the at least one selected concentration valueC(k=k1) as a most likely value of said concentrationvalue for said target molecule no. kl in said new test gas.

6. The system of claim 5, wherein said computer is furtherprogrammed:

to minimize said error function EQCk}) with respect to achoice of value of each of said selected concentrationvalues Ck (1 -k-K) for said new gas;

to compute a value of said error function EQCk})_E(min)using said selected concentration values for said newgas; and

when the value Efrain) is no greater than a selected errorfunction threshold, E(thr), interpreting satisfaction ofthis condition as indicating that target gases are theprimary components of said new gas.

7. The system of claim 1, wherein a frequency of transmis-sion of a first reporting cycle for said first sequence changesmonotonically relative to a frequency of transmission of asecond reporting cycle for said second sequence in responseto change of a length of said first reporting cycle relative to alength of said second reporting cycle.

8. The system of claim 7, wherein said computer is furtherprogrammed:

to minimize said error function EQCk}) with respect to achoice of value of each of said selected concentrationvalues Ck (1 -k-K) for said new gas;

to compute a value of said error function EQCk})_E(min)using said selected concentration values for said newgas; and

when the value E(min) is greater than a selected errorfunction threshold, E(thr), interpreting satisfaction ofthis condition as indicating that at least one gas that is notone of said target gases is also a primary component ofsaid new gas.

9. The system of claim 1, wherein said frequency of trans-mission of said first reporting cycle decreases monotonically,relative to said frequency of transmission of said secondreporting cycle, as said length of said first reporting cycleincreases relative to said length of said second reportingcycle.

10. The system of claim 1, wherein said sensors, saidmultiplexer and said wireless transmission module togetherconsume no more than about 60 mWatt electrical power inoperation.

113. The system of claim 1, wherein:at least one of said sensors is exposed to said test gas to be

interrogated for presence or absence of one or more ofsaid K target chemicals in said test gas (k=1, ... ,K;K? 1);

measurement values EPV(n;test) are provided of a selectedelectrical parameter for said test gas for each said ofsensor no. n=1, ... , N (N?2), for each of said K targetchemicals and for each of a selected sequence of con-centrations Ck (k=1, ... , K; K--':N) of the target mol-ecule no. k; and

wherein said computer is further programmed:to provide at least N+1 functional relationships

K

EPV(G, f ; meas; n) = ao ,n + Y, ak,nCk(ref),k =1

EPV(G, f ; meas; n; C = 0) = ao,n,

relating a reference gas configuration, G,. fCj({Ck(ref)}), inwhich target chemical no. k, is present in a reference gas witha known, non-negative concentration value Ck(ref)(k=1, ... , K), to a test measurement value EPV(G r fineas;n)of the reference gas that is sensed by said sensor no. n, wherea,,,, and a,,,,, are known calibration coefficients that are inde-pendent of concentration values present, and ao n is an EPVvalue that would be measured by said sensor no. n in a test gaswhen all of the concentration values Ck are 0;

to receive or provide a sequence of test measurement val-ues, V„ (G;meas) and V,(G;meas), on said test gas in agas configuration G—GQCk}), where at least one of theconcentration values Ck of a target chemical isunknown;

to provide a numerical-valued error function EQCk}) asso-ciated with the test measurement values V„ (G;meas),where the error function is defined by

N P

2_-({Ck}) _ wn EPV(C; meal; n) — EPV(C; meal; 0)_

Y, a,,,kCk

n=1k =1

united states patent us 7,968,054 b1 li - nasa · mu uuuu ui iiui iiui mu uui lull uui imi iuu mui...

Documents