A modified cryostat for photo-electrical characterization of porous materials in controlled atmosphere at very low gas dosage

We developed an integrated system for photo-electrical characterization of materials for sensing applications in strictly controlled environment conditions. The peculiar aspect of this setup is the capability of a fine-tuned gas dosage and a fast dynamic chamber pressure control, coupled with current and voltage sensing within a modified cryostat. To illustrate the capabilities of our system we have characterised both p+-type mesoporous silicon (meso-PS) membranes and nano-crystalline mesoporous titanium dioxide (nc-TiO2) films. In particular, as a main topic is presented a well-resolved characterization of mesoporous silicon electrical conductivity changes induced by presence of ethanol. At low pore filling level adsorbate-shunted conduction is avoided, while dielectric screening effects on frozen doping centres are observable. Beside we presented observation of mesoporous titanium dioxide photo-conductivity as a function of different gas pressure reporting opposite effects of relatively low- and high-pr...

We developed an integrated system for photo-electrical characterization of materials for sensing applications in strictly controlled environment conditions. The peculiar aspect of this setup is the capability of a fine-tuned gas dosage and a fast dynamic chamber pressure control, coupled with current and voltage sensing within a modified cryostat. To illustrate the capabilities of our system we have characterised both p + -type mesoporous silicon (meso-PS) membranes and nano-crystalline mesoporous titanium dioxide (nc-TiO 2 ) films. In particular, as a main topic is presented a wellresolved characterization of mesoporous silicon electrical conductivity changes induced by presence of ethanol. At low pore filling level adsorbate-shunted conduction is avoided, while dielectric screening effects on frozen doping centres are observable. Beside we presented observation of mesoporous titanium dioxide photo-conductivity as a function of different gas pressure reporting opposite effects of relatively lowand high-pressure regimes. High reproducibility provided by the system is discussed as a final remark. C

I. INTRODUCTION
Nanostructured semiconducting materials of large specific surface (100-500 m 2 cm −3 ) have been widely studied in recent years for several applications, one of them being sensing, with increasing the need for performing measurements in highly controlled ambient conditions [1][2][3][4] Dealing with such materials and systems, even small fluctuations in ambient conditions induce large variation in their response when the measurement precision is significant, making difficult a fine and reproducible characterisation. Since in some cases sensitivity of ppb level has been reported, a very accurate control of the experimental environment will be necessary in any future activity to approach lower and lower detection levels. 5,6 Moreover, in this kind of materials, given the large surface to volume ratio, extrinsic phenomena like electronic transport through adsorbates easily overwhelm bulk effects. This is a strong limiting factor when investigating intrinsic properties of materials.
In order to be able to control such phenomena, we developed an integrated system for electrical characterization of materials for sensing applications in strictly controlled environment conditions modifying a commercial cryostat. To highlight the potentiality of this instrument, we investigated the reactivation of doping centers in mesoporous Silicon by dielectric screening of ethanol condensed within its pores. This is an example of bulk effect induced by the presence of a polar compound

A. System description
The system we implemented is based on a Janis ST-100 H cryostat equipped for fine-tuned gas injection driven by a Bronkhorst F201CV mass-flow controller (Ar calibrated), which can be connected to both a gas cylinder or a liquid container. The latter can be heated to achieve higher vapour pressures and to increase the stability of the mass-flow controller. F-201CV can reach a maximum flow of 10 sccm (see Figure 1 for a schematic description of the modified instrument). This model of cryostat was provided with four UV-grade fused silica optical windows. Two of them have been suppressed to gain a direct and independent access to the chamber for both gas injection and pumping (see Figure 1(a)), while the two remaining windows are available for any compatible optical purposes, even though, in principle, windows of different materials can be mounted, according to the wavelength of the employed radiation. The windows mounting position can be varied, in order to have them one in front each other, on opposite sides of the chamber, or sharing the same corner (angle of π /2) for optical applications in transmission or in reflection geometries, respectively. The system is additionally provided of a dry vacuum line to avoid back streaming of pollutant vapours from the pumps, a Varian SH-100 scroll pump and an Alcatel ATP 80 turbo-molecular pump. The pumping line is equipped with a manual Edwards Speedy-Valve TM just at the exhaust of the cryostat. This way the pumping efficiency can be also manually modulated in order to attain a wider pressure range. For measurements performed in vacuo, the system guarantees on the sample a vacuum down to 3.5 · 10 −5 mbar.
Both an E type thermocouple and a platinum Pt-100 resistive temperature sensor (DIN 43760 and IEC 751 Standard calibration down to 70 K) provide temperature sensing. An additional Pt-100 can be placed to perform temperature studies in remote mode (on the sample other than at the heater element). This additional sensor served also to test the temperature gradient between the sample holder and the electrical wiring late. A Lake-Shore 331 controller coupled with a liquid nitrogen transfer-line provides the sample holder temperature control and monitoring. The temperature controller features a standard P.I.D. feedback action. The temperature measurement precision was taken as 70 mK in the 77-550 K range. This depends on the electronic chain and the intrinsic limits of the sensor's calibration. The sample holder at the end of the cold finger column has been modified to host a wider variety of samples. Two shafts with three degrees of freedom were added to the sample holder in order to mechanically fasten samples in the most suitable way. These end with screws provided of a teflon-covered tip. No blind screw-holes that could limit the pumping speed are present.
Instead of the original tip-probe system, the electrical wiring inside the chamber ends at the sample holder side with eight mechanically fastened pin connectors. Wires are AWG 42 gauge, 0.63 mm diameter, kapton sleeved were plug directly into a drilled fiberglass-polyester board (red plate in Figure 1(c)). This way there are no restrictions to the sample contact geometry, because the only request is the presence of two (or more) macroscopic pads on the sample support, while the in-situ contact geometry can be engineered free of constraints, only depending on the sample itself.
The system is also provided with a Keithley 2400 SourceMeter TM for electrical measurements in different wiring configurations. Instead of the original tip-probe system, the electrical wiring inside the chamber (AWG 32 leads) ends at the sample holder side with a 8 pin male connector. This way there are no restrictions to the sample contact geometry, because the only request is the presence of two (or more) terminal pads for each sample on the support, and the on-sample contact geometry can be engineered free of constraints, only depending on the sample itself. Then it is possible to make connections between sample and measurement leads directly by means of small gauge gold wires. An ad hoc software interface has been realised within the LabVIEW TM environment. The flowmeter is driven by its own Flow-DDE TM server, which can be linked to LabVIEW by means of a proper library of subVis. The module devoted to reach and maintain a target pressure has a sampling rate of about 250 ms. The flow is controlled according to the target pressure by a feedback software module implemented within LabWIEW in which the readings from the vacuum gauges drive the mass-flow set-point. The solenoidal valve is driven by the Bronkhorst programmable proprietary P.I.D. feedback.
As an example of the ability of the system to react to successive requests of variation of the equilibrium pressure within the chamber is reported in Figure 2. Starting from an initial equilibrium condition of 0.1 mbar of N 2 , the first target pressure was set to 5 · 10 −2 mbar. The target pressure is reached after 100 s, with oscillations of ±1 · 10 −2 mbar. The amplitude of the latter is progressively dumped in time and after 400 s from the transient they are below ±1.5 · 10 −3 mbar. Once the equilibrium has been reached, another step toward 2.5 · 10 −2 mbar follows, where the system reacts in a similar way. To reach and maintain the target pressure, the mass-flow controller set point automatically changes according with software parameters, see blue curve in Figure 2. Another example of the chamber conditions control is reported in Figure 3. The chamber pressure is 0.100 ± 0.05 mbar of ethanol at 220 K. As the temperature decreases to 210 K some ethanol condenses on the sample holder and the pressure tends to drop (see art-line in Figure 3 as a guide). To keep the constant target pressure the mass-flow controller varies the modulation and value of its set-point until a new equilibrium conditions is reached. As the temperature-induced pressure drop occurs (time = 6600 s) the feedback reacts and the target pressure is restored in about 100 s at the initial level within oscillations of ±0.05 mbar.
The software is designed to perform linear and logarithmic Current-Voltage or Current-Time measurements while simultaneously controlling temperature and pressure. The Keithley 2400, Lake Shore 331 communicate by standard GPIB 448.2 protocol, while a RS232 serial protocol drives the Edwards gauge controller.
A 6W UV tube lamp notch filtered at 365 nm have been used to expose titania thin films to band-gap light illumination. Light intensity on the sample was 20 μW cm −2 . Gases dosed during photoconductivity measurements on nc-TiO 2 was supplied from cylinders of laboratory grade gas (purity of ppb).

B. Sample preparation
Porous Silicon membranes have been prepared by electrochemical etching of p + -type Si (Boron doped, resistivity 8-12 m cm, 100 oriented). A teflon cell was filled with 25% HF aqueous etching solution (HF[50% in H 2 O] 1:1 in ethanol) and several etching-rest cycles were performed (J = 325 mA cm −2 current density for 1 s and J = −0.1 mA cm −2 for 10 s). These membranes, 10 μm thick, were detached from the bulk silicon by a short electropolishing step at a current density of 1 A cm −2 , and moved on quartz plates. This kind of support prevents leakage current through the substrate during dc measurements. The porosity was calculated to be about 70% by gravimetric measurements on sacrificial samples. Planar electrical contacts at a distance of 300 μm were made by means of a stencil mask with colloidal silver (Pelco 187). Crosscheck measurements to test the effect of bulk ethanol as a shunt resistance have been performed on bulk silicon samples. These samples consisted in p-type silicon with 90 nm of thermal SiO 2 on the polished side. The conductivity of these oxide films is comparable to meso-PS, so it is possible to discriminate the effect of the sole ethanol in conductivity enhancement due to shunt resistance. Electrical contacts in planar configuration were painted with colloidal silver (Pelco 187) at a distance of 500 μm.
TiO 2 samples consisted in commercial electrodes for dye-sensitized solar cells (DSSC). 14 These were 15 μm thick mesoporous films of mostly anatase phase titanium dioxide deposed on conductive glass that served as bottom electrode (porosity around 60%). Measurements were conducted in stack geometry after the deposition of a graphite cement top electrode. For reproducibility considerations, similar anatase films, but with gold contacts in planar configuration, were used. These were obtained by screen printing of commercial anatase colloidal dispersion (Solaronix), and were 3.5 μm thick. The planar contact geometry allowed to host three samples per support (bare quartz). Detailed description of measurement preparation protocol for this reactive material are reported elsewhere. 15

III. RESULTS AND DISCUSSION
This experimental set-up guarantees a fine control on the chamber pressure, better than 1% in the range of 10 −2 mbar and even better at higher pressures. As already discussed in the experimental (Section II, Figure 3) by acting on the temperature of the sample holder it is possible to reach the gradual condensation of ethanol on the sample while maintaining a constant pressure in the chamber. Ethanol, in meso-PS experiments, was chosen for several reasons. We already performed spectroscopic studies on the same topic with ethanol filled meso-PS 10 and ethanol has one of the largest dielectric constants among polar solvents ( d = 25). Moreover its condensation in the pores significantly increases the effective dielectric constant eff of mesoporous Silicon membranes. Finally its freezing point at pressures of 0.1 mbar is at temperatures low enough (160 K) to achieve condensation and even pore filling, but no solid transition, within a broad temperature range. Another intriguing aspect is the possibility to verify that liquid-solid transitions for confined liquids is downshifted. 16 In order to get rid of injection-or ionization-related effects, dc measurements on both meso-PS and TiO 2 samples were carried out at small applied voltages (electric field smaller than 10 2 V cm −1 in the present sample configurations).

A. Dielectric screening in meso-PS
In Figure 4 are reported two Arrhenius type plots. Diamonds represent the temperature dependence of conductivity for a porous silicon sample in vacuum after a thermal annealing (450 K for 30 min, 1 K min −1 ) to clean the surface from adsorbed water from ambient. In meso porous silicon the electrical transport can be considered to be thermally activated as in standard semiconductors, because the crystalline structure is large enough (10-20 nm) to prevent quantum confinement effects. 17,18 A typical Arrhenius behavior is expected in absence of adsorbates.
Circles represent the characteristic for the same sample, at a pressure of 0.1 mbar of ethanol. Three main features are present. The first consist in a steeper decrease of conductivity for exposed meso-PS at high temperature (see Figure 4 inset). This can be related to the effect of donorlike surface states induced by adsorbed ethanol molecules. 9,19 Following this hypothesis, in this temperature range, the amount of ethanol adsorbed in the channels of the meso-PS is not enough to induce an effective dielectric screening within the bulk skeleton of meso-PS. The dominant effect is the formation of more donor-like states at surface that compensates the holes concentration.
The second feature is a jump (labeled "j" in Fig. 4 of the exposed curve, at about 210 K. The conductivity of the ethanol-exposed meso-PS sample crossing this temperature is 3.4 · 10 −13 S cm −1 , while in absence of ethanol it was 7.4 · 10 −14 S cm −1 . We assign to this feature the doping centers reactivation by dielectric screening of ethanol. In Eq. (1) is reported the expression for thermal activated electrical conductivity the pre-exponential factor σ 0 depends on free carrier concentration The ethanol-exposed curve features two significant jumps labeled "j" and "J", as we refer to these in the text.
(p) and mobility (μ): Note that the slope of the exposed sample curve in Figure 4 at 200-180 K exhibits a slope very similar to the unexposed one (and to itself before the first jump ("j"), indicating that only the pre-exponential factor of conductivity was significantly affected by ethanol. Hence the transport mechanism can be considered unchanged. FTIR measurements published elsewhere 10 showed a carrier reactivation due to ethanol dielectric screening on similar samples, but electrical transport measurements could also in principle account for mobility effects due to reduced coulomb blockade within transport paths. 20 The third interesting feature is the huge jump (labeled "J" in Fig. 4 at 180 K of the exposed curve. We attribute this effect to the saturation of ethanol within the pores with a consequent shunting of the meso-PS porous network. To justify our assignments, we made some calculations based on a Bruggerman effective medium approximation. 21 Under this effective medium approximation, for a porosity of 70%, the d.c. conductivity of ethanol-saturated meso-PS at RT would be about 1.4 · 10 −6 S cm −1 , against a common reported value for bare meso-PS conductivity 22 of about 10 −10 S cm −1 . This means a difference of 4 orders of magnitude between unfilled and saturated meso-PS. Since in the present case the experimental conductivity of the exposed sample at 210 K is only a factor of 4.5 larger than the conductivity of the unexposed one, while at lower temperature occurs a huge jump, a complete saturation of pores at 210-180 K could be excluded and the enhancement is assigned to additional free carriers. Note that the reported conductivity at RT for ethanol in literature is lower than the value we used here for calculations. 23 We measured the conductivity of ethanol employed in this experiment in a two-electrode cell. However, a possible conductivity enhancement due to the typical impurities and moisture contamination of laboratory grade ethanol cannot be excluded. The situation is rather different for the third reported transition ("J"), around 180 K. Here the conductivity of the unexposed meso-PS 2.8 · 10 −14 S cm −1 while the exposed one it is 2.4 · 10 −10 S cm −1 , with indeed a 4 orders of magnitudes increment. As already pointed out, some confinement effects could be highlighted. Note that even at temperature lower than 160 K there is not an abrupt change in conductivity for exposed meso-PS. This suggests, in our opinion, that ethanol is not yet frozen on the surface nor within the pores os meso-PS.
The undershoots in σ (T) at 220 K and 180 K are due to the time for current to reach equilibrium at the beginning of the jumps of characteristics. We decided to set a maximum settling time of 1000 s for each point acquired to avoid excessive exposure to ethanol condensing vapour.
A crosscheck test performed on a bare Silicon oxide sample, confirmed that at low temperature, when exposed to ethanol vapour, the conductivity suffers an abrupt change with a jump of 3 orders of magnitude. This should confirm the nature of the third reported transition (Fig. 4 -"J"), and allowed us to assign the second (Fig. 4 -"j") one to dielectric screening effects on frozen doping centers.

B. Mesoporous TiO 2 photoconductivity under gas dosage and effects of background ambient
We performed experiments on mesoporous TiO 2 films in dark, for pressures below 5 mbar of dry-air (1 mbar O 2 ), and those did not highlight any substantial effect on the conductivity, while at 10 3 mbar the electron-scavenger nature of adsorbed oxygen have been observed clearly.
Concerning the experiments under illumination the effect of small gas dosages was instead remarkable. Under the reported illumination conditions, conductivity rose of about 5 orders of magnitude compared to dark conductivity at the same bias voltage (0.5 V), with extremely slow transients, reported in Figure 5 to reach equilibrium. This behaviour is consistent with other works on similar topics. 24, 25 Figure 6 reports the perturbation induced by pressure steps under constant illumination. In region "a" the pressure is kept constant at a minimum average value of 1.5 · 10 −2 mbar and the current is slowly decreasing after a preview run of the experiment. For our purposes it is not a concern this not completely steady current. In region "b", at a pressure of 4 mbar of dry-air, the conductivity is degraded with a steep slope decay characteristic. The duration of the exposition to dry-air have been limited in order to limit the recovery time (region "c"). Conversely at lower dry-air pressure levels (region "d") the conductivity is enhanced. The last region "e" of the plot shows the slow relaxation transient after the small dosage and is consistent with region "a", in fact the initial current decrease (T < 4000 s) is the result of a preview small dosage like the one of region "d" as the sequence have been repeated during each run of the experiment. In the case of small dosage the exposition time was longer to attain a clear enhancement of photoconductivity because this effect is weaker than the one due to higher dry-air pressure.
Some considerations are necessary at this point. At any degree of vacuum achievable from the presented system, the TiO 2 films at a point reach equilibrium with the background chamber ambient. Background ambient, in absence of other sources (dosages from the mass-flow controller), is due to chamber leaks and pumps back-stream, roughly composed by N 2 , O 2 and water similarly to the external ambient. So on the surface of TiO 2 a certain amount of adsorbed oxygen (O 2(ads) ) and hydroxyl groups (OH) are always present and affect concentration and mobility of excess photoelectrons. In particular electron trapping at O 2(ads) sites (Eq. (2)) and interaction with holes quickly trapped at OH sites (Eq. (3)) favour recombination. 26 Surface charge induced band banding leads to bottlenecks at nano-crystals interface. 27 Once dry air is injected the background ambient is brought out of equilibrium with respect to TiO 2 surface. Photoconductivity enhancement at lower exposure (0.1 mbar of dry-air) seems to indicate that the TiO 2 surface is partially depleted of recombination centers, the surface band banding is reduced or both.
To better understand this fact other experiments were performed. A sample was initially left to equilibrate with background ambient in vacuum at 10 −4 mbar (see Fig. 7 for t < 1000 s). Then the turbo-molecular pump was stopped and the ambient conditions changed accordingly. At the same time the current dropped. This can be attributed to oxygen coming from the chamber leaks and pump back-stream. No other sources were present. Moreover as reported in Figure 7 at about 2600 s, nitrogen were dosed (0.2 mbar) and the current rose until N 2 was pumped out again.
Because nitrogen is expected to be inert on TiO 2 its effect on TiO 2 surface population should be different compared to oxygen's one. In our opinion the effect of nitrogen may be related to the dilution of the photo-adsorbed oxygen at surface with consequent reduction of oxygen-related effects on photo-conductivity. At the moment the same group is working on this effect on different prepared titania surface.
In summary it seems possible that different phenomena are responsible for similar photoconductivity enhancement effects on illuminated TiO 2 mesoporous films. The important point within the context of this work is the possibility of clearly observing and repeatedly reproducing them by the present experimental set-up.

C. Reproducibility
As already pointed out, one crucial aspect of working with large surface area materials is the reproducibility of both sample preparation and characterization. In the following we present some data to better clarify the reliability of the present system about that. Series of samples were prepared and characterised in different moments, hence the starting conditions of both sample surface and vacuum chamber could be different in principle (pollutants on sample surface, ambient humidity). In this case support (a, b) hosts three nominally identical samples (A, B ,C), which were characterised in stable conditions after a common thermal treatment mandatory to remove most of the adsorbed water. 15 In Figure 8 are reported two sets of data from in dark electrical measurements on anatase thin films. The strong reproducibility between the sets of data confirms the similarity of the samples (a necessary but not sufficient condition), moreover shows that even for independent loading and preparation steps the system led likely to the same surface condition, and allowed for highly reproducible electrical characterization.

IV. CONCLUSIONS
We developed an integrated system for the electrical photo-characterization of sensing-oriented materials in strictly controlled experimental conditions. The reported experimental results show its capability of resolve the different effects of respectively adsorbed, condensed and bulk ethanol within the pore volume of free-standing meso-PS membranes. We remark that IR techniques, though selective and versatile for free carrier-related investigations, do not easily allow distinguishing these different phenomena. Moreover typical gas dosage in spectroscopy starts from a few mbar, while this system allows for dosage of fractions of mbar. Our system also allows for discriminating competitive surface phenomena affecting TiO 2 porous films photo-conductivity by means of finetuned gas dosages and well-resolved current measurements.