It was important to have a reliable basis for the synthesis of samples, so an investigation was undertaken to find the optimum calcination and sintering
methodologies. To this end, samples were synthesised with different calcining and sintering times in order to determine which produced the least impurity phase, as has been reported in the literature [1].
Initial calcine times were obtained from references [2, 3], and these were then modified to give the results with least impurity phase. The ramp rates were all a uniform 3◦C per minute, both for heating and cooling. This was also the ramp rate used for sintering the samples. Once calcined, the samples were crushed into a fine powder using a pestle and mortar. It was found that the calcined powder was not phase pure, which can be seen in Figure 3.1, showing an example calcined
Figure 3.1: 45 minute x-ray diffraction measurement of 20% KBT calcined powder. The blue triangles correspond toBi12T iO20and the red diamond corresponds toBi5T i3F eO15.
Figure 3.2: A comparison of sintering times and temperatures in 20 mol% KBT BFO-KBT presented as a stacked plot of logarithmic values. These were taken over 45 minutes. From top to bottom, these were sintered at 1000◦C for 4 hours, 1000◦C for 3 hours, 1000◦C for 2 hours, 950◦C for 2 hours and 900◦C for 2 hours.
powder of 20% KBT. Extra phases can be seen highlighted around 27.5◦ 2θ,and 33◦ 2θ in blue corresponding to Bi12T iO20 and around 31◦ 2θ corresponding to Bi5T i3F eO15in red. Impurity phases like these were found in all calcined powders; it proved impossible to find a phase pure calcined powder. Well defined impurity peaks can be resolved in Rietveld refinement, so do not have a significant impact on the quality of a refinement, but it is better to start with as close to a phase pure sample as possible.
A batch of the calcined powders were then sintered. Since it was possible to generate almost phase pure powders by sintering, more emphasis was put on ensuring good sintering values. In Figure 3.2, five examples of sintered powders can be seen with varying sintering temperature and time. It was found from c, d and e that the impurity phase was decreased with increasing temperature, but it was already known that sintering at 1050◦C would result in the sample melting, meaning that 1000◦C was the optimum temperature reached for 20 mol% KBT. For higher KBT content samples, a higher temperature was possible and thus was used.
Different lengths of time for sintering were investigated. Between 2 and 3 hours, impurity phase around 27◦ 2θ was decreased slightly. However it then increased slightly by sintering for 4 hours, so the final decision was to use the 1000◦C 3 hour sintered sample. The impurity phase around 27◦ 2θ the samples was found to be Bi2O3, despite bismuth being reported in the literature as the element most commonly lost upon heating from other sources due to volatility [4, 5, 6, 7]. In addition, in KBT it was known that potassium volatility causes problems in synthesis [8]. As such, a selection of samples were weighed before and after calcining, and before and after sintering. It was found that the majority of mass loss was moisture, with dry mass loss around 0.3% found for calcining and mass loss around 0.3% found for sintering. The samples were synthesised from a stoichiometric mix of reagents. This would be unlikely to result in a stoichiometric final sample because of the volatility issues discussed, but the SEM EDX measurements from Section 3.2.2 did not observe any large systematic variations from a stoichiometric sample.
Once sintered, the powders were then re-crushed into a fine powder with a pestle and mortar. These initial experiments made clear that there was a danger of the samples melting, which would result in an unusable sample. Because of this, some samples required a lower calcine and sintering temperature, which was offset by calcining or sintering for longer. Table 3.1 lists the calcine temperatures and times, and the sintering temperatures and times.
The heating processes were conducted in alumina crucibles. These crucibles were necessary to ensure a sealed environment for the reaction, preventing the loss
Table 3.1: BFO-KBT sample table with calcine time/temperature and sintering time/temperature for each sample. Different sintering times and temperatures were experimented with to produce the best samples possible. It was found that 1050◦C was an acceptable temperature away from the end members, allowing samples with little impurity phase to be produced and not resulting in melted samples, but closer to the end members was likely to result in melting. The samples indicated with * were synthesised by project students, K. Zahra and S. M. Johari, using the same equipment, powders and methodology.
Composition (%KBT) Calcine Temperature (◦C) Calcine Time (Hours) Sinter Temperature (◦C) Sinter Time (Hours) 0 790 2 800 5 10* 850 4 900 2 15 900 2 950 5 20 900 2 1000 3 25 900 2 1050 2 30 900 2 1050 2 33 900 2 1050 2 40 900 2 1050 2 50 900 2 1050 2 60* 900 4 1050 2 70* 900 4 1050 2 80* 850 4 1050 2 90* 850 4 1075 2 100 850 12 1030 20
of any volatile compounds. In order to prevent the reaction with the crucibles which was found at high temperature, all samples were placed on a platinum plate inside the alumina crucibles. This prevented any contact with the alumina crucibles, ensuring the purity of the sintered powders. Over time, it was found that the alumina crucibles would become stained with BFO-KBT. Multiple crucibles were used to prevent contamination of the samples. In order to ensure they were completely clean between samples, the platinum plates were cleaned with hydrochloric acid.
3.2.2 Scanning Electron Microscopy
A representative selection of the final sintered powders were examined with scanning electron microscopy (SEM) to check the stoichiometry through Energy-Dispersive X-Ray Spectrometry (EDX) using an EDAX system. The selection of samples studied provided a baseline for the expectations for the final composition compared to the nominal composition of the synthesised samples.
Gold coating of the samples was required to attain meaningful results from the SEM equipment because the powder was found to charge quickly which made measurements unreliable. Since BFO is known to be a good conductor, this could be an effect of the possible non-stoichiometry of the samples. The gold coating allowed the charging to be circumvented as long as the electron beam was not focussed on any single point for longer than 30s; sufficient time for long scans, as long as a wide enough area was scanned. The samples were mounted on a carbon pad stuck on an aluminium base, which meant that powders investigated would generally also show gold, carbon and aluminium peaks.
To determine the composition of the powders, several methods were used in tandem. The first and simplest was to compare the ratios of the A site atoms and B site atoms to see how well that ratio matches the predicted amount. In other words, the ratio of Bi:K and Fe:Ti, which would be expected to vary by composition.
For a 30% powder, as an example, the ratio of Fe:Ti should be 70:30, and for Bi:K 85:15. These ratios provided a quick check that the powders were
approximately correct, and could easily be calculated while data were being collected.
A more sophisticated method was also used, looking at each atomic
quantity in isolation and comparing this with a known quantity of that atom which should exist in the whole. Examples of this can be seen in Table 3.2. Here, the compositional percentage from the EDX measurements are compared with a look up table calculated for the expected compositional percentages for each element for each mol% KBT composition. The observed mol% KBT compositions from these
Table 3.2: Two examples for the 30% KBT powders. The left hand table shows the results from a powder which was formed as a part of a pellet, while the right hand table shows results from a loose powder. In both cases, they can be seen to be close to the nominal compositions of 30% (errors calculated as 1 standard deviation of a reliable selection of data)
Pellet Powder
Element Composition % KBT % Element Composition % KBT %
Bi 42.7(4) 30 Bi 41.4(4) 34
K 8.4(4) 34 K 7.3(4) 30
Ti 15.6(8) 31 Ti 14.7(8) 29
Fe 33.3(9) 33 Fe 36.6(9) 27
tables are then listed. For example, at 100% KBT, there would be 25% Bi, 25% K, 50% Ti. At 0% KBT, there would be 50% Bi, 0% K, 0% Ti, 50% Fe. So if the compositional percentage obtained from the EDX measurements for Bi was 50%, this would correspond to a mol% KBT of 0%.
Comparing each atom’s percentage in the EDX measurements with the known values for that particular atom in a given mol% KBT, it is possible to discern the mol% KBT that the measured atomic percentage corresponds to, meaning that a mol% KBT was calculated individually for each atom. Once these were completed for each of the four atoms observed, they were compared with one another and the nominal composition expected.
So, drawing both of these methods together, the example on the left of Table 3.2 would be in the range 30-34% KBT, with a nominal composition of 30%. The ratios given from the other method for this powder give approximately 33% by Bi:K and 32% by Fe:Ti, so this powder is around 32%±2%.
For the example on the right of the table, the range is from 27% to 34%. Bi:K gives a composition around 30%, Fe:Ti gives a composition around 29%, so this was considered to be 30%±4%.
While the errors on the calculations were generally in the range of 2% - 5%, the average values for the powders investigated in this way (15%, 20%, 30%) were within 2% of the nominal compositional values. From this, the methodology for generating the powders was considered to produce good enough results that the nominal compositions could be used in calculations and occupancies when analysing the data.
In addition, to investigate whether the sintered powder was made up of a single mixed material or grains of BFO and grains of KBT, elemental mapping was employed to search for differences in composition; if there were individual grains, then some regions of the map would have been found to be potassium and titanium
(a)SEM image of the 15% KBT powder, looking at individual grains.
(b)Bismuth elemental map (c)Iron elemental map
(d)Titanium elemental map (e)Potassium elemental map
Figure 3.3: Elemental maps of the constituent parts of the 15% BFO-KBT and the raw SEM image of the scanned area for comparison. Note that all elements appear to coexist
rich and iron deficient, while others would have been found to be iron rich and deficient in titanium and potassium. Figure 3.3 shows an example of one of these SEM measurements, conducted on 15% KBT. Figure 3.3a shows that the sample range chosen was made up of multiple individual grains, the majority of which would be expected to be BFO with the remainder KBT in a mixed powder rather than a mixed material. Figure 3.3b shows the bismuth elemental map. Since bismuth is present in both BFO and KBT, this provides a useful baseline for the other graphs, showing the areas that would be expected to be picked up by the system. Figure 3.3c shows the iron elemental map, which can be seen to cover the same areas as the bismuth map, though since bismuth has the strongest response by being the heaviest element present, the other graphs are slightly less well defined. However, it is still clear that they cover the same area. Figure 3.3d shows the titanium elemental map, which is completely analogous to the iron map. As confirmation, the potassium map is included as Figure 3.3e, which is also completely analogous to the other maps. From these it can be concluded that the BFO-KBT powder is not a mix of BFO and KBT powder grains.
Focussing on any single grain was impractical, so elemental maps of single grains were not obtained. This was due to the effects of charging; a result of low conductivity in the material: an encouraging sign for the piezoelectric properties of the material, but a complicating factor for SEM measurements.