Sensors and Actuators B121(2007)
18–35
Metal oxide-based gas sensor research:How to?
N.Barsan,D.Koziej,U.Weimar∗
Institute of Physical and Theoretical Chemistry,University of T¨u bingen,Germany
Available online27October2006
Abstract
The paper critically reviews the state of the art in thefield of experimental techniques possible to be applied to the study of conductometric gas sensors based on semiconducting metal oxides.The used assessment criteria are subordinated to the proposed R&D approach,which focuses on the study,and subsequent modelling,of sensors’performance in realistic operation conditions by means of a combination of phenomenological and spectroscopic techniques.With this viewpoint,the paper presents both the to-date achievements and shortcomings of different experimental techniques,describes–by using selected examples–how the proposed approach can be used and proposes a set of objectives for the near future.©2006Elsevier B.V.All rights reserved.
Keywords:Metal oxide;Gas sensor;Transduction;Spectroscopy;Operando
1.Introduction
Conductometric gas sensors based on semiconducting metal oxides are actually one of the most investigated groups of gas sensors.They have attracted the attention of many users and sci-entists interested in gas sensing under atmospheric conditions due to the:low cost andflexibility associated to their production; simplicity of their use;large number of detectable gases/possible applicationfields[1–4].The initial momentum was provided by thefindings of metal oxide-gas reaction effects of Heiland[5], Bielanski et al.[6]and Seiyama et al.[7]and the decisive step was taken when Taguchi brought semiconductor sensors based on metal oxides to an industrial product(Taguchi-type sensors [8]).Nowadays,there are many companies offering this type of sensors,such as Figaro,FIS,MICS,UST,CityTech,Applied-Sensors,NewCosmos,etc.[9–13].Their applications span from “simple”explosive or toxic gases alarms(see information pro-vided by the gas sensors manufacturers on their homepages)to air intake control in cars[14]to components in complex chem-ical sensor systems[15].
On the side of the R&D work the most visible result is a large number of publications,generally reportin
g excellent indi-vidual gas sensing performance.The latter is obtained mainly by measuring the signals of laboratory samples(change of sample/sensor’s electrical resistance)in quite unrealistic envi-
∗Corresponding author.
E-mail address:upw@ipc.uni-tuebingen.de(U.Weimar).ronments from the viewpoint of real sensors’working condi-tions,id est.in the absence of changing background conditions (e.g.humidity,presence of interfering gases,temperature,etc.). Sometimes,especially when the understanding of the sensing is targeted,some spectroscopic input is also provided.This type of approach,which is still dominant for the time being,is at the basis of most of R&D shortcomings and explains why,in spite of so many excellent laboratory results,the choice of devices to be used in real applications is still rather limited.It also explains why the modelling of gas sensing with metal oxide-based gas sensors is still in its infancy.
To understand what is wrong with the above-mentioned approach we need to realise that the reasons for high sensitivity to a particular gas and,simultaneously,low selectivity are related to the metal oxides-based sensors working principle. The cause of the change of sensor resistance(sensor signal)can be traced down to a ionosorption process and explained in terms of a free charge carriers(electrons)transf
er from the semicon-ductor to adsorbed surface species or the other way around.The adsorption process that is responsible for the sensor signal is strongly influenced by the presence of the pre-adsorbed species (like ionosorbed oxygen,hydroxyl groups,carbonates,etc.) and by only measuring the change of resistance upon exposure to the target gas we will only record the overall electrical effect of quite complex surface reactions;summarizing,by only measuring the resistance change we do not have the needed discrimination for the correlation between surface species and their electrical effect.In principle,the discrimination we are missing should be provided by the results obtained by applying
0925-4005/$–see front matter©2006Elsevier B.V.All rights reserved. doi:10.1016/j.snb.2006.09.047
N.Barsan et al./Sensors and Actuators B121(2007)18–3519
spectroscopic techniques;there is a wealth of data provided by surface physics studies performed on metal oxides[16–18].The problem here is related to the mismatch between the optimal conditions for measuring sensor performance and acquiring spectroscopic data.Spectroscopic acquisition techniques are extremely powerful for the characterization of metal oxides since they can provide details about the active sites and reveal insight into the reaction process.Unfortunately,the standard spectroscopic inves
tigations are mostly performed in conditions far away from the ones normally encountered in real sensors applications,namely:in UHV[19];at low temperatures[20,21]; required preconditioning of the samples at high temperatures, quenching and exposure to high concentrations of reactive gases;conducted on simplified systems(transmission mea-surements on powders[22,23],crystals,thin layers).The latter handicap is quite relevant also because it was demonstrated that the performance of a sensor is very much dependent on the sen-sors fabrication technologies(type of substrate and electrodes, thickness and morphology of the sensing layer,etc.),which indicates that the spectroscopic input should also be gained on actual sensors and not on sensing material samples.It is possible to take a different approach as it was recently proposed[24]. This one is based on the fact that by applying simultaneously several complementary methods(id est.FTIR and Raman spectroscopy,CEM,KP,ac,dc,conversion)in conditions as close as possible to the real sensors’working conditions,a synergetic effect can be obtained,namely the identification of the sensor effect of the different surface reactions.
The actual contribution aims to provide a critical review of the phenomenological and spectroscopic techniques(and of the interpretation of their results)that can be used in the R&D work dedicated to metal oxides-based gas sensors.The techniques that provide structural input will not be addressed here.
2.Objectives and approach of the R&D work
Thefinal objective of the R&D activities is the design and fabrication of good gas sensors id est.suited for solving a certain application.It is important to keep in mind that the quality of a sensor is almost impossible to be defined without understand-ing the application needs,which besides the target gas/gases, possible cross-interferences and environmental conditions also relate to the cost/price restrictions of the instrument using the sensors.The latter factor is generally described as the fourth S –suitability–that is more and more considered in addition to the three classical ones(sensitivity,selectivity,stability).
In order to understand the challenges of the research in the field,we should have a look at the way in which the sensor signal is generated.A sensor element normally comprises the following parts:
•Sensitive layer deposited over a
•Substrate provided with
•Electrodes for the measurement of the electrical characteris-tics.The device is generally heated by its own •Heater;this one is separated from the sensing layer and the electrodes by an electrical insulating layer.
Such a device is normally operated in air,in the presence of humidity and residual arbon dioxide).It is gener-ally accepted that in such conditions,at working temperatures between200and400◦C,at the surface of the sensitive mate-rial–the metal oxide–various oxygen,water and carbon dioxide-related species are present.Not all of them have a direct influence on the sensor resistance id est.are involved in free charge carrier exchanges with the metal oxide(ionosorption); for a general discussion on the different types of adsorptions, see[25].Some species will form bonds by exchanging electri-cal charge with specific surface sites(surface atoms),meaning that they may form dipoles;the latter will not affect the con-centration of free charge carriers so that they will not have an impact on the resistance of the sensitive layer.Those situations are described in Fig.1for the simplified case of adsorbed oxy-gen ions,as electron traps,and hydroxyl groups bound to the metal,as dipoles,at the surface of a n-type metal oxide semi-conductor;expressed in the energy bands formalism for the metal oxide,the effect of the former is a band bending while the effect of the latter is a change of the electronic affinity
when Fig.1.Schematic representation of(left)flat band in n-type semiconductor and(right)band banding model illustrating adsorption at the surface of n-type semicon-ductor.The changes of the work function( Φ)are determined by band bending(qVs—due to ionosorption)and changes the electron affinity( χ)due to building of dipoles at the surface(Mδ+–OHδ−).
20N.Barsan et al./Sensors and Actuators B121(2007)18–35reaction paper to metaphor
compared to the situation existing before the adsorption took place.
Changes in the band bending,induced he reaction of the oxygen ions with carbon monoxide,will be translated into changes of the overall electrical resistance/conductance of the sensitive layer.It looks simple but it is not and the simple proof is the well-known influence of the ambient humidity on the sensor signal upon CO exposure;this indicates that the reaction cannot be that simple and there should be some involvement of species that do not have an obvious,measurable electrical resistance effect.A lot of questions need to be answered on both the side of the surface reactions and the side of their transduction in an electrical signal especially when one examines the influence of different fabrication technologies:
•How does the reaction between the target gas,in ase CO,and oxygen really take place?Does it involve atomic oxygen ions or molecular ones?Can it be that the reaction does not involve oxygen ions?Can we measure sensor effects in the absence of ambient oxygen?
•What is the role of the other pre-adsorbed species?Are they involved in the reaction?Is the overall equilibrium influenced?•Does the reaction take place in the same way in the whole sensing layer?Is the concentration of the gas to be measured the same all over in the sensitive layer?Are there more reac-tive regions?Does the presence of the electrodes and/or the electrode material play a role?The one of the substrate?•What will be changed if one uses noble metals additives?How much should one
use?
•What does one measure?Free charge carriers concentration changes?Charge carriers mobility effects?
•How important is the sensitive layer morphology in translating microscopic reactions into macroscopic signals?Is the non-linearity of the sensor response a consequence of the surface chemistry or of the transduction?Can that be changed?•Do the electrode–metal oxide contacts play a role in the over-all resistance?
Getting the right answers is not only of academic interest;the modelling/understanding of the sensing is crucial for the sen-sor design because it will allow the tuning of the performance towards the desired one in a rationale manner.This will help a lot keeping in mind how many parameters can be changed and the amount of work that needs to be invested.If one agrees that the understanding of the sensing is important,it is clear that by only measuring the resistance one can not get enough informa-tion.There is a clear need to add both phenomenological and spectroscopic techniques and to apply them,as much as possi-ble,simultaneously in order to gain inputs that will complement each other.When this is not possible we will have to make sure that the samples we are studying are quite similar allowing us to gather information that can be used together.
Also,it is clear that one has enough reasons to believe and convincing proofs[24],that the sensor performance does not only depend on the sensitive material;thefinal performance is due to the entire device so it should be studied as a whole.If one agrees that it is important to study sensors then one should also make sure that we are preparing them in a reproducible manner,which in fact asks for the evolution from the study of ’unique’samples to the use of samples fulfilling industrial stan-dards.There are investigation techniques that one cannot or it would be prohibitively complicate to apply on sensors(TEM is a simple example);nevertheless the input they provide can be very important so,in such cases,the studies will have to be performed on simplified samples.There are also cases in which it will be important to have,for the sake of completeness or as a control mean,also measurements he sensitive material itself [26–28].Nevertheless,most of the effort should be dedicated to measurements of sensors,prepared in a reproducible manner and combining as much as possible complementary experimen-tal investigations simultaneously applied.It is also crucial to make sure that the conditions in which the experiments are per-formed and the ranges in which the parameters will be varied are relevant to the targeted applications.In most of the cases that means to work:at ambient pressure;in the presence of humidity and,possibly,of interfering gases;with sensors heated at appli-cation required temperatures;in the concentration range of the target gases expected in practical applications.There are situa-tions in which it will be important to get away from the most encountered a
pplication conditions in order to understand some underlying basic phenomena;the most obvious examples are the interaction with oxygen and water vapour in the study of which it makes sense to start from oxygen and/or humidity free con-ditions.Even in such cases it is important to keep the feet on the ground meaning to en in nitrogen at ambient pressure and avoid UHV.
3.Experimental techniques
In this section,the most important types of experimental techniques–for the investigation of conductometric gas sen-sors based on semiconducting metal oxides–will be critically reviewed.In each case we will present their inherent possibil-ities and limitations and the most important results gained so far,according to the authors’opinion.In some cases,when it will be necessary for the interpretation of the results,also tar-geted basic information will be provided.The techniques will be grouped under the Phenomenological and Spectroscopic head-lines;to make things clear,by the former we understand all methods that will not provide microscopic knowledge.The kind of knowledge they provide gives access only to the macro-scopic effects–e.g.change of sensor resistance or change of the composition of the ambient atmosphere that follow the detec-tion of CO–of the elementary reaction steps that can be the reaction of CO with ionosorbed oxygen or surface hydroxyl groups.Questions such as“Is the resistance change dominated by free charge concentration or mobility changes?”should be possible to be answer
ed on the basis of the inputs provided by the phenomenological techniques.Discriminating between CO reaction with oxygen ions or hydroxyl groups,if at all possible,cannot be made without spectroscopic input.The other way around,ascribing a sensor effect to a certain sur-face reaction is not possible on the basis of spectroscopic input alone.
N.Barsan et al./Sensors and Actuators B121(2007)18–3521
4.Phenomenological experimental techniques
The typical measurement technique for the here-discussed sensors is the one of their conductance or resistance;in fact,in almost all cases it is the resistance/conductance of the sensors, which is linked to the composition of the surrounding atmo-sphere that is used for gaining the desired chemical information. The measurements are performed in different modes of operation spanning from constant operation temperature and permanent polarization to modulated operation temperature and periodic dc tests(with longer or shorter pulses).
In whatever conditions the tests are performed,one has to understand what is measured and how the measurement condi-tions will influence their result.The former question is related to the transduction of the surface reactions into a change of the electrical resistance of the sensor.The latter has to do with t
he need of avoiding that,for the same changes of the ambient atmo-sphere composition,the recorded resistance changes will differ depending on the measurement conditions.In the attempt of making the picture clear we can start by examining the two sit-uations presented in Fig.2.There,two types of sensitive layers, id est.a compact and a porous one,for an n-type material,are described in a simplified manner(for details,see[29]).Even if in practice the sensitive layers are polycrystalline,for the case of the compact layer the grains and the corresponding grain–grain Schottky barriers were not shown because,not being accessible to the gases,they will not change when the ambient atmosphere composition changes.In thefigure are presented the contribu-tions to the overall resistance corresponding to:the changes in the band bending at the material/grain surface and the poten-tial barriers that are appearing due to the metal/metal oxide contact(the former depend on the changes in the ambient atmo-sphere,the latter ones not[29];they are still discussed because they represent an add-on to the sensing material properties due the presence of the electrodes).The image of the resistance of the sensitive material being dominated by the gas-dependent grain–grain potential barriers(the height of which modulates the concentration of free charge carriers allowed to travel between the electrodes)is widely accepted but,possibly,over-simplified.For grains that are small enough it is possible to have an influ-ence on the mobility of the free charge carriers;such effects are reported,modelled but their interpretation is difficult[30,31]. Anyhow,the main idea one can deduce fro
m Fig.2is that the contribution of the gas sensing process to the overall resistance of the sensor depends not only on the gas sensing material prop-erties but also on various sensor characteristics(sensitive layer morphology,electrode parameters,etc.).
It is also important to note that in systems that can be described as a series of potential barriers the very value of the measured resistance can depend on the measurement parame-ters,more precisely on the measurement voltage/current;the potential drop that is used for the measurement will be dis-tributed over the different series resistances in the layer and will be concentrated over the high resistance elements(grain–grain and electrode–grain barriers).In fact,such dependencies are reported in the literature either as such(polarization effects) [32,33]or as“varistor”effects[34–37];they are also well known to all scientists that are using multimeters operated in auto-range mode and have difficulties to connect experimental points acquired in different ranges(due to the different measurement conditions that are range-specific).A good way to forecast the risk of this type of problem occurrence is to guesstimate the potential drop per individual barrier(from the grain size and electrode spacing information)and to compare it to the k B T, which is the thermal energy at the operation temperature.A value of the grain–grain potential drop above the one of the thermal energy indicates trouble.
The ideas discussed above are related to pure electronic effects.Besides them,one should consider the chemical ones that can be influenced/determined by the nature of the electrodes and different polarizations.The influence of electrodes on the gas sensing performance is well documented[38–44].There are also reports suggesting that the polarization has something to do with the chemistry at or in the vicinity of the electrodes[45]. An additional problem is related to the fact that in many cases the gas detection is a gas conversion from reactive(CO,C x H y, NO2,O3,etc.)towards inert or less reactive species(CO2,H2O, N2,etc.)[46–48].That increases the complexity of the
prob-Fig.2.Schematic representation of a porous(a)and a compact(b)sensing layer with energy bands.Schematic representation of compact and porous sensing layers with geometry and energetic bands,which shows the possible influence of electrode-sensing layers contacts.RC resistance of the electrode–SnO2contact,Rl1 resistance of the depleted region of the compact layer,R1equivalent series resistance of Rl1and RC,equivalent series resistance of and RC,R gi average intergrain resistance in the case of porous layer,E b minimum of the conduction band in the bulk,qV S band bending associated with surface phenomena on the layer,and qV C also contains the band bending induced at the electrode–SnO2contact.
22N.Barsan et al./Sensors and Actuators B121(2007)18–35
lem by adding on uncertainties:is the sensitive layer and/or the gas concentration in the layer homogeneous;does the sensitiv-ity depend on the geometry of the electrodes or of the sensitive layer;how many gas sensors can be tested simultaneously;are the test conditions in a dynamic set up comparable to the ones in a real application.
To summarize,one can say that the so-easy-to-measure sens-ing effect of a gas sensor can be rather complicated to understand and that in the optimization and characterization of the sensor performance
one should carefully consider the role of many fac-tors.This asks for investigation methods that are able to provide the appropriate answers and it is clear that the sensor resistance measurement alone is not sufficient.
In the following,examples will be given on the kind of infor-mation the different experimental techniques are able to yield. The evaluation of the sensor performance is thefirst step to be considered.It is important,as already described,to check the influence of the measurement parameters on the results.Fig.3 [49,50]presents such results obtained in humid air and under exposure to CO and NO2in humid air(the tested sensors were thickfilm SnO2-based ones).It is interesting to observe that the R(V)dependence is different for different test conditions.One can still identify a polarization range in which the influence on the test result is minimal(below500mV);in the latter case,the recorded sensor resistance will depend only on the target gas concentrations and can be used as input in modelling.There is additional information to gain from the influence of the polar-ization,as it will be described later on.
One of thefirst questions to be answered is about the role played by the contacts in the overall sensing and,as already discussed,it is possible to encounter both electrical and elec-tro/chemical aspects.An extremely useful tool for
determining Fig.3.Influence of measurements voltage V on the sensors resistance in differ-ent gas atmospheres.For details see[50].
the electrical contribution of contact resistance is the use of transmission line measurements(TLM);for more details,see [49–51].In such experiments one uses substrates with different geometric parameters,like the ones presented in Fig.4b and the results could look like the ones presented in Fig.5[24,50].There, the dependence of the sensor resistance is plotted as a function of the spacing between electrodes(between10and2000m)for different ambient conditions.Thefit of the experimental data can be made by using the formula describing the resistance R of an ideal,rectangular shaped layer of homogeneous resistivityρ, length l,thickness t and width w:
R=ρ
l
tw
(1) Fig.4.Different configurations of the electrodes.
版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。
发表评论