ZnO Nanowire Field Effect Transistor for Biosensing: A Review

The last 19 years have seen intense research made on zinc oxide (ZnO) material mainly due to the ability of converting the natural n-type material into p-type. For a long time, the p-type state was impossible to attain and maintain. The review focuses on ways of improving the doped ZnO material which acts as a channel for nanowire field effect transistor (NWFET) and biosensor. The biosensor has specific binding which is called functionalisation achieved by attaching a variety of compounds on the designated sensing area. Reference electrodes and buffers are used as controllers. Top-down fabrication processes are preferred over bottom-up because they pave way for mass production. Different growth techniques are reviewed and discussed. Strengths and weaknesses of the FET and sensor are also reviewed.


Introduction
Zinc Oxide (ZnO) material has been known as a semiconductor for over seventy (70) years, with some of the first literature being reported as early as in 1944 [1]. It was never put to use like other semiconductors (GaN, Si) because it is difficult to dope. The past nineteen (19) years have seen a revival on the research and use of material because of new and emerging ways of doping it. The material is naturally n-type [1]- [4] and by controlling the conditions of growth, the donor concentration can be controlled. The growth conditions include: temperature, diethyl zinc (DEZ) reactant, O2 or H2O reactant and pressure. P-type material [1]- [4] is difficult to grow and tends to slowly revert back to n-type. Researchers [5]- [14] that managed to deposit the p-type material have shown that it converts back to n-type within a few days. Maximum time period shown on p-type ZnO was a few months [5]- [14].
ZnO has practical advantages that make it an attractive semiconductor from an industrial point of view. It has low costs, abundant, non-toxic, transparent, large excitonic binding energy of 60 meV, soluble, compatible with intercellular material, and wide and direct band gap of 3.37 eV making it highly sensitive. It is well known that semiconductors have a small band gap which allows switching between conduction and off-states. The larger the band gap, the better is the semiconductor able to switch states and to insulate leakage currents. Band gap affects sensitivity because a device that possesses a wider band gap allows for higher currents to travel but also preventing leakage currents which results in more sensitive and accurate readings. With low-temperature fabrication processes, high-quality devices can be fabricated using conventional processing technology, thereby making it suitable for low cost mass-production. It has potential applications in opto-electronics, transparent electronics, and spintronics. ZnO and its alloys has versatile electrical and optical properties for applications in thin film or nanowire transistors, light emitters, biosensors and solar cells. The nanowire biosensor has a high surface-to-volume ratio, enabling real time and label-free detection [1]- [4], [15]- [17].
Currently, the main commercial application for ZnO (and/or IGZO) material is in displays, with companies like Sharp and Samsung putting IGZO into mobile phone displays [18]- [20]. IGZO displays out-perform other semiconductor displays such as amorphous silicon and organic semiconductors by providing improved resolution and reduced power consumption. This is possible because IGZO has a x20 to x50 times higher mobility than amorphous silicon and polymers, which allows for device scaling without affecting performance [18]- [20]. Higher mobility values can also be achieved with amorphous silicon technology, but it needs to be laser annealed which is expensive.

Growth Techniques of ZnO
ZnO films can be grown using three methods: gas transport (vapour-phase deposition), hydrothermal synthesis, and/or melt process. Melt growth techniques are a problem due to high vapour pressure of ZnO. Growth using gas transport is difficult to control for large film layers and is normally used for bottom-up ZnO nanostructures. Hydrothermal synthesis is therefore preferred as a method of growth. Thin films can be produced through: chemical vapour deposition, metalorganic vapour phase epitaxy, electrodeposition, pulsed laser deposition, sputtering, sol-gel synthesis, atomic layer deposition, spray pyrolysis, etc. All the mentioned techniques fall under hydrothermal synthesis and one of the preferred methods is atomic layer deposition (ALD). The ALD process is capable of producing highly conformal and quality films [21]. The process is cyclic and is based on the number of reactants. Fig. 1 shows that the ALD process for ZnO films is cyclic and depends on two reactants: metallisation and oxidation. Fig. 1 Schematic diagram illustrating a single cycle of ZnO deposition using the ALD tool [42] Metallisation uses diethyl zinc (DEZ) as the zinc (Zn) metal precursor. Purge and pump steps are used to separate the execution of the reactants and to remove by-products. Before deposition: the wafer (substrate) is pre-heated at a temperature that will be used for deposition and it is also cleaned with O2 plasma so as to remove any polymer layer. During the metallisation step, the DEZ (Zn (C2H5)2) is absorbed onto the surface of the wafer and the residual Zn (C2H5)2 is removed from chamber. 'R' in Fig. represents C2H5.Then on another step, water or O2 is delivered to react with the absorbed DEZ [22]- [24]. These steps are executed separately and to ensure this, purge steps are introduced in between the steps.
When water is used instead of O2 for oxidation, the process is called thermal ALD. This process tends to produce films similar to chemical vapour deposition (CVD) techniques [24], [43], [44]. When O2 is used instead of water, then the process needs plasma energy. Remote plasma atomic layer deposition (RPALD) is a fairly new process which is why it is still not in used. It is better than the other deposition techniques as it tends to produce films close to epitaxial layers. The layers are crystalline but tend to be non-uniform to the underlining layer which is why they are not called epitaxial layers. It is a process with great potential for depositing highly conformal and quality films. The process is better than thermal ALD in terms of conformity and quality, but both processes do not generally produce epitaxial layers due to non-uniformity to the underlining substrate. The plasma assisted ALD method has the following advantages: reduction of OH impurity, allows more freedom in processing conditions and provides wider range of material properties. The OH impurity is not desired as it affects the conductivity of the semiconductor and induces defects in the dielectrics. Table 1 compares various growth techniques and how they affect NWFET output characteristics. Chemical vapour deposition (CVD) is the most popular technique for bottom-up nanowire processes. There are two growth techniques classified under CVD which are vapour-liquid-solid (VLS) and vapour-solid (VS) deposition techniques. CVD normally give the highest mobility as they produce crystalline wires with the only flaw being from the catalysts that guide the growth. VS produces better quality nanowires than VLS as it uses no catalysts but instead uses very high temperatures (> 900 °C). The problem with VS is that it is usually harder to control the size and morphology of the nanowires. Table 1 also shows that atomic layer deposition (ALD) is an attractive technique because it deposits high quality films at low temperatures between 120 °C and 210 °C [42], [45]. The problem with ALD is that it has only this window for good quality conducting films. At temperatures below 120°C the deposition can be incomplete or experience condensation depending on growth rate. At temperatures above 210 °C the deposition tends to experience desorption or it decomposes towards CVD deposition. Nonetheless, it is one of the best techniques toward growing films close to epitaxial growth (crystallinity is achievable whereas uniformity is still difficult to achieve) [42], [45]. The tool has shown potential by achieving high values of field effect mobility > 30 cm 2 /Vs with excellent crystallinity.

Native point defects
There are three (3) types of defects in a crystal lattice: point defects, area defects, and volume defects. Point defects which are caused by native elements and impurities are the major problem for ZnO semiconductor. Native point defects for ZnO include the following: zinc interstitial (Zni), zinc antisite (Zno), zinc vacancy (VZn), oxygen interstitial (Oi), oxygen antisite (OZn), and oxygen vacancy (Vo). Over the years, a lot of research advocated them as the major cause for the n-type behaviour. Oxygen defects are seen as the main contributors toward the n-type behaviour [3], [15]. There are some researchers [1]- [4] who hypothesise that impurities (not the native point defects) are the main cause of the n-type behaviour because they tend to be shallow donors whereas Zn and O2 defects tend to be deep donors [1]- [4]. The two theories have not been proven so currently the main cause of the natural n-type behaviour of ZnO [1]- [4] is not certain.  Donor Impurities for the n-type ZnO can either be shallow or deep. Fig 2 shows shallow donors compared to deep donors. Shallow impurities require little energy to ionize (this is energy typically around the thermal energy or less). These donor impurities possess energy close to the band edgethe extra valence electron of these impurities are loosely bound and occupy effective-mass states near the conduction band maximum-CBM-at low temperatures. Deep impurities on-the-other-hand require energy greater that the thermal energy to ionize. These donor impurities possess energy far from the band edge (CBM) making them very hard to ionize. Their presence within the semiconductor tends to contribute only a small fraction of free carriers. Deep donors are also called traps because they act as effective recombination centres in which electrons and holes fall and annihilate each other. Grain boundaries (GB) are main source of deep state impurities and they adversely affect transistor performance. ZnO is a wide Bandgap material and research suggests [3], [4], [16]

Top-down fabrication of ZnO nanowire FETs
There are four main methods capable of producing nanometer features using top-down approaches: UV stepper lithography, e-beam lithography [68], focused ion-beam lithography [69] and spacer method [45], [70]. UV lithography is the standard industrial method for fabricating nano devices. Ebeam and focused ion-beam lithography are often used and can pattern devices down to 5 nm, but the equipment is very expensive and the pattern writing is very slow. These two instruments resemble scanning electron microscope (SEM) in terms of operation. Whereas SEM is used to focus a beam of electrons to image samples within a chamber, these instruments are used to create patterns on the samples. The difference between e-beam and focused ion-beam is that the latter uses an ion beam to pattern wafers and hence does not require photoresist. Their advantage over optical UV lithography is the small features they reach. For low cost applications such as biosensors, the problem with these two methods is that they are expensive.
The spacer technique is a low cost fabrication method for fabricating nanowires. It was first reported in 2005 by H. Ge, et al., [71] and other researchers [41], [72], [73] have since carried it forward. The technique has great potential in shaping nanometer features using conventional, low cost photolithography. Fig. 3 shows the concept of the spacer technique. It uses first anisotropic etch to create a vertical pillar on an insulating layer (SiO2), then after deposition of a semiconductor layer (ZnO) and a second anisotropic etch to create nanowires made up of the semiconductor layer. This method allows nanowire features with controllable dimensions to be developed. The ICP tool is usually used for anisotropic etching and produces surface roughness less than < 1.5 nm. Other tools such as RIE and ion beam etch produce roughness greater than > 5 nm. The fabrication process for the complete ZnO NWFET structure is as outlined in [134].

Background on FETs
The ZnO field effect transistor (FET) has been around for decades. The success of the device in meeting the technological demands has largely been dominated by the shrinking size of its physical geometry. It has an advantage as a junctionless (no p-n junctions) FETs compared to conventional FETs [17], [21]- [24], [43], [44], [46], [47]. There has been an introduction of new materials and heterojunction structures developed so as to move away from conventional silicon devices. High-K dielectrics have been introduced to replace the conventional SiO2 which should help maintain acceptable dielectric thicknesses while keeping gate leakage currents low [17], [21]- [24], [43], [44], [46], [47].
Even with so many improvements being made to the device, the limits of FET scaling are approaching. The thickness of the oxide (tox) cannot be less than 1nm due to high tunnelling current and significant operational variation. The substrate doping is also very high which creates leakage and tunnelling currents that are unacceptable to device operation.

ZnO thin film transistors (TFTs)
TFTs have also been fabricated using ZnO, mainly thin film transistors for application in displays. Fig. 4 compares twenty ZnO TFTs fabricated by different authors [44], [46], [48]- [65] using a variety of fabrication methods over the last five years. The graph is a plot of field effect mobility versus subthreshold slope which are two of the main parameters that describe the performance and efficiency of a device. The best device was fabricated by Bayraktaroglu, et al. [64] with a SiO2 insulator and pulsed laser deposited ZnO active channel layer. The device had a field effect mobility 110 cm 2 /Vs and an excellent subthreshold gate voltage swing of 109 mV/decade. This value of mobility is much higher than the value of around 1 cm 2 /Vs that is typically achieved with amorphous silicon TFTs in production displays. It is clear therefore that ZnO TFTs have considerable potential for application in high performance displays.

Nanowire field effect transistors (FETs)
Emerging non-planar devices [17], [21] are being researched to prolong the future progress for FETs. Devices based on quasi-one-dimensional nanostructures are still at an embryonic stage from an industrial point of view. These nanostructures include the following: nanowires, nanobelts, nanoribbons, and nanoneedles [66], [67]. This review is interested in nanowire FETs which are also being researched for application in biosensors because the high surface to volume ratio provides high sensitivity.

Comparing ZnO NWFETs
Fig. 5 compares fifteen (15) different ZnO NWFETs fabricated by different authors using a variety of methods [42], [74]- [86]. The graph is plotted with field effect mobility against the subthreshold slope which are two important device parameters that determine ZnO NWFET performance. The nanowires were fabricated using top-down and bottom-up (self-assembled) processes. Self-assembled processes tend to display very high field effect mobility which is normally above 200 cm 2 /Vs; whereas the top-down have lower mobility values. Most of the top-down fabricated devices have mobility < 1.0 cm 2 /Vs with around three papers giving a mobility > 10.0 cm 2 /Vs. The difference in the mobility may be due to the fact that self-assembled nanowires are single-crystal, whereas topdown nanowires are polycrystalline. Nonetheless, top-down techniques are desirable as they currently pave way for mass production and will be pursued in this research investigation.

Biosensors
A biosensor is defined by the International Union of Pure and Applied Chemistry (IUPAC) as "a self-contained integrated device that is capable of providing specific quantitative or semi-quantitative analytical information using a biological recognition element (biochemical receptor), which is retained in contact direct with a transduction element" [87]. A biosensor is a "more-than-Moore device" because it incorporates functionalities that do not necessarily scale according to Moore's law.
Under the roadmap, the device falls under the category of sensors and actuators. Other categories include analogue/RF, passives, HV power and biochips [88], [89]. Typical structure of a biosensor. The bio-molecules are contained within an analytic solution that attach themselves to immobilised enzymes or immune-agents on the receptors. The transducer then converts the energy signal produced into an electrical signal which is then transmitted for data processing [42] Fig . 6 shows a typical structure of a biosensor [90]- [92]. The bio-molecules are contained within an analytic solution that attach themselves to immobilised enzymes or immune-agents on the linkers. Linkers in turn are attached to the transducer. The transducer then converts the charge on the analyte into an electrical signal which is then transmitted for data processing. Biosensors can be considered as part of the research field known as 'chemical sensors' in that a biological mechanism is used for analyte detection within an analyte solution [93]- [95]. Quasi-one-dimensional nanostructures have a greater surface-to-volume ratio compared to planar structures and are therefore expected to be more sensitive than planar sensors [93]- [95].
Nanowires are the same as nanorods. The words can be used interchangeably [80]. These have received enormous attention due to their suitable properties for designing novel nanoscale biosensors. For example, the dimensions of ∼1-100nm are similar to those of many biological entities, such as nucleic acids, proteins, viruses, and cells [79]. In addition, the high surface-to-volume ratios for nanomaterials allows a large proportion of atoms in the bio-analyte to be located at or close to the surface. Moreover, some nanowire materials have surfaces that can easily be chemically modified which makes them significant candidates for biosensors [79], [80]. There are a number of nanostructure-based electrical biosensors which include: single-wall carbon nanotubes (SWCNT), nanowires, nanogaps, nanochannels, and nano-electromechanical (NEM) devices. The project will focus on nanowire-based devices as they have considerable potential for electrical biosensing that offer the possibility of portable assays in a variety of point-of-care environments [70], [90], [96].

Silicon biosensors
Over the past decade, silicon nanowires have been the most researched for application as biochemical sensors [97]- [108]. Silicon nanowires are of interest for a number of reasons, for example, the material is well known and is compatible with CMOS integrated circuits for the development of sensor systems [97]- [108]. The nanowire is expected to have high surface-to-volume ratios which gives high sensitivity and the electrical sensing will give real-time label-free detection without the use of expensive optical components. Mass manufacturing is also a main advantage for silicon and is critically important for nanowire biosensor applications because of the widespread uptake of biosensors in 'point-of-care' settings, the biosensor needs to be disposable [97]- [108].
A number of fabrication methods are well established for silicon nanowires which utilise both bottom-up and top-down methods (These methods are called hybrids). It still remains that bottom-up techniques have the advantage of simplicity [97]- [108]. Bottom-up methods are still limited due to the alignment problem. The hybrid methods require further nanowire technologies to achieve alignment, such as electric field or fluid-flow-assisted nanowire positioning to locate the nanowires between lithographically defined source and drain electrodes. The technique is interpreted as a hybrid between bottom-up and top-down. Top-down methods overcome these problems, and several researchers have used advanced lithography techniques to fabricate single-crystal silicon nanowires on silicon-on-insulator (SOI) substrates. SOI wafers are expensive and to overcome the problem some researchers [109] have devised alternatives to SOI. The electrical output characteristics of silicon nanowires are good and they are well-suited for biosensing applications. The sensitivity range for most silicon-nanowire based biosensors is between 50 and 400 mV [97]- [108].

Comparing ZnO nanowire biosensors
ZnO is investigated as it is expected to be more sensitive than Si due to its wider band-gap [109]. This is observed by comparing Table 2 with Table 3. ZnO devices show results comparable to silicon devices; especially looking at response time and limit of detection. It is required that biosensors should have the liquid reference electrode. There are many different types of ZnO nanostructures being used for sensing application and Table 2 compares the ZnO nanostructures such as nanotetrapods, nanocombs and nanorods used for biosensing [110], [111]. Nanotetrapods [112] are like nanorods but with four single crystalline legs. Most of the ZnO devices were synthesized by vapour-phase method and then transferred on Au electrode to form a multi-terminal network for the sensor receptors. Like all other bottom-up ZnO nanostructures discussed here, they are transferred to a surface of a working electrode to form a thin layer to modify the transducer. The devices have low sensitivity but the nanotetrapods exhibit good detection limit down to ~1.0 nM. The researchers [112] did not explained why the nanostructures possess low sensitivity but its three dimensional features have the potential for multi-terminal communication applications [112].
In nanocombs [113] design, each comb has between 3 to 10 rods connected to one another by a single rod. ZnO nanocombs were used as the channel for sensing glucose [113] and as label-free uric acid biosensor based on uricase [114]. The functionalised ZnO nanorods showed thermal stability, anti-interference capability and direct electron transfer (DET) between enzyme electroactive sites and external electrodes. The activity of the enzyme and the sensitivity can be increased by introducing a lipid film between the channel and the enzyme. Another uric acid biosensor [115] example is based on uricase-functionalised ZnO nanoflakes, which was hydrothermally prepared at low temperatures on Au-coated glass. The sensor produced a sensitivity based on sub-threshold slope of ~66 mV/decade. Bottom-up ZnO nanorods [116] were also used as lactate oxidase (LOD) biosensor using glutaraldehyde cross-linkers. The device had a sub-threshold sensitivity of ~41 mV/decade, with maximum detection of 0.1 µM. To test for cholesterol, porous ZnO mirco-tubes [117] were constructed using 3D assembled porous flakes. ZnO nanorods [118] were grown on Ag electrode to make a cholesterol sensor.

Conclusion
Most researchers use bottom-up approaches to fabricate the ZnO biosensors because of the straightforward synthesis process. However, these bottom-up devices have variable electrical performance due to the lack of geometrical dimension control and addressing the nanostructures for sensing application. So far, there is limited research reported on top-down ZnO biosensors and previous work demonstrated the viability of top-down ZnO NWFET for biosensor applications. In the work, however, there was no passivation layer on the ZnO nanowires, which led to dissolution of the material. This made the device unstable and the sensing results were not reproducible. There exists a need to develop a passivating layer technology and optimise the fabrication process for biosensor applications. That way, reliable measurement of sensitivity for the non-specific and specific sensing of lysozyme and bovine serum albumin (BSA) can be achieved.