Chemical Vapor Deposition of Oxide Ceramics

Chemical Vapor Deposition of Oxide Ceramics

DESIGN, SYNTHESIS AND CHARACTERIZATION OF PRECURSORS FOR CHEMICAL VAPOR DEPOSITION OF OXIDE-BASED ELECTRONIC MATERIALS OLIVER JUST*, BETTIE OBI-JOHNSON*, JASON MATTHEWS*, DIANNE LEVERMORE*, TONY JONES**, AND WILLIAM S. REES, JR*. *School of Chemistry and Biochemistry and School of Materials Science and Engineering, and the Molecular Design Institute, Georgia Institute of Technology, Atlanta, GA 30332-0400 **InorgTech, 25 James Carter Road, Mildenhall, Suffolk, IP28 7 DE, United Kingdom ABSTRACT Ferroelectric and other high dielectric constant metal oxides currently are sought-after for a variety of applications in the electronics industry. To meet the demand of preparation of these interesting materials in a manner compatible with traditional silicon-based fabrication procedures, chemical vapor deposition routes are desired for film growth. Compounds displaying high vapor phase stability are necessary as precursors for these applications. Additionally, in general, it is preferred to utilize compounds in a liquid state, due to the more rapid re-establishment of equilibrium at a liquid-vapor interface, compared to that present at a solid-vapor interface. This combination of desired molecular properties, in turn, presents a great challenge to the coordination chemist. Several of the metals of interest for these uses reside in groups 2-5. Common design features are emerging for the ligands best suited for attachment to these metals for subsequent utilization in the deposition of metal oxides. In order to achieve coordinative saturation of the relatively high ionic radii exhibited by most of these elements, multidentate, monoanionic ligands are relied upon. In the past, most often, homoleptic ligand sets have been employed, thereby reducing the chance for ligand scrambling to occur during the growth process. Such disproportionation processes have been credited, in previous work, with the observation of a temporal decay in vapor pressure of heteroleptic compounds. In some interesting new developments, it has been found that heteroleptic compounds possess sufficient vapor phase integrity to permit their evaluation as CVD precursors. These, and related, results are presented herein. INTRODUCTION One of the greatest challenges in materials chemistry is to close the loop between evaluation of final device performance and the design of precursors, which enter into processes, utilized in device manufacture. This Holy Grail of "post-mortem" detection of failure devices, and its integration into the "pre-embryonic" design of molecular precursors, has attracted substantial interest from researchers in recent years. As shown in Figure 1, it is incumbent on researchers in the area of precursor development to take a broad view of what are considered as inputs and outputs to the overall area of precursor design. Frequently, it is viewed that the singular input is design and the only output is a CVD precursor. Design includes the components of cost, technical specifications, equipment limits, and process parameters, each of which must be independently considered and weighed against one another in decisions regarding precursor design. The output is not only the compound itself, but also additionally equipment, and process recommendations to accompany all chemistries, which have been developed. In this vein, one may have discovered a compound which is not amenable to delivery by traditional (vapor phase) modes. An example of this is the emergence of liquid delivery systems to accommodate precursors, which are not useable in processes relying exclusively on traditional vapor delivery schemes. Mat. Res. Soc. Symp. Proc. Vol. 606 2000 Materials Research Society 3

INPUT Cost OUTPUT CVD Precursors Technical, > Specifications / Equipment ' \ Limits ( Process / Parameters y PRECURSOR DEVELOPMENT / v. ^< \ I * Equipmei / Recommendat ^ Process Recommendations Figure 1: Inputs and Outputs for MOCVD precursors. This manuscript follows the two themes of design and characterization in the following sections. DESIGN Statement of the Challenge Many desired dielectric, insulating, and other electronic materials contain elements residing in groups 1-5 of the periodic chart. The heaviest representatives among these elements have the smallest known charge/size ratio among the entire periodic chart. Thus, this problem is among the most difficult for a coordination chemist to tackle. Additionally, there are substantial chemistry knowledge gaps present in these s block and early d block transition elements. Therefore, the wide pyramid base which was present in p block chemistry, and contributed to the early development of alternative precursors in III-V compound semiconductors, is absent in this region. Furthermore, organometallic chemistry (which is known for the early transition elements) is often not directly applicable to the growth of metal oxides, which are necessary for most modern electronic materials. Therefore, substantial basic research effort must be invested to compensate for these fundamental chemistry knowledge gaps among the elements, which are vital to the preparation of the next generation of electronic devices. RECENT RESULTS As shown in Figure 2, magnesium with a fc(/?-diketonate) ligand is four coordinate. The magnesium being divalent, binds two monovalent ligands, each ligand being bidentate, with the net result being a coordination number two less than the optimum number of six for magnesium. As shown in the Figure, the material picks up two additional diethyl ether molecules in axial positions to become octahedral, and, therefore, six coordinate. The diethyl ether ligands are intermolecular in nature, and their weak coordination is capable of becoming disassociated in vapor phase transport.

Figure 2: Ball-and-stick representation of Mg(tmhd)2(Et2O)2. In order to compensate for this loss, recently intramolecularly coordinating ligands have been designed. In the specific case of magnesium, these were tridentate monoanionic ligands. A structure of one of these resultant products is depicted in Figure 3, Figure 3: Ball-and-stick representation of bis(5-n-dimethylamino-2,2,7-trimethyl-3-octanato) magnesium. and the thermogravimetric analysis plot is presented in Figure 4.

100 75 50 25 100 200 300 400 TEMPERATURE C Figure 4: Representation of the thermal behavior of bis(5-n-dimethylamino-2,2,7-trimethyl-3- octanato) magnesium. As can be observed from the TGA, the material goes to 0 wt% at one atmosphere of pressure at approximately 260 C in a single step; therefore there is neither solid nor vapor state decomposition prior to sublimation and during transport of this material, respectively. This example of intramolecular coordination satisfying the high coordination number of low-valent metal cations has been successfully employed in several research groups, notably those of Rees and Marks for the group 2 elements. 1 Overall, there are several different approaches to looking at metal-ligand interaction systems. These include: i) one metal and one ligand; ii) multiple metals and one ligand; iii) one metal and multiple ligands; and iv) multiple metals and multiple ligands. The traditional one metal/one ligand approach has been used for decades in the preparation of compound semiconductors. It is the one, which is most frequently the entry point into a new materials system. Thus, when an initial result appears in chemical vapor deposition of a new material it is inevitably using off-the-shelf precursors, which are well known to be one metal/one ligand compositions. In the chemical sense, these are referred to as homoleptic compounds. The last one mentioned above (multiple metals and multiple ligands) is generally utilized in the sol-gel processing of electronic materials and, to the best of our knowledge, has yet to be met with success in the area of chemical vapor deposition of processing of electronic materials. The notion of having multiple metals and one ligand is also primarily (at this stage) reserved for the use of sol-gel processing. The final remaining one (one metal and multiple different ligands), referred to in a chemist's vocabulary as a heteroleptic compound, is one which has received limited attention, until recently. There are several recent success stories in this area, which are highlighted below. One early example of this heteroleptic approach was the use of a single metal mixed ligand system from Gordon's group at Harvard. In this approach, a combinatorial batch of ligands was prepared and then used for direct combination with the metal. This mixed ligand, single metal system was an ambient condition liquid. Although not highlighted by the authors in their original work, a key advantage of this approach is that the ligand purification step has been circumvented. Therefore, the frequently time-consuming and difficult process of purifying an organic compound to absolute homogeneity prior to being mixed with a metal has been skipped. This avoids the inevitable loss of material that occurs during purification, as well as the concomitant increased cost of the final product. Another recent example in this area is the dimeric heteroleptic zirconium isopropoxide /?- diketonate, which has been prepared at InorgTech, and was used in a liquid injection system. It

has a higher vapor pressure than the homoleptic tetrakis-tmhd zirconium complex, as indicated in Figure 5. Temp f*cj Figure 5: Comparative thermal decomposition data for Zr 2 (O I Pr) 6 (thd) 2 and Zr(thd) 4. The deposition rate of this heteroleptic compound is very similar to that of lead to-tmhd as shown in Figure 6. Figure 6: Metal oxides growth rates as a function both of substrate temperature and utilized precursor. The evaporator is reported to remain residue-free after long-term use of this material. 3 The design of this particular material is such that the bridging ligands between the two zirconiums are both alkoxides. Purely from consideration of a traditional coordination chemistry point of view, and looking at the electrostatics present within the various anionic ligands, it is predicted that alkoxides would bridge, and the /?-diketonates would function as terminal ligands in this case. Despite the above highlighted successes in one metal/mixed ligand heteroleptic systems for chemical vapor deposition precursors, there are nevertheless numerous concerns, which remain unsolved with these approaches. Ligand scrambling may lead to a temporal instability in vapor pressure. If such scrambling is present, then congruent evaporation of the resulting temporally changing evaporator contents cannot be compensated for. There may be inconsistent thermal decomposition profiles resulting from the differing bond association energies, which are present amongst the ligand set. In some respects, this may be turned to the advantage of the chemist, because they can then have a selective loss of ligand attachment of metal to the

surface and subsequent thermal decomposition of the remaining ligands on the metal in a different regime. Additionally, there are challenges to in-situ feedback control monitors when multiple ligands are present in the flowing reaction co-products. Frequently, these monitors are optimized for single point detection of species. Lastly, heteroleptic compounds may have different kinetic stabilities in the presence of other vapor reactants, which lead to the possibility of premature vapor phase reactions. CHARACTERIZATION In characterization of compounds in the precursor area for chemical vapor deposition, there are two different domains, which frequently are discussed. The first one is directly intended towards application in organometallic vapor phase epitaxy. Separate from that, there is the issue of gaining chemical insights through characterization of compounds, which is entered back into the feed cycle to generate a better design for subsequent generations of precursors. In this particular approach, the motive may be summed up as follows. In order to achieve improvements on existing source compounds, one must understand the beneficial and detrimental components of their decomposition mechanisms. This may be regarded as studying failure cases to examine the mode(s) of failure, thereby eliminating the identified failure mode(s) from subsequent molecules. This is not to be confused with designing the perfect precursor. In the same vein, it is elimination of unproductive growth pathways, which practitioners are after, not the creation of exclusively perfect growth pathways. The characterization of compounds should be focused on the critical properties for organometallic vapor phase epitaxy precursors. There are three such properties. Precursors must have suitable vapor pressure, vapor phase stability, and condensed phase stability. Each of these issues is addressed individually. In the case of suitable vapor pressure, all too frequently it is simply a one atmosphere TGA plot that is relied on for this data. In reality, very few manufacturing processes are operated at one atmosphere, thus it is at reduced pressure that the most important measurement of usefulness can occur. As shown in Figure 7, an automated vapor pressure determination apparatus has been constructed at Georgia Tech. a.) Figure 7: Photographs of the VPDA apparatus a.) outside the oven (turbo pump and cold cathode gauge are not shown), b.) inside the oven.

Figure 8 presents a CAD view of the system. To Reference Ba nitron To Pump Figure 8: The CAD view of the VPDA system. Representative data obtained for yttrium, copper and barium tmhd compounds are shown in Figure 9a and for barium and strontium tmhd compounds in Figure 9b. 5 4 S 3 2-1 Cu(tmhd)2 Y(tmhd)3 * Ba(tmhd)2 0.00170 0.00190 0.00210 0.00270 0.00290 Figure 9a: Graphical depiction of 1/T versus lnp for Cu(thd) 2, Y(thd) 3 and Ba(thd) 2. [Sr(thd)2]3 [Ba(thd)2]4 Linear([Ba(thd)2]4) Linear([Sr(thd)2]3) 0.0 0.0024 0.0025 0.0026 0.0027 0.0028 Figure 9b: Graphical depiction of 1/T versus lnp for [Ba(thd) 2 ]4 and [Sr(thd) 2 ]3.

Once having constructed an instrument specifically designed for accomplishing vapor pressure determination, which addressed the key issue for suitable vapor pressure for organometallic vapor phase epitaxy precursors, the next question was to address vapor phase stability of these materials. A system was designed based on work of Desisto at the Naval Research Laboratories 4, which utilizes in-situ vapor phase UV spectroscopy. Such a system is depicted in Figure 10, and a photograph of the apparatus is shown in Figure 11. Figure 10: Illustration of the MOCVD apparatus with in-situ vapor phase UV monitoring. Figure 11: Photograph of the UV-MOCVD apparatus. This allows the experimenter to evaluate the input UV-VIS vapor phase signal, as well as the output signal, and thereby obtain the kinetics for the overall process. The single constant demanded for acquiring this data resides in the determination of the molar absorbtivity in the vapor phase for all species present. Since one cannot simply look up these values in tables, they must be measured independently, prior to initiation of these experiments. Having dealt with suitable vapor pressure and vapor phase stability as key issues, the third issue to be concerned with in precursor characterization is condensed phase stability. It generally is regarded that changes in the condensed phase are irreversible processes occurring by ligand 10