Towards low energy consumption data storage era using phase-change probe memory with TiN bottom electrode

1 Introduction

There is no doubt that today’s world is in the age of big data where everything strongly related to the daily life of citizens is linked to a data source. The advent of the age of big data can be attributed to the prosperity of the social networks such as Facebook, Twittter, and Instagram, as well as the digitization of the daily service. Thanks to the age of big data, the total amount of global digital data has been sharply proliferating during the last 5 years, which has surpassed 4.4 ZettaBytes in 2013 and is predicted to outpace 44 ZettaBytes by 2020 [1]. This indeed necessitates the requirement for aggressively increasing the recording capacity of the conventional data storage devices. Unfortunately, the conventional mass storage devices have suffered from their respective physical limits, such as superparamagnetism for hard disk, optical diffraction for optical disc, and scaling limits for flash memory [2]. In this case, it is timely for new, emerging memory device to enter the storage field.

In comparison to other concurrent memory devices, phase-change probe memory (PCPM) has been considered as one of the most promising candidates due to its ultra-high density, short switching time, long data retention, and low cost [3]. PCPM usually makes use of a conductive scanning probe with nanoscale size and a tri-layer stack that comprises a Chalcogenide alloy (particularly Ge₂Sb₂Te₅) sandwiched by a capping layer and a bottom layer. During the recording process, a write current is injected to Ge₂Sb₂Te₅ media through the conductive probe tip to either heat the media to crystalline temperature for crystallization or to the melting temperature followed by a rapid cooling to generate amorphization. The readout process is achieved by detecting the difference on the sensing current between highly conductive crystalline state and highly resistive amorphous state. Due to the attractive traits of PCPM, detailed investigations have previously been dedicated to the writing and readout performance of PCPM, resulting in emergence of various phase-change stacks [3], [4], [5]. In spite of the diversity, a capping layer that can protect Ge₂Sb₂Te₅ media from oxidation and wear and a bottom layer utilized to collect write current are unanimously required by any phase-change stacks. In addition to the aforementioned functions, the electrical and thermal conductivities of the capping layer also need to be relatively low so as to prevent the current leakage and the heat dissipation. In this case, diamond-like carbon (DLC) film seems to be the most suitable material for capping layer, which has been extensively adopted by researchers worldwide [3], [4], [5], [6]. In contrast to the capping layer, the physical properties of the bottom layer has received less attention, as it is only used to collect writing current. However, as the bottom layer acts as an electrode, it is desired to be more electrically conductive than the capping layer, thus enabling a reduction on the required current for recording as well as the resulting energy consumption. Nevertheless, to the best of the authors’ knowledge, the sole material used for the bottom electrode of PCPM to date is the DLC that, however, exhibits a relatively higher resistivity [7], causing extra energy consumption. In this case, it would be necessary to explore alternative materials with higher electrical conductivity than DLC to target low energy consumption.

2 Experiment

In addition to DLC, titanium nitride (TiN) has recently been considered as another prospective coating material because of its excellent thermal and mechanical properties, low electric resistivity, and metallurgical stability at high temperature [8], [9]. Today TiN has already received extensive applications, ranging from hard and protective coating in the ornament industry to diffusion barrier layers in the semiconductor industry [10]. Besides the above applications, TiN has also been adopted as the main constituent of the heater for phase-change random access memory (PCRAM) during the last decade [11], [12]. The TiN heater that bridges the top electrode and the bottom electrode allows writing current to pass through the phase change materials and thus induce the phase transformation at the heater to phase change materials contact because of the resulting high current density. However, the practicality of using TiN film as the bottom electrode of any type of phase-change memories including PCRAM and PCPM still remains mysterious. As a result, the electric resistivity of TiN film with respect to thickness and exposure time at room temperature is measured, which is compared with the DLC film. In our experiment, TiN films are deposited in a silicon (Si) substrate by DC reactive magnetron sputtering system from a 5 cm high purity (99.999%) titanium (Ti) target in a mixture of high purity argon (99.999% Ar) and nitrogen gas (99.999% N₂). The dynamic base pressure of 2.2×10^-5 Pa and the power of 5 kW are used for experiments. The Ti target is subjected to presputtering process for 2 min in Ar atmosphere prior to sputtering to remove the surface oxide and nitride layer of the target. Similar to Ti film, DLC films are also DC sputtered in argon on Si substrate with a distance of 3.2 cm from the target. The sputter rate and the sputter pressure are 3 kW and 50 sccm, respectively. This deposition condition for DLC film is chosen in a manner such that the lowest electrical resistivity can be secured for a given thickness and a given exposure time. Electric resistivity of TiN and DLC films are measured using four-point probe techniques with pads of indium used when required to obtain a low and consistent contact resistance.

3 Results and discussions

The thickness and time dependence of resistivity for DLC film and TiN film are shown in Figure 1A and B, respectively.

Figure 1:

The thickness and exposure time dependent resistivity of (A) DLC film and (B) TiN film.

As can be seen from Figure 1A, for the case of zero exposure time (blue curve), the resistivity of DLC film exhibits strong thickness dependence, as the resistivity varies from 0.7 ohm m to 0.02 ohm m with the thickness ranging from 5 nm to 30 nm. The fact that lower resistivity can be obtained by increasing the film thickness may imply that it is preferable to have a thick DLC film as the bottom electrode of the PCPM. Besides, Figure 1A also reveals the stability of the resistivity of the DLC film over exposure time. For a given thickness of 5 nm, it is found that the DLC resistivity changes from 0.7 ohm m to 3 ohm m, as the exposure time varies from 0 h to 480 h. However, the dependence of the resistivity on the exposure time seems to be degraded along with the increase of the film thickness, and there is no clear time dependence found for the thickness of DLC film >20 nm. The dependence of the resistivity on film thickness is also found on TiN film, as the resistivity varies from 8×10^-5 ohm m to 0.2×10^-5 ohm m with the thickness ranging from 2 nm to 25 nm. However, for TiN thickness more than 5 nm, the resistivity almost remains constant. Compared to DLC film, the resistivity of TiN films shows less time dependence, as the resistivity varies from 8×10^-5 ohm m to 2×10^-4 ohm m with the exposure time between 0 h and 480 h. Moreover, there is no eminent time dependence observed when the thickness of the TiN film exceeds 5 nm. According to the findings above, the capability of TiN films to have much lower resistivity with better stability than DLC film has rendered it an attractive contender for the bottom electrode of the PCPM.

Considering the superior characteristics of TiN to DLC film, we redesign the previous PCPM architecture by replacing the DLC film with the TiN layer as the bottom electrode. The modified phase-change stack consists of a Ge₂Sb₂Te₅ layer sandwiched by a DLC capping layer and a TiN layer, which are deposited on Si substrate. The thickness of the Ge₂Sb₂Te₅ layer and the capping layer are optimized to be 10 nm and 5 nm, respectively, to reduce the required threshold voltage based on previous results [13]. The thickness of the TiN bottom electrode is chosen to be 25 nm so as to minimize the resistivity of the TiN layer, thereby reducing the required writing pulse. In addition, the thermal conductivity of the TiN layer at 25 nm is reported to be around 0.5 W m^-1 K^-1, which can substantially suppress the heat dissipation through the Si substrate [14]. A silicon dioxide (SiO₂) encapsulated Si tip with platinum silicide (PtSi) at the tip apex is employed as the recording means. The use of PtSi mitigates the conduction and the anti-wear properties of the tip itself, while SiO₂ encapsulation increases the physical diameter of the tip and therefore effectively reduces the pressure between tip and sample [15]. Based on the new PCPM system, the writing of crystalline bits is experimentally performed using a 2 V triangular pulse with a 50 ns rising time and a 50 ns trailing time. After the recording process, a readout potential of 1 V is directly applied to the conductive tip to measure the readout current to establish the size of the written bit, giving rise to Figure 2A. The readout resistance image illustrated in Figure 2A shows that a written bit with a diameter of around 48 nm is formed, thus resulting in an ultra-high recording density. Note that smaller tip contact areas would result in smaller written bits, demonstrating the capability of this technique in entering Terabit-per-square-inch density regime. The corresponding current-voltage (I-V) curve, shown in Figure 2B, suggests a threshold voltage of 1.8 V and a maximum writing current of 100 μA, while for DLC bottom electrode, a 5 V writing pulse of 200 ns is usually required to yield the similar sized bit accompanied with a writing current of 286 μA. Moreover, the maximum readout current, as illustrated in Figure 3, is found to be around 10 μA. It is also indicated in Figure 3 that the readout current induced when tip is in contact with crystalline phase is much higher than the amorphous phase, thus readily discriminating the crystalline bit from the surrounding amorphous background.

Figure 2:

(A) The crystalline bit obtained from PCPM with TiN bottom electrode using a 2 V pulse of 100 nss duration and (B) the resulting I–V curve.

Figure 3:

Variation of readout current along with tip position.

In addition, the writing rate for this system is approximately 10 Mbit/s which, however, can be readily boomed to 10 Gbit/s by simultaneously using 1000 probes. Furthermore, the use of TiN bottom electrode allows for a lower crystallization voltage than DLC electrode, thus heavily mitigating the cycling performance of the whole system. Most importantly, the required written energy for the resulting bit is about 20 nJ, much lower than the energy consumed using the similar system but with DLC bottom electrode (around 300 nJ) [16]. This implies that the energy consumption for the use of 100 tips and 1000 tips will be 2 μJ and 20 μJ, lower than any other probe memory devices [17]. However, a fact that mutual migrations between Ti atoms from Ti-rich TiN film and Te and Sb atoms from the Ge₂Sb₂Te₅ layer may occur at TiN-Ge₂Sb₂Te₅ interface is likely to impair the thermal conductivity of the storage device. In this case, the Ti/N atom ratio needs to be determined carefully [18]. The characteristic performances of the PCPM systems with different capping and under layer configurations are shown in Table 1. Obviously, the PCPM device with DLC capping layer and TiN under layer exhibits the best writing performance, heavily prevailing over other competitors. As a result, the feasibility of using TiN media as the bottom electrode of the PCPM system to provide much lower energy consumption than DLC layer in conjunction with the capability of having ultra-high density and large data rate has been demonstrated in this paper.

4 Conclusions

The practicality of using TiN to replace DLC as the bottom electrode of the PCPM system is investigated. The electric resistivity of both TiN and DLC show a thickness and exposure time dependence, which gradually attenuates along with the increase of the film thickness. Compared to the DLC film, TiN film exhibits a much lower electric resistivity with better stability over exposure time. As a result, a new PCPM architecture with TiN bottom electrode is fabricated where the writing of crystalline bits is experimentally conducted. Results show that the improved system with TiN bottom electrode is able to secure ultra-high recording density and large data rate while at the cost of much lower energy consumption than the system with DLC electrode.