Influence of thickness of SiO
 2
 layer on the performance of SINW sensors

In recent decades, silicon nanowire, one of the most widely used 1D nanomaterials, has been considered as the promising sensitive material with its unique advantages in the sensor area [1–5], owing to its distinct mechanical, electronical and optical properties [6–10]. The ultrahigh surface to volume ratios leads to the super-sensitivity and extremely low detection limits of silicon nanowires [11, 12]. Moreover, silicon nanowire can be easily modified by various biological and chemical molecules that can be used to specifically detect target molecules, including a wide range of chemical and biological species [13–15]. In addition, it is compatible with the scale of target biochemical molecular for silicon nanowires with diameters less than 100 nm, whose unique advantages in biochemical detection is given to silicon nanowires [16, 17]. In recent years, the applications of silicon nanowires in various fields have been widely studied, including solar cells, photodetectors, resonators, thermoelectric generator and so on [18–22]. Especially, silicon nanowires are designed as biosensors for the detection of various targets, which could detect nucleic acids, proteins, virus, and chemical species [23–28]. Owing to the application of MEMS technology, the silicon nanowires devices obtained advantages like small size, low price and easy to batch fabricated [29–31]. However, some drawbacks existed in the preparation process, in which the silicon nanowires were wrapped by silicon oxide due to selflimiting oxidation method [32–34]. Since silicon nanowires were employed as sensitive elements of sensors, the silicon nanowires must be exposed by etching away the silicon oxide using the buffered oxide etch (BOE) solution before being used for tests [35, 36]. This corrosion process, which was difficult to precisely control, would greatly influence the performance of silicon nanowire sensors. The corrosion process of silicon nanowire devices could be recognized as one of the most important challenge to control the silicon oxide layer thickness. In order to ensure high performance of devices for subsequent test experiments, this influence of silicon oxide thickness need to be investigated carefully. In this paper, we have investigated the influence of BOE etching process, i.e. silicon oxide layer thickness, on silicon


INTRODUCTION
In recent decades, silicon nanowire, one of the most widely used 1D nanomaterials, has been considered as the promising sensitive material with its unique advantages in the sensor area [1][2][3][4][5], owing to its distinct mechanical, electronical and optical properties [6][7][8][9][10]. The ultrahigh surface to volume ratios leads to the super-sensitivity and extremely low detection limits of silicon nanowires [11,12]. Moreover, silicon nanowire can be easily modified by various biological and chemical molecules that can be used to specifically detect target molecules, including a wide range of chemical and biological species [13][14][15]. In addition, it is compatible with the scale of target biochemical molecular for silicon nanowires with diameters less than 100 nm, whose unique advantages in biochemical detection is given to silicon nanowires [16,17]. In recent years, the applications of silicon nanowires in various fields have been widely studied, including solar cells, photodetectors, resonators, thermoelectric generator and so on [18][19][20][21][22]. Especially, silicon nanowires are designed as biosensors for the detection of various targets, which could detect nucleic acids, proteins, virus, and chemical species [23][24][25][26][27][28].
Owing to the application of MEMS technology, the silicon nanowires devices obtained advantages like small size, low price and easy to batch fabricated [29][30][31]. However, some drawbacks existed in the preparation process, in which the silicon nanowires were wrapped by silicon oxide due to selflimiting oxidation method [32][33][34]. Since silicon nanowires were employed as sensitive elements of sensors, the silicon nanowires must be exposed by etching away the silicon oxide using the buffered oxide etch (BOE) solution before being used for tests [35,36]. This corrosion process, which was difficult to precisely control, would greatly influence the performance of silicon nanowire sensors. The corrosion process of silicon nanowire devices could be recognized as one of the most important challenge to control the silicon oxide layer thickness. In order to ensure high performance of devices for subsequent test experiments, this influence of silicon oxide thickness need to be investigated carefully.
In this paper, we have investigated the influence of BOE etching process, i.e. silicon oxide layer thickness, on silicon  [4,37,38]. SEM and I-V characteristics were used to find the relationship between the performance and the morphological change during the etching process. It was found that the thickness of silicon oxide layer enclosed nanowires was gradually reduced until completely etched in few minutes to expose the silicon nanowires. Only when the silicon oxide thickness was close to zero, the sensing characteristics could be acquired by silicon nanowire, which provide an effect way to control the fabrication process of high performance devices in subsequent biochemical tests.

FABRICATION
Our silicon nanowires device was fabricated by employing the MEMS technology, which possessed various advantages like small size, low prize and easy to batch fabricated [39,40]. The manufacturing process of our silicon nanowires device was shown in Figure 1(a). The device was fabricated on a (111) silicon-on-insulator (SOI) wafer with 10 μm top silicon layer. First at all, a 100 nm silicon nitride layer was deposited on the wafer by LPCVD technology. Designed etched windows were transferred to silicon wafer by photolithography, and the exposed silicon nitride was etched using reaction ion etching (RIE) to form designed rectangular windows arrays. Then the exposed top silicon was removed by deep reaction ion etching (DRIE), taking shape the etched rectangular cavities arrays. In controllable anisotropic etching solution such as KOH, silicon walls arrays were appeared among the rectangular etched cavities arrays. The size of the silicon nano-walls can be controlled by appropriate etching time. After the self-limiting oxidation process of silicon nano-walls arrays [32][33][34], silicon walls would convert to silicon oxide walls with small portion of silicon remained. The residual silicon formed silicon nanowires with an inverted-triangle shaped cross-section, which located on the top centre of silicon oxide walls. The silicon nanowires were wrapped by silicon oxide, which protected from possibly damage in subsequent processes. The doping windows were FIGURE 1 Fabrication, modification and response mechanism of silicon nanowire devices etched at designed source and drain locations, and boron was doped at the windows position to form good contact between silicon and metal. Subsequently, the source electrodes, drain electrodes and top-grid electrodes were deposited on the respective locations by magnetron sputtering technology and metal corrosion, before the isolation grooves were fabricated to insulate the source and drain using RIE and DRIE process. Finally, the device was manufactured completely after removing silicon dioxide walls to expose silicon nanowires.
a. Schematic illustration of our silicon nanowire device fabrication process. (i) A 100 nm Si 3 N 4 layer was deposited on the wafer by LPCVD technology. (ii) The designed Si 3 N 4 windows were etched by RIE and the exposed silicon was etched by DRIE. (iii) The silicon oxide walls were fabricated by controlled KOH corrosion, and the silicon nanowires were prepared by self-limiting oxidation method. (iv) The doped locations were exposed and were doped by boron, and the electrodes were deposited. (v) The isolation grooves were fabricated by RIE and DRIE. (vi) The silicon nanowires were exposed by etched away the silicon oxide b. The surface modification process of our silicon nanowires c. Response mechanism of silicon nanowires to pH In particular, the silicon nanowires must be modified to connect specific probes before detection. To detect pH solutions, a two-step modified method was used to modify nanowires in Figure 1(b) [4,38,41]. At first step, the silicon nanowires were treated by oxygen plasma. On one hand, the nanowires can be cleaned by plasma, on the other hand, hydroxyl groups were modified on nanowires surface, which produced a large amount of silanol termination (Si OH). The second step, the devices were immersed in 2% 3-aminopropyltriethoxysilane (APTES) solution for hours, so that the silanol groups can bind to amine groups ( NH 2 ).
The response mechanism of pH detection was shown in Figure 1(c), when pH solution was added dropwise to silicon nanowires, for acidic pH solutions, hydrogen ions would combine with the amino groups on silicon nanowires. This process can be called "protonation", which made uncharged amino groups to positive charged ammonia ion and decreased the conductance of p-type device. Similarly, for alkaline pH solutions, hydroxide ions would combine with the hydroxyl groups on silicon nanowires. This process can be called "deprotonation", which made the uncharged hydroxyl groups to negative charged oxygen ions and increased the conductance [42]. In addition, these two reactions are reversible.

EXPERIMENTS
The morphology of silicon nanowires device was characterized by SEM. The images of completed device and silicon nanowire were shown in Figure 2(a,b), displaying that the nanowire diameter was about 73.7 nm. The small size of silicon nanowires device was exhibited by the SEM images. And it demonstrated the characteristic of easy to batch fabricated that more than 800 devices were prepared on one silicon wafer. The working mechanism of silicon nanowires device was the field effect principle, in which nanowires working as conductive channel were modulated by the gate voltage to control their state [43,44]. The device modulation characteristic of gate voltage to source-drain current was exhibited in Figure 2(c), revealing the p-type field effect property.
In Figure 2(d) and inset, the I-V characteristics before corrosion were relatively similar, but there appeared obvious difference after BOE corrosion, which revealed the significant influence of BOE corrosion on the performance of silicon nanowires device. Consequently, the influence of corrosion on device performance would be investigated by sensing capability, morphology and electrical properties.
There was no additional gate voltage which was applied to the silicon nanowire sensor. The gate voltage of the silicon nanowire was entirely derived from the changes in the electric potential generated by the surface molecular modification or the combination with the target molecule. When the silicon oxide thickness was very thick, the gate voltage would be shielded by silicon oxide, then the current of silicon nanowire was difficult to be modulated by gate voltage. The shielding effect of silicon oxide on the gate voltage would decrease as the silicon oxide thickness becomes thinner. When the thickness of silicon oxide was very thin, the gate voltage will obviously modulate the current of silicon nanowires. To realize the modulation of silicon nanowires by the gate voltage, the silicon oxide outside silicon nanowires needed to be corroded in the preparation process, and the change of silicon oxide during the corrosion process should be studied.
In order to study the influence of BOE etching on device performance, its effect on sensing capabilities should be focused firstly. The pH standard solutions were selected to test the performance of silicon nanowire devices in different corrosion time. Standard solutions with pH of 4 and 7 were repeatedly added to the devices to test the responses. Silicon nanowires devices that had been etched for different time were used to perform this detection experiment. The different responses for this series of devices could be obtained by measuring the current change after contacting pH solution, which may demonstrate the influence of corrosion process to device performance.
The structure or morphology of silicon nanowires was an intuitive representation of etching results and could be used to describe the changes in corrosion process, which can be observed by SEM images. However, the silicon nanowires are hidden under silicon nitride film and gate electrode. To observe silicon nanowires, the silicon nitride film and the metal must be etched away. However, it is difficult to complete corrosion without damaging the devices due to its small size. In order to solve these difficulties, the silicon nitride film was etched away after self-limiting oxidation. These completed devices were divided into different groups and were etched in different time. Afterwards, the morphologies of these etched devices were characterized by SEM images to reflect the influence of corrosion process.
The device I-V characteristic was defined by the change of source-drain current with voltage when the gate voltage was zero. The corrosion influence on silicon nanowires electrical properties was also investigated to supplement by measuring the devices I-V curves before and after different time corrosion.

RESULTS AND DISCUSSIONS
The response to pH can be used to represent the sensing capability of silicon nanowires device. The pH solutions responses of devices that had been etched in different time were exhibited in Figure 3 and Figure S1 in the Supporting Information, and it can be easily seen that the responses of devices etched for less than four minutes were almost negligible. However, there were significant responses to pH solutions in devices that etched for more than five minutes. The obvious response that occurred in devices etched for five minutes indicated that the silicon nanowires had undergone a decisive change during four to five minutes etching time. Therefore, the transformation of silicon nanowires at this stage, which was the key to explain the difference in pH test results. The etching rate of silicon oxide in BOE solution was about 100 nm per minute, and the etching rate of silicon in BOE solution was much slower than that of silicon oxide. In the early stage of the etching process, the silicon oxide was not completely etched, and the silicon nanowires were protected from the corrosion by the silicon oxide. When the etching time exceeded 5 min, the silicon nanowires were exposed to the etching solution. However, due to the extremely slow silicon corrosion rate of BOE solution, the main structure of silicon nanowires would not be significantly damaged in a short etching time. At this time, the BOE solution may corrode the defects on the surface of the silicon nanowires. When the defects were completely etched, even if the etching time was extended, the silicon nanowires were less affected by the corrosion.
The morphological changes of silicon nanowires device were an important way to describe this stage and explain their differences, which was observed by SEM. The structural changes of silicon nanowires in etching process were illustrated in the SEM images of Figure 4 and Figure S2 in the Supporting Information, which were the top view of silicon nanowires. In uncorroded state, the silicon nanowires were wrapped in silicon oxide walls. When the devices were immersed into BOE solution, the silicon oxide walls were gradually thinned as corrosion time increased, revealing that the silicon oxide was being etched. And the silicon oxide was etched totally in about five minutes, when the silicon nanowires were completely exposed at this time. These results The I-V characteristics before and after corrosion were measured to supplementally describe the influence of etching. By comparing the curves before and after corrosion, the I-V characteristics did not change much after less than four minutes corrosion, until them were started to change from being etched for five minutes and continued to change up to saturate in fifteen minutes, which was similar to the change trends of pH tests and SEM characterization. With the help of SEM, it was found that the thickness of silicon oxide layer was changed during the BOE corrosion process. According to this result, the influence of corrosion on devices could be recognized as the change of silicon oxide layer thickness. Hence, the effect of corrosion was considered as the influence of silicon oxide thickness on the performance of device. So, the different response of the devices could be explained by the difference of the silicon oxide thickness caused by the different corrosion time.
The relationship between pH response and silicon oxide thickness could be explained by combining the results of pH test and SEM characterization. For non-corroded devices, silicon nanowires were wrapped by thick silicon oxide layer, and the process of modification and reaction would take place on the surface of silicon oxide layer. Since the surface was too far from nanowires, the potential change caused by reaction was difficult to affect the carrier concentration of silicon nanowires, leading to no response to be detected. The thickness of silicon oxide was still thick for devices etched short time, and their responses were similar to the results of uncorroded devices. When etched for about five minutes, the silicon nanowires were exposed after silicon oxide was etched totally. At this time, the reaction would occur on or near the surface of nanowires, where the nanowires conductance could be obviously modulated by the external electrical signal. As a result, the exposed silicon nanowires with longer etching time demonstrated significant responses. The electrical signals of silicon nanowires could be changed obviously by external stimuli only when the silicon oxide almost disappeared. It revealed the significant influence of silicon oxide thickness on the device response.
Subsequently, the thickness of silicon oxide layer and response of pH solutions in different corrosion time were shown in Figure 5 to more intuitively show their relationship. The influence of silicon oxide thickness on the performance of silicon nanowires could be explained more clearly by this graph, which affected the response of silicon nanowires to external signals. Only when the silicon oxide thickness was close to zero, the sensing characteristics could be acquired by silicon nanowire that can act as a transducer. It revealed the influence of silicon oxide thickness on the device performance.
In order to more clearly show the effect of corrosion on I-V characteristics, the measured curves were analysed by data fitting. Since the Fermi level distribution of semiconductor was related to the Boltzmann exponential function e − E k B T , so the Boltzmann distribution formula I = I − −I + 1+e V − + I + can be used to fit the I-V characteristics curves, which could more intuitively demonstrate the change trend with corrosion time by parameters analysis. In formula, I+ and I− distributions were the positive and negative saturation currents, and α and β were the relevant parameters of device itself. The change trend of these four parameters at different etching times was exhibited in Figure 6. The significant changes for these parameters basically occurred after five minutes corrosion and increased obviously when the corrosion time continued to increase. It can be considered that the critical point of silicon nanowires corrosion was approximately five minutes, which was consistent with the results of pH tests and morphological characterization.
The response of silicon nanowires to pH solutions are used to characterize the influence of silicon oxide thickness on the sensing properties of silicon nanowires. It compares the change of sensing performance with the thickness of silicon oxide. However, the response of silicon nanowires to pH solutions is difficult to represent their actual sensing performances. Therefore, the sensing performances of silicon nanowires could be demonstrated by testing the response of quantitative target substance. In our reported work, we detected the response of silicon nanowires to TNT solutions and compared it with other detection methods [45]. It can be seen that silicon nanowires showed the low limit of detection concentration of TNT, indicating the advantages of silicon nanowires in sensing performances.

CONCLUSION
In conclusion, we have studied the influence of SiO 2 thickness on the performance of silicon nanowires in details. With the help of pH detection, the discernable responses were acquired in the devices that had been etched for more than five minutes. The morphological and electrical properties changes of silicon nanowires were characterized by SEM and the I-V characteristics during BOE corrosion, indicating that the silicon oxide surrounding nanowires was slowly etched and thinned in etching solution and basically etched totally in about five minutes. The effect of corrosion on silicon nanowire devices could be recognized as the change of silicon oxide thickness. Therefore, the effect of corrosion could be equivalent to the influence of silicon oxide thickness on device performance. It provides an effect way to control the fabrication process of high performance devices. However, there are some challenges of silicon nanowire sensors needed to be concerned, such as miniaturization and integration of devices, influence of complex samples with ionic strength, and long-term stability. The performance of silicon nanowire sensors could be improved from the targeted aspects in future research, including the integration of test and signal output system on chip, novel sensing methods to avoid the influence of ionic strength, and the like. Silicon nanowire sensors will have broad and exciting prospects in the field of biomedical and hazardous substances detection, and even realize commercial application in the near future, which will widely applicate for clinical diagnosis, early screening for disease or virus, trace explosive detection and so on.

MATERIAL AND EQUIPMENT
The SOI wafers were purchased from OKMETIC Inc. The LPCVD equipment is DF550-4 of Semco Inc. The RIE equipment is RIE-101iPH of Samco Inc.