The proposed procedure for the fabrication of cantilever arrays combines anodic oxidation of aluminium foil with a conventional photolithography process ().

Figure 1 SEM images of a six-cantilevers array (2 μm thickness) obtained through chemical photolithography: (a) general view, (b) porous structure of AAO cantilever with 30 nm average pore diameter and 105 nm average interpore distances and (c) pore size distribution.

Using the lithography approach for AAO pattern formation mostly becomes possible due to selective and anisotropic etching of anodic alumina films on aluminium in basic solutions. On the other hand anodic oxidation itself enables easy control over channel diameter and interpore distance of the porous film over a wide range (pore diameter from 15 to 200 nm; interpore distance from 50 to 500 nm, film thicknesses up to 200 micrometers) [3]. Moreover the thickness of AAO films can be regulated precisely by using conventional electrochemistry methods by controlling the total electric quantity and assuming a current efficiency of 90% [12]. Thus, the proposed approach enables easy control over porosity, surface area, and geometric characteristics of cantilever arrays which provide wide opportunities for the design of micromechanical sensors with specific mechanical response.

To predict the behavior of the cantilever in the gas/liquid phase we investigated alterations of the amplitude–frequency characteristics when changing pressure and humidity. The frequency response of elastic beam is strongly dependent on the fluid it is vibrating in [2]. In vacuum, when the vapors are absent, the measurement of the resonance frequency is routinely made compared to measurements in viscous media. At first we emphasized the damping effect on cantilever vibration for porous AAO and standard Si cantilevers explored in the real system (air) and in model (vacuum). The amplitude–frequency profiles of rectangular cantilevers made of Si and porous anodic aluminium oxide are provided in . Their characteristics are summarized in . The *Q* factor for both types of beams significantly increases after change air media to vacuum. Natural shifts to high-frequency region are observed for the porous cantilever as well for the Si cantilever.

Figure 2 a) Frequency responses of Si rectangular cantilever (2 μm thick, 100 μm long and 30 μm wide) in the air (black line) and in vacuum (red solid line). The frequency shift is 780 Hz. b) Frequency responses of AAO cantilever (2 μm thick, 800 μm long and 100 μm wide) in the air (black line) and in vacuum (red solid line). The shift here is 3900 Hz.

Table 1 Shifts of resonance frequency and quality factors *Q* for Si and AAO cantilevers under different pressure (vacuum and air, humidity 22%).

For comparison we also studied the effect of the water content on the vibration of the cantilever from the AAO membrane obtained at 40 V in 0.3 M oxalic acid. shows the experimental results using a cantilever with a length of 800 μm and 10% porosity.

Figure 3 Frequency responses of AAO rectangular cantilever (2 μm thick, 800 μm long and 100 μm wide) at relative humidities of 10 and 21% (1120 Hz shift).

The resonant frequency decreases from 117.7 kHz at 10% to 116.5 kHz at 21% humidity. Changing the humidity from 21 to 10% increases the resonant frequency by 1120 Hz and quality factor of the AAO cantilever increases from 61 to 82, respectively ().

Using and the Young’s modulus of AAO of 340 GPa one can evaluate the quantity of water adsorbed onto the anodic alumina surface. The calculation gives a result of Δ*m* = 20 pg at a sensitivity Δ*f*/Δ*m* of 56 Hz/pg. On the other hand, the amount of absorbed water can be estimated from the Langmuir monolayer absorption isotherm. Taking into account the total area of the cantilever of 1.74·10^{−6} m^{2} (1.58·10^{−6} m^{2} interior and 1.63·10^{−7} m^{2} exterior area) and the density of the active sites of 4.8 nm^{−2} the maximal mass of a water monolayer adsorbed onto cantilever surface is 250 pg. This results in an increase of the experimental water surface coverage of about 8% within the experimental humidity levels.

From the Langmuir sorption isotherm a humidity change from 10 to 22% should result in growth of the surface coverage by 8.9%, which fits well to the obtained value. However, despite the good agreement between the estimations we should stress that they can be far from being realistic due to a strong chemical interaction of water molecules with AAO surface and pore curvature.

illustrates the resonant frequency dependence on air humidity. The experimental points were obtained by measuring the resonant frequency with decreasing humidity levels from 22 to 10% (at 298 K).

Figure 4 Resonance frequency of the AAO cantilevers (2 μm thick, 800 μm long and 100 μm wide) as function of relative humidity from 10 to 22%. Red line: approximation of the linear part as *f* = 118.9972·(1 + 0.00198*H*)^{−1/2}. Inaccuracy of frequency detection is below the size of the symbols.

After each humidity decrement the system was left to reach equilibrium for 10–30 min. Equilibration criterion was resonant frequency creep below 1 Hz/min. The curve indicates a linear behavior approximated by *f* = 118.99·(1 + 0.00198*H*)^{−1/2} (red line), which correlates well with . The humidity sensitivity of the AAO cantilever in the linear range estimated by the slope of the experimental curve equals about 100 Hz/%. Hysteresis was not observed for the considered range close to equilibrium. The resonance frequency for the same value of humidity measured after a cycle of increasing and decreasing humidity is well reproduced within the error of 30 Hz. We believe this error corresponds to an inaccuracy of setting up the desired humidity value and a slight deviation from equilibrium during measurements. The plateau at low humidity levels demonstrates a moisture detection limit of 10%. This likely occurs due to residual native water absorbed on the AAO surface, which cannot be removed without heating. Linear behavior of environmental humidity on the resonance frequency of the cantilever oscillation allows us to use AAO cantilevers as humidity sensors at least in the humidity range of 10–22%.