Inter-dependency between surface morphology and sensitive low RH detection: exploration of an intricate mechanism to extend the lower detection limit

The water vapor molecular dynamics inside a pore structure control both molecular adsorption and desorption processes and the limit of minimum detection (LOD). Pore morphology design, and a higher concentration of electrolyte-driven anions, in accordance with the kinetics of water vapor molecules, is reported here, as the ultimate answer to extremely low relative humidity (RH) detection. In this report, a series of samples were prepared by anodization in different voltage windows, related to specific electrolyte solutions. The sensing attributes comprised: (i) a LOD of ∼3 RH%, (ii) excellent response time (6 s) and recovery time (54 s), and (iii) a hysteresis loss of ∼0.36%, with sustained stability over the period of one year; all these occurring in a sample with a pore diameter ∼5 nm ±3 nm. Interestingly, the LOD extend towards a lower RH% with a decrease in pore diameter; and a suitable explanation is given for the entire range of humidity level, in terms of the molecular mean free path, loss of kinetic energy due to scattering inside the pores, and subsequent overall loss of Brownian energy of the molecules. It is inferred from the sensing response characteristics that pore morphology and lower detection limit are interrelated; therefore, a further extension in LOD from extremely low RH% to trace levels, needs careful engineering of the pore morphology and parameters related to molecular kinetics.


Introduction
Humidity measurement is playing an ever-increasing role in industrial, laboratory and process control. Over the years, a wide variety of humidity sensing devices have been developed. Most commercially available sensors are based on polymers or ceramics. [1][2][3][4][5][6][7][8] Among ceramics, alumina is the most extensively employed material for humidity sensing applications, owing to its high thermal and chemical stability, hygroscopic nature and mechanical strength. 9 Table 1 compares the performance of commercially available and recently reported humidity sensors. Several techniques are available to make porous alumina, such as chemical vapor deposition (CVD), sputtering and sol-gel processes. 10 However, all of these techniques have a common issue: i.e. the lack of control over the developed morphology. The electrochemical anodization method, on the other hand, has a niche advantage in developing a tailor-made porous anodic alumina (PAA) nanostructure where the pore diameter, pore depth and pore widening may be controlled easily. 9 In terms of structural aspects, the incorporation of anions and the design of electrodes work together synergistically to govern the sensing mechanism in a nano-structured PAA.
The measurement of high relative humidity (RH) is not an issue at all, but it certainly is for low RH. Lots of R&D work is underway globally on precursor materials, surface morphology and engineering aspects to check for improvements in the lower detection limit (LOD), sensitivity, long term dri, and hysteresis. Investigations on PAA-based humidity sensors are plentiful and have been reported elsewhere. [19][20][21][22][23][24][25] Y. Kim et al. investigated the effect of electrode design on humidity sensing, 22 L. Yao et al. 23 explained the effect of isotropic etching of pores on sensing behavior, and Nahar 24 studied the performance degradation of PAA sensors upon exposure to high humidity. S. W. Cheng et al. 18 reported sensitivity enhancement by application of a magnetic eld. Presently, none of the reported humidity sensors shows any signicant change in capacitance/ impedance below 30 RH%. [19][20][21][22][23][24][25] Even when sensing with high sensitivity, the fastest response-recovery, high resolution and stability also arouse great concern for both high-and low-level RH. Again, the quantitative sensing response is a matter of concern, irrespective of sensing level. Therefore, a complete analysis is required to bridge the gap to appreciate the interplay of pore morphology and sensor performance.
In our previous research article, 9 the PAA nanostructure was optimized for humidity sensing applications; and the impacts of various surface pre-processing parameters were discussed in detail. In continuation of that, the present investigation reports a study of the surface morphology dependent humidity sensing response of PAA. For this, a series of samples were prepared by anodization in different voltage windows, related to specic electrolyte solutions. The underlying principles of PAA fabrication, morphology design and internal molecular dynamics, control the adsorption and desorption processes, and the sensing attributes of the sensor.

Fabrication of PAA
Aluminium sheets of high purity (99.997%, Alfa Aesar) were annealed and electropolished in a 1 : 4 ratio of perchloric acid, and ethanol solution. The process of electropolishing and its optimization were explained earlier in our previous research article. 9 The electropolishing was carried out in galvanostatic mode at 500 mA for 80 seconds at 3 C to create a reproducible surface nish. The aluminium sheets were then ultrasonicated in acetone in order to remove loosely bound impurities, if any, present on the surface of the Al sheet.
2.1.1. Anodization. The pore diameter of PAA depends directly on the anodizing voltage. Therefore, samples with different surface morphologies were obtained by anodizing at different anodic polarizations. However, anodization voltage is electrolyte specic; there exists a narrow window of voltage in which the generation of a uniform, symmetrical and selfassembled porous structure with a specic range of pore diameters can occur. Table 2 summarizes the anodization voltage window and the related pore diameter range for the electrolytes used in the present investigation.
It can be observed from Table 2 that the voltage range is lowest for sulfuric acid, medium for oxalic acid and maximum for phosphoric acid solution among the reported electrolytes in the present investigation. In each electrolyte solution, two samples were anodized (as detailed in Table 3) and thus a total of six samples with six different morphologies were prepared and investigated for their humidity sensing response. Samples SH05 and SH25 were anodized at 5 V and 25 V in sulfuric acid solution; samples SO38 and SO50 were anodized at 38 V and 50 V in oxalic acid solution and nally samples SP60 and SP90 were anodized at 60 V and 90 V in phosphoric acid solution. All the samples were anodized for 3 h at temperatures below 5 C.
Aer anodization, the samples were treated with alumina etchant solution to widen the nano-pores. The etching rate relies on the chemical composition of the porous oxide layer; that again depends on the anodizing electrolyte and its concentration. Several experiments were performed to optimize  the pore-widening step for the three electrolyte solutions. Table  3 gives the optimized values of time and temperature for different electrolyte solutions. The sheets were thoroughly rinsed with DI water and ultrasonicated in acetone for ve minutes at 27 C in order to remove any loosely bound chemicals le on the surface. Aer ultrasonication, the sheets were again rinsed with DI water.

Sensor designing and fabrication
Parallel plate capacitive sensors were developed where the anodized lm works as the sensing layer. The aluminium at the base acts as the bottom electrode. A porous gold layer of 3 nm in thickness was deposited on the anodized alumina. This gold electrode serves the purpose of the second electrode of a parallel plate capacitive sensor. Fig. 1 schematically describes the formation of the capacitive sensor step by step. Electrical connections were taken out from the top and the bottom electrodes using conductive silver epoxy. The top gold electrode design is deliberately made to cover most of the area of the active PAA layer to serve two purposes: rst, it should act as one of the capacitive electrodes, and second, the top Au layer should be porous to allow water vapor to enter through it to be adsorbed on the dielectric, i.e. the PAA layer. Therefore, careful optimization of the Au lm thickness is very necessary to satisfy both the purposes mentioned. The top gold electrode, deposited via a sputtering technique is highly porous and water permeable in nature. We have checked that a gold lm thickness of $3 nm or slightly thicker proved to be the best for sensing performance. Fig. 2 displays the FESEM images of bare and gold-coated PAA. It can be interpreted from the FESEM image that if the optimized thickness of gold layer is deposited, it will not cover the pores beneath it and thus allows water vapor molecules to permeate through it.

Humidity sensing setup
The humidity sensing characteristics of the as-prepared samples were measured in a home-made humidity sensing setup (Fig. 3), specially designed to make real-time measurements in a saturated humidity environment. The sensor chamber is divided into two unequal parts: part-A and part-B; the volume of part-A is kept to a bare minimum. Humidity of a specic concentration is generated in part-B, and allowed to rise until it attains a pre-set level of humidity concentration to be measured. A commercial sensor is connected to chamber B to monitor the concentration of the saturated humidity level. Once it is achieved, the shutter is opened in a pulsed manner, to allow water vapor ux to directly hit the developed sensor kept in part-A. Dry commercial grade synthetic air (moisture content $4 ppm) is used as a carrier gas,  and its ow is divided into two paths, as shown in Fig. 3. The dry gas and moist gas (coming via the bubbler route) are combined before the gas is allowed to pass through the sensor chamber. Both the chambers, A and B, of the sample chamber are purged separately for 12 h, in order to achieve a stable baseline capacitance, followed by exposure to different humidity concentrations. Sensing chamber A is connected to a data acquisition system (Semiconductor Characterization System, SCS, Keithley 4200), where a test signal of 30 mV (r.m.s) was used to excite the sensor at frequency 5 kHz with no DC bias.
The moist gas then passes through a commercial RH sensor (Vaisala HMT 330). The measurement of sensing parameters, such as the humidity sensing response, response-recovery time, hysteresis and long-term stability, is carried out at room temperature (300 K) and at atmospheric pressure (1.03 bar). The ow rate of the carrier gas is kept xed at 2 lpm (liter per minute) while performing the experiment.

Fabrication parameters of PAA
The anodic polarization has a direct inuence on the diameter of the pores. 25 Depending on the applied voltage, different morphologies with distinctive pore size, interpore spacing, anodic alumina lm thickness and electrolyte-driven species, were observed in the PAA nanostructure. 31 These parameters are easily tunable, and their effectiveness in sensor performance may be checked. Fig. 4(a) shows the variation in current density with time, which is very similar to that of conventional anodizing conducted in potentiostatic mode. It can be observed from Fig. 4(a) that the current density saturates aer some time ($300 seconds). However, the value of the saturation current density (J sat. ) is different for each sample. Even for samples anodized in the same electrolyte, the current density is greater for the sample anodized at a higher voltage, than for the sample which is anodized at a low voltage: e.g. J sat. , which inherently enter into the anodic oxide layer during anodizing. The plots in Fig. 4(a and b) clearly signies that the concentration of acid-driven anions is greater for the samples These anions assist the adsorption of water vapor molecules on the porous oxide structure; hence a signicant change in capacitance is observed for the samples with a high concentration of acid-driven anions. This is explained and  Nanoscale Advances Paper shown later in Section 3.5 ( Fig. 9 and 10) except for samples SH05 and SH25, where the pore diameter dominates over the electrolyte-driven anion concentration to govern the characteristics of the sensor.

Surface morphology analysis
The surface morphology of as-prepared PAA samples was observed using FESEM (FEI, NOVA NANOSEM 450). The samples were generally coated with a thin layer of gold prior to This journal is © The Royal Society of Chemistry 2020 Nanoscale Adv., 2020, 2, 2564-2576 | 2569 Paper Nanoscale Advances imaging, to avoid any charging effect caused by the insulating nature of PAA. For imaging, the samples were placed on an aluminum stub with carbon tape and positioned inside the sample chamber. Fig. 6 shows the FESEM images of samples SH05, SH25, SO38, SO50, SP60 and SP90. The micrographs are displayed at two different scales (Fig. 6); emphasizing the long-range ordering at the micronscale, and the size of the pores at the nanometer-scale (the images in the insets); in order to precisely analyze the effect of pore size and the overall surface morphology on the humidity sensing response of PAA. The samples anodized in sulfuric acid solution at 5 V and 25 V, i.e. SH05 and SH25, are shown in Fig. 6(a and b) with average pore diameters of 5 AE 3 nm and 10 AE 3 nm, respectively. Fig. 6(c, d) and (e, f) show micrographs of samples anodized in oxalic acid (at 38 V and 50 V) and  The pore diameter plays a major role in deciding the sensor characteristics, especially the lower detection limit and sensitivity. 33 However, the overall surface morphology, e.g. the self-assembled long-range ordering and uniformity of the porous structure, also affects the sensing characteristics. [26][27][28][29] It has been observed that, amongst samples with the same pore diameter but differences in long-range ordering and uniformity, a sample with a better pore arrangement gives better results than a sample with an unordered porous structure, in terms of sensitivity and dening the lower detection limit. Therefore, the Nanoscale Advances aforementioned parameters must be optimized for a better humidity sensing response. In the rst study, aer testing, sensors SH05, SO50 and SP60 were kept at a humidity of 50 RH% continuously for 3 months at room temperature ( Fig. 7(a)-(c)). It can be observed that the morphologies of the samples remain almost the same and no degradation in the porous structure is observed. Thus it can be concluded that no degradation in morphology is observed aer exposing the sensor continuously to medium-range humidity (0-50 RH%).

Sensor stability and structural integrity analysis
In the second study, the developed sensors were tested periodically over the complete RH range: i.e. 3-90 RH% for 12 months at room temperature. Fig. 7(d)-(f) and (g)-(i) display the FESEM images of bare and gold-coated sections of sensors SH05, SO50 and SP60, respectively. A minor variation in pore size for samples SH05 and SO50 can be observed in the respective FESEM images for bare (Fig. 7(d) and (e)) and goldcoated ( Fig. 7(g) and (h)) sections of sensors SH05 and SO50, respectively. However, negligible deterioration was found in their sensing data and this is reected in Fig. 13, a plot of their aging effect/shelf life.
It can be concluded that upon exposure to high humidity that a minor variation in porous structure is observed for samples anodized in sulfuric and oxalic acid solution, whereas no degradation in structure is observed for samples anodized in phosphoric acid solution. Therefore, a sensor developed from anodization in phosphoric acid solution is found to be particularly useful for applications where continuous monitoring of a high humidity level (>50 RH%) is required. On the other hand, a sensor developed from anodization in sulfuric acid or oxalic acid solution is suitable for monitoring low and medium ranges of humidity (0-50 RH%).

Phase determination of PAA
The phase of PAA fabricated by anodization was determined by carrying out X-ray diffraction measurements (Rigaku, Smartlab) using Cu-Ka radiation, where l ¼ 1.54Å. Fig. 8 displays the XRD spectrum of PAA, conrming its amorphous nature. 30,31 Alumina exists in many stable and meta-stable phases; however, only two crystalline phases, i.e. a-Al 2 O 3 and g-Al 2 O 3 , are important for humidity sensor fabrication. Most of the reported humidity sensors were found to exist in the g phase of Al 2 O 3 . 32 The g-Al 2 O 3 phase is liable to change to g-Al 2 O 3 $H 2 O which leads to surface degradation upon exposure to high humidity. 23 The phase is mainly amorphous in nature with a small crystalline content. 33 The a-Al 2 O 3 is expected to be a promising candidate owing to its high thermal stability. Chen in 1991 reported an a-Al 2 O 3 based humidity sensor fabricated by an anodic spark deposition technique. 34

Humidity sensor parameter measurement
The humidity sensing performance of PAA samples with different surface morphologies was analyzed by measuring the capacitance of the sensor, as a function of humidity. It can be observed from Fig. 9 that capacitance increases exponentially with increase in humidity. The relationship between sensor capacitance (y) and relative humidity (x) is represented by the equation where the values of y o , A, R o are given in the plots (Fig. 9) for each sample. The capacitive response of the as-prepared samples is found to be in excellent agreement with eqn (1) with a value of R 2 $ 0.9987; as given separately in each plot (Fig. 9). The capacitive responses of samples SH05 and SH25 with a small pore diameter, SO38 and SO50 with a medium pore diameter and SP60 and SP90 with a large pore diameter in nanodimensions, are displayed in Fig. 9(a, b), (c, d) and (e, f), respectively. The insets in Fig. 9 show the capacitive response for the low RH range, where generally the rise in capacitance is almost negligible, as in published reports. [18][19][20][21][22][23][24] Sample SH05 shows an exceptionally high change in capacitance for a low RH range, and the lower detection limit is found to be extended to 3 RH%. We suppose, in the rst instance, that these are directly related to its exceptionally small pore diameter. An optimum pore size comparable to the mean free path and size of a water vapor molecule is desirable for adsorption in a porous dielectric material; which is given by the Kelvin equation as where R K is the Kelvin radius, g is the surface tension, M w is the molar mass and r is the density of water vapor molecules. R is the universal gas constant, T is the temperature, P H 2 O is the partial pressure of water and V p is the vapor pressure of water. At a low RH level, especially below 10 RH%, the size of the cluster of water vapor molecules is very small, and, therefore, a very small pore diameter is desirable for moisture detection at a low RH range. Sample SH05, with the smallest pore diameter of 5 nm AE 3 nm among the prepared samples, satises the condition and is able to detect at 3 RH% ( Fig. 9(a)).
It can be concluded from the plots that, as the pore diameter increases, the maximum change in capacitance decreases and LOD shis towards a high RH value. This happens because water molecules at low RH level can easily escape due to less scattering with the pore walls or with other molecules, leading to a large mean free path, less loss of kinetic energy and subsequently a high molecular Brownian energy. 35 Therefore, the higher the pore diameter, the greater the Brownian energy, and the greater is the chance that vapour molecules, at low RH, will not adsorb on the pore walls; and it is evident from Fig. 10 that the minimum detection limit pushes towards a high RH with an increase in pore diameter. Another observation is the effect of anionic concentration on the anodic structure. A sample with a low concentration of incorporated anions in the anodic structure show less change in capacitance; except for samples with exceptionally small pore diameters, e.g. SH05 and SH25.
The LOD and sensitivity are dominated by pore diameter. The smaller the diameter of the pores, the lower is the LOD, and the better the sensitivity. However, sensitivity also depends on electrolyte concentration; the larger the concentration of incorporated anions, the larger is the sensitivity. So in order to move towards trace-level detection or detection between trace and RH, an exceptionally small pore diameter along with a higher concentration of electrolyte-driven anions is required.
The comparative normalized capacitive responses of the developed sensors are shown in Fig. 10(a), whereas Fig. 10(b) exhibits the maximum change in capacitance versus LOD. Table  4 summarizes the effect of the fabrication dynamics of PAA, e.g. electrolyte anion incorporation and pore diameter, on sensor performance.
In order to determine the response time (T r ) and recovery time (T c ), the sensor is exposed to a step change in humidity: i.e. 0-50 RH%. The response time is dened as the time taken to go from 10% to 90% of the maximum change in sensor output; whereas the recovery time is dened as the time taken to recover from 90% to 10% of the maximum change in sensor output. The transient curves exhibiting the response and recovery time of the samples are shown in Fig. 11.
The prepared samples were found to be extremely fast; response and recovery times were found to be within 20 seconds respectively. The recovery times of samples SH05 and SH25 are relatively high, which is attributed to the small pore diameter of the samples, as it takes a longer time for desorption of water molecule from nano-order pores; where water vapor molecules are initially bound chemically to make a chemisorption layer on PAA. 35 Hysteresis loss is one such parameter that negates any sensor, however good it is on other fronts. It is measured for all sensors, working on adsorption and desorption phenomena; and is dened as the percentage deviation in capacitance while humidity increases or decreases in a given RH range, i.e. 0-50 RH%. It is evident from Fig. 12 that all the prepared samples return to the original capacitance value to a large extent; showing that the sensors are stable with negligible baseline Paper Nanoscale Advances dri. Fig. 12 displays the hysteresis plots of samples SH05, SO38 and SP60, with hysteresis losses of 0.36%, 0.6% and 0.13% at 45% RH, respectively. Dri is an another important sensor parameter, which is useful as far as the long-term utility of the sensor is concerned, and that arises because of an aging effect. The prepared samples were safely stored aer testing, and the dri in sensing prole was checked every 3 months for a year. Samples with the smallest pore diameter were found to be the most stable with negligible dri in the sensor output vis-à-vis the others. The long-term stability plots of samples SH05, SO38 and SP60 are shown in Fig. 13.

Conclusions
We have reported a comprehensive as well as transparent understanding of the humidity sensing mechanism in the developed sensors, taking into account the effect of the diameter and uniformity of self-assembled nano-pores, along with electrolyte-driven anions in the grown PAA nanostructure. The underlying principle of PAA fabrication and its impact on sensor performance were thoroughly investigated; and a mechanism to further reduce the minimum detection limit with improved sensing characteristics was suggested. All the prepared samples exhibit an exponential change in capacitance with an increase in humidity, which is given by the equation y ¼ y o + Ae R o x . Sample SH05 with a pore diameter less than 5 nm AE 3 nm, was found to be extremely sensitive and the lower detection limit (LOD) was reduced even further to 3 RH%, an exceptional outcome indeed. Besides, the developed sensor (SH05) was found to be extremely fast with response and recovery times of $6 s and $54 s, respectively; however, the recovery times of samples SH05 and SH25 were found to be a little longer, which is directly related to the molecular kinetics of vapour molecules in an exceptionally small pore diameter (<11 nm). Sensors prepared via anodization in phosphoric acid solution were found to be useful for monitoring high humidity levels (>50 RH%) whereas sensors developed via anodization in sulfuric or oxalic acid solution are useful for monitoring low and medium ranges of humidity (0-50 RH%). The long-term dri was also investigated and found to show a negligible shi in capacitance value, ensuring the long-term stability of the developed sensor.

Conflicts of interest
There are no conicts to declare.  This journal is © The Royal Society of Chemistry 2020 Nanoscale Adv., 2020, 2, 2564-2576 | 2575

Paper
Nanoscale Advances