We want the electronic gadgets of tomorrow to be smaller and lighter, but also faster and more powerful: Whether MP3 players, camera mobile phones, navigation systems or notebooks, all have to be compact but also able to store increasingly large amounts of music, images, films or maps, and process them quickly. Innovative new solid state memories would contribute greatly towards making electronics smaller and more powerful, especially if they were able to save information permanently, even without being connected to an electric power source, but still process data as quickly as the working memories of present-day computers.

To realize such advanced solid state memories, new concepts are being developed worldwide, among them magnetic (MRAM), phase-change (PCRAM) and ferroelectric (FRAM) random access memories. The latter have attained a clear advantage - they already achieved the state of commercial products, however, so far only with a rather low memory density of, e.g., 64 Mbit. This is not sufficient for present-day working memories of PCs, which require memory densities in the Gbit/inch² or even Tbit/inch² range. On the other hand, the target is worth the efforts: A notebook with a non-volatile FRAM as high-density, quick working memory would work without the slow and heavy harddisk, and the annoying booting procedure would become a matter of the past, too. Physicists and material researchers worldwide work hard to overcome the present limits of ferroelectric memories [1]. Problems which need to be solved are both material- and process-related. Among the ferroelectric materials under study are complex oxides like PbZr_0.2Ti_0.8O₃ (PZT) or BaTiO₃. Chemical stability, oxygen stoichiometry, and processibility are matters of concern. To achieve a memory density in the range of 100 Mbit/inch² by help of an array of ferroelectric "dots", i.e. small features consisting of a ferroelectric material, the lateral repeat distance - the pitch - of the dots needs to be smaller than 100 nm, which requires each dot to be even smaller in lateral size. Ferroelectricity being a collective phenomenon, a certain minimum number of unit cells is required to be involved.

It was not clear from the beginning whether a ferroelectric dot of far less than 100 nm size has enough unit cells and thus would still show the required ferroelectric properties, or not. Another problem is related to the surfaces and interfaces: In a small object with a volume of, say, 0.0001 µm3, a considerable part of the atoms are close to the surface or the interfaces with other materials. These atoms may behave differently from those in the bulk, thus affecting the properties. Moreover, the quality of the surface is also related to the preparation and processing methods which have been used to obtain the small ferroelectric object: If they include harsh steps, like mechanical (imprinting) or chemical (etching) removal of the material around, the surface of the remaining small ferroelectric object may be damaged, which may result in considerably degraded properties.

We have been working on these problems using different approaches during recent years [2]. Recently we found a method to prepare arrays of ferroelectric dots of PZT which are embedded between bottom and top platinum electrodes, thus forming tiny capacitors [3]. Each of the capacitors has a lateral size below 100 nm, even going down to 40 nm. In the latter case the pitch is as small as 60 nm, which is equivalent to a memory density of 176 Gbit/inch². In spite of these small sizes, the PZT material still preserves its ferroelectric properties, and thus the array of nano-capacitors can be used as a solid state memory. Due to the small sizes of our nano-capacitors, various microscopies had to be applied in order to look into the details of structure and properties, viz. scanning (SEM) and transmission electron microscopy (TEM), and scanning force microscopy both in tapping mode (AFM) and in the so-called piezoresponse mode (PFM).

The performance of our arrays of nano-capacitors is a result of the principle on which they are based. In the ferroelectric material PZT, all unit cells have permanent electrical dipoles. The positive and negative poles of a permanent electrical dipole can be interchanged quickly. Cooperating with colleagues from Max Born Institute Berlin, we were recently able to show that this process requires an intrinsic time as small as one picosecond [4]. PZT can therefore save data permanently like a hard drive, but process it as quickly as an advanced present-day working memory.

Figure 1 shows a scheme of the process. A thin perforated mask made of anodic aluminum oxide (AAO) is being prepared, which contains an ordered array of small holes with sub-100 nm diameter. This mask is obtained by electrochemically oxidizing an aluminum surface, a method known as the eloxal process. A basic version of the latter has been used for decades to provide aluminum components with a protective coating and to give aluminum tableware a matt-metallic sheen. In the basic version of this process, pores generally eat into the aluminum oxide in a random pattern. However, by carefully selecting the temperature, pH level and chemical composition during oxidation, in our case the pores were forced into a hexagonal arrangement where each pore is surrounded by six others [5]. The mask was then placed on a single crystalline magnesium oxide (MgO) plate heated to 650°C. This plate is coated with an epitaxial platinum layer which serves as support, and later as bottom electrode. The material PZT is deposited through the holes of the mask onto the platinized MgO plate by a process called pulsed laser deposition (PLD) [6, 7], thus obtaining small PZT dots with a thickness of, say, 70 nm. A thin platinum cover completes the capacitors, in which now the platinum layers serve as electrodes, and the PZT is the dielectric. Finally the thin alumina mask is removed, leaving the array of ferroelectric nano-capacitors easily accessible.