Zinc-Oxynitride TFTs: Toward a New High-Mobility Low-Cost Thin-Film Semiconductor

Demands for high-performance, low-cost, and low-energy-consumption displays continue to drive the development of new semiconductor materials.  The success of the metal-oxide semiconductor IGZO for display backplanes has triggered even more activity, and zinc oxynitride is proposed as a possible solution.

by Yan Ye

THE CARRIER MOBILITY of a semiconductor material, which is a measure of how fast mobile electrons or holes can move in the material under an applied electric field, is a key parameter for determining the performance of an electronic device.  For active-matrix thin-film-transistor (TFT) displays, high mobility is related to high resolution, high refresh rate, low energy consumption, and high manufacturing yield.

This article will look at the achievable performance of indium-gallium-zinc-oxide (IGZO) as well as that of other metal-oxide semiconductor materials for TFT applications.  Efforts to make IGZO TFTs prevalent in display manufacturing are still ongoing as are searches for even better semiconductor materials that will match the high performance of low-temperature polysilicon (LTPS) TFTs and the low cost of amorphous-silicon (a-Si) TFTs.  Zinc oxynitride (ZnON) is described here as one possible path to realizing these performance and cost goals.

a-Si and LTPS

a-Si is widely used as a TFT channel material.  Although the field-effect mobility of an a-Si TFT is below 1 cm2/V-sec, the low cost of the material and its TFT fabrication set a benchmark for cost-effective production.  LTPS typically has a mobility of 50–150 cm2/V-sec.  However, because the cost to fabricate LTPS TFTs is much higher than that for a-Si TFTs, LTPS is used today mostly for high-end small-area display products.  Recently, amorphous indium-gallium-zinc-oxide (a-IGZO), one of the metal-oxide semiconductors developed in the last decade, has been used to successfully make display backplanes with a mobility higher than that of a-Si TFTs at a cost lower than that of LTPS TFTs.

It is of interest to know whether the mobility of an IGZO TFT can reach a level similar to that of an LTPS TFT through process optimization.  A wide range of measurements of the field-effect mobility of IGZO TFTs, which are extracted from I–V measurements, have been reported in the literature.  Although TFT field-effect mobility is a key parameter for validating channel-layer performance, its intrinsic limit is difficult to establish based on existent data.  In fact, accurate field-effect-mobility assessment relies on precise measurements of the I–V transfer curve, channel width and length, capacitance, etc.  It is sensitive to TFT fabrication, measurement, calculation, and human error.

Hall mobility, the carrier mobility attained from Hall-effect measurement, is a more direct measurement of the material itself.  Therefore, Hall mobility is more suitable for comparing different materials, although certain limitations or errors can also exist.  Figure 1 plots the Hall mobility of amorphous- IGZO of a composition of about 1:1:1:4 attained in a test.  These data suggest that mobility achieved by a-IGZO should be higher than that of a-Si, but not LTPS.


Fig. 1: The Hall mobility of a-IGZO films is shown before and after annealing at different temperatures.

Single-Metal Oxides

Several mechanisms limit electron or hole transport inside n- or p-type semiconductors.  Single-metal oxides, such as ZnO, In2O3, and SnO2, have been studied intensively as transparent conductive oxides (TCOs) for decades and as semiconductors for TFTs in the last decade.  Figure 2 shows the Hall mobility at different carrier concentrations for doped and undoped ZnO films.  It is clear that two mechanisms limit carrier transport in the material: grain-boundary-limited transport and ionized impurity scattering.  ZnO is normally polycrystalline.  The boundaries between the crystalline grains can trap charges, deplete the carriers around them, and interrupt the electron transport across them, presenting potential barriers with respect to the conduction-band minimum [Fig. 3(a)].  The barrier height decreases with increasing carrier concentration until Coulombic ionized impurity scattering dominates as the mechanism-limiting electron-carrier transport.

As shown in Fig. 2, the transition point at which the ionized impurity scattering mechanism becomes dominant is at a carrier concentration of ~1020 cm-3 in ZnO.  It is worth noting that in single-crystal Si, the transition point at which ionized impurity scattering becomes dominant compared to phonon scattering is at a carrier concentration of ~1017 cm-3.

For other single-metal oxides, it is observed that high mobility can be achieved at high carrier concentration where the material is more like a conductor than a semiconductor.  For example, In2O3:H films attain a mobility above 100 cm2/V-sec at a carrier concentration over ~1020 cm-3, characterizing as conductors.  In fact, it is difficult to make a single-metal oxide film with a high mobility and a low carrier concentration as desired for TFTs.  Therefore, it is necessary to search for other options.


Fig. 2: Mechanisms limiting Hall mobility are shown based on the Hall-mobility measurement of various doped or undoped ZnO films at different carrier concentrations.


Multi-Metal Oxides

The grain-boundary-limited transport problem encountered in single-metal oxides can be addressed effectively by forming multi-metal oxides, as shown in Fig. 3(b).  Since multiple metals are involved in forming the crystalline structure of the film during deposition, under a balanced ratio, preferential growth of a crystalline structure from any metal used can be interrupted by the presence of other metals.  a-IGZO, with a composition of about 1:1:1:4, is a good example in this category.

However, it is observed that even though there is no clear grain boundary in a-IGZO, electron transport is still limited by a potential barrier just as it results from grain-boundary-limited transport, showing an increase in mobility with an increase in carrier concentration [Fig. 3(d)].  The barrier remains the dominant mechanism limiting the mobility at carrier concentrations up to ~1020 cm-3 with no signs of Coulomb scattering or defect reduction taking over.  Therefore, the mechanism resulting in this barrier is strong.  Notably, the mobility-carrier concentration relationship remains the same even for single-crystalline IGZO as shown in Fig. 3.  A percolation conduction model has been used to explain the barrier.

Conceptually, the potential barrier in IGZO may also be explained based on the model of s-orbital wave-function overlap.  In an ideal case for a multi-metal oxide, as illustrated for IGZO in Fig. 3(b), the distance of the interaction between s-orbitals of neighboring metal cations should be larger than that of cation sites.  However, it may not occur everywhere in reality.  Even by excluding the complexity of randomness in an amorphous phase, it may not be thermodynamically favorable to pack cations of different metals at an equal distance, as described in the ideal case.  In addition, different metals may interact with others differently, making the interaction distance between cations from different metals more complicated.  Therefore, it is very likely that there will be some locations where the site distance is larger than the interaction distance between cations, causing interruptions on the conduction-band minimum that will give rise to potential barriers.  This model may also explain why mobility is reduced in a highly doped single-metal oxide, even though no significant grain-boundary-limited transport is present, and why a ternary compound such as IGO or IZO may result in a higher mobility as reported due to less complexity than other quaternary compounds.

It is possible to achieve a higher mobility by changing the composition of a multi-metal oxide, i.e., making an IGZO with a different ratio.  When making a film with an unbalanced ratio, crystalline structures grow relatively easily if the defect level is kept low, and thus barriers associated with grain-boundary-limited transport return.  Besides, barriers associated with multi-metal cations still exist, just as in a ternary compound or a highly doped binary compound.  Therefore, the improvement in mobility thus far with a carrier concentration suitable for TFTs is limited.


Fig. 3: The barrier resulting from grain boundary commonly encountered in a single-metal oxide (a) can be suppressed by forming either a multi-metal oxide (b) or a single-metal oxynitride (c).  Data plotted in (d) show that the mobility of a-IGZO or c-IGZO increases as carrier concentration increases, indicating that a potential barrier, similar to that due to grain-boundary-limited transport, exists.  Data plotted in (e) show that several as-deposited ZnON films attained from a 50°C process follow a different trend compared with IGZO.


Zinc Oxynitrides

The barriers resulting from both grain boundary and multi-metal cations can be suppressed by forming a single-metal cation (Zn) and multi-anion (O and N) metal oxynitride such as zinc oxynitride (ZnON), as shown in Fig. 3.  Since both oxygen and nitrogen are used as anions, which require different crystalline structures in the compound, the arrangement of zinc cations in the film is disordered.  The film can be deposited through a reactive sputtering process using a metallic zinc target and reactant gases such as oxygen and nitrogen.  Composition and crystalline structure of the film can be varied by controlling the competition of reactions between zinc and nitrogen and between zinc and oxygen, which can be achieved by adjusting the flow rate ratio of oxygen and nitrogen along with the associated pressure, temperature, and power used in the process.

Figure 4 shows films deposited at different oxygen flow rates under a high nitrogen flow rate of 500 or 300 sccm in a reactive sputtering process.  When no oxygen gas is introduced and oxygen unintentionally remaining or is carried into the chamber is minimized, a Zn3N2 film is produced.  Zn3N2 film is polycrystalline as observed by X-ray diffraction (XRD) and grazing-angle XRD (GXRD).  When a small amount of oxygen gas is introduced, the growth of the Zn3N2 crystalline structure is interrupted because the reaction between zinc and oxygen starts to compete with the reaction between zinc and nitrogen.  For a flow rate of oxygen between 5 and 40 sccm in the test (Fig. 4), crystalline structures for both Zn3N2 and ZnO crystalline structures have been significantly suppressed and films deposited are either amorphous or highly disordered nanocrystalline ZnON films.  The amorphous fraction and composition of ZnON films vary depending on reactions involved in the film-deposition process.


Fig. 4: Shown is the change in Hall mobility and crystalline structure in the films deposited at different oxygen flow rates in a reactive sputtering process using a Zn- or Al-doped Zn target. In the sputtering process, the flow of nitrogen gas remains high. By adjusting a dominant reaction from between Zn and N to between Zn and O, zinc nitride (Zn3N2), zinc oxynitride (ZnON), nitrogen-doped zinc oxide (ZnO:N), and zinc oxide (ZnO) films can be produced. High mobility is attained in the films in which crystalline structures of Zn3N2 and ZnO are suppressed.


As the oxygen flow rate is increased to about 50 sccm, the reaction rate of zinc with oxygen becomes higher than that of zinc with nitrogen, and ZnO crystalline structures emerge.  However, nitrogen is still embedded in the film, as indicated by the shift of ZnO-crystalline peaks in XRD and GXRD measurement.  Since the film has a clear zinc-oxide crystal-line structure, it is often called nitrogen-doped zinc oxide (ZnO:N).  The reaction between zinc and oxygen becomes dominant as oxygen flow further increases or nitrogen gas decreases, and the film becomes a ZnO film even though a lot of nitrogen is still present in the process.  It is evident that the reaction rate between zinc and oxygen is much higher than the reaction rate between zinc and nitrogen in the reactive sputtering process.  This also explains why an abnormally high ratio of nitrogen over oxygen is often needed to produce a ZnON film, why the transition regime from Zn3N2 to ZnON is very small, and why a Zn3N2 crystalline structure can be suppressed even by a small amount of residual oxygen.

Although the films made from the process can be Zn3N2, ZnON, ZnO:N, or ZnO, Hall-effect measurements show all of the films produced are n-type semiconductors.  Therefore, the mobility of the films should rely on the conduction band of the materials.  The result in Fig. 4 shows a clear trend that the higher mobility is achieved in the films in which the Zn3N2 and ZnO crystalline structure is suppressed.  The variation of mobility in a ZnON film depends on its composition, amorphous fraction, and possible defects in the film.  Table 1 shows the mobility of a ZnON film before and after annealing.  It is clear that mobility increases significantly from 38 to 135 cm2/V-sec, while the carrier concentration is reduced, most likely due to defect reduction in the film through annealing.


Table 1: Hall mobility and carrier concentration of a ZnON film are shown before and after annealing.
Film Resistivity Ohm-cm Hall Mobility cm2 V-1 S-1 Carrier Concentration cm-3
ZnON as deposited (50°C process) 2.7E-02 38.5 -9.06E+18
After anneal at 400°C for 1 hour 1.6E-02 135 -2.82E+18


Another interesting difference between IGZO and ZnON is illustrated in Fig. 3(e).  Several ZnON films tested have shown an increase in Hall mobility as carrier concentration decreases in the region to below 1020 cm-3, which is different than the trend reported for IGZO.  This indicates that in the ZnON films, ionized impurity scattering remains dominant even in a region where it is relatively weak, which, in turn, indicates that the potential barrier on the conduction-band minimum in the ZnON films is relatively low.

Actually, ZnON should be closer to the ideal model shown in Fig. 3(b) if the disordering from the multi-anions is not too severe.  Since only Zn serves as the cation in the ZnON films, the interactions between neighboring cations are more uniform, and the distribution of metal cations is less complex, despite their amorphous arrangement.  This increases the probability of maintaining an interaction distance larger than the distance between the cation sites.  Therefore, the degree of interruptions on the conduction-band minimum, or potential barriers, in the single-metal oxynitride should be reduced compared to that of IGZO film.  As a result, a higher mobility than IGZO can be achieved.

ZnON has some unique characteristics compared to other metal oxides.  For example, the wet-etch rate of ZnON is at least 10 times faster than that of IGZO.  However, ZnON is more resistant to dry plasma.  Under typical plasma dielectric etching or metal-etch process conditions, the etch rate of ZnON is about 10 times slower than that of IGZO.

It is also observed that the surface of ZnON is more hydrophilic than other oxides.  It can absorb moisture and pollutants in the air and form weak acids or bases that accelerate oxidation of the film from the top surface down in a catastrophic way.  Typically, the shelf life of ZnON is about several weeks if it is exposed to air without any passivation or protection.  Annealing, however, can significantly extend the shelf life of ZnON film from several weeks to several years.  ZnON TFTs are often less sensitive to ambient conditions tested against other TFTs of the same structure.  However, the threshold voltage, Vth, of ZnON TFTs is often more negative compared to other metal-oxide TFTs.  As experienced with any other new semiconductor, the integrated process to fabricate ZnON TFTs will be different from that for IGZO, and some changes will have to be implemented.  Some of these features are challenges that need to be addressed, but some of the unique film properties can be utilized to improve the TFT stability and reduce the costs for manufacturing TFT backplanes.  So far, ZnON active-matrix TFT backplanes for a 3.8-in. QVGA display have been tested with yields close to 100%.  A field-effect mobility of about 100 cm2/V-sec has also been achieved by groups with ZnON TFTs.

In summary, IGZO TFTs have successfully demonstrated higher mobility than a-Si TFTs and lower cost than LTPS TFTs.  Even though efforts to fully implement IGZO TFTs in display manufacturing are still ongoing, searches for even better semiconductor materials have clearly already started.  In order to make TFTs with a performance as high as that for LTPS TFTs and as low cost as that for a-Si TFTs, breakthroughs are needed.  ZnON TFTs have been demonstrated here as one candidate to achieve this goal, but more progress is expected in the future. •


Yan Ye has been working on the development of new technology at Applied Materials for more than 20 years.  He can be reached at yan_ye@contractor.amat.com.