Progress and opportunities in atomic layer deposition
05/01/2003
Overview
A new era in film deposition control in semiconductor technology is being ushered in with atomic layer deposition. Milestones include the extension of cylinder and deep trench DRAM capacitors <100nm using new dielectric materials. Additionally, the fabrication of advanced gates for low-power transistors with several orders of magnitude lower leakage than SiON has been achieved to signal the migration to new materials for logic. The manufacturing of thin film reader heads using Al2O3 ALD in data storage is now a fact. Other applications using ALD are in the wings. Several key results illustrating process control and the migration to liquid precursors are presented and future opportunities and challenges are discussed.
Atomic layer deposition (ALD) is reaching the commercial stage for capacitor applications in semiconductor chip making and is already used commercially for thin film head reader devices in the data storage market. Several technical milestones have been achieved in the semiconductor segment, driving applications to pilot production. These results are focused on new materials for capacitors and gates at ≤100nm.
Notably, ~100% conformality in 60:1 aspect ratio DRAM trench structures at feature sizes <100nm has been achieved [1]. The integrated chip performance with enhanced capacitance, lower leakage, and good reliability of high-quality Al2O3 dielectrics has also been reported [1, 2]. While Al2O3 represents a first-generation ALD "insertion" film for capacitors, HfO2 and alloys or nanolaminates of HfOx and other materials are useful for sub-100nm capacitors [3, 4].
 Figure 1. Deep trench capacitor with 60:1 aspect ratio and ~100% step coverage of 40Å Al2O3. From Gutsche et al. [1] |
null
 Figure 2. Dielectric constant and crystallization temperature. From Londergan et al. [7] |
null
While capacitor applications are clearly riding the ALD commercial insertion wave for semiconductor use, gate applications are also receiving a lot of attention. Low-power advanced MOSFET capabilities have been demonstrated with leakage values several orders of magnitude <SiO2 controlled to an equivalent oxide thickness (EOT) of 1.5nm [5]. Ultimate success for scaled gates for high-performance transistors requires the simultaneous achievement of low leakage (<10–2A/cm2) at an EOT < or ~1.0nm, and importantly, with no degradation of MOSFET carrier mobility. While progress has been made for low-power capability, the challenge remains an active area of development for advanced, high-performance transistors needed circa 2006 for the ~65nm node and below [5].
In 2002, ALD production began in the data storage market for thin film magnetic head (TFMH) devices because of the ability to achieve high-quality insulating gap layers, the performance of which has eclipsed attempts to scale reactively sputtered Al2O3 [6].
Capacitors: Deep drilling and high-rise construction limits
As the semiconductor DRAM segment scales <100nm, the ability to scale capacitance density by the method of increasing area is reaching a limit. In the trench architecture (etching or "drilling" deeper trenches) and in the cylinder stack capacitor architecture (making taller structures), the limiting factors include cost-to-make and mechanical stability. With respect to these limits, the strategy is to implement higher-k dielectrics to avoid cost and yield issues of extreme topology structures.
Advanced capacitor scaling requirements include increasing the capacitance while simultaneously lowering leakage. Al2O3, which has a larger k value than SiO2 and SiON, has lower leakage capabilities than both the conventionally used materials for the same equivalent thickness. Figure 1 shows an advanced deep trench bottle structure [1]. Using ALD, it is conformally coated with Al2O3 inside the bottle even though the feature is re-entrant. Additionally, the film is uniform to the angstrom level within the bottle trench and from wafer-to-wafer. Electrical requirements for implementation have been met, such as leakage currents, retention times, and bias temperature stress specifications. The stressed film data provide device-operating lifetimes of 10 years [1]. Fully integrated 128-Mbit trench DRAMs, using Al2O3 as high-k node dielectric in silicon-insulator-silicon (SIS) capacitors, were recently fabricated [2].
Below 90nm, the dielectric constant must be increased above that offered by Al2O3. Hence, one consideration is the use of alloys of HfO2 because of their thermal stability. The dependence of dielectric constant and the crystallization temperature as a function of composition between Al2O3 and HfO2 is shown in Fig. 2 [7]. It is possible to use compositions near the middle of the phase diagram and achieve process integration with useful results. ALD also offers a high degree of process control. Variance in the deposition rate/cycle with respect to process temperature, process pressure, pulse timings, and purge timings has indicated an overall control capability on the order of <1% for Al2O3 and HfO2 using Genus equipment and processes [7].
 Figure 3. Nanolaminate using Al2O3 and alloys of HfO2/Al2O3. From Sneh et al. [8] |
null
Equipment and processes have been developed for nanolaminates of higher-k values than Al2O3. A TEM micrograph shows the deep corner of a 15:1 aspect ratio trench that is 100nm wide (Fig. 3). The trench is conformally coated with alternating layers of Al2O3. and alloys of Al2O3. and HfO2 [8]. The control at the deepest regions of the corners is excellent, and the result is characteristic of the self-limiting nature of the ALD process. This type of multilayer structure is useful for the development of dielectrics for 70nm capacitors and beyond and potentially for gates.
Cylinder capacitors for use at ≤100nm have been developed [3]. In this case, nanolaminates of Al2O3. and HfO2 similar to that in Fig. 3 were used.
One of the key commercial issues for ALD is the availability of precursors with adequately high vapor pressure. TMA and water, or O3, are adequate for good precursor delivery and hence good throughput for Al2O3. . However, HfO2 has been traditionally deposited with HfCl4, which has relatively low vapor pressure leading to limited conformality and throughput for high-aspect structures. The saturation characteristics and properties of HfO2 made with tetrakis (ethylmethylamino) hafnium (TEMAH) using water as well as ozone have recently been characterized [9]. The exposure pulse times are 3–4x lower than for HfCl4. The TEMAH-based HfO2 is starting to find applications in advanced capacitors for use below 100nm; high conformality, low leakage, and device lifetime of 10 years are some of the benchmark characteristics [4].
Blazing trails for cooler gate solutions
The most critical problem facing silicon device scaling today is the limitation caused by gate leakage in scaled MOSFET transistors. SiO2 (ON) gates have been scaled to their tunneling current limits (at a thickness ª 20Å), where leakage currents are ~1A/cm2 at 1V. With billion-transistor counts forecast for future logic chips, the power dissipation due to the gate leakage of CMOS transistors must be reduced. By 2006, SiO2 or SiON will have to be replaced with thicker, higher-k materials, but today there are no known solutions that meet all the requirements for high-performance transistors. Unlike capacitors, there are many other requirements that must be achieved for making this transition. These include low-interface-state density, low boron diffusion, and low charge (coulomb) scattering, which reduce the current carrying capacity of the transistor.
The advanced gate solution is a most challenging one. Many materials have already been eliminated from consideration. Ta2O5, a high-k material of interest for advanced capacitors for years, was excluded because the electronic band alignment to silicon electrodes has a "band offset" that is too low for low leakage in NMOS transistors. Additionally, both Ta2O5, and ZrO2 have poor thermal stability characteristics relative to HfO2. HfO2 has an excellent band offset value for both n- and p-type electrodes, and somewhat better thermal stability characteristics, making it suitable as a core material for advanced CMOS [10].
There are two best-known results: HfAlON [5] and HfSiON [11]. Al, Si, and N elements are used to stabilize the structure and to control boron diffusion and improve carrier mobility. In the HfSiON case, in the results reported, the films were made with sputter methods and will not be discussed here, except to say that ALD HfSiON is of high interest due to the potential of layer-by-layer film deposition control [11]. One of the assets of ALD, high conformality, is not needed for near-term transistor applications, but may well be needed when 3-D transistor technology is introduced.
HfAlON has been developed for MOSFET transistors for advanced low-power logic [5]. The results are encouraging for nearer-term use for low-power applications, with an EOT of ~1.5nm, leakage reduced two orders of magnitude for comparable EOT SiON, and mobility at 80% that of SiON. The 80% mobility is referenced to the low-field case, and high-field mobility is equal to the SiON control (Fig. 4).
These results were obtained with HfO2 and Al2O3 ALD deposition methods, followed by post-deposition anneal to engineer the interface by adding N and O to obtain higher mobility [5]. These results are arguably the best-known, high-k, low-power, transistor MOSFET results reported to date. However, more work must be done to prove usefulness at the chip level and for high-density integration.
 Figure 4. Gm and leakage current density. From Jung et al. [5] |
null
Thin film head reader devices using ALD in production
Al2O3 is now used in production at several thin film head manufacturers for data storage. Today, devices are being shipped with ALD films in them as "gap isolation" layers. For these devices, high breakdown voltages of >9MV/cm were reported on films with thicknesses from 75–500Å [6]. Compared to conventional physical, sputtered Al2O3-processed films, ALD displays excellent step coverage (needed for the upper-gap dielectric), better thickness uniformity, higher film density, and pin-hole-free films. The Al2O3 ALD film is expected to meet reader thin film head specifications from today's production at 40GB/in2 through several future generations to ~100GB/in2.
Conclusion
ALD has reached the commercial stage. Production has started in TFMH devices for reader sensors in data storage. Many semiconductor development programs are under way; the transition to production for semiconductors is built, in part, on the success of recent demonstrations, such as film conformality for high-topology DRAM capacitors and layer-by-layer film control for advanced transistors. Going forward, for ultrathin conformal films, ALD will be the technology of choice.
Tom Seidel, Ana Londergan, Jerald Winkler, Xinye Liu, Sasangan Ramanathan, Genus Inc., Sunnyvale, California
References
1. M. Gutche et al., "Capacitance Enhancement Techniques for Sub-100nm Trench DRAMs," IEDM Tech. Digest, p. 411, 2001.
2. M. Gutsche et al., FUTURE FAB International, Wafer Processing, Issue 14, 2003.
3. J-H. Lee et al., "Practical Next-generation Solution for Stand-alone and Embedded DRAM Capacitor," VLSI Digest of Tech. Papers, IEEE Cat. No. 01CH37303, 9.2, p. 84, 2002.
4. J-H. Lee et al., "Mass Production-worthy HfO2-Al2O3 Laminate Capacitor Technology Using Hf Liquid Precursor for Sub-100nm DRAMs," IEDM Tech Dig., paper 3.1, 2002.
5. H-S. Jung et al., "Improved Current Performance of CMOSFETs with Nitrogen Incorporated HfO2-Al2O3 Laminate Gate Dielectric," IEDM Tech Dig., paper 34.2, 2002.
6. Wei Xiong et al., "Ultra-thin ALD Alumina Reader Gaps for High-density Recording Heads," Session HP-03, MMM Conf., 2001.
7. A. Londergan et al., "Process Optimization in Atomic Layer Deposition of High-k Oxides for Advanced Gate Engineering," Proc., Vol. 2002-11, The Electrochemical Soc. Inc., pp.163–175, 2002.
8. O. Sneh et al, "Equipment for ALD and Applications for Semiconductor Processing," Thin Solid Films, Vol. 402/1-2, pp. 248–261, Jan. 2002.
9. X. Liu et al., "ALD of HfO2 from TEMAH with Ozone and H2O," ALD-02, Seoul, Korea, American Vacuum Society.
10. H.Y. Yu et al., "Energy Gap and Band Alignment for (HfO2)x(Al2O3)1-x on (100) Si," Appl. Phys Lett., 81, No. 2, July 8, 2002.
11. A. Rotondaro et al., "Advanced Transistors with a Novel HfSiON Gate Dielectric," VLSI Digest of Tech. Papers, IEEE Cat. No. 01CH37303, 15.2, p. 148, 2002.
For more information, contact Tom Seidel, Genus Inc., 1139 Karlstad Dr., Sunnyvale, CA 94089; ph 408/747-7140, ext. 1175, fax 408/747-7198, [email protected].