Improvements to Oxide Films
This invention relates to oxide films, especially ceramic films. The invention has application to thin ceramic films that are particularly suitable for non-volatile electronic memory applications. More specifically, the invention relates to thin ceramic films of doped lead zirconate titanate (PZT) or doped lead titanate (PT).
PZT films are typically deposited onto silicon substrates that are themselves coated with layers of electroding metals such as Pt/Ti to act as contact electrodes to the ceramic films. Undoped PZT films exhibit very poor fatigue characteristics, especially when deposited onto metal electrodes such as Pt.
Ferroelectric thin films have attracted a great deal of attention for use in non-volatile semiconductor memories because of their many advantages over Si-based devices. Lead zirconate titanate Pb(Zr, Ti)O3 (PZT) thin films have been widely studied and are recognized to be highly promising because of their excellent ferroelectric properties. Megabit-scale stacked memories have been produced using PZT or bismuth layered perovskites.
PZT thin films have the advantage that they form at significantly lower temperature than the bismuth layer structure. However, PZT thin films on Pt electrodes for ferroelectric random access memories exhibit degradation problems such as severe polarization fatigue after long bipolar switching pulses.
The bismuth layer structure ferroelectric thin films such as SrBi2Ta2O9 and (Bi3.25Lao.75)Ti3Oι2 (BLT) have been found to exhibit essentially no fatigue and low coercive field with Pt bottom electrodes.. However, they are generally prepared at relatively high processing temperatures, and show low remnant polarization.
C NMM ATION COPY
There has been some success in finding methods to mitigate fatigue in PZT thin films, mainly through the use of metal oxide electrode materials.
A wide variety of technologies have been used to process high quality PZT thin films, such as RF magnetron sputtering, pulsed laser ablation and metal-organic chemical vapour deposition. One of the favored techniques is "sol-gel" processing. Organo- metallic precursor compounds of the desired ceramic oxides are mixed and dissolved in a suitable solvent. The resultant solution is then hydrolysed to form a structured solution or gel containing organo-metallic polymers or macroclusters. Additives, such as ethylene glycol, formamide, can control the viscosity and surface tension of the sol gel solution.
Films are prepared by either spin, dip or spray coating or painting onto an appropriate substrate. The coated substrate undergoes an intermediate firing or baking to remove organic material and a post-fire heating step is usually performed to fully develop the final ceramic structure. Subsequent layers can be deposited to build-up a required layer thickness.
The sol gel process has several advantages over other fabrication methods. It is simple, more economically feasible and permits coating of complex geometry. Usually ceramic films up to about 0.1~0.5 μm can be deposited in a single layer. A final film thickness of around 2 μm or more can be achieved without cracking.
In the preparation of PZT thin films by solution deposition, metalorganic starting reagents such as alkoxide (M(OR)n where M is a metal and R is an alkyl group), carboxylate (M(OOCR)n and β-diketonate (MOx(CH3COCHCOCH3)n) compounds of lead, zirconium, and titanium are employed. The most frequently used solution- preparation approaches may be grouped into three categories:
1. Sol-gel processes that use 2-methoxyethanol as a reactant and solvent. The 2- methoxyethanol process is perhaps the most widely utilized solution-deposition process. When properly carried out, the methoxyethanol process offers controllable and reproducible chemistry, and nonhydrolysed solutions exhibit minimal aging effects. The primary disadvantages of the process are the rather involved chemistry, particularly for non-chemists, and toxicity concerns regarding 2-methoxyethanol. Handling this teratogenic solvent presents a safety concern.
2. Hybrid processes that use chelating agents such as acetic acid to reduce alkoxide reactivity. Compared to the 2-methoxyethanol process, hybrid processes offer the advantages of relatively simple solution synthesis. Involved distillation and refluxing strategies are not required. However, while the process is simple and rapid, the chemistry involved in solution preparation is quite complex due to a number of side reactions. This results in a diminished ability to control precursor structure and properties. Another drawback of these processes is that continued reactivity in the precursor solution following synthesis can result in a change in precursor characteristics over time and thereby a degradation in film properties. This occurs because substituent groups such as acetate, while less susceptible to hydrolysis than alkoxy groups, may still be attacked by water. This and other reactions result in continued oligomerization and eventually precipitation. In spite of these disadvantages, very high-quality thin films have been prepared by this approach and a number of research groups routinely use it as their primary method of film fabrication.
3. Metalorganic decomposition (MOD) approaches that use large, water-insensitive carboxylate compounds. The basic approach consists of dissolving the metalorganic compounds in a common solvent, usually xylene, and combining the solutions to yield the desired stoichiometry. Since the starting compounds are water-insensitive, they do not display the oligomerization behavior, and the precursor species that exist in solution retain a strong resemblance to the starting materials. Solution synthesis is straightforward, and the approach allows for rapid compositional mapping of material systems. While the process is straightforward,
it does possess a number of limitations. First the large organic ligands of the most commonly used starting reagents may cause cracking during thin-film processing. Second, because the characteristics of the precursor species can exhibit dramatic effects on thin-film properties, the inability to "tailor" the properties of the low- reactivity starting compounds through reactions such as chelation, hydrolysis, and condensation restricts process flexibility. Hence control of structural evolution and thin-film microstructure becomes limited to variations in deposition and heat- treatment conditions.
The three classes of solution-preparation approaches all have in common the synthesis of a precursor solution that is used to deposit an amorphous thin film, which is subsequently crystallized into the desired perovskite phase by heat treatment.
The present invention has been made from a consideration of the problems and disadvantages of existing oxide films and methods of manufacturing same.
Thus, the present invention seeks to provide a thin oxide film having improved fatigue characteristics.
The present invention also seeks to provide a new solution-preparation approach which is based on the sol-gel process.
According to a first aspect of the present invention, there is provided a thin oxide film for use in non-volatile electronic memories whose composition is described by the formula:
(Fbι+δ(&ι-xΗ ι«- θzθ3) (where 0.06 < x< l, 0 < z < 0.2, 0 <b~ <0.2)
We have surprisingly found that the inclusion of manganese as a dopant in a ferroelectric PZT thin film with the perovskite crystal structure gives the film
excellent fatigue characteristics when the film is used in a non-volatile electronic memory.
Thus, switching fatigue in such materials is commonly believed to be caused by the migration of oxygen vacancies in the perovskite lattice. This is particularly pronounced when the film is deposited onto a metal electrode such as platinum. According to normal understanding of the way dopants work in PZT, manganese should substitute onto the (Zr,Ti) site of the lattice and increase the number of oxygen vacancies. This should make the switching fatigue worse. In fact, according to the invention described herein, the inclusion of manganese dramatically improves the switching fatigue. Thus, it is totally unexpected that such a dopant would provide for excellent switching fatigue characteristics.
In a preferred embodiment, the oxide film of the present invention has a formula in which δ = 0.1, x= 0.7 and z = 0.01
According to a second aspect of the invention there is provided a process for producing a crack-free polycrystalline ceramic film of manganese doped PZT on a substrate, the process including the following steps: a) mixing selected individual organo-metallic precursors and dissolving these selected organo-metallic precursors in a mutual solvent or a hybrid solvent so as to produce a uniform stable solution of manganese doped PZT. b) applying said stable solution to a selected said substrate so as to provide a coating: c) drying said coated substrate; and d) firing said coated substrate so as to remove organic constituents and produce a stable crack-free polycrystalline metallic oxide film on said substrate; wherein steps (b)~(d) may be repeated to build the film up to a required thickness, for example up to 1 μm thick - this may need needs 11~12 layers - and using the resulting film for a non- volatile electronic memory application.
Step (a) may include stirring and/or gently warming if necessary to assist dissolving the selected precursors.
Step (c) may be carried out at a temperature of 200°C-350°C, preferably 200°C- 300°C.
Step (d) may be carried out at a temperature of 410°C-600°C, preferably 450°C- 600°C and more preferably 450°C-530°C.
According to a preferred embodiment a liquid precursor is first made by dissolving lead acetate trihydrate in methanol with Mn dopant in the presence of ethanolamine or diethanolamine with gentle warming and without heating the solution to a sufficiently high temperature to remove the water contained in the lead acetate trihydrate; this solution being subsequently added with mixing to a solution of zirconium n- propoxide and titanium n-butoxide in ethanol mixed with a stoichiometric amount of acetic acid and for using the resulting precursor solution to deposit a film of doped lead zirconate titanate oxide onto a substrate by applying a layer of said precursor onto a substrate, for example by spin, dip or spray coating, which is then heated. Other alternative precursors include Ti(OEt)4 or TiCO'P and Zr(OjPr)4. The Mn dopant precursor used was manganese acetate. Others are described below.
The method employs similar starting metal reagents to the previously described methods of producing PZT sols but, instead of using a toxic solvent (2- methoxyethanol), ethanol and methanol are used as solvents. Unlike the previously- described methods, the water contained in hydrated lead acetate precursor is not expelled out of the system by distillation at any stage during the whole process of the synthesis. Hence the product is no longer moisture-sensitive. Distillation and refluxing strategies are definitely not required. There is no special care needed for the protection of the solution from being hydrolyzed. Because the final precursor solution
is not moisture-sensitive, the solution aging that normally accompanies acetic acid- methanol based PZT sols is no longer a problem.
This new solution-preparation approach overcomes all the drawbacks described in the above solution-preparation approaches. The approach to synthesis of manganese doped PZT stock solution may also be employed to synthesise undoped PZT stock solution by omitting the manganese dopant and the following description includes use of the approach to produce manganese doped PZT and undoped PZT stock solutions and a comparison of oxide films produced from such solutions.
The invention will now be described in more detail with reference to a specific embodiment whereby the preferred film composition is fabricated using a sol gel process and with reference to the accompanying drawings wherein: -
Figure 1 shows XRD Patterns of PZT and Mn-doped PZT thin films, thickness=3 OOnm;
Figure 2 shows hysteresis loops of non-poled and poled a) PZT and b) PMZT thin films, film thickness=300nm. A dc bias was applied to the films for poling with positive voltage linked to the top electrodes.
Figure 3 shows fatigue characteristics of the ferroelectric capacitors of (a) PMZT (30/70) thin film and (b) PZT (30/70) thin film, thickness = 300nm.
Figure 4 shows hysteresis loops of a PMZT thin film capacitor before and after fatigue test with a bipolar pulse switching up to 1010 cycles.
Figure 5 shows retention characteristics of the ferroelectric capacitor of Au PMZT/Pt. P*T is polarization transferred out of the capacitor from 0 N to 10 N (read voltage) at
the end of a retention period; P*rT is remnant polarization in the capacitor from ION (read voltage) to 0 N during the first read pulse after a retention period; -P* is polarization transferred out of the capacitor from 0 N to -10 N (read voltage) during the second read pulse; and -P*r is remnant polarization from -10 V (read voltage) to 0 V during the second read pulse.
Figure 6 shows hysteresis loops of the Au/PMZT/Pt capacitor before and after a retention time of 3 x 104 s with ION write/read voltage.
For convenience, in the sections which follow, the generic term "PMZT" is used to described manganese doped PZT compositions and, in the absence of the manganese dopant, the term "PZT" is used.
Manipulation of air sensitive solutions
Air sensitive precursors (Titanium n-butoxide and Zirconium n-propoxide) are stored under dry nitrogen in a dry box. The raw materials are transferred into reaction vessels under dry nitrogen using dry box techniques.
Purification of raw materials
Pb(OAc)23H2O (A.R. grade) was purchased from Fisher Scientific Chemicals, UK. It was further calibrated and found to be 99.99% pure. Ti(OnBu) (97%) and Zr(OnPr)4PrnOH (70 wt.%) were purchased from Aldrich Chemical Company and were used without further purification, but calibrated before use. Methanol (anhydrous, 99.8%), ethanol (anhydrous, 99.99%), acetic acid, ethylene glycol, diethanolamine and ethanolamine were also purchased from Aldrich Chemical company and were used without further drying.
PZT precursor solution manufacture
Examples of different processes which can be used according to the present invention for the manufacture of PZT film precursor solutions will now be described.
Process 1:
Pb(CH COO)23H2O (13.90g ) is dissolved in a mixed solution of methanol (15ml) and ethanolamine (l.lg) or diethanolamine (l.lg) with gentle warming (40~50°C) for lOmin. A colorless solution is formed.
Zr(OnPr)4 nPrOH (76%) ( 4.30g ) is mixed with Ti(OnBu)4 (97%) (7.94g) in a glove box. The mixture containing the Zr and Ti precursors is stirred for 3min. under nitrogen. Ethanol (40ml) and acetic acid (2ml) is added to the Zr/Ti mixed solution. The solution is stirred at room temperature (20°C) for 1-2 hours.
Pb solution is added to the Zr/Ti mixed system in air. Then the new mixture is stirred at room temperature for 2h. Acetic acid (30ml) is added in the solution. The byproducts (10ml) formed in the synthesis are distilled off the system by heating the solution at 60-70°C. A light yellow solution (0.4M) is formed and filtered in air through a 0.2μ filter (ZapCap-CR filters, hydrophilic nylon membrane, Aldrich). Ethylene glycol (lg) is added to the solution. The PZT solution so- formed can be stored in air at room temperature or in a refrigerator at ca 0°C. The properties of the solution are listed in Table 1.
Table 1. Properties of PZT solution made by Process 1.
Process 2:
Pb(CH3COO)23H2O (13.90g ) is dissolved in a mixed solution of methanol (15ml) and ethanolamine (l.lg) or diethanolamine (l.lg) with gentle warming (40~50°C) for lOmin. A colourless solution is formed.
Zr(OnPr)4 nPrOH (76%) ( 4.30g ) is mixed with Ti(OnBu)4 (97%) (7.94g) in a glove box. The mixture containing the Zr and Ti precursors is stirred for 3min. under nitrogen. Ethanol (30ml) and acetic acid (2ml) are added to the Zr/Ti mixed solution. The solution is stirred at room temperature (20°C) for 1-2 hours.
The Pb solution is added to the Zr/Ti mixed system in air. Then the new mixture is stirred at room temperature for 2h. Acetic acid (30ml) is added in the solution. The by-products formed in this synthesis are not distilled off the system. (This is the major difference from Process 1). A light yellow solution (0.4M) is formed and filtered in air through a 0.2μ filter (ZapCap-CR filters, hydrophilic nylon membrane, Aldrich). Ethylene glycol (lg) is added to the solution. The PZT solution so- formed can be stored in air at room temperature or in a refrigerator at ca 0°C. The properties of the solution are listed in Table 2.
Table 2. Properties of PZT solution made by Process 2.
Process 3
Pb(CH3COO)23H2O (13.90g ) and Mn(OAc)2 (0.058g) are mixed and dissolved in a mixed solution of methanol (15ml) and ethanolamine (l.lg) or diethanolamine (l.lg) with gentle warming (40~50°C) for lOmin. A dark brown solution is formed.
Zr(OnPr)4 nPrOH (76%) ( 4.30g ) is mixed with Ti(OnBu)4 (97%) (7.94g) in a glove box. The mixture containing the Zr and Ti precursors is stirred for 3min. under nitrogen. Ethanol (30ml) and acetic acid (2ml) is added to the Zr/Ti mixed solution. The solution is stirred at room temperature (20°C) for 1-2 hours.
The resulting Pb/Mn solution is added to the Zr/Ti mixed system in air. Subsequently, the new mixture is stirred at room temperature for 2h. Acetic acid (30ml) is added to the solution. The by-products formed in this synthesis are not distilled off the system. A yellow solution (0.4M) is formed and this is filtered in air through a 0.2μ filter (ZapCap-CR filters, hydrophilic nylon membrane, Aldrich). Ethylene glycol (lg) is added to the solution. The Mn-doped PZT solution so-formed is stored in air at room temperature. The properties of the solution are listed in Table 3. In the discussion which follows, Mn doped PZT films will be referred to as PMZT.
Table 3. Properties of PZT solution made by Process 3.
Coatings
Examples will now be given of the methods for coating substrates with the sols whose manufacture is described above. The concentration of coating solution made by any one of processes 1-3 was 0.4M. The substrate was Pt/Ti/SiO2/(100)Si (layer thicknesses 1000A/50A/4000A). Spin coating equipment was a photo resist spinner (Model 1-EC101D-R790, Headway Research Inc). Spin coating conditions were 3000rpm for 30 seconds at room temperature. Two hot plates were used to dry and decompose PZT films. One hot plate was set at 200°C and the other set at a higher temperature (CEE Custom Model 1100 Hotplate, Brewer Science, Inc). Both hot plates were calibrated by a SensArray's Process Probe wafer (Intertrade Scientific). Coatings were put on the 200 °C hot plate for 30 sec. for drying and then firing at higher temperature. The firing time depends on the temperature. At 500°C, it needs 7min and at 530°C, it needs 5min for each layer. More layers can be built on by
repeating above procedure if required. For example, a lμm thick crack-free film needs 11 - 12 layers.
Examples of films and their properties follow:
Example 1 (PZT films)
PZT precursor solution by Process 1 was deposited onto planar substrates used to produce crack free PZT films with a composition (Pbι+δ(Zn.xTix)ι-zMnzO3) (where x=0.7, z=0, δ=0.1) that were 300 nanometres (mn) thick. The films were found to be insulating indicating that they were pinhole free.
Example 2 (PMZT films)
PZT precursor solution doped with 1 at % of Mn by Process 3 was deposited onto planar substrates and used to produce crack free PMZT films with a composition (Pbι+δ(Zrι.xTix)ι.zMnzO3) (where x=0.7, z=0.01, δ=0.1) that were 300nm thick. The films were found to be insulating indicating that they were pinhole free.
Physical and electrical property summary of PZT and PMZT films
Figure 1 shows the x-ray diffraction (XRD) patterns for undoped (Example 1) and Mn-doped PZT (Example 2) thin films on Pt/Ti SiO2/Si substrates annealed at a temperature of 530°C. All the films were single-phase perovskite and possessed a [111] preferential orientation. A small percentage of [100] orientation was observed in the doped-PZT thin film. The sharp XRD peaks suggested that the films were well crystallized.
Fatigue, retention and imprint under various conditions are recognized as the most important reliability characteristics for memory applications. The imprint is described simply as the preference of a certain polarization state over the other in ferroelectric
bistable states and eventually leads to a failure when retrieving the data stored. The preferred state can be developed intrinsically at fabrication process as well as from capacitor structure.
Figure 2 shows the hysteresis loops of PZT and PMZT thin films before poling and after poling at 5 V. The small shift of the hysteresis loop on the non-poled PZT film toward the direction having |Ec(+)| > |Ec(-)| is caused by either the asymmetric top and bottom electrodes or "self polarization of the film. The direction of the corresponding "internal bias field" was the same in all measured samples directing from the top to the bottom electrode. This "internal bias field" increased when Mn was doped in PZT film. Its magnitude varied from about 5-10 KV/cm for the PZT film to about 25 -30 KV/cm for the PMZT film.
Before poling, both P-E loops of PZT and PMZT thin films showed asymmetry with |Ec(+)j>|Ec(-)|. After poling with the positive end of dc bias connecting to top electrode (Au), the P-E loop of PZT film shifted towards the negative electric field having |Ec(-)|>|Ec(+)|. This indicated that the poling direction is opposite to the "self polarization" of the films. The self-polarization direction was pointing to the Pt bottom electrode. This shift of the P-E loop was even more obvious in the case of poled PMZT film. The shape of the hysteresis loop becomes more "square". The total coercive field (|Ec(+)|+|Ec(-)|) increased. Its magnitude, (|Ec(+)| +|Ec(-)|), is about 223 and 263 kv/cm for PZT and PMZT thin films, respectively. The increase of the coercive field in PMZT films is most likely to be due to the large crystal strain in the tetragonal phase and small grain size as the tetragonality increases with Mn addition. It was observed that if repeated polarization reversals were given the P-E loop of PMZT films would shift towards a symmetric position though a completely symmetric P-E loop was not ultimately reached.
Ferroelectric fatigue is the loss of switchable polarization with repeated polarization reversals, which is due to pinning of domain walls, inhibiting switching of the domains. Polarization fatigue tests were performed using a square wave electric field of300kV/cm at 60 kHz.
Figure 3 shows the normalized polarization as a function of polarization switching cycles, where P and PΛ represent the switching polarization between two opposite polarity pulses and non-switching polarization between the two same polarization pulses, respectively. The values of P - PΛ denote the switchable polarization. The switching cycles in terms of logarithm on the jc-axis are generally used to evaluate the fatigue characteristics. For Au/PZT/Pt capacitor, three stages are found in the cumulative switching cycles.
(1) Slow fatigue state: this stage corresponds to the cumulative switching cycles ranging from 1 to lxlO4 cycles. No obvious polarization degradation was observed at this stage.
(2) Logarithmic fatigue stage: this stage ranges from lxlO4 to lxlO8 cycles and has
+ Λ been recognized by many researchers. P - P decreases continuously with the increase of cycle number.
(3) Saturated stage: this stage corresponds to at switching cycles above 108. The P* - P tends toward a saturated value which is about 67% of the original value. For Au/PMZT/Pt capacitor, no fatigue behaviors were observed at least up to 108 cycles as shown in Figure 3. The PMZT thin films almost shows a fatigue-free behaviour at 1010 cycles, at which point the P - PA value only had a reduction of 7% of the original value.
Figure 4 shows the hysteresis measurements of a PMZT thin film capacitor before and after fatigue test. There is no obvious reduction of switchable polarization at 1010 cycles. This indicates that Mn can greatly improve the fatigue behaviours of ferroelectric PZT thin films. It is worth noting that after the cycles of 1010, the P-E loop shifted to a relatively symmetrical position and toward the self-polarization direction.
Ferroelectric retention properties were measured at room temperature. The test pulse sequence used for retention measurement was as follows: at first, a triangular pulse of -10V was applied to write a known logic state; then, after a predetermined time, the logic state was sequentially read by applied two triangular pulses of +10V (read#l) and -10V (read#2). The pulse width for all triangular pulse is 0.5ms. The time delay between the write pulse and the first read pulse is called the retention time.
Figure 5 shows the polarization decay of capacitor Au/PMZT/Pt measured at the above conditions over a range of retention time 1-30 000s. As can be seen, the polarization of this capacitor almost remained constant, indicating that the written logic states kept well after 3x104s. A retention test has also been performed using the write voltage of +10 and a read voltage -10V and +10V as a function of retention time and similar retention properties were obtained.
Figure 6 shows ferroelectric hysteresis loops of the Au/PMZT/Pt capacitor before and after a retention time of 3 x 104 s with a 10V write/read voltage.
It will readily be appreciated that the fatigue results for the PMZT films presented here represent a major improvement relative to the undoped PZT thin films and thus offer great potential for non-volatile memory applications. Moreover, while the invention has been particularly described with reference to doped PZT films, it will be understood that the invention includes doped lead titanate (PT) films and in particular manganese doped PZT (PMZT) films and manganese doped PT (PMT) films. It will also be appreciated that manganese doped PT films may be produced with processes similar to those described herein for the production of manganese doped PZT films by appropriate selection of the starting materials as will be understood by those skilled in the art.
It will also be appreciated that while the sol gel process is described here in great detail, it is only one method by which such films can be grown. Thin films of these oxides can be deposited by a wide range of other techniques, including RF magnetron
sputtering, pulsed laser ablation and metal-organic vapour phase deposition. Such manganese-doped films grown by these other methods would be expected to show similarly good fatigue characteristics, if the composition is in the range defined by the formula of the present invention.
It will also be appreciated that, while the oxide described in the above are grown onto metal layers such as Pt/Ti as electrodes, conductive oxides such as LaNiO3 and YBa2Cu3O7.g can equally well be used as the electrodes.