Nano Structural

Materials Today Volume 18, Number 3 April 2015 RESEARCH

nsot

nee

uststr

on

RESEARCH:Rev

iewspectrometric techniques of high sophistication and novel sensing materials become available. Here

advances in these technologies in connection to breath analysis are critically reviewed. A number of

breath markers or tracer compounds are summarized and related to different diseases, either for

diagnostics or for monitoring. Emphasis is placed on chemo-resistive gas sensors for their low cost and

portability highlighting their potential and challenges for breath analysis as they start to be used in

studies involving humans.

IntroductionThe idea using information in breath to determine the physiologi-

cal state of humans has its origins at the time of Hippocrates (460

370 BC). Much later, in 1782/83, Lavoisier reported the first

quantitative analysis of CO2 in the breath of Guinea pigs and

demonstrated that CO2 is a product of combustion in the body [1].

Modern breath analysis appeared in the late 1960s and early 1970s

together with modern analytical chemistry when Pauling demon-

strated by gas liquid partition chromatography that over 200

gaseous compounds appear in human breath [2]. At the time,

these compounds were not identified. Since then several reviews

[36] have shown the importance and potential of breath analysis

as one of the least invasive techniques for clinical diagnostics,

disease state monitoring and environmental exposure assessment.

This review describes the current state of breath analysis focus-

ing on the potential use of inexpensive and portable chemo-

resistive gas sensors that are already used for explosive detection

[7], air-quality monitoring [8] and as alcohol breath analyzers [9].

Such sensors can complement or serve as an alternative to more

sophisticated spectrometric systems for breath analysis and speci-

ation, especially for clinical diagnostics and monitoring. Here, the

current knowledge in breath components as illness markers is first

summarized, along with a classification of applied techniques.

Next, the main focus of this review, metal oxide chemo-resistive

gas sensors, is discussed in detail. Finally the review summarizes

the potential of such sensors for breath analysis and discusses

related challenges in connection to sensors properties and appli-

cations to humans.

Breath analysisThe bulk of breath consists of nitrogen, oxygen, CO2, water, inert

gases, and traces of volatile organic compounds (VOCs) at con-

centrations ranging from parts per trillion (ppt) to parts per

million (ppm) [10]. However, our understanding of the human

breath VOC spectrum is far from complete. A compendium of

1765 VOCs with a chemical abstract number (CAS-number) of

healthy humans has been published recently [11]. Most VOCs are

however, not generated in the body (endogenous), but derive from

food ingestion, exposure to an environmental pollutants (exoge-

nous) or from metabolization of a drug [12]. To monitor metabolic

or pathologic processes in the body, endogenous substances are

particularly attractive.

Breath markersSome VOCs present in the breath, the so-called breath markers,

correlate to specific diseases [13,14]. Furthermore, few pathologies

are related to a pattern of VOCs rather than a specific marker while

some breath components are related to more than one disease.*Corresponding author: Pratsinis, S.E. ([email protected])

1369-7021/ 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.mattod.2014.08.017163Breath analysis by naoxides as chemo-resiMarco Righettoni1, Anton Amann2,3 and S

1 Particle Technology Laboratory, Department of Mechanical and Process Engi2Univ.-Clinic for Anesthesia, Innsbruck Medical University, A-6020 Innsbruck, A3Breath Research Institute of the University of Innsbruck, A-6850 Dornbirn, Au

Recently breath analysis has attracted a lot of attentiostructured metaltive gas sensorsiris E. Pratsinis1,*

ring, ETH Zurich, CH-8092 Zurich, Switzerland

riaia

for disease monitoring and clinical diagnostics as

RESEARCH Materials Today Volume 18, Number 3 April 2015

RESEARCH:ReviewFigure 1 summarizes potential breath markers and their typical

concentrations in healthy humans (green bars) compared to those

during disease (gray bars). Each of these molecules is produced

endogenously (sometimes also by bacteria in the gut or in the oral

cavity) and has potential clinical relevance. Here few markers are

highlighted.

Acetone is one of the most abundant components in the breath

and has been studied in detail and related to metabolic disorders.

The acetone concentration in breath is mainly governed by its

concentration in the blood vessels supplying the upper airways

and by its respective blood:breath partition coefficient of 340[15,16]. The acetone concentration increases in patients with

FIGURE 1

Select breath markers with their average concentrations for healthy andunhealthy humans: acetone (diabetes) [17,21], ammonia (kidney disease)

[21,22], carbon monoxide (lung inflammation) [27,28], dimethyl sulfide (liver

disease) [34,35], ethane (schizophrenia) [38,39,42], hydrogen cyanide(bacterial infection) [43,44] and nitric oxide (asthma) [3033].uncontrolled diabetes (Fig. 1) [12,17] but also during normal

overnight sleep in healthy individuals [18] and during prolonged

fasting [19]. The acetone contained in the exhaled breath is a

metabolic product of the body fat breakdown and so it is expected

to be a good indicator of fat-burning [20].

Ammonia is present at relatively high concentrations in the

human breath (830 ppb) [21] and its increase might be related tobacterial production in the oral cavity, to kidney disease (Fig. 1)

[22], hepatic encephalopathy or infection with Helicobacter pylori

[23]. An accurate real-time monitoring of the renal function via

ammonia and trimethylamine [24] could improve the assessment

of treatment response and dialysis efficiency for end-stage renal

disease and acute kidney injury [22].

Exhaled carbon monoxide (CO) and nitric oxide (NO) are breath

markers to pulmonary diseases such as asthma, chronic obstruc-

tive pulmonary disease (COPD) and bronchiectasis [25]. Elevated

levels of exhaled CO (56.2 ppm) have been reported in untreated

asthma patients. Upon treatment with inhaled corticosteroids the

breath CO returns close to the healthy levels (1.61.8 ppm) [26].

The difference in exhaled CO between healthy [27] and asthmatic

humans [28], however, is much less than in exhaled NO. Fractional

exhaled nitric oxide (FENO) is the most extensively studied breath

marker as abnormal NO levels in the exhaled breath have

been documented in several lung diseases [25], particularly asthma

164[2931]. Increased levels of exhaled NO have been widely docu-

mented in patients with asthma [32,33].

Sulfur-containing compounds like dimethylsulfide are respon-

sible for the characteristic odor in the breath of cirrhotic patients

[34,35]. For healthy people the concentration of sulfur-containing

compounds in the breath is very low (few ppb) while for sick it is

slightly increased. For example, the dimethyl sulfide concentra-

tion increased from about 10 to 60 ppb in the breath of liver

disease patients [34]. Dimethylsulfide is also produced by bacteria

and fungi which are related to lung infection (e.g. Streptococcus

pneumoniae, Haemophilus influenzae) [36].

Many studies have demonstrated that breath ethane [37,38] is

related to lipid peroxidation [39] and oxidative stress [40]. Elevated

ethane levels have been noted also in other diseases [41] such as

breast cancer, heart-transplant rejection, rheumatoid disease,

acute myocardial infarction and schizophrenia [42].

Hydrogen cyanide (HCN) is a common metabolite in human

breath and an increase in its concentration might be attributed to

cystic fibrosis [43]. HCN may originate from endogenous produc-

tion, bacteria (e.g. Pseudomonas aeruginosa) [44], foods containing

cyanogenic glycosides, smoking and intoxication via inhalation

[43].

In addition to exhaled breath, volatile compounds can be

measured directly in the headspace of cell cultures [45]. Around

50 volatile compounds released or consumed by human cells have

been identified by spectral library match and retention time. Most

of these compounds have also been observed in exhaled breath,

skin emanations, urine headspace or in feces. In investigations of

lung cancer patients it turned out that there is no single volatile

biomarker, but much more a panel of compounds which differ-

entiates lung cancer from healthy volunteers [46].

As shown in Fig. 1 most breath markers show ranges of con-

centrations and may still vary considerably in the group of healthy

persons or even in one particular volunteer when observed longi-

tudinally. There are indeed variations between individuals that

depend on multiple factors, such as inter-individual physiological

differences, dietary dependencies, as well as sampling method and

measurement technique. Nevertheless, identifying the concentra-

tion ranges for each specific breath marker is a requirement for the

development of analytical tools (e.g. sensors).

Methods for breath analysisBefore addressing techniques for breath analysis it is important to

briefly address breath sampling and collection. This is important as

breath collection has not reached standardization yet and this

contributes to the variability of analytical results among different

studies [47]. Therefore, optimized procedures are a key step toward

optimization of breath analysis and standardized sampling meth-

ods should be implemented [48], such as for the determination of

the amount of NO in exhaled breath, for which standardized

procedures are already available (American Thoracic Society Docu-

ments 2005).

There are several techniques for breath component detection

and measurement that are categorized mostly into three groups:

methods based on gas chromatography or mass spectrometry (1),

laser-absorption spectroscopic techniques (2) and chemical sen-sors (including arrays or electronic noses) (3). Gas chromatography

(GC) is the most common method. Various detection methods can

be coupled to GC [49] to identify breath components, such as mass

spectrometry (MS) [10], flame ionization detector (FID) [40] and

ion mobility spectrometry (IMS) [50].

In addition, more sophisticated analytical methods have been

developed recently for real-time quantification of several trace

gases in both air and breath such as proton transfer reaction mass

spectrometry (PTR-MS) [51] and selected ion flow tube mass spec-

trometry (SIFT-MS) [52]. Furthermore chemical sensors based on

nanomaterials have shown promising results for clinical diagnos-

tics by breath analysis [53]. Most notably, electronic noses (e-

noses) consisting of sensor arrays offer a cheaper and simpler

alternative to GC and MS based techniques [54,55]. The most

common type of sensor arrays is based on conductive inorganic

nanomaterials such as metal nanoparticles [56], single wall nano-

tubes [57] and polymers [58]. E-noses are already widely deployed

in food processing, perfumes, chemical industry and environmen-

tal monitoring [59]. Furthermore, the use of sensor arrays in

medical diagnosis [60] has already shown promising results for

detection or monitoring of kidney disease (62 volunteers) [61],

diabetes (three diabetics) [58], Alzheimer (15 patients) [62], Par-

kinson (30 patients) [62] and lung cancer (62 patients) [56].

The numbers of papers published related to breath analysis had

Materials Today Volume 18, Number 3 April 2015

RESEARCH:Rev

iewan exponential increase recently (Fig. 2), with most of them

coming in the last decade. The number of publications on breath

analysis (including reviews and medical reports, Fig. 2: blue) has

more than doubled since 2000. Breath analysis has become in-

creasingly significant due to its potential applications in clinical

diagnostics and technological progress in gas chromatography and

mass spectrometry (yellow), spectroscopy (red) and sensors

(green). It is a non-invasive technique and is particularly attractive

for patients requiring monitoring of daily parameters such as

glycaemia and urea. Nevertheless, other than the most common

breath ethanol test for blood ethanol (law enforcement), only a

few types of breath tests have been used successfully as clinical

FIGURE 2

Number of publications related to breath analysis in general (blue) and

specifically on gas chromatography (GC) and mass spectrometry (MS,yellow), spectroscopy (red) and chemical sensors (green) for breath analysis

(Web of Science, January 2014).diagnostics, such as the 13/14C urea breath test in the diagnosis of

H. pylori infection [63], the CO2 detection by capnography [64] and

NO breath test in the diagnosis of airway inflammations [65].

Therefore, the blood assay remains the gold standard for diag-

nostic decisions [47]. According to the American Diabetes Associ-

ation, for instance, the glycemic control for diabetic patients is

performed primarily by two techniques: daily self-monitoring of

blood glucose (SMBG) and interstitial glucose, and of HbA1c (every

three months) [66]. However, a non-invasive measurement tech-

nique such as breath analysis would provide a major relief, espe-

cially for diabetic patients who have to give blood samples thrice a

day.

Metal oxide chemo-resistive gas sensorsAs described in the previous section, numerous emerging technol-

ogies are developed for use in non-invasive gas diagnostics. Che-

mo-resistive gas sensors based on metal oxides have been used

already quite extensively for several applications. In comparison to

other methods (e.g. GC and MS based techniques) such sensors

offer several advantages such as simplicity, high miniaturization

potential [67], low power and low cost production [68,69] that

make them attractive for routine clinical tests though they do not

provide the speciation of GC/MS. These sensors are based on the

electrical properties of metal oxide semiconductors. Such chemo-

resistive gas sensor films can be produced by a variety of techni-

ques [70] that may affect their sensing performance with respect to

sensitivity to ppb (Fig. 1), selectivity, reproducibility and long term

stability [69].

To apply such sensors to breath analysis some of their char-

acteristics (size, phase composition, structure) need to be tailored

during their synthesis and processing. First, sensors should exhibit

a high sensitivity to the low concentrations of analyte gases that

are present in the breath, ranging from ppt to ppm (Fig. 1). Second,

selectivity is crucial for the detection of a specific analyte (breath

marker) due to the large amount of similar compounds present in

the human breath. Third, the sensors also need to be able to work

at the high relative humidity (rh) of the breath (90%) [71] and tobe robust to its fluctuations. Last but not least, the sensors must

exhibit rapid response and recovery times for fast and on-line

measurements. Since sensor performance is closely related to

material characteristics and operational conditions, it becomes

crucial to investigate and optimize these parameters to fulfill the

above requirements.

Sensing mechanismChemo-resistive metal oxide gas sensors rely on changes of elec-

trical conductivity due to a change in the surrounding atmo-

sphere. Gas detection is related to the reactions between

ionosorbed surface oxygen and target analyte gas [72]. There is

a shift in the state of equilibrium of the surface oxygen reaction

due to the presence of a target analyte (receptor function) [73]. The

resulting change in chemisorbed oxygen is registered as a change

in the sensing material conductivity (transducer function) [73].

For example, when a reducing gas like CO or H2 comes into contact

with the sensors surface (Fig. 3, receptor function), it changes the

density of the ionosorbed oxygen. This is detected as a change

RESEARCHin the sensors conductance. Typically the sensing layer consists

of single crystalline grains that are loosely connected (Fig. 3,

165


RESEARCH:Reviewtransducer function) so the target gas can diffuse through the

porous sensing layer.

The usual operating temperature of metal oxide gas sensors is

from 200 to 500 8C [74]. To avoid long term changes, metal oxidebased gas sensors, should be operated at temperatures low enough

so that significant bulk variations are prevented [69], and high

enough so that gas reactions occur within the desired response

time.

The sensor response is determined by the efficiency of the

catalytic reactions with the analyte taking place at the sensor

surface. High catalytic reactivity of the sensor surface, and espe-

cially selectivity of this reaction to the target analyte are advanta-

geous. For this reason the control of catalytic activity of a new

sensor material is often used as the main method for a preliminary

estimation of its suitability as a sensor along with its optimal

operation temperature (Fig. 3c,d). The maximum catalytic activity

is obtained as the maximum sensor response (Fig. 3c) at different

temperatures for different gases for different sensing materials.

Therefore optimization of the operating temperature could be also

used to increase selectivity toward a specific analyte [75].

FIGURE 3

Receptor and transducer function of a semiconductor gas sensor (adapted from as a function of crystal size (adapted from [82]). (b) Sensor response of SnO2 and

Pd surface additives (at 0 [1], 0.12 [2] and 1.1 wt% [3]) on SnO2 gas sensor respo

sensors on sensor response and response time (adapted from [69]). Influence of

(adapted from [69]) and (f ) SnO2 film resistance (adapted from [78]).

166Despite the obvious similarities between chemical sensing and

heterogeneous catalysis, the choice of the material for gas sensor

applications is not determined only by catalytic activity. For

example, by surface modification of SnO2 films by noble metals

(e.g. Pd) a shift in the sensor response maximum toward lower

operating temperatures can be obtained together with an increase

in sensor response at an optimal additive content (Fig. 3c) [76]. The

surface modification (Fig. 3) creates optimal conditions for both

electron and ion (spillover effect) exchange between noble metal

and metal oxide support [75]. However, excessive increase in the

additive loading leads to a reduction in sensor response attributed

to a catalytic conversion without electron transfer, as exemplified

by Pt-doped SnO2 sensors during CO detection [77].

The sensor film thickness affects the response time (Fig. 3e).

Additionally, thin sensor films exhibit enhanced sensitivity [69].

For instance, the response of a SnO2-based sensor to ethanol was

increased five times by decreasing the film thickness from 600 to

50 nm [78] and this was attributed to the deeper penetration of the

analyte within the sensing film [79]. However, decreasing the film

thickness excessively, can dramatically increase the film resistance

[73]). Parameters influencing sensor performance: (a) SnO2 sensor response WO3 as a function of Si-content (adapted from [83,106]). (c) Influence of

nse (adapted from [76]). (d) Influence of the operating temperature of In2O3SnO2 film thickness on gas sensors (e) response and response time

(Fig. 3f) [78] that should be as low as possible to minimize power

consumption and increase measurement accuracy.

Therefore optimization of these parameters is essential in order

to obtain high performance sensors for breath analysis. The sens-

ing mechanism, briefly discussed here and more in detail in other

reviews [74,80], could be extended to the sensing mechanism

toward not only a single gas but to a much more complex mixture

as the human breath. However, it is important to take into account

the effect of having different gases at different concentrations on

the sensing element. There might be additional effects, such as

competition mechanisms between gases reaching the grain surface

and different diffusion rate between gases through the porous

sensing layer.

Sensor structure and morphologyNanomaterials can improve the sensitivity of gas sensors [81].

During sensor synthesis, the size as well as the shape of nanoma-

terials can be controlled, for example, in the form of particles

(spherical), rods, wires, and coreshell structures, which determine

their chemical, optical, magnetic, and electrical properties. For

instance, SnO2 gas sensors exhibited higher sensitivity toward H2and CO when their crystallite size was decreased (Fig. 3a) [82]. The

gas sensitivity increased steeply as the crystallite size decreased and

sensitivity depends on the surface area of the sensing particles [74].

Spheres are inferior in terms of surface-to-volume ratio to quasi

one (1D) and two (2D) dimensional geometries like rods and

plates, respectively. Therefore considerable effort has been made

during the last two decades to optimize the geometry of produced

nanostructured particles to increase their surface-to-volume ratio

(Fig. 4) [85].

One dimensional nanostructures have attracted considerable

interest due to their high surface-to-volume ratio and unique

electro-conductive properties, making them promising candidates

for highly sensitive and potentially low temperature gas sensors

(Fig. 4). Depending on the aspect ratio, one can differentiate

several 1D architectures in ascending order: nanorods [86], nano-

tubes [87], nanobelts [88], nanowires/nanofibers [89]. A major

advantage of nanobelts is their lack of crystallographic defects

due to their rectangular shape, promoting ideal sensing [84]. For

example, vanadium pentoxide (V2O5) nanobelts (Fig. 4f) showed

to be highly selective to ethanol against different gases even at

relatively low temperature (150 8C), which is advantageous forcheap and low power device applications [88].

Figure 5 shows the response to H2S of WO3 nanoparticles

(dotted line) in comparison with 2D platelets (dashed line) and

1D wires (black line), respectively [90]. Possibly due to the men-


d W

RESEARCH:Rev

iewbecame comparable to about twice the Debye length (6 nm forSnO2) [82]. As a result, the sensitivity to ethanol could be en-

hanced by tuning accordingly the SnO2 grain diameter through Si-

doping of SnO2 sensors [83].

Taking into account the complexity of the gas-sensing mecha-

nism, it becomes clear that one has to consider the influence of

various structural parameters of the metal oxide matrix on gas

sensor performance [84]. The sensing mechanism of chemo-resis-

tive gas sensors is mainly driven by reactions of the analyte gas

with chemi- and physisorbed surface oxygen species. Therefore

FIGURE 4

TEM images of, (a) WO3 nanoparticles, (b) ZnO nanorods [86], (c) Au decoratenanofibers [89], (f ) V2O5 nanobelts [88], (g) ZnO nanoplatelets (sheets) [92], (h) Z

Reproduced with permission from the corresponding journals.tioned reduction in aspect ratio and the associated reduction in

surface-to-volume ratio, nanoplatelets showed inferior sensing

performance compared to 1D nanostructures but still preferable

to not specifically designed nanoparticles at all temperatures [90].

The increased sensor sensitivity for increasing surface-to-volume

(SV) ratio is attributed to the increased adsorption of analyte gas

molecules onto the sensor surface [85]. Furthermore, the WO3nanowire and platelet sensors exhibited constant sensitivity when

increasing the relative humidity up to 60%. However, upon in-

creasing the rh up to 90% the sensitivity slightly decreased [90].

O3 nanorods [91], (d) In2O3 carbon nanotubes (CNT) [87], (e) SnO2

nO hollow spheres [96], and (i) Au decorated SnO2 hollow spheres [97].

167


FIGURE 6

Sensing performance was affected by different structures: (a) Decoration ofWO3 nanorods with Au nanoparticles increased the sensitivity and

selectivity to hydrogen (adapted from [91]). (b) Gas sensor response to H2S,

acetone and toluene of pristine and Pt-functionalized WO3 hemitubes(adapted from [93]). (c) Comparison of sensitivity to several analytes

(adapted from [96]) between commercial ZnO particles and hollow spheres

(Fig. 4g). (d) Decoration of SnO2 hollow spheres with Au nanoparticles

increased the sensitivity and selectivity to ethanol against acetone (adaptedfrom [97]).

RESEARCH:ReviewAdditionally these structures can be decorated with sensing en-

hancing elements, as for example catalytically active palladium or

gold, supporting and facilitating the gas sensing mechanism as

illustrated in Fig. 6a where WO3 nanowires were decorated with Au

nanoparticles (Fig. 4c) increasing both sensitivity and selectivity

[91].

Decreasing the aspect ratio and maintaining rectangular cross-

section yields two-dimensional structures like nanoplatelets [90].

Gas sensors made of two-dimensional ZnO nanosheets of about

1020 nm thickness (Fig. 4g) detected selectively acetone and

gasoline at 360 and 180 8C, respectively [92]. Thin-walled WO3hemitubes made by polymeric fiber-templating were used as sen-

sors for possible diagnosis of halitosis and diabetes, through the

detection of H2S and acetone (down to 120 ppb), respectively [93].

These structures exhibited very high H2S selectivity (Fig. 6b) at

high (85%) relative humidity, mandatory for breath analysis

FIGURE 5

WO3 sensor response to 1000 ppm H2S for different structures as a functionof temperature. The WO3 nanowires exhibited better sensitivity to H2S than

nanoplatelets or nanoparticles at all temperatures (adapted from [90]).applications. However, even though the sensor response was

relatively decreased by Pt-functionalization, such Pt/WO3 hemi-

tubes exhibited superior acetone sensing with negligible response

to H2S and toluene (Fig. 6b). This was attributed to the spillover

effect [94] of the Pt catalyst, which can effectively dissociate

adsorbed oxygen molecules into ionized oxygen (O, O2) onthe WO3 surface, accompanied by the capture of electrons from

the WO3 conduction band [93].

As controlled mass transfer toward the active sites is of crucial

importance in catalysis, gas sensing or battery applications, a great

effort has been made to compose films of hollow multilayered

porous spheres maximizing mass transfer throughout the bulk

films and maintaining a high surface-to-volume ratio [95]. Such

composed hollow spheres showed enhanced performance com-

pared to usual nanoparticles in gas sensing and lithium-ion bat-

teries [95]. Figure 6c shows the response toward various analytes by

a gas sensor consisting of ZnO hollow spheres (Fig. 4h) compared

to a sensor based on commercial ZnO [96]. The sensor response

and selectivity to NO2 was strongly enhanced with the hollow

sphere based gas sensors (Fig. 6c). Additionally, and similarly to

Fig. 6a,b, decorating SnO2 hollow spheres with Au (Fig. 4i) en-

hanced their sensitivity and selectivity to ethanol over bare SnO2

168(Fig. 6d) [97]. Recently, even more complex structures have been

prepared to obtain high performance sensors such as urchin,

flowers, cubes and various hollow structures [95].

Sensor compositionTin oxide is the most common metal oxide for gas sensing [68,81].

It is sensitive to many gases and organic compounds [68] but this is

also its major drawback: selective sensors based on sole SnO2 are

not available yet. Great effort has been made to improve that, by

varying crystal structure or morphology, adding dopants, chang-

ing the operating temperature, etc. [69]. The application, however,

of SnO2 as a gas sensor for breath analysis has been limited by its

cross-sensitivity to humidity (major component of the human

breath). Additional efforts have been made to overcome this, such

as the embedding of active filters placed above the SnO2 semicon-

ductor layer in order to exclude undesired gases [98]. Flame-made

SnO2 doped with about 5% TiO2 resulted in a solid solution

material with reduced cross-sensitivity to humidity and increased

sensitivity to ethanol [99].

A further challenge is the relatively poor stability, intrinsic to

nano-sized materials at the usually high operating temperature.

This might result in drifts of the baseline resistance and sensor

RESEARCH:Rev

iewresponse. A common practice to improve thermal stability is the

use of dopants [100], that may decrease, however, the overall

performance [69]. For example, Si-doping increased thermal sta-

bility and enhanced sensing performance for ethanol detection

down to 100 ppb (Fig. 3b) [83]. Furthermore one dimensional

SnO2 structures enhanced sensitivity and selectivity [85].

Tungsten oxide was used for the first time as a H2 detector in

1967 for safety applications [101]. It has attracted a lot of interest

due to its various crystal structures: monoclinic (e phase), triclinic(d phase), monoclinic (g phase), orthorhombic (b phase) and

tetragonal (a phase) [102]. A highly sensitive gas sensor based

on a WO3 thick film has been developed for the detection of trace

amounts of aromatic hydrocarbons [103]. That WO3 sensor had

enough sensitivity to detect traces of VOC gases, especially aro-

matic hydrocarbons, at the ppb level with excellent selectivity

[103].

As with SnO2 also WO3 is sensitive to many gases. However, its

different phases have different sensing properties. In particular, its

e-WO3 phase is highly sensitive and selective to acetone [104]. Onthe other hand, g-WO3 is promising for selective NO detection

[105]. Epsilon-WO3 that was thermally stabilized by Cr- [104] or Si-

doping [106] selectively detected acetone down to 20 ppb at 90%

rh [107]. Furthermore, the Si-doped WO3 sensor response was only

decreased by 4% when increasing the rh from 80 to 90% indicating

sufficiently precise detection regardless of humidity fluctuations

[107].

Other materials such as TiO2, ZnO, MoO3, etc. have been

investigated also for gas sensing. Titania-based gas sensors are

particularly attractive for their lower cross-sensitivity to humidity

than SnO2 sensors [99]. Titania sensors were applied to detection

of organic compounds such as acetone and isoprene and showed

high sensitivity down to 1 ppm [108]. Zinc oxide has been studied

for its large scale and low-cost synthesis by hydrothermal methods

resulting in 1D structures (Fig. 4c) [86] like ZnO nanorods stable

and highly sensitive sensors, especially for ethanol detection with

short response time. Similarly to WO3, also MoO3 can be synthe-

sized in different crystal structures for the selective detection of

specific gases such as orthorhombic MoO3 [109]. Most notably

promising results were obtained by MoO3 films for selective detec-

tion of NH3 in simulated exhaled breath that contained CO2, NO2and isoprene [110]. Additionally the NH3 selectivity could also be

improved by optimizing the operating temperature (400450 8C)[111].

Applications to human breathToday only a few studies involving human subjects have been

made with chemo-resistive gas sensors. To start using sensors for

breath analysis, their measurements need to be benchmarked

against spectroscopic techniques that provide breath speciation.

For instance WO3-based thin films have been used for quantitative

detection of NO in the human breath of four people collected in

Tedlar bags (off-line) and compared with a chemiluminescence

NO analyzer for validation [112]. The sensor prototype demon-

strated the capability of breath NO monitoring in the range

between 0 and 100 ppb. However, this was accomplished by first

oxidizing NO to NO2 by a catalyst (KMnO4) and the response to

Materials Today Volume 18, Number 3 April 2015 other organic compounds was reduced by non-polar molecular

sieves (e.g. silicalite) [112].Portable Si-doped WO3 gas sensors accurately monitored

breath acetone concentrations of healthy humans both on-line

[113] and off-line (Tedlar bags) [114] with proton-transfer-reac-

tion time of flight mass spectrometry (PTR-TOF-MS) [115]. On-

line sensor measurements enabled the monitoring of acetone

concentrations during rest and physical activity of five human

subjects, in good agreement with PTR-MS (five healthy persons)

[113]. Off-line measurements of eight healthy volunteers showed

high correlations between acetone, sensor response and blood

glucose after overnight fasting [114]. This highlighted the poten-

tial of gas sensors to replace glucose testing, at least, after over-

night fasting.

Nevertheless clinical studies with several patients are required in

order to assess such sensors further and make them a clinical

reality. Recently, a prototype portable breath acetone analyzer

was developed for on-line monitoring fat burning and was tested

on 17 healthy adults [116]. The acetone concentrations were

validated successfully by GC measurements of breath collected

into Tedlar bags. The device included two sensors (SnO2 and WO3)

that enabled the acetone concentration to be calculated while

taking into account other background gases present in the breath

such as ethanol, hydrogen and humidity [116].

An interesting alternative to metal oxide nanomaterials is

represented by conducting polymers. Polyaniline (PANI), for ex-

ample, displays high sensitivity at room temperature that is ad-

vantageous for reduction of power consumption and increased

portability [117]. For example, very promising results have been

obtained by inkjet-printed PANI nanoparticle sensors for breath

ammonia detection (down to 40 ppb), related to dysfunction of

the kidney and liver, and operated at room temperature [118]. A

cohort of 20 hemodialysis patients breath samples was evaluated

and showed good correlation between breath ammonia and blood

urea nitrogen levels. These sensors were validated by photoacous-

tic laser spectroscopy (PALS) [118]. Additionally, inorganic nano-

particles, such as metal oxides, are combined with conductive

polymers in a sensing layer to exploit synergies in terms of the

sensing performance [119] while reducing the sensor operating

temperature [118].

Potential and challengesRecent studies have led to increased sensor portability, decreased

power consumption (low operating temperature), lower limit of

detection and in some cases even increased selectivity. Further-

more, due to their miniaturization potential sensors could be

easily integrated in daily used commodities, such as mobile

phones. The first step toward the use of such devices for breath

analysis remains the validation of sensor with well-established and

selective techniques, such as GCMS, PTR-MS or similar as de-

scribed above. Figure 7 shows an example of the different sampling

methods for that purpose. The breath sample can either be directly

analyzed [113] (1) (2) or first sampled (3) (4) in a container (e.g.

Tedlar bag) [114]. Additionally, the portion of the breath can be

either controlled (e.g. end tidal fraction) for example with a CO2trigger or the whole lung capacity can be considered. Then the

same breath sample should be analyzed by both sensor and an

analytical device. Again it is of crucial importance that the sam-

RESEARCHpling procedure is performed accurately. For example, to avoid

condensation and thus alteration of the breath composition, all

169


RESEARCH:Reviewthe lines and the Tedlar bags need to be heated at 40 8C during off-line breath sampling (Fig. 7).

The majority of the studies on metal oxide gas sensors are

carried out in dry air conditions, far away from the high relative

humidity of the breath (90%). Therefore there is the need toperform additional studies at realistic conditions to asses those

sensors for breath analysis and come a step closer to their imple-

mentation in clinical studies. Moreover the restricted selectivity is

yet the most limiting factor, especially for trace concentrations

(ppt) of target analytes. Several strategies in addition to tailoring

the material properties, addition of dopants and optimizing the

operating temperature have been used to improve the selectivity,

such as pre-treatments (e.g. sodium hydroxide filter) [120], layer

filters (e.g. Pd/Al2O3) [121] or even additional sensors to monitor

humidity or background gases (e.g. CO2). For instance these

methods improved the sensor performance at high relative hu-

midities [120] and the selectivity to NH3 against water and CO2[120] or to CH4 against water, CO and ethanol [121].

FIGURE 7

Schematic of breath analysis (on-line and off-line) by gas sensors and to

spectroscopic methods. The breath can be either directly analyzed on-line

where the whole breath pulse is sampled (1) or (2) only a portion (e.g. endtidal fraction) after being separated by a control unit. Similarly for off-line

measurements the whole breath (3) or a portion of it (4) was sampled in

bags or rigid canisters (3).Up to now only a few chemo-resistive gas sensors have been

applied to breath analysis and most of recent studies, especially for

cancer research, were obtained by sensor arrays [60], therefore,

avoiding the problem of selectivity. For example, recently, it has

been demonstrated that a breath test based on a nanoscale artifi-

cial nose could distinguish patients having breast, lung, colon, and

prostate cancer and healthy controls [122]. The e-nose was based

on an array of highly cross-reactive gas sensors [56] that could

identify and separate different traces. Each sensor showed an

individual response to all (or to a certain subset) of the volatile

biomarkers that make up the cancer-odor. The odor was identified

by pattern recognition of the sensor signals [122]. In general

pattern recognition algorithms can be used to obtain information

on the overall breath composition [56] as well as concentration of

specific breath compounds [58]. Moreover due to the fact that the

accelerated metabolism of tumor cells may produce a number of

VOCs that could differ both qualitatively and quantitatively from

VOCs released by healthy subjects, in most of the studies for cancer

research a pattern of VOC instead of a single one has been studied

and therefore sensor arrays that are not limited by a single gas

detection are required [123].

170Finally, the possibility of using this sensor at room temperature

and therefore avoiding the need of heating, reducing power

consumption and facilitating miniaturization, would be advan-

tageous in terms of device portability. This would avoid long term

instability of the nanoparticle layer as resistance baseline and

sensor response might drift [124]. Yet the major drawbacks of

room temperature sensors are usually the increased response and

recovery times [69] that are critical for on-line breath analysis.

However, reduction in the sensor response time can be achieved

by reduction in film thickness [69], appropriate morphologies

(e.g. nanowires) [85] or addition of catalytic particles (e.g. noble

metals) [77].

Summary and outlookBreath analysis has attracted increasing interest among the scien-

tific and clinical communities. It is a powerful technique for non-

invasive and rapid (on-line) provision of information about the

disease state/progression and monitoring of therapy. Currently

only a small fraction of the exhaled gases from the human breath

can be related to specific physiological conditions. For those

components, however, chemo-resistive gas sensors have a great

potential for the development of devices that could dramatically

reduce the medical costs and improve the quality of life. The

technological improvement of gas sensors is therefore crucial in

terms of selectivity, sensitivity and response time at high relative

humidity. Different materials, structures, morphologies and even

different crystal phase properties of specific materials could en-

hance further the performance of chemo-resistive gas sensors and

therefore allow the selective detection of specific breath markers

related to different diseases. Nevertheless, considerable progress

and research is still required to identify such specific breath

markers or patterns of VOCs. The majority of diseases, including

various types of cancers, is related to several breath markers and

therefore might require a multiple component analysis provided

for example by sensor arrays (e-noses) that will benefit of high

performance chemo-resistive gas sensors. Though a lot needs to be

learned, enough is known to start using systematically chemo-

resistive gas sensors in clinical studies, with large number of

patients for specific illnesses.

AcknowledgementsThis research was supported by the Swiss National Science

Foundation, grant 200021_130582/1, and the European Research

Council under the European Unions Seventh Framework Program

(FP7/2007-2013, ERC grant agreement n8 247283).

References

[1] A. Donovan, Antoine Lavoisier: Science, Administration and Revolution, Cam-

bridge University Press, 1996.

[2] L. Pauling, et al. Proc. Natl. Acad. Sci. U. S. A. 68 (10) (1971) 2374.

[3] A. Manolis, Clin. Chem. 29 (1) (1983) 5.

[4] W. Miekisch, et al. Clin. Chim. Acta 347 (12) (2004) 25.

[5] W.Q. Cao, Y.X. Duan, Clin. Chem. 52 (5) (2006) 800.

[6] A. Amann, et al. Mini-Rev. Med. Chem. 7 (2007) 115.

[7] L. Senesac, T.G. Thundat, Mater. Today 11 (3) (2008) 28.

[8] J.S.G. Dos Santos-Alves, R.F. Patier, Sens. Actuator B: Chem. 59 (23) (1999) 69.

[9] J.J. Ho, et al. Sens. Actuator B: Chem. 50 (3) (1998) 227.

[10] M. Phillips, et al. J. Chromatogr. B 729 (12) (1999) 75.[11] C.B. de Lacy, et al. J. Breath Res. 8 (1) (2014) 014001.

[12] J.D. Pleil, A.B. Lindstrom, Clin. Chem. 43 (5) (1997) 723.

[13] A. Amann, et al. Expert Rev. Mol. Diagn. 11 (2) (2011) 207.

[14] H. Haick, et al. Chem. Soc. Rev. 43 (5) (2014) 1423.

[15] J.C. Anderson, et al. J. Appl. Physiol. 100 (3) (2006) 880.

[16] J. King, et al. J. Math. Biol. 63 (2011) 959.

[17] C.H. Deng, et al. J. Chromatogr. B 810 (2) (2004) 269.

[18] J. King, et al. Physiol. Meas. 33 (3) (2012) 413.

[19] M. Statheropoulos, et al. J. Chromatogr. B 832 (2) (2006) 274.

[20] S.K. Kundu, et al. Clin. Chem. 39 (1) (1993) 87.

[21] C. Turner, et al. Physiol. Meas. 27 (4) (2006) 321.

[22] S. Davies, et al. Kidney Int. 52 (1) (1997) 223.

[23] D.J. Kearney, et al. Dig. Dis. Sci. 47 (11) (2002) 2523.

[24] B. Grabowska-Polanowska, et al. J. Chromatogr. A 1301 (2013) 179.

[25] S.A. Kharitonov, P.J. Barnes, Biomarkers 7 (1) (2002) 1.

[69] G. Korotcenkov, Sens. Actuator B: Chem. 107 (1) (2005) 209.

[70] A. Tricoli, et al. Angew. Chem. Int. Ed. 49 (2010) 7632.

[71] L. Ferrus, et al. Respir. Physiol. 39 (3) (1980) 367.

[72] N. Barsan, et al. Fresenius J. Anal. Chem. 365 (4) (1999) 287.

[73] N. Yamazoe, Sens. Actuator B: Chem. 5 (1) (1991) 7.

[74] N. Barsan, U. Weimar, J. Electroceram. 7 (3) (2001) 143.

[75] N. Yamazoe, et al. Sens. Actuator 4 (2) (1983) 283.

[76] G. Korotcenkov, et al. Thin Solid Films 436 (1) (2003) 119.

[77] L. Madler, et al. J. Nanopart. Res. 8 (6) (2006) 783.

[78] A. Tricoli, S.E. Pratsinis, Nat. Nanotechnol. 5 (2010) 54.

[79] T. Becker, et al. Sens. Actuator B: Chem. 77 (12) (2001) 55.

[80] A. Gurlo, R. Riedel, Angew. Chem. Int. Ed. 46 (21) (2007) 3826.

[81] H. Ogawa, et al. J. Appl. Phys. 53 (6) (1982) 4448.

[82] C. Xu, et al. Sens. Actuator B: Chem. 3 (2) (1991) 147.

[83] A. Tricoli, et al. Adv. Funct. Mater. 18 (13) (2008) 1969.


RESEARCH:Rev

iew[26] K. Zayasu, et al. Am. J. Respir. Crit. Care 156 (4) (1997) 1140.

[27] S.E. Deveci, et al. Respir. Med. 98 (6) (2004) 551.

[28] I. Horvath, et al. Thorax 53 (10) (1998) 867.

[29] L.E. Gustafsson, Eur. Respir. J. 11 (1998) 49S.

[30] A.F. Massaro, et al. Am. J. Respir. Crit. Care 152 (2) (1995) 800.

[31] C.P. McSharry, et al. J. Allergy Clin. Immunol. 116 (1) (2005) 88.

[32] S.A. Kharitonov, et al. Lancet 343 (8890) (1994) 133.

[33] S.A. Kharitonov, et al. Eur. Respir. J. 21 (3) (2003) 433.

[34] S. Van den Velde, et al. J. Chromatogr. B 875 (2) (2008) 344.

[35] A. Tangerman, et al. Clin. Chim. Acta 130 (1) (1983) 103.

[36] W. Filipiak, et al. Microbiology (UK) 158 (2012) 3044.

[37] M. Kazui, et al. Free Radic. Biol. Med. 13 (5) (1992) 509.

[38] M.A.E. Larstad, et al. Acta Physiol. 189 (1) (2007) 87.

[39] P. Paredi, et al. Am. J. Respir. Crit. Care 162 (2) (2000) 369.

[40] T.H. Risby, et al. J. Appl. Physiol. 86 (2) (1999) 617.

[41] B. Buszewski, et al. Biomed. Chromatogr. 21 (6) (2007) 553.

[42] B.K. Puri, et al. Prog. Neuro-Psychopharmacol. Biol. Psychiatry 32 (3) (2008) 858.

[43] K. Stamyr, et al. Biomarkers 14 (5) (2009) 285.

[44] B. Enderby, et al. Pediatr. Pulmonol. 44 (2) (2009) 142.

[45] W. Filipiak, et al. Cancer Epidemiol. Biomarkers Prev. 19 (1) (2010) 182.

[46] A. Bajtarevic, et al. BMC Cancer 9 (1) (2009) 348.

[47] T.H. Risby, S.F. Solga, Appl. Phys. B: Lasers Opt. 85 (2) (2006) 421.

[48] W. Miekisch, et al. J. Breath Res. 2 (2) (2008) 026007.

[49] R. Teranishi, et al. Anal. Chem. 44 (1) (1972) 18.

[50] H. Lord, et al. Anal. Chem. 74 (21) (2002) 5650.

[51] A. Hansel, et al. Int. J. Mass Spectrom. Ion Processes 149 (1995) 609.

[52] D. Smith, P. Spanel, Rapid Commun. Mass Spectrom. 10 (10) (1996) 1183.

[53] U. Tisch, H. Haick, J. Breath Res. 8 (2) (2014) 027103.

[54] Y.Y. Broza, H. Haick, Nanomedicine 8 (5) (2013) 785.

[55] C. Di Natale, et al. Anal. Chim. Acta 824 (2014) 1.

[56] G. Peng, et al. Nat. Nanotechnol. 4 (10) (2009) 669.

[57] H. Haick, et al. ACS Nano 3 (5) (2009) 1258.

[58] J.-B. Yu, et al. Sens. Actuator B: Chem. 108 (12) (2005) 305.

[59] F. Rock, et al. Chem. Rev. 108 (2) (2008) 705.

[60] A.P.F. Turner, N. Magan, Nat. Rev. Microbiol. 2 (2) (2004) 161.

[61] O. Marom, et al. Nanomedicine 7 (5) (2012) 639.

[62] U. Tisch, et al. Nanomedicine 8 (1) (2013) 43.

[63] J.P. Gisbert, J.M. Pajares, Aliment. Pharmacol. Ther. 20 (10) (2004) 1001.

[64] S.Y. Liu, et al. Chest 102 (5) (1992) 1512.

[65] R.A. Dweik, et al. Am. J. Respir. Crit. Care 184 (5) (2011) 602.

[66] American Diabetes Association, Diabetes Care 36 (2013) S11.

[67] C. Hagleitner, et al. Nature 414 (6861) (2001) 293.

[68] G. Eranna, et al. Crit. Rev. Solid State Mater. Sci. 29 (34) (2004) 111.[84] G. Korotcenkov, Mater. Sci. Eng. Rep. 61 (1) (2008) 1.

[85] E. Comini, G. Sberveglieri, Mater. Today 13 (78) (2010) 28.

[86] X. Jiaqiang, et al. Sens. Actuator B: Chem. 113 (1) (2006) 526.

[87] N. Du, et al. Adv. Mater. 19 (12) (2007) 1641.

[88] J.-F. Liu, et al. Adv. Mater. 17 (6) (2005) 764.

[89] K.Y. Dong, et al. Sens. Actuator B: Chem. 157 (1) (2011) 154.

[90] C.S. Rout, et al. Sens. Actuator B: Chem. 128 (2) (2008) 488.

[91] Q. Xiang, et al. J. Phys. Chem. C 114 (5) (2010) 2049.

[92] H.Q. Fan, X.H. Jia, Solid State Ionics 192 (1) (2011) 688.

[93] S.-J. Choi, et al. Anal. Chem. 85 (3) (2013) 1792.

[94] S. Matsushima, et al. Jpn. J. Appl. Phys. 27 (Pt 1) (1988) 1798.

[95] J.H. Lee, Sens. Actuator B: Chem. 140 (1) (2009) 319.

[96] J. Zhang, et al. Sens. Actuator B: Chem. 139 (2) (2009) 411.

[97] J. Zhang, et al. J. Mater. Chem. 20 (31) (2010) 6453.

[98] T. Sahm, et al. J. Mater. Res. 22 (4) (2007) 850.

[99] A. Tricoli, et al. Nanotechnology 20 (31) (2009) 315502.

[100] C. Xu, et al. J. Mater. Sci. 27 (4) (1992) 963.

[101] P.J. Shaver, Appl. Phys. Lett. 11 (8) (1967) 255.

[102] P.M. Woodward, et al. J. Solid State Chem. 131 (1) (1997) 9.

[103] K. Kanda, T. Maekawa, Sens. Actuator B: Chem. 108 (12) (2005) 97.

[104] L. Wang, et al. Chem. Mater. 20 (15) (2008) 4794.

[105] P.I. Gouma, K. Kalyanasundaram, Appl. Phys. Lett. 93 (24) (2008) 244102.

[106] M. Righettoni, et al. Chem. Mater. 22 (10) (2010) 3152.

[107] M. Righettoni, et al. Anal. Chem. 82 (9) (2010) 3581.

[108] A. Teleki, et al. Sens. Actuator B: Chem. 119 (2) (2006) 683.

[109] A.K. Prasad, et al. Thin Solid Films 436 (1) (2003) 46.

[110] P. Gouma, et al. J. Breath Res. 5 (3) (2011) 037110.

[111] D. Mutschall, et al. Sens. Actuator B: Chem. 36 (13) (1996) 320.

[112] B. Fruhberger, et al. Sens. Actuator B: Chem. 76 (13) (2001) 226.

[113] M. Righettoni, et al. Anal. Chim. Acta 738 (2012) 69.

[114] M. Righettoni, et al. J. Breath Res. 7 (3) (2013) 037110.

[115] J. Herbig, et al. J. Breath Res. 3 (2) (2009) 027004.

[116] T. Toyooka, et al. J. Breath Res. 7 (3) (2013) 036005.

[117] H. Bai, G.Q. Shi, Sensors 7 (3) (2007) 267.

[118] T. Hibbard, et al. Anal. Chem. 85 (24) (2013) 12158.

[119] S.M. Hicks, A.J. Killard, Sens. Actuator B: Chem. 194 (2014) 283.

[120] K. Toda, et al. Anal. Chem. 78 (20) (2006) 7284.

[121] T. Sahm, et al. Sens. Actuator B: Chem. 127 (1) (2007) 63.

[122] G. Peng, et al. J. Brit, Cancer 103 (4) (2010) 542.

[123] N. Queralto, et al. J. Breath Res. 8 (2) (2014) 027112.

[124] N. Yamazoe, Sens. Actuator B: Chem. 108 (12) (2005) 2.171

Breath analysis by nanostructured metal oxides as chemo-resistive gas sensorsIntroductionBreath analysisBreath markersMethods for breath analysis

Metal oxide chemo-resistive gas sensorsSensing mechanismSensor structure and morphologySensor compositionApplications to human breathPotential and challenges

Summary and outlookAcknowledgementsReferences

Documents

Nano Structural