Behaviour of single-phase materials

Variation o f baseline in air with composition o f the material

The variation o f both conductance (Fig 7.5.) and conductance activation energy (Fig 7.6.) with composition taken together with the variation o f surface composition (Fig 7.4.) show that there is some subtle interplay o f different effects which it is necessary to interpret before interpreting the gas behaviour. The behaviour o f pure Sn02 has been interpreted as that o f an extrinsic semiconductor in which the trap states controlling the conductivity are oxygen species at the surface o f the crystallites. As outlined in the Introduction (Chapter 1.) and developed further in Chapter 6. The model, at the two extremes, is either a surface trap limited conductance or a Schottky barrier controlled conductance. The activation energy for conduction is then either the energy gap between surface state and conduction band edge (Æ s in Fig 6.5.) or the surface barrier height whilst the charge carrier concentration is controlled by the concentration o f bulk donor states. Given that the band gap o f Sn02 and TiOz are respectively 3.6 [21] and 3.0 eV [22] and that the bulk donor states Sn sn and T i’ji are respectively « 0.15 and 0.8 eV below the conduction band edge in the pure oxides Sn02 and Ti02 (110) ([23]; and in Fig 6.5.), this interpretation is reasonable.

So, for the undoped materials, the activation energy for conduction decreases linearly with increase o f T i02 mole fraction (Fig 7.6.), from 1 eV for pure Sn02 to 0.4 eV for pure Ti0 2. This is consistent vsath a surface trap limited conductivity with the position o f the surface trap level with respect to the valence band edge (oxygen p- states) remaining approximately constant with change o f composition. Hence, as the band gap decreases from Sn02 to Ti0 2, AEs decreases also.

The effect o f Sb addition was markedly to decrease the activation energy for (Sn- Sb)0 2. The interpretation is straightforward: as noted in Chapter 4., increase o f Sb concentration increases the donor concentration which decreases both the Debye length (so that a surface trap-limited model becomes inappropriate) and the surface barrier height. In this case, the conduction activation energy is taken to be the surface barrier height.

For the undoped materials, the decrease in conductivity with addition o f TiOi to SnOi implies a decrease in bulk donor density; to a minimum when Xjioi « 0.8. Trace impurities in the oxides (e.g. trivalent elements such as Fe and Al), at different concentration in the two oxides, might cause this effect. More spectacular is the effect o f addition o f Ti02 to the Sb-doped SnÛ2, where the addition o f small amount o f Ti caused a decrease in conductivity by 2-3 orders o f magnitude. Since the Sb/Sn is maintained constant, addition o f Ti would dilute the Sb and hence decrease the charge carrier concentration. The effect is not, however, linear in Ti02 mole fraction. It seems that Ti substitution enhances the surface segregation o f Sb (Fig 7.4.) and thus decreases the bulk donor density more that expected by the simple effect o f dilution. The surface might even be saturated in Sb. The sharp decrease in activation energy also caused by Ti substitution into the doped material might be interpreted as an effect o f the surface segregation upon the surface barrier height, perhaps because there would be consequentially a non-uniform spatial distribution o f bulk donor states. For the Sb-doped materials, therefore, for compositions where (Sn-Sb) O2 is the major constituent, the behaviour should be dominated by the distribution o f Sb between bulk and surface. This effect would explain the plateau in conductivity for XTi02 = 0.1-0.5. When TiÛ2 becomes the major constituent, the behaviour o f the undoped and doped materials becomes similar.

Effects o f composition variation upon gas sensitivity

Surface enrichment demonstrated that the gas response o f these compounds was mediated at a surface with a quite different composition to the bulk. Thus, the effect o f dopant density and surface composition are intimately linked. Hence, the cation

composition o f the lattice had an effect on the dopant density by changing the electron occupancy o f non-bonding states localised on the metal, and altering the dopant density by changing the Sb concentration changed the surface composition as a result o f surface segregation.

The effect o f surface acceptor-state density on the conductivity o f n-type porous metal oxides in the limit o f low bulk-donor density (surface-trap limited model) has been formulated previously [19] and developed in Chapter 6. The conductivity can be expressed in terms o f surface acceptor-state density,

(T / g = + #2)) / (7.1)

Here, equilibrium between surface acceptors (oxygen species) with the conduction and the valence bands can be expressed as:

S~ o S + e';

exp(-AE_y / kT) and

«S' <=> iS" + /?*

= Ny exp[-(A£'g - ) / kT]

where and Nj^ denote the bulk concentrations o f acceptor and donor states, respectively. These are identified as empty and filled metal d-orbitals, so that

{Nj^+Nj^) is constant. Here, also, Ny and Nc are the densities o f states at the valence and conduction band edge, represents the band gap, AE^ is the surface state energy with respect to the conduction band edge and denotes the total surface state density. If the majority o f charge carriers are electrons then the first term o f equation (7.1.) dominates, and if holes then the second.

As in Chapter 6., the dependence o f the conductivity on the surface trap concentration, i.e. the sensitivity (Sgas) is obtained by differentiating equation (7.1.) vyith respect to N^.

6/(0- / g) / A ^ 3 ( + # ^ ) / / (AT^ + ) (7.2)

For n-type semiconducting oxides such as Ti(SnSb)0 2, the sensitivity according to this compensated surface-trap limited model will be determined by the first term o f equation (7.2.) and will show a strong dependence on N^. The magnitude o f Ks is

controlled by the surface acceptor-state energy, E^. Hence, since the value o ï K3

varies exponentially with small changes in the surface acceptor-state energy could have a large effect on the sensitivity.

Equation 7.2. explains the broad features o f the variation in response to methane and carbon monoxide with variation o f composition, in the range Xxi02 = 0.1-1. Taking the undoped m aterials first. Fig 7.5. was interpreted as showing a decrease in No with composition, in the range Xxi02 = 0-0.8, followed by an increase in with further increase in xxio2- Fig 7.6 was interpreted as showing a linear decrease o f with increase o f Xxio2 over the whole range. A linear decrease o f EE^ with increase o f Xxi02 would cause an exponential decrease in K3. Fig 7.7. shows a steady decrease in

sensitivity to CO and CH4 with increase o f Xxi02 in the range 0.1 to 0.8; with minimum at Xxi02 = 0.8 and then an increase with further increase o f Xxi02- In this, the variation o f sensitivity exactly parallels the variation o f conductivity and the variation o f Nq. According to equation 7.2 the range xxio2 ^ 0.1-0.8, the decrease in both Nd and K3 would act to decrease the sensitivity. Over the range Xxio2 = 0.8-1,

increases markedly but the change in is not great. Qualitatively, therefore the increase in sensitivity is explained.

What is not explained by equation 7.2. is the marked increase in sensitivity from Xxi02

= 0 to Xxio2 = 0.1. Furthermore, Sb doping o f SnOz increases Nd but markedly decreases sensitivity. Both facts draw attention to the assumption o f a surface trap- limited conductivity in fully depleted crystallites, which underlies equation 7.2. The behaviour can be interpreted both by Schottky barrier-limited conductivity model and by the effects o f charge in doing upon barrier height and Debye length as elucidated in Chapter 4. Increasing donor density decreases surface barrier height and Debye length. Both effect give a decrease in sensitivity. Decreasing donor density would then increase sensitivity. Fig 7.4. and 7.5. have been interpreted by a marked decrease in donor density caused by Ti addition to the doped materials. Therefore, the increase in sensitivity in the range Xxio2 - 0-0.1 is interpreted as the effect o f decreasing donor density causing an increase in Debye length and leading to a transition from a surface barrier to a surface limited regime.

C om parison o f se n sitiv ity to w a ter va p o u r w ith th at to oth er g a se s

The notable difference between the effects o f water vapour and those o f CO and CH4

is that the increase in sensitivity for Xxi02 = 0.8-1 was not observed. Equation 7.2. gives a framework for interpretation. The increase insensitivity to CH4 and CO was interpreted above as an effect o f variation o f N^, dominating an effect o f variation o f

K3. However, if the binding site for water were different to the reaction site for other

gases, that is if K3 were different, then the result might be different. It was shown in

both Chapter 3. and 6. that an assumption o f a different value o f K3 for states

associated with water, K3^h2o, would be plausible. It was further shown that an

assumption o f K3^h2o which was sensitive to the surface composition would also be

plausible. Therefore, the monotonie decrease o f Sh2o with xjioi is interpreted, within the framework o f equation 7.2., as the result o f a monotonie decrease in Æj, ^ 2 0 with

dominates over the effect o f increasing for xxio2 > 0.8.