Flat-Panel-Display Backplanes: LTPS or IGZO for AMLCDs or AMOLED Displays?
After more than 20 years as the flat-panel-display backplane TFT material of choice, a-Si:H is running out of steam. The two contending replacement options are LTPS and IGZO. Which is better and how does this material choice impact the race between AMLCD and AMOLED-display front-plane technologies?
by John F. Wager
UNTIL RECENTLY, it was obvious that low-temperature polysilicon (LTPS) would eventually replace hydrogenated amorphous silicon (a-Si:H) as the thin-film-transistor (TFT) channel layer material for display backplane switching applications. After all, silicon always wins. Just ask the gallium arsenide integrated-circuit guys. Then the upstart indium gallium zinc oxide (IGZO) appeared on the scene. What’s going on here?
Amorphous Oxide Semiconductors
IGZO is one example of a relatively new class of materials, amorphous oxide semiconductors (AOSs).1-4 AOSs were originally formulated for transparent-conductive-oxide (TCO) applications. Now they are being considered for TFT channel layers and other applications.
The key to appreciating the desirable attributes of an AOS is to understand how it is designed. Hosono et al. proposed that the design of an AOS involves (i) forming an oxide using multiple cations and (ii) selecting these cations from the portion of the periodic table shown in Fig. 1.5 Multiple cations are employed in AOS designs because simple binary oxides, e.g., ZnO, SnO2, and In2O3, have a strong tendency to crystallize. In contrast, the use of multiple cations confuses the lattice with respect to which microstructure it should adopt, thereby frustrating crystallization. Moreover, cations that are selected from the portion of the periodic table shown in Fig. 1 possess conduction bands that are derived from large-ionic-radius spherically symmetric 4s, 5s, or 6s electron orbitals.5 This means that an electron can readily move from one orbital to another, rapidly propagating through the material. In other words, the nature of these orbitals means that an AOS is expected to have a relatively high electron mobility compared to that of a-Si:H (see Table 1), independent of whether the microstructure is crystalline or amorphous.
Fig. 1: This portion of the periodic table was identified by Hosono et al.5 for selecting amorphous-oxide semiconductor cations.
In addition to high electron mobility, a commercially viable TFT channel semiconductor must provide favorable characteristics for integration into a robust and cost-effective semiconductor manufacturing process. Although other AOS compositions are potentially useful for display backplane applications, IGZO offers a compelling combination of performance and process attributes and is the AOS leader in early commercialization. Thus, our subsequent AOS discussion focuses exclusively on IGZO.
In summary, IGZO is amorphous, like a-Si:H, but has a superior electron mobility compared to that of a-Si:H.
AMLCD Backplanes: Performance or Cost?
The architecture of an AMLCD backplane pixel is very simple, consisting of a single voltage-controlled switch, usually a TFT. Three AMLCD TFT technology options – a-Si:H, LTPS, and IGZO – are compared in Table 1.
Since AMLCDs were commercially introduced in around 1990, a-Si:H has until now been the AMLCD backplane TFT champion. a-Si:H was selected for AMLCD applications because it had adequate performance, process simplicity, the lowest cost, and could be readily scaled to large-area meter-sized dimensions. Three a-Si:H advantages highlighted in Table 1 – threshold-voltage (VT) uniformity, mobility uniformity, and process complexity – are direct consequences of the amorphous nature of a-Si:H. Of the three a-Si:H liabilities included in Table 1 – poor stability, poor mobility, and NMOS – mobility is the most important consideration likely to inhibit the use of a-Si:H for upcoming AMLCD commercial applications because that mobility is inadequate for the higher anticipated refresh rates required for future products and limits the ability to reduce TFT size, as needed for small pixels in high-resolution mobile displays.
After the mobility of a-Si:H was recently determined to be a limiting factor, LTPS was considered the obvious a-Si:H replacement. With respect to mobility, LTPS is the clear winner in Table 1. Additionally, LTPS TFTs have much better stability than a-Si:H TFTs. Finally, the availability of complementary metal-oxide-semiconductor (CMOS) TFTs using LTPS means that row and column drivers, for example, or other peripheral circuits can be integrated onto the glass substrate.
In a word, the LTPS advantage is performance.
In contrast, since p-channel TFTs are not available using IGZO, CMOS is not possible. Also, the electron mobility of IGZO is less than that of LTPS. However, IGZO is amorphous. Thus, it possesses the same manufacturing/scaling/cost advantages as a-Si:H with respect to threshold-voltage uniformity and mobility uniformity. Additionally, as is the case for a-Si:H, IGZO processing is simple. In fact, IGZO source/drain contacts can be formed directly by simply patterning the contact metal (or TCO) directly onto an IGZO channel layer. No channel layer contact doping or deposition of an additional doped contact layer between source/drain and active channel is required, as is the case for a-Si:H and LTPS. This simplifies IGZO processing, potentially eliminating one or more process steps. However, IGZO surfaces tend to be highly sensitive, so development of a back-channel etch process such as that currently used in advanced a-Si:H TFT manufacturing appears to be challenging. First-generation IGZO technology has been implemented using etch-stop processing, thus requiring an extra masking step compared to that of advanced a-Si:H TFT processing, but still fewer than required for an LTPS TFT process.
Additionally – and this is important – since a-Si:H and IGZO process flows are quite similar, it appears that for a relatively modest capital investment, an operating a-Si:H TFT fab can be retrofitted for IGZO by replacing the a-Si:H plasma-enhanced chemical vapor deposition (PECVD) channel layer process with physical vapor deposition (PVD) IGZO and the SiNx PECVD gate dielectric process with PECVD SiO2. This would save money. In contrast, going to LTPS will generally require construction of a new fab rather than retrofit of an existing a-Si:H plant.
In a word, the IGZO advantage is cost.
The Meaning of Performance
Until recently, the LTPS vs. IGZO debate largely boiled down to a discussion of performance vs. cost, as just presented. However, Sharp seems to have confused things a bit with respect to performance.6 The company’s basic argument may be elucidated with the assistance of Fig. 2.
Figure 2 shows an idealized comparison of log (ID) – VG transfer curves for a-Si:H, IGZO, and LTPS TFTs. As indicated in Fig. 2, increasing mobility and decreasing leakage are two primary transfer-curve considerations that determine the suitability of a TFT for an AMLCD switching application. As mentioned previously, LTPS is the clear winner in terms of mobility (see Table 1), although IGZO offers significant mobility improvement compared to a-Si:H. A higher channel-layer mobility is attractive because TFTs may be reduced in physical size and yet still supply the required current, and the TFT response time will be faster, enabling increased display refresh rates.
Fig. 2: The illustration shown depicts an idealized drain-current or gate-voltage [log (ID) – VG] transfer-curve comparison of a-Si:H (red), IGZO (blue), and LTPS (green) TFTs.
Sharp’s important contribution to the discussion of TFT performance is to point out that off-state drain-current leakage considerations are also pertinent when evaluating a TFT for its suitability for AMLCD switching. Figure 3 clarifies that there are two primary contributions to off-state drain-current leakage, involving leakage in the channel and/or through the gate insulator.
Fig. 3: Arrows indicate the directions of electron flux for the primary leakage paths giving rise to the off-state drain-current leakage in a bottom-gate TFT. The horizontal (vertical) arrow corresponds to channel (gate insulator) leakage.
In terms of off-state drain-current leakage, IGZO is the clear winner. IGZO TFTs have lower leakage across the channel because IGZO is a wide-bandgap [i.e., EG (IGZO) = 3.25 eV] unipolar semiconductor. In contrast, LTPS and a-Si:H have significantly narrower bandgaps [i.e., EG (LTPS) = 1.1 eV, EG (a-Si:H) = 1.7 eV] and are bipolar so that channel inversion occurs at sufficiently large reverse gate bias. Under reverse-bias operation, leakage through the gate insulator may also contribute to the measured off-state drain-current leakage. IGZO TFTs tend to have relatively low gate leakage since they employ a high-quality SiO2 gate insulator (superior to that of SiNx used in a-Si:H TFTs) and they have smooth surfaces so that a uniform electric field develops across the gate insulator/IGZO interface (not the case for LTPS because grains give rise to pronounced roughness at an insulator/LTPS interface). A lower leakage is desirable because less power is dissipated when a TFT is off and the TFT switch can retain an internal pixel charge for a longer period of time so that display refresh rate may be reduced. This leads to reduced power dissipation and in the case of touch-enabled displays, improved touch-screen capability (due to less noise/interference with touch detection since the display refresh and touch-sensing cycles may be interleaved rather than run simultaneously).
In three words, the off-current performance is a key IGZO advantage, largely unappreciated until it was pointed out by Sharp.
LTPS or IGZO?
Table 2 highlights the relevant strengths and weaknesses of both LTPS and IGZO technologies. Clearly, if an application requires use of a TFT with an electron mobility higher than that obtainable with IGZO (i.e., ~30 cm2/V-sec) and/or CMOS circuit functionality, LTPS is an excellent choice and is, in fact, the only choice. However, those advantages are largely irrelevant for AMLCD backplane applications. Basically, this AMLCD application requires the availability of an inexpensive TFT that can function as a voltage-controlled pixel switch, which has an electron mobility at least an order of magnitude greater than that of a-Si:H and which can be scaled to large areas, compatible with Gen 10 (or larger) glass processing. LTPS scaling to large areas is difficult and expensive, largely due to the fact that ion implantation and excimer-laser recrystallization is required. Thus, IGZO appears to be a very attractive choice, especially when the off-state performance advantages of IGZO are recognized. However, IGZO is still in an early stage of commercialization, and it is a fair question to ask whether all of the bugs have been worked out yet.
In the literature, the most likely IGZO “bug” that may need to be worked out appears to be the negative-bias illumination-stress (NBIS) instability (involving a threshold-voltage shift to negative voltages for an IGZO TFT subjected to a large negative applied gate voltage and simultaneous near-bandgap optical excitation). The physical mechanism responsible for the NBIS instability is controversial, but appears to involve the IGZO top surface and/or IGZO subgap electronic states.1-4,7 Various materials and processing fixes are being explored to control NBIS, such as top-surface passivation and post-deposition annealing. If NBIS cannot be adequately controlled via materials fixes, aggressive light-shielding measures may be required. The fact that Sharp and LG are now shipping IGZO-based products demonstrates that these NBIS challenges can be successfully overcome.
Thus, in deciding between LTPS and IGZO for AMLCD backplane applications, the more attractive choice seems to be IGZO.
AMLCD or AMOLED?
Up to now, only AMLCDs have been considered. An LCD pixel is basically a valve that controls the transmitted intensity of light incident from a backlight source, i.e., an LCD is a non-emissive (transmissive) display. In contrast, an OLED is an emissive display. An emissive display offers many advantages, including wider inherent viewing angle; higher contrast ratios (in dark ambient conditions); faster response time; lower power consumption; and a sleeker, lighter, and thinner form factor. Thus, an AMOLED offers a potentially superior viewing experience.
Recall that an AMLCD backplane pixel is very simple, consisting of a single voltage-controlled switch. The architecture of an AMOLED display pixel is more complex because an OLED is a current-controlled device. Consequently, active-matrix current control is more difficult to accomplish in an AMOLED display, requiring more than one TFT per pixel. In addition to requiring multiple TFTs to provide the basic current-control function, current-control circuit architectures are very sensitive to variation and drift in TFT parameters, particularly threshold voltage.
The simplest possible AMOLED display pixel architecture consists of two TFTs and one capacitor (2T + 1C) in which one TFT is used for selecting and charging a storage capacitor during addressing while the second TFT functions as a current source to drive the OLED display. An attractive version of the 2T + 1C pixel, as shown in Fig. 4, was proposed by Sony.8,9 Since Write-Select and Drive-Select lines can be independently controlled, this circuit may be useful in compensating for TFT and/or OLED threshold-voltage drift. Alternatively, other more complex AMOLED pixel architectures employing compensation may be required, e.g., 4T + 1C.10
In choosing between AMLCDs and AMOLED displays, pixel architecture appears to be a key factor. If the simple 2T + 1C pixel architecture shown in Fig. 4 provides adequate compensation, then the superior viewing experience offered by AMOLED technology may lead to its eventual dominance. Alternatively, if an appreciably more complex pixel architecture is required for successful AMOLED commercial implementation, it is hard to imagine that AMOLED technology will be able to compete with AMLCD technology. After all, the AMLCD is the flat-panel-display beast, defeating all comers.
Fig. 4: Shown is a two-transistor one-capacitor (2T + 1C) AMOLED pixel architecture proposed by Sony in 2012.8,9
AMLCD or AMOLED? Although I am not a betting man (odd, given my last name), if I had to bet, I would go the safe route and choose AMLCD. However, I am rooting for AMOLED! Either way, I prefer IGZO to LTPS because it seems to me that cost will be the main driver, at least for the forseeable future.
Acknowledgments
This article is based on a Keynote Address entitled “Exciting Developments in Oxide TFT Technology,” presented at the Society for Information Display International Symposium held in Vancouver, Canada, May 21, 2013, and upon work supported by the National Science Foundation under the Center for Chemical Innovation Grant No. CHE-1102637.
References
1T. Kamiya, K. Nomura, H. Hosono, “Present status of In-Ga-Zn-O thin-film transistors,” Sci. Technol. Adv. Mater. 11, 1–23 (2010).
2J. K. Jeong, “The status and perspectives of metal oxide thin-film transistors for active matrix flexible displays,” Semicond. Sci. Technol. 26, 034008-1-034008-10 (2011).
3J. S. Park, W-J. Maeng, H-S. Kim, J-S. Park, “Review of recent developments in amorphous oxide semiconductor thin-film transistor devices,” Thin Solid Films 520, 1679–1693 (2012).
4J. F. Wager, B. Yeh, R. L. Hoffman, and D. A. Keszler, “An amorphous oxide semiconductor thin-film transistor approach to oxide electronics,” J. Curr. Opin. Solid State Mater. Sci.(in press).
5H. Hosono, M. Yasukawa, H. Kawazoe, “Novel oxide amorphous semiconductors: transparent conducting amorphous oxides,” J. Non-Cryst. Solids 203, 334–344 (1996).
6Sharp, “IGZO: vision for the future,” http://online.wsj.com/ad/article/vision-breakthrough
7J. F. Conley, “Instabilities in amorphous oxide semiconductor thin-film transistors,” IEEE Trans. Device Mat. Rel. 10, 460–475 (2010).
8T. Arai, “Oxide-TFT technologies for next-generation AMOLED displays,” J. Soc. Info. Display 20, 156–161 (2012).
9N. Morosawa, Y. Ohshima, M. Morooka, T. Arai, T. Sasaoka, “Novel self-aligned top-gate oxide TFT for AMOLED displays,” J. Soc. Info. Display 20, 47–52 (2012).
10A. Nathan, G. R. Chaji, S. J. Ashtiani, “Driving schemes for a-Si and LPS AMOLED displays,” J. Display Technol. 1, 267–277 (2005). •