DMMP (dimethyl methylphosphonate) was purchased from Qindao Hanhua Fireproofing Material Ltd., China. Ferric trichloride (FeCl₃·6H₂O), ammonia solution, anhydrous methanol, anhydrous alcohol trichloromethane and n-hexane were all AR grade. The HC-49/U AT-cut 6.0 MHz quartz crystals were from Hosonic International (Hangzhou) Ltd., China.

Recently, Buathong and co-workers reported a template-free synthesis of nano-Fe₂O₃ particles with diameters of about 39 nm, but the synthesis has to be carried out at 250–300 °C under argon protection [21]. Rangaraju and co-workers synthesized iron oxide nanotubes on a pure iron substrate by an electrochemical anodization method. The grain size is in the range of 80–110 nm. When sodium tetraborate and sodium tetrafluoroborate were used, annealing at 500–550 °C was indispensable [22]. The synthesis including a 200–450 °C heating treatment process was reported in the references [23,24] as well. Instead of room temperature, the best detection temperature was 230 °C, as described in the reference [25].

For the experiments reported in this paper, the highest temperature for synthesis was as low as 100 °C, and all detection measurement were carried out at room temperature, so both the synthesis and measurement temperatures were much lower than those reported in the references [23,24]. Furthermore, several environmentally friendly concepts were introduced throughout our experiments such as the relatively low temperature reaction, the recycle of the clear upper liquid and the detection at room temperature. The synthesis quality could be improved by further study and the synthesis conditions could also be optimized to obtain uniform grains.

The XRD patterns of the hollow and solid ball-like Fe₂O₃ samples are shown in Figures 1(a) and 1(b), respectively.

The diffraction peaks in 33.1, 36, 40.8, 49.4, 54 and 57.4 degrees as labeled in Figure 1(a,b) are well matched with the characteristic peaks of α-Fe₂O₃ [24], so the spectra indicate the samples are mostly made of Fe₂O₃. Due to the drying at room temperature in air, the samples suffered from poor crystallinity and humidity. As a result, background peaks can be seen in Figure 1. But this will not affect the sensing tests, because the specific surface area and morphology of the material are the most important factors. Better XRD patterns could be obtained by annealing [24,25], but the QCM would suffer from such high temperatures. In order to save energy and simplify the processes, the same experimental conditions were used in both XRD and QCM gas sensing experiments.

In Figure 2(a,b), Fe₂O₃ samples exhibit a hollow and solid ball-like morphology, respectively. The hollow balls have a diameter range from 20 nm to 120 nm, and the solid balls are of 200–800 nm. Figure 2(a,b) also illustrates that most of hollow ball samples are of nanometer scale, but their solid ball counterparts are not. The solid balls are just used for the comparison. The larger the diameter, the smaller is the specific surface area, so the wide range of the solid ball diameters will not affect the comparison results. Because the hollow and solid balls have the same total mass on sensors' films, it is obvious that the hollow balls have larger specific surface area when they have the same diameter. Moreover, for ball-like Fe₂O₃, the larger diameters mean less specific surface area for the same mass, so the specific surface area of the hollow ball-like nano-Fe₂O₃ film is larger than the solid ball-like one when the hollow ball's diameter is less than that of the solid ones.

For the sake of simplicity, the sensors with the sensing films of hollow ball-like nano-Fe₂O₃ and solid ball-like Fe₂O₃ are called SENSOR No.1 and SENSOR No.2, respectively. SENSOR No.1 has a rapid (25 s) and high (115 Hz) response to 4 ppm DMMP, as shown in Figure 3.

According to the report of Tasaltin and coworkers, their QCM sensor's response time to 12 ppm DMMP was 18 minutes and response amplitude to 6 ppm DMMP was lower than 100 Hz [7]. Also the response time of our SENSOR No.2 is about 20 s, but the amplitude is a little bit lower than 20 Hz. As a result, SENSOR No.1 has a sensitivity of 29 Hz/ppm, which is at least 500% more sensitive than SENSOR No.2. Ying and coworkers reported that their DMMP sensor's sensitivity was 3.19 Hz/ppm [9]. Since the responses of the two sensors, SENSOR No.1 and SENSOR No.2, are repeatable, N₂ is a successful desorption gas. Because adsorption of N₂ is physical, namely, unspecific surface absorption, the successful desorption of DMMP suggests that the DMMP adsorption in the sensing film is not chemical, or it would not be desorbed well by N₂, so the intermolecular force between the gas and the sensing material must be the main interaction force, which can explain why the SENSOR No.1 can have higher response amplitude. Because of the larger specific surface area, SENSOR No.1 has more surface energy than SENSOR No.2.

Figure 4 shows the responses of SENSOR No.1 to DMMP with concentrations from 1 to 10 ppm, which illustrates that the responses are gradually enhanced with the increasing concentrations of DMMP. It is important to note that there is a linear response from 1 to 6 ppm.

In Figure 5, the smallest response is observed when the relative film thickness (μg/mm²) of the hollow ball-like nano-Fe₂O₃ is 15 μg/mm². This is different from the result in reference [19], which reported that the trend of response amplitude to relative film thickness is firstly increasing to a maximum and then dropping. Our test results in this paper are due to the hollow property of the nano-Fe₂O₃. The 15 μg/mm² film can only cause monolayer saturated gas adsorption. Meanwhile, the 5 μg/mm² film leads to a direct multilayer adsorption. The 20 μg/mm² film is initially monolayer saturated adsorption and then multilayer one. Because the 40 μg/mm² film has the largest mass comparing to the others, with the similar adsorption feature to 20 μg/mm² film, it has the highest response amplitude to DMMP. Nevertheless, the 40 μg/mm² sensor can cause a long recovery time of the sensor, which has been observed in our experiments. Considering both the response amplitude and the recovery time, 20 μg/mm² film (a balanced condition) was adopted as the optimized thickness based on our experimental results.

Figure 6 depicts the response amplitude of the SENSOR No.1 to eight different kinds of gases with 10 ppm concentration.

In the graph bar 1 is methanol, 2 is ethanol, 3 is formaldehyde, 4 is acetaldehyde, 5 is ethyl acetate, 6 is n-hexane, 7 is chloroform and 8 is DMMP. The response amplitude of DMMP is much higher than those of other gases, which indicates that the sensor has an excellent selectivity towards DMMP. Since the P=O function group may easily interact with the -OH groups of the nano-Fe₂O₃ surface by hydrogen bond [12,15], the response of the sensor to DMMP is better than the others.