Imec researchers have developed a novel technique – termed conductive atomic force microscopy tomography (or scalpel C-AFM) – that enables a three-dimensional characterization of emerging logic and memory devices.
BY UMBERTO CELANO, imec, Leuven, Belgium
Umberto Celano, using the novel scalpel C-AFM tool.
With the introduction of three-dimensional devices (such as FinFETs) and stackable architectures (such as vertical NAND Flash memories), there is a growing need for 3D characterization techniques. These techniques should not only be capable of probing in three dimensions and examining the topological properties. They should also enable an analysis of the electrical properties of the 3D nano-sized volumes.
A shining example illustrating the need for this technique are conductive bridging random access memory (or CBRAM) devices. These devices belong to the emerging class of resistive RAM (or RRAM) memories which exhibit a fast operation, low power consumption, high endurance and high scalability. They are currently seen as a candidate memory technology for application in storage class memories and embedded non-volatile memories. Their operation basically relies on the formation of a highly conductive path, the conductive filament, in a poorly conductive medium. But the formation of this filament in an integrated device has so far never been observed with the techniques available today. A full 3D characterization of the conductive filament would considerably enhance our understanding of the filament growth dynamics and the underlying physical mechanisms. And it would enable a further optimization of the memory device.
Scalpel C-AFM, extending the 2D capabilities of C-AFM
A well-known characterization technique for advanced logic and memory devices is scanning probe microscopy (or SPM), where a sharp tip slides on a flat surface.
The 2D-maps of electrical properties provided by this technique have for many years enabled the understanding and development of advanced planar technologies at the nanoscale. SPM comes in several flavors, such as scanning tunnel microscopy (STM), atomic force microscopy (AFM), and a whole range of secondary analysis modes such as conductive AFM (or C-AFM). C-AFM is based on contact-mode AFM using a (biased) conductive tip. The topography is measured in contact-mode, while the current flowing between the biased sample and the tip is recorded simultaneously.
Researchers at imec have now evolved the C-AFM technique into a 3D characterization tool, suited to probe very confined volumes at the nanoscale. The new method consists in collecting the C-AFM images of the sample at different depths. The sectioning is induced by a controlled material removal. This is done by applying a strong pressure (GPa) between the (biased) conductive-diamond tip and the sample during the C-AFM scan. This way, sub-nm vertical removal rates are obtained. Since the diamond tip acts as a scalpel, the new method is referred to as scalpel C-AFM. The technique can be used for a wide variety of materials, and can be extended to other contact-mode AFM methods such as scanning spreading resistance microscopy.
Case: CBRAM memory devices
The imec researchers have used the scalpel C-AFM technique for studying the conductive filament formation in CBRAM memory devices. In these devices, an abrupt change in electrical resistance occurs when the device is subjected to a voltage pulse. The different resistance states are induced by the formation or dissolution of a highly conductive filament into a poorly conductive medium.
The heart of the CBRAM memory cell is a thin dielectric (e.g., Al2O3) that is sandwiched between the active electrode (Cu or Ag) and an inert counter electrode (e.g., TiN). When a positive voltage is applied to the active electrode, a field-assisted injection and transport of cations begins. This leads to the creation of the conductive filament inside the Al2O3 oxide layer. The presence of this filament dramatically lowers the resistance of the device, leaving it in a low resistive state (LRS). The conductive filament can be dissolved by applying a negative voltage to the active electrode and thus restoring a high resistance state (HRS). The two different resistance states are used as the logic values 1 or 0 for data storage applications. The overall performance of the device is highly related to the properties of the conductive filament, which has so far not been observed in 3D on scaled devices.
Observation of the conductive filament
The memory device under investigation is a Cu/5nm Al2O3/TiN-based memory, integrated in a one-transistor-one-resistor configuration. The device is placed at the cross-point between the bottom and top electrode. The scalpel C-AFM technique was applied to memory devices programmed in both the low and high resistive state. An in-house fabricated conductive- diamond tip was used for probing and removing the material.
By using the scalpel C-AFM technique, the researchers were able to observe, for the first time ever, the conductive filament formation which is responsible for the resistive switching behavior in CBRAM devices (FIGURE 1). The observed conductive filament, embedded in the Al2O3 oxide, shows a conical shape: it shrinks moving from the active electrode (Cu) towards the inert electrode (TiN). The low resistive state is created when the conductive filament eventually shorts the two electrodes.
FIGURE 1. CBRAM device: Cross-section transmission electron microscopy (TEM) image of the CBRAM memory device (left) and the device stack (middle), and AFM image of the cross-point area (right).
The experiments suggest that the dynamics of the conduction filament growth are limited by the mobility of the Cu cations in the electrolyte (FIGURE 2). When the bias is reversed, a Joule-heating assisted electro-chemical reaction is responsible for the rupture of the conductive filament (the high resistive state).
The study also demonstrates the close correlation between the programming current, the physical volume of the conductive filament and the resistance. A larger programming current induces a larger physical volume and a lower resistance value of the conductive filament. Hence, by controlling the programming current, the resistance can be modulated. This opens the possibility of creating multiple resistance sates in one single memory cell, which can considerably enhance the memory density of non-volatile CBRAM devices.
Scalpel C-AFM will rapidly find applications in other emerging technologies as well. At imec, the technique is currently being used for investigating vertical NAND Flash memory devices and oxide-based RRAM memory devices.
FIGURE 2. filament growth model: Illustration of the eletrochemical processes during resistive switching. (1) First, the Cu oxidizes and Cu+ ions are injected in the Al2O3. Second, the high electric field might lead
to the formation of oxygen vacancies in the dielectric layers (white balls in the cartoon). (2) The slow
migration of Cu+ ions in the switching layer implies
that a reduction reaction occurs before the Cu+ reaches the inert-electrode. (3) The conductive filament (CF) growth continues and the CF eventually shorts the two electrodes thereby creating the low resistive state. (4) When the bias is reversed, a Joule-heating assisted electrochemical reaction is responsible for the rupture of the CF in the point of max power dissipation, that is, CF constriction.
Suggested additional reading
‘Three-dimensional observation of the conductive filament in nanoscaled resistive memory devices’, U. Celano et al., Nano Letters, 2014. http://pubs.acs.org/ doi/abs/10.1021/nl500049g.
‘The memory roadmap, a paradigm shift from 2D to 3D’, interview with imec’s Jan Van Houdt in imec magazine, March issue. http://magazine.imec.be/ data/57/reader/reader.html#preferred/1/package/57/ pub/63/page/4.
UMBERTO CELANO is PhD student in the Material and Component Analysis (MCA) group at imec, Leuven, Belgium.
