Molecular Engineering of Silicon Phthalocyanine to Improve the Charge Transport and Ammonia Sensing Properties of Organic Heterojunction Gas Sensors

Novel organic heterostructures fabricated with a bilayer consisting of an axially substituted silicon phthalocyanine (R2‐SiPc) derivative and lutetium bis‐phthalocyanine (LuPc2) are investigated for their ammonia sensing properties. Surface and microstructure characterization of the heterostructure films reveal either compact or highly porous surface topography in (345F)2‐SiPc and Cl2‐SiPc‐based heterostructures, while electrical characterization reveals a strong influence of the axial substituent in R2‐SiPc on NH3 sensing capabilities. Electrical characterization further demonstrates an apparent energy barrier for interfacial charge transport, which is higher in the (345F)2‐SiPc/LuPc2 heterojunction device. In‐depth charge transport studies by impedance spectroscopy further reveal a resistive interface in (345F)2‐SiPc/LuPc2 and faster bulk and interfacial charge transport in Cl2‐SiPc/LuPc2 heterojunction devices. Different interfacial charge transport capabilities and surface topographies affect NH3 sensing properties of the two heterojunction devices, in which (345F)2‐SiPc/LuPc2 reveals a fast and non‐linear response with a limit of detection (LOD) of 310 ppb, while Cl2‐SiPc/LuPc2 exhibits a slow, and linear response to NH3 with LOD of 100 ppb. Finally, different metrological parameters of the two sensors are correlated to the respective gas‐material interactions, in which adsorption and diffusion regimes are modulated by the surface topography and hydrophobicity of the sensing layer.


Introduction
Organic semiconducting materials have gained attention in industrial and academic settings because of their low manufacturing cost, numerous available processing methods, and broad applicability in organic electronic devices. In some cases, devices based on these materials have demonstrated comparable efficiencies to devices fabricated using inorganic counterparts. Gas sensors based on device architectures, including electronic noses, have recently become more widely explored, having applications in air quality monitoring, food production, and healthcare. [1][2][3] In such devices, organic semiconductors can be used as the active sensing layer, [4] owing to their aforementioned advantages. Moreover, because of the ability of these materials to be incorporated in flexible electronics and processed in solution by existing printing technologies, miniaturized and portable gas sensing devices to detect various redox gases of industrial and environmental relevance are being developed and have been reported. [5,6] Over 175 million tons of ammonia (NH 3 ) are produced globally each year as an essential industrially produced chemical. [7] In addition to its widespread use in agriculture, pharmaceutical development, and food production, NH 3 has received increased attention as a refrigerant, capable of replacing chlorofluorocarbons. Moreover, it is being increasingly used as a source of green hydrogen [8] and an important biomarker in disease monitoring. [9] Despite its widespread use, NH 3 is highly toxic, corrosive, and damaging to human health and the environment. Short-term (10-15 min) exposure limits to NH 3 are typically set at 35 parts per million (ppm) while longer (8-10 h) limits are set at 25 ppm. [10] To enable a proactive response to NH 3 exposure and prevent serious negative health outcomes, NH 3 gas sensors must have a limit of detection (LOD) below these thresholds, with rapid response and regeneration times, and with high selectivity for NH 3 in environments containing other gaseous analytes, especially water vapor. Many sensing technologies have been reported previously for NH 3 detection, such as chemiresistors, [11] diodes, [12] and field effect transistors (FET), [13,14] which revealed high sensitivities to NH 3 . However, conventional chemiresistors based on metal oxides often require high operating temperatures (typically over 100°C), have long response/regeneration times, and suffer from intrinsic drift in the signal baseline of the sensor. [15][16][17] FET-based sensors require relatively expensive fabrication methods, such as micromachining and clean-room requirements, limiting their upscaling for large area coverage.
Bouvet and co-workers introduced new sensor architectures based on organic heterostructures, combining two organic semiconductors with different work functions and charge carrier concentrations. [15,16] The sensors had a planar bilayer architecture where the bottom layer was either a continuous layer or only localized around the electrode. Sensors based on these architectures benefitted from good interfacial charge transport, owing to organic heterojunction effects, in which opposite mobile charges (e− and h+) accumulate at the organic-organic interface, [17] as well as the unique configuration of the heterostructure, in which larger interelectrode distance from the organic layer promotes charge transfer through the interface. [18] As a consequence, sensor performance toward NH 3 gas demonstrated an improvement in sensitivity, reproducibility, response/recovery kinetics, and operational stability compared to existing chemiresistors. [19,20] A wide range of molecular semiconductors and polymers have been investigated as active sensing materials in organic heterojunction sensors for NH 3 detection including metal phthalocyanines (MPc), polyporphyrins, [21,22] perylene derivatives, [23] and triphenodioxazine. [24] MPcs are thermally stable conjugated macrocycles with favorable semiconducting properties making them ideal for use in organic heterojunction sensors, besides other areas of organic electronics such as organic thin-film transistors (OTFTs) and organic photovoltaics (OPVs). [25,26] MPcs are particularly interesting due to their ability to be synthesized with a variety of metal cores and peripheral substituents, effectively tuning their optical and electrical properties. [27][28][29] These properties have been exploited in developing wide ranges of organic heterojunction sensors for NH 3 detection. For instance, the effect of the central metal atom in MPc/LuPc 2 heterostructure (M: Cu, Co, Zn) was studied by modulating the semiconducting polarity and measuring its impact on NH 3 sensing performance, revealing the highest sensitivity for CoPc/LuPc 2 , while an ambipolar behavior for ZnPc/LuPc 2 -based sensors. [30] Similarly, the effect of peripheral substituent was investigated in an R 8 NiPc/LuPc 2 heterostructure toward NH 3 sensing properties, in which NiPc bearing hexyl sulfanyl groups revealed p-type behavior and the most stable and the highest sensitive response. [31] Silicon phthalocyanines (R 2 -SiPc) are a class of MPc that have shown promise as a molecular semiconductor in n-type or ambipolar OTFTs. [32,33] Additionally, due to the multiple oxidation states of silicon, R 2 -SiPc can be functionalized with axial ligands, besides peripheral substitution, to tune their crystalline structure in thin films, their chemical affinity, and their electronic properties. [34,35] A large catalog of axially substituted R 2 -SiPc derivatives have been reported, which are processable into devices by physical vapor deposition (PVD), [34] solution casting, [36] or both, [37] making them suitable for applications in organic electronics. R 2 -SiPcs have shown great promise in heterojunction devices, having been incorporated into both planar heterojunction (PHJ) [26,38] and bulk heterojunction (BHJ) [39] OPVs as non-fullerene acceptors. Despite exhibiting high performances in PHJ and BHJ, the use of R 2 -SiPcs as an active material in gas sensors has been scarce. Recently, carbon nanotube crosslinked to R 2 -SiPcs via axial substitution was used in chemiresistive sensors for detecting NH 3 and H 2 . [40] A conductometric sensor based on a Langmuir-Blodgett coated R 2 -SiPc, having tertbutyl as a peripheral substituent and ethyloxy substituents as axial ligands, was reported to detect NO 2 at 5 ppm. [41] Elsewhere, the gas-sensing properties of polymeric SiPc films were investigated at high temperatures to detect NO 2 and Cl 2 vapors. [42] However, to the best of our knowledge, there is no SiPc-based heterostructure reported as gas sensors. Considering their good conductivity performance in PHJ and BHJ photovoltaic devices and their applications in chemiresistor gas sensors, they are an attractive class of materials for incorporation into organic heterojunction gas sensors.
The present work explores the potential of R 2 -SiPcs in a bilayer organic heterojunction device (Figure 1) by combining it with a highly conducting LuPc 2 and investigates its NH 3 sensing properties. Two different R 2 -SiPc derivatives, axially substituted with chlorine and trifluorophenol, respectively (Figure 1a), were studied in heterostructures as the sublayer in order to modulate the interfacial charge transport and NH 3 gas sensing properties of resulting devices.
The chemical functionality and microstructure of the bilayer films are characterized by UV-vis and Raman spectroscopies, while their surface topographies are investigated by atomic force microscopy (AFM). The electrical properties of the heterojunction devices have been studied by current-voltage (I-V) measurements, while in-depth charge transfer studies have been performed by impedance spectroscopy. NH 3 sensing properties of the heterojunction sensors are finally evaluated at room temperature and correlated to a thin film and molecular structure.

Microstructure Characterizations of Bilayer Heterostructures
The normalized optical absorption spectra of each layer, as well as the resulting bilayer heterostructures, are displayed in  the R 2 -SiPc sublayer that correspond to Q-band between 650 and 750 nm, [43] with peaks in Cl 2 -SiPc/LuPc 2 and (345F) 2 -SiPc/LuPc 2 at 749 and 670 nm, and 723, 693, and 673 nm, respectively. A band of LuPc 2 (670 nm) overlapped with the Q-band of the R 2 -SiPcs. The weak band in both bilayer heterostructures at 467 nm corresponds to the transition between a filled orbital toward the semi-occupied molecular orbital (SOMO) of the LuPc 2 . [44,45] At higher energy levels, both bilayers, experienced strong absorption peaks at 378 and 341 nm (Cl 2 -SiPc/LuPc 2 ), and 360 and 342 nm ((345F) 2 -SiPc/LuPc 2 ), which are attributed to the Soret or B-bands of MPcs in both layers. Notably, Q-and B-bands of both bilayers experienced splitting into two or three sub-bands, which are ascribed to Davydov splitting, originating stronger intermolecular interactions in the thin film of MPc heterostructures.
Raman spectroscopy measurements were performed on powders of Cl 2 -SiPc, (345F) 2 -SiPc, LuPc 2 , and their corresponding bilayer heterostructures. The Raman spectra of the bilayer clearly show a combination of peaks that correspond to the chemical fingerprints of the LuPc 2 layer and the Cl 2 -SiPc or (345F) 2 -SiPc layer (Figure 2c,d). While the Raman signature of Cl 2 -SiPc and (345F) 2 -SiPc have not previously been reported, the peaks are consistent with common MPcs including CuPc and ZnPc. [46] In the Cl 2 -SiPc/LuPc 2 heterostructure, the peak at 592 cm −1 is consistent with the Pc breathing mode. Additional peaks at 756, 965, 1005, 1048, 1188, and 1525 cm −1 and the strong peak at 1558 cm −1 belong to Cl 2 -SiPc. In the (345F) 2 -SiPc/LuPc 2 heterostructure, peaks at 594 and 652 cm −1 correspond to Pc breath-ing. Peaks at 1047 and 1196 cm −1 correspond to C─H bending, and peaks at 1615 cm −1 and the strongest peak at 1553 cm −1 belong to (345F) 2 -SiPc. The results are compiled in Table S1, Supporting Information. Most of the peaks from LuPc 2 overlap with peaks from the sublayer, resulting in a change in peak intensity, for example, the peak at 682 (strong in LuPc 2 ) overlaps with peaks at 679 (medium in Cl 2 -SiPc), and appears at 680 cm −1 in Cl 2 -SiPc/LuPc 2 bilayer heterostructure. Both Cl 2 -SiPc and (345F) 2 -SiPc are highly active in Raman spectra compared to the LuPc 2 layer. However, peaks at 1174 (Cl 2 -SiPc/LuPc 2 ), or 1177 ((345F) 2 -SiPc/LuPc 2 ), and 1512 cm −1 are unique to LuPc 2 . The results clearly illustrate a simple addition of peaks corresponding to each individual layer without the introduction of additional peaks, suggesting that no degradation of the Cl 2 -SiPc or (345F) 2 -SiPc layer is observed due to the subsequent deposition of LuPc 2 during the fabrication of bilayer devices. Due to the charge transport being primarily at the bilayer interface, it is critical to prevent degradation of the bilayer interface as a result of the deposition of LuPc 2 .
To further probe the surface topography of the bilayer heterostructures, films were characterized by AFM. Figure 3 depicts the AFM micrographs of Cl 2 -SiPc/LuPc 2 and (345F) 2 -SiPc/LuPc 2 bilayer films coated on glass substrates. The two heterostructures present different surface topographies, as evidenced by a highly rugged and porous film of the Cl 2 -SiPc/LuPc 2 heterostructure, as opposed to the compact packing of molecular crystals in (345F) 2 -SiPc/LuPc 2 heterostructure. Taking into account the common top layer in the bilayer heterostructures, it is clear that the organization of the sublayer provides a template for the sub- sequent growth of the top layer, thus influencing the surface topography of the bilayer film. [47] The evolution of different surface topographies of Cl 2 -SiPc and (345F) 2 -SiPc can be attributed to different molecular structures, affecting the solid state packing and evaporation sticking coefficients, [26,43] which would determine their microstructural organization in the thin film. The distribution of surface pores of up to 50 nm in Cl 2 -SiPc/LuPc 2 film is evident in the surface height variations profile (Figure 3c) deduced from a line scan in the AFM image. Notably, the surface depth is similar to the thickness of the sum of the evaporated thickness of both layers, indicating that the sublayer does not completely cover the substrate, leaving behind holes to be filled by LuPc 2 . Further analysis of the total image area resulted in the mean roughness (RMS) of 15.6 nm. The surface of (345F) 2 -SiPc/LuPc 2 film is mainly composed of closely packed elongated grains of size ≈160 nm, while the RMS roughness of the total image area is estimated to be ≈7.3 nm. The presence of fewer but larger grains can be observed as bright spots in the image, with a surface feature height increasing by up to 40 nm (Figure 3d) in the selected line scan. The different surface topographies as confirmed by AFM images clearly have implications on the nature of R 2 -SiPc/LuPc 2 interfaces formed within the heterostructures. Higher RMS roughness and the porous and rugged surface of the Cl 2 -SiPc/LuPc 2 film indicate that there may be a non-continuous interface between Cl 2 -SiPc and LuPc 2 layer. Conversely, the relatively smooth and compact surface of the (345F) 2 -SiPc/LuPc 2 bilayer predicts a continuous interface between the constituent organic layers. Such significant differences in the R 2 -SiPc/LuPc 2 interface in the two heterostructures are expected to influence the electrical properties and sensitivity of resulting gas sensors.

Electrical Properties of Heterojunction Devices
Electrical properties of the Cl 2 -SiPc/LuPc 2 and (345F) 2 -SiPc/LuPc 2 bilayer heterojunction devices were investigated by performing I-V measurements in the applied bias range of −10 to +10 V. Figure 4a,b depicts the recorded I-V curves, which are non-linear and symmetric. Such features are characteristic of bilayer organic heterojunction devices, originating due to the accumulation of mobile charges at the interface. [18] However, the degree of non-linearity and magnitude of the current at ±10 V are different in the two devices, which are attributed to the nature of interface formation between the two organic layers and the charge transport resistance in the R 2 -SiPc sublayer. The nonlinearity in the I-V curves is quantified by estimating the apparent energy barrier (U th ) by tracing a tangent on the I-V curves and extrapolating it to the x-axis. The calculated U th is 2.2 and 5.1 V for Cl 2 -SiPc/LuPc 2 and (345F) 2 -SiPc/LuPc 2 devices, respectively, clearly indicating that the (345F) 2 -SiPc/LuPc 2 device has a higher energy barrier/resistance for the transport of mobile charges, although the work function of the sublayer materials are similar (4.1 and 4.4 eV). [43] The low U th and higher current at +10 V in Cl 2 -SiPc/LuPc 2 can be assigned to the non-continuous interface formation between Cl 2 -SiPc and LuPc 2 , because of the porous nature of the bilayer film, causing weaker heterojunction effects.
To further understand the electrical characteristics of the two heterojunction devices, in-depth charge transport studies were performed by impedance spectroscopy in a wide frequency range from 10 Hz to 10 MHz at different AC and DC bias. The resulting Nyquist plots of both heterojunction devices yielded two depressed semicircles (Figure 4c,d), in which the one at high frequency (HF) does not change, while the other at low frequency (LF) gets smaller, with the increase of DC bias from 0 to 10 V. This confirms the presence of an organic-organic interface in the organic heterostructures and the semicircles at HF correspond to the bulk charge transport, while the semicircles at LF are associated to interfacial charge transport. [48] This is further validated by the Nyquist plot of the LuPc 2 resistor, exhibiting only one depressed semicircle, whose size remains independent of the applied DC bias ( Figure S1, Supporting Information). Since the size of the depressed semicircle in the Nyquist plot is inversely proportional to the charge transfer resistance, it implies that bulk charge transport remains constant and interfacial charge transport increases with increasing bias in both devices. However, the magnitude of the bulk and the interfacial impedances are lower in Cl 2 -SiPc/LuPc 2 devices than in (345F) 2 -SiPc/LuPc 2 devices, implying that both bulk and interfacial charge transport are faster in the former.
To quantitatively determine different charge transport parameters, the experimental Nyquist plots were fitted with constant phase element (CPE) based circuits, taking into account distributed microstructural properties of the heterostructures. [49] CPE is an imperfect capacitor, which impedance is defined by Equation (1), where Q i , , and are non-ideal capacitance, frequency, and a constant varying between 0 to 1, respectively. CPE turns into a perfect capacitor or a perfect resistor, when equals 1 or 0, respectively.
The Nyquist plots of both heterojunction devices were modeled by two-component Ri-CPEi elements connected in series as shown in the inset of Figure 4c. A contact resistance Rc is also added in series to take into account the interface between the electrode and the sublayer. Here, the R1-CPE1 describes the semicircles at HF, while the R2-CPE2 represents the semicircles at LF, and they are correlated to bulk and interfacial charge transport, respectively. The fitting resulted in the determination of R1, 1, and C1 eff (effective capacitance; defined in Equation (S1), Supporting Information) as bulk charge transport parameters and R2, 2, and C2 eff as interfacial charge transport parameters, whose variations with DC bias are depicted in Figure 5.
From the variations of bulk resistance (R1) (Figure 5a) and interfacial resistance (R2) (Figure 5b) of both devices as a function of DC bias, it is noted that R1 remains nearly independent of applied bias, while R2 yields an exponential decay with increasing applied bias. This observation infers that interfacial charge transport gets faster with increasing applied bias, while bulk charge transport remains unchanged, which is a typical characteristic feature of bilayer heterojunction devices. [50] The rapid decrease in the R2 is attributed to the arrival of injected charges from the electrode to the R 2 -SiPc/LuPc 2 interface, which further populates it with accumulated mobile charges, thereby filling the traps and enhancing the charge mobility along the interface. Conversely, the mean variation of the mobile charges concentrated in the large bulk region does not change significantly with changing bias, keeping charge transport unchanged. Notably, R1 and R2 of the (345F) 2 -SiPc/LuPc 2 bilayer device are 66 and 2760 times higher at 1 V than R1 and R2 of Cl 2 -SiPc/LuPc 2 , respectively, which are also commensurate with the higher magnitude of current and lower U th of the latter in the I-V curve. Such a large R2 in (345F) 2 -SiPc/LuPc 2 device can be attributed to the small number of mobile charge carriers accumulated at the interface between (345F) 2 -SiPc and LuPc 2 because of their work function differences. Moreover, the bulk of (345F) 2 -SiPc microstructure is highly resistive. On the other hand, the interface between Cl 2 -SiPc and LuPc 2 layers is not continuous because of the porous surface topography, which reduces the length of interfacial and bulk charge transport pathways, thus reducing its resistance. The alpha parameters are related to dispersion in the R-CPE circuit, such that a value of 1 corresponds to a homogeneous CPE or an ideal capacitor, while a decrease from 1 to 0 indicates an increase in heterogeneity in the CPE. In other words, as alpha changes from 1 to 0, the capacitive contribution decreases and the conductive contribution increases in the device's total electrical output. In the present case, 1 of both heterojunction devices is initially greater than 0.93 and increases toward 1 as bias increases. This data indicates that CPE associated with the bulk region is highly homogeneous, which is consistent with the negligible variation in the electrical properties of the bulk with a change in bias. On the contrary, 2 experienced a significant decrease from 0.95 to 0.81 in (345F) 2 -SiPc/LuPc 2 heterojunction devices and from 0.83 to 0.69 in Cl 2 -SiPc/LuPc 2 heterojunction devices. This indicates that the interfacial charge region becomes more heterogeneous with increasing bias in both devices, which is attributed to the arrival of a higher concentration of mobile charges at the interface, thus extending the conducting networks by filling the traps. Notably, the CPE becomes more heterogeneous in Cl 2 -SiPc/LuPc 2 heterojunctions compared to (345F) 2 -SiPc/LuPc 2 heterojunctions, which can be explained by the porous surface topography of the heterostructure forming a non-continuous interface, and whose electrical heterogeneity increases at a higher bias. The variations of effective capacitances of the heterojunction devices with DC bias are presented in Figure 5e,f, exhibiting the bulk capacitance (C1 eff ) in the range of pF, while interfacial capacitance (C2 eff ) is in the range of nF. Notably, interfacial capacitance is much higher (of the order 1000) than the bulk capacitance, which is commensurate with the thin interface region compared to the bulk, taking into account that capacitance is inversely proportional to the layer thickness. Previous reports for CuPc/F 16 CuPc bilayer heterojunction demonstrate that the interface region with opposite charges accumulation was much thinner than the bulk of the layer.

Ammonia Sensing Performance of Heterojunction Sensors
Both heterojunction devices were characterized for their NH 3 vapor sensitivity by exposing under 90 ppm of NH 3 for 10 min followed by recovery under synthetic air for 30 min. The current variations with time (I-t) (Figure 6a,b) exhibit a decrease in current under exposure, while an increase in current during the recovery period. Such behavior of the heterojunction sensors suggests p-type nature, taking into account the electron-donating tendency of NH 3 . It is important to note here that even though (345F) 2 -SiPc is known to exhibit n-type performance in organic thin film transistors, [33] in the present experimental conditions, particularly in air and combined with LuPc 2 in a bilayer heterojunction, it exhibits p-type behavior. This can be attributed to the hole injection from the LuPc 2 , which is facilitated by the higher work function of the R 2 -SiPc derivatives compared to LuPc 2 (Figure S2, Supporting Information). Moreover, the adsorption of O 2 and H 2 O when operated under ambient conditions leads to traps for the mobile electrons. [51] Notably, the baseline of both I-t curves demonstrated a slight drift, which is attributed to incomplete desorption of the NH 3 during the recovery step. Nonetheless, sensors recover more than 90% of their signal during the desorption period. Interestingly, the shapes of the response curves associated with adsorption and desorption phases present different profiles in Cl 2 -SiPc/LuPc 2 and (345F) 2 -SiPc/LuPc 2 heterojunction sensors. In Cl 2 -SiPc/LuPc 2 heterojunction sensors, a continuous decrease/increase in current is noticeable throughout the exposure/recovery cycles, while in the (345F) 2 -SiPc/LuPc 2 heterojunction sensors, a plateau in the I-t curve is observed in the exposure/recovery steps. These results imply that NH 3 vapor diffuses within the volume of the Cl 2 -SiPc/LuPc 2 layer, and accesses adsorption sites within the bulk, which is also expected taking into account its highly porous topography as confirmed by AFM imaging. On the other hand, NH 3 adsorption remains confined to the surface in (345F) 2 -SiPc/LuPc 2 bilayer heterojunctions, in which limited adsorption sites are available. Consequently, sorption equilibrium is attained immediately in the exposure/recovery cycles for (345F) 2 -SiPc/LuPc 2 devices. This sorption behavior of NH 3 on (345F) 2 -SiPc/LuPc 2 bilayer is also consistent with its smooth surface, as visualized by AFM. The implication of variable adsorption/desorption profiles of NH 3 on the two sensing layers is evident in the response/recovery times (time taken to change 90% of the current in the respective step) of the two sensors, which are estimated as more than 7 min and less than 1 min for Cl 2 -SiPc/LuPc 2 and (345F) 2 -SiPc/LuPc 2 devices, respectively.
To further evaluate the NH 3 sensing performances of the two sensors in a wide concentration range, they were subjected to short exposure/recovery cycles of 1 min/4 min at different NH 3 concentrations in the range of 10 to 90 ppm. Figure 7a,b depicts the response curves of Cl 2 -SiPc/LuPc 2 and (345F) 2 -SiPc/LuPc 2 sensors, respectively, exhibiting current decrease and increase under changing exposure/recovery cycles. Notably, the current variations are highly reversible during the exposure and recovery cycles and increase as a function of the increasing NH 3 concentration. Moreover, the sensors' responses at a given NH 3 concentration are highly repeatable, as similar variation in current is observed in four consecutive exposure/recovery cycles recorded  at each concentration. It confirms the good reproducibility of the two sensors in an extended range of NH 3 exposure.
To quantitatively assess the variations in sensor response based on R 2 -SiPc material selection, the relative response (RR) of the two sensors was calculated using Equation (2), and their variations as a function of NH 3 concentration, referred to as calibration curve, is depicted in Figure 7c,d for Cl 2 -SiPc/LuPc 2 and (345F) 2 -SiPc/LuPc 2 heterojunction devices, respectively. In the equation, I 0 and I f are the currents at the beginning and the end of the exposure period, respectively.
The calibration curves of the two sensors present different profiles, exhibiting a linear decay in the magnitude of their response Adv. Sensor Res. 2023, 2, 2200030 www.advancedsciencenews.com www.advsensorres.com from −1.28% to −15.5% for the Cl 2 SiPc/LuPc 2 heterojunction device, and an exponential decay in the response from −3.1% to −11.4% for (345F) 2 -SiPc/LuPc 2 heterojunction device, as NH 3 concentration increases from 10 to 90 ppm. The evolution of different calibration curves of the two sensors is attributed to different adsorption regimes, associated with the surface topography of the heterojunctions. The porous layer of Cl 2 SiPc/LuPc 2 imparts access to larger adsorption sites to NH 3 molecules, as they can diffuse within its volume, thus yielding a large relative availability of the adsorption sites for the increasing number of NH 3 molecules in the range of 10 to 90 ppm. Conversely, the compact topography of (345F) 2 -SiPc/LuPc 2 films provides a relatively limited number of adsorption sites, mainly on the surface, and restricts the NH 3 diffusion within the bulk of the sensing layer. Consequently, the limited number of surface adsorption sites are saturated with increasing NH 3 concentration, exhibiting a Langmuir-type adsorption isotherm. The two heterojunction sensors were also characterized under an NH 3 concentration in the range of 1 to 9 ppm, in which the (345F) 2 -SiPc/LuPc 2 -based sensor presented a stable and repeatable response ( Figure S3, Supporting Information). It is also notable that (345F) 2 -SiPc/LuPc 2 heterojunction sensors can experimentally detect NH 3 vapor at a concentration of 1 ppm. LOD is an important metrological parameter of a gas sensor, which is estimated for the heterojunction sensors using Equation (3). Here, N is the noise of the sensor signal, estimated by determining the standard deviation of the sensor blank signal (in synthetic air), S is the sensitivity given by the slope of the sensor calibration curve, and I 0 is the sensor baseline current.
The LOD of Cl 2 SiPc/LuPc 2 and (345F) 2 -SiPc/LuPc 2 heterojunction sensors are estimated as 100 and 310 ppb, respectively. The obtained LOD values of the sensors are below the environmental and occupational guidelines recommended for NH 3 exposure, making them suitable for applications in commercial and occupational settings. The obtained sensitivity and LOD are also either comparable or better than the previously reported heterojunction sensors (under similar operating conditions), as shown in Table 1.
Humidity can interfere with the real environment application of NH 3 sensors and can experience significant variation over short timescales. Therefore, the behavior of the two heterojunction sensors was investigated under changing relative humidity (RH) conditions. Figure 8 depicts the measured current of the two sensors over time as RH is increased from 30% to 60% and then decreased from 60% to 30%. The I-t curves of the two sensors revealed distinct features. In Cl 2 SiPc/LuPc 2 heterojunction sensor, the current continuously increases during the monotonic increase of RH from 30% to 60% and remains constant during the monotonic decrease of RH to 30% (Figure 8a). The increase of measured current in the sensor can be attributed to the trapping of mobile electrons in the sensing layer, facilitating the hole transport in the sublayer. [51] On the contrary, the (345F) 2 -SiPc/LuPc 2 heterojunction sensor reveals a rapid increase in current followed by a plateau-like feature during each step of RH increase and a similar decreasing pattern during monotonically decreasing in RH value. Different responses of the two sensors to changing RH can be correlated to the surface topography of the respective sensing layer, imparting different abilities to adsorb H 2 O. The highly porous layer of Cl 2 SiPc/LuPc 2 film results in the diffusion of H 2 O molecules in the bulk of the heterostructure, causing a large increase in the sensor current during monotonically increasing steps of RH change. However, H 2 O molecules remain trapped in the porous morphology of the film, causing the sensor current to stay constant during the decreasing cycles of RH. In the case of (345F) 2 -SiPc/LuPc 2 heterojunction sensor, H 2 O interaction remains confined to its surface because of its highly compact topography. Consequently, water adsorption/desorption cycles have a rapid and reversible effect on the sensor current. Moreover, fluorination of the phenoxy groups of the (345F) 2 -SiPc layer likely repels adsorbed water molecules more easily than Cl 2 SiPc, preventing H 2 O molecules from diffusing into the sublayer. Thus, the results suggest that the choice of axial functional groups in SiPc plays a significant role in the moisture sensitivity of the sensor. For the real-world implementation of these sensors, an association of a humidity sensor in the measurement device will be necessary to minimize the interference of RH variations.

Conclusion
We have demonstrated the application of silicon phthalocyanines (Cl 2 -SiPc and (345F) 2 -SiPc) in thin film bilayer heterojunction sensors for NH 3 detection for the first time. The choice of axial groups in R 2 -SiPcs imparts different thin film morphologies and electronic properties, which strongly influence the interfacial charge transport in the associated heterojunction device and sensing performance toward NH 3 . The bilayer heterostructures of R 2 -SiPc derivatives in combination with LuPc 2 revealed characteristic absorption bands and vibrational Raman peaks of phthalocyanines, confirming that each macrocycle preserves its electronic identity in the heterostructures. The two heterostructure films exhibited distinct surface topography as revealed by the AFM imaging, such that Cl 2 -SiPc/LuPc 2 films had a highly porous and rugged surface, while the (345F) 2 -SiPc/LuPc 2 film possesses a compact surface. Different topographies of the two heterostructures have a measurable impact on their electrical and charge transport properties. Accordingly, Cl 2 -SiPc/LuPc 2 heterojunction devices exhibit a smaller U th and faster interfacial and bulk charge transport than (345F) 2 -SiPc/LuPc 2 . Both sensors exhibit p-type behavior upon exposure to NH 3 vapor, revealing a quasi-reversible response in the alternate exposure/recovery steps. However, (345F) 2 -SiPc/LuPc 2 reveals a short response/recovery time (less than 1 min), while Cl 2 -SiPc/LuPc 2based sensor presents a long response/recovery time (more than 7 min). Moreover, the calibration curve of the Cl 2 -SiPc/LuPc 2based sensor is linear, while the one of (345F) 2 -SiPc/LuPc 2 is exponentially decaying, in a range of NH 3 concentrations of 10 to 90 ppm. The different sensing properties of the two heterojunction sensors have been correlated to different regimes of surface adsorption and bulk diffusion, associated with their surface topography. Both sensors can detect NH 3 vapor in the sub-ppm range, making them suitable for implementation in real environmental conditions.
Sensor Fabrication and Electrical Measurements: Organic heterojunction sensors were fabricated on interdigitated electrodes (IDEs), lithographically patterned on a glass substrate. The IDE consisted of 16 pairs of ITO digits, having a width of 75 μm and spacing between digits of 75 μm. Electrodes were properly cleaned through stepwise ultrasonication in CH 2 Cl 2 , acetone, ethanol, and water, followed by drying in an oven for 1 h at 100°C. The bilayer heterostructures were coated on the IDE by thermal evaporation in UNIVEX 250 thermal evaporator under a secondary vacuum (≈10 −7 mbar). During the evaporation, 50 nm-thick layers of R 2 -SiPc and LuPc 2 were sequentially deposited on top of the electrodes at a rate of ≈1 Å s −1 . The sublimation temperature for Cl 2 -SiPc and LuPc 2 were noted in the range of 475-495°C and 400-450°C, respectively, while (345F) 2 -SiPc was deposited in the range of 390-410°C. A glass plate was also placed next to the IDE, in order to achieve identical bilayer heterostructures, to be used for different surface and microstructural characterizations. I-V characterization and sensing measurements of the heterojunction sensors were performed at room temperature (18-22°C) using a Keithley 6517B electrometer, at bias ranging from −10 to +10 V in steps of 0.1 V. Applied bias was started and ended at 0 V to avoid any polarization effect. Impedance spectroscopy measurements were performed on a Solartron SI 1260 impedance analyzer. The interface of the bilayer heterojunctions was studied over a wide range of frequencies (from 10 Hz to 10 MHz) with fixed AC (200 mV) and varying DC bias from 0 to 10 V. Obtained data were treated using Zview software from Ametek.
Characterization Techniques: Electronic absorption spectra of different R 2 -SiPcs, LuPc 2 , and the bilayer films were registered using a Cary 50 (Varian) UV-vis spectrophotometer, between 280 and 800 nm. Raman spectral measurement of different films was conducted using Renishaw inVia Raman microscope, utilizing a 473 nm laser as an excitation source.
The surface topography of the bilayer heterostructures coated on glass was evaluated by AFM imaging, on Bruker Icon 2 nanoDMA equipment by employing nanoDMA peak force tapping mode. Scan of the surface was performed using a silicon probe (DNP SA or ScanAsyst-HR) (tip radius <5) at a peak force frequency of 2 kHz and amplitude in the range of 50-100 nm. The images were finally analyzed by Nanoscope analysis software.
Gas Sources and Sensing Experiments: Ammonia (NH 3 ) gas, at 985 and 98 ppm (mol/mol) in synthetic air was used from standard commercial cylinders, purchased from Air Liquide, France. The total flow was in the range of 0.5-0.55 NL min −1 depending on NH 3 concentration and the volume of the test chamber was 8 cm 3 . Gas flow in different fluidic lines was controlled by mass flow controllers and electronic valves, which were interfaced with homemade software. [55] Detailed configurations of the measurement setup are reported in our previous work. [50] Pre-Processing of Data: UV-vis and Raman raw data were normalized and compared on the same axis by OriginLab software. Analysis of AFM images and calculation of surface roughness were performed by Nanoscope software. Nyquist plots from impedance measurements were analyzed by Zview software and comparison graphics were made by OriginLab.
Data Presentation and Sample Size: The roughness of the surface was estimated by taking the average over the entire image area by Nanoscope software and is presented with obtained standard deviation (sample size: 3). Sensor calibration curve is presented by taking an average of the four responses at each concentration. The standard deviation at each data point is included. For the impedance analysis, Z-View simulation was performed three times on each experimental dataset.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.