Reversible Phase Change Characteristics of Cr-Doped Sb2Te3 Films

Jul 26, 2016 - State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology, and. §. Sh...
0 downloads 0 Views 9MB Size
Download PDF
Research Article www.acsami.org

Reversible Phase Change Characteristics of Cr-Doped Sb2Te3 Films with Different Initial States Induced by Femtosecond Pulses Qing Wang,†,‡,§ Minghui Jiang,‡,∥ Bo Liu,*,†,§ Yang Wang,*,∥ Yonghui Zheng,†,‡,§ Sannian Song,†,§ Yiqun Wu,∥ Zhitang Song,†,§ and Songlin Feng†,§ †

State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology, and Shanghai Key Laboratory of Nanofabrication Technology for Memory, Shanghai Institute of Micro-system and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China ‡ University of the Chinese Academy of Sciences, Beijing 100049, China ∥ Key Laboratory of High Power Laser Materials, Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Shanghai 201800, China §

ABSTRACT: As a kind of chalcogenide alloy, phase change material has been widely used as novel storage medium in optical disk or electrical memory. In this paper, femtosecond pulses are used to study the reversible phase transition processes of Cr-doped Sb2Te3 films with different initial states. The SET processes are all induced by multiple pulses and relate to the increase of crystallized partial in the irradiated spot. When the Cr concentration is 5.3 at % or 10.5 at %, the crystallization mechanism is still growth-dominated as Sb2Te3, which is beneficial for high speed and high density storage, whereas the necessary crystallization energy increases with more Cr-dopants, leading to higher amorphous thermal stability. RESET results by multiple pulses show that Cr-dopants will not increase the power consumption, and the increase in Cr-dopants could greatly increase the antioxidant capacity. Single-pulse experiments show that the RESET process involves the competition of melting/amorphization and recrystallization. The reversible SET/RESET results on different initial states are quite different from each other, which is mainly due to the different surroundings around the irradiated spot. Crystalline surroundings provide higher thermal conductivity and lead to easier crystallization, whereas amorphous surroundings were the reverse. All in all, Cr-doped Sb2Te3 films with suitable composition have advantages for storage with high density, better thermal stability, and lower power consumption; and the suitable initial states could ensure better reversible phase transition performances. KEYWORDS: phase change, femtosecond pulse, Cr-doped, Sb2Te3, initial state

1. INTRODUCTION

The possibility of nonthermal phase change has been theoretically and experimentally proposed for nucleationdominated Ge2Sb2Te5 (GST) alloys11−19 and growth-dominated AgInSbTe alloys.18,19 Cr-doped Sb2Te3 alloy with suitable composition has been proved to be a novel phase change material with high speed and good thermal stability used in PCRAM.20 In this paper, femtosecond pulses are used to study the reversible phase change processes of Cr-doped Sb2Te3 films with selected compositions, which are in different initial states, in order to get more information about their crystallization mechanism, power consumption for amorphization, and the influences of different initial states.

The concept of phase change data storage was demonstrated by S. R. Ovshinsky more than 40 years ago.1 It is based on a rapid and reversible phase transition of amorphous-to-crystalline structure of chalcogenide phase change alloys, which is induced by controlled heating and cooling, and the two structure states generally have distinguishable differences of optical and electrical properties.2−5 In phase change optical disks, for example, blu-ray discs, heat is delivered to the film by a short laser pulse, and the data are also read optically, based on the reflectivity difference between amorphous and crystalline phases.3,6 In electrically addressed storage device, which is socalled phase change random access memory (PCRAM), the phase transition is induced by controlled electrical pulses and the data is also read electrically based on the resistance difference between the two phases.2,7 More recently, ultrashort (ps and fs) laser pulses were employed to obtain more detailed information on phase transitions by nonthermal methods.8−19 © XXXX American Chemical Society

Received: June 4, 2016 Accepted: July 26, 2016

A

DOI: 10.1021/acsami.6b06667 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces

process for group 1−2 is SET process with multiple pulses; for group 3−5 it is RESET process with multiple pulses; for group 6−7 it is RESET process with single pulse. Before the studied SET/RESET processes, the films were pretreated with different methods, as shown in Table 2. The pretreated films are defined as in initial states, which were labeled in Table 2 for convenience. In these labels, As stands for as-deposited state, C stands for crystalline state after annealing at 300 °C for 5 min, with Ar atmosphere as protection, RE or S stands for RESET or SET operation on the irradiated spot with optimized femtosecond pulse number and fluence, 4m in Group 5 means that the samples were exposed in atmospheric environment for 4 months before any other pretreatment. In addition, 80 nm GST film was also studied in Group 1 as comparison. A pair of microtopographies of a CST_10.5 film (40 nm) in As initial state and S final state (related to group 1) were taken by transmission electron microscopy (TEM), in order to intuitively show the structure transition that is mainly responsible for the reflectivity change.

2. EXPERIMENTS 2.1. Sample Preparation. Eighty nanometer Sb2Te3 and Crdoped Sb2Te3 films were deposited on Si (100) substrates by magnetron cosputtering using separate Cr and Sb2Te3 targets. Details of the deposition conditions and composition analysis can be found in ref 20. Four different Cr-doped Sb2Te3 films with Cr concentrations from 2.4−17.5 at % were studied in ref 20, whereas we selected 5.3− 10.5 at. % as a suitable concentration range after a trade-off of amorphous thermal stability, composition stability, and data resolution, when used in PCRAM. So in this paper, the two representative compositions, together with pure Sb2Te3, were studied in detail by ultrafast laser pulse irradiation. For convenience, the composition ratios of them are shown in Table 1, which are the same as in ref 20.

Table 1. Composition Ratios (%) of Different Films and the Related Labels20 power of Cr/ Sb2Te3

Cr (±0.7 at %)

Sb (±1 at %)

Te (±1 at %)

label

0/20 W 7/20 W 10/20 W

0 5.3 10.5

42.4 41.7 40.7

57.6 53.0 48.8

Sb2Te3 CST_5.3 CST_10.5

3. RESULTS AND DISCUSSION 3.1. Sample Quality Detection. STEM experiments tell that the element distributions are almost uniform in all the studied areas, whether the films are in As state or C state. Most of the STEM results are not shown here; Figure 2a shows only the element distributions in one typical area of CST_10.5 crystalline film as a representative. No visible cluster or segregation is found. In Figure 2b, the atomic percentages of 8 spectrum points randomly selected from Figure 2a are shown in detail. It also indicates a compositional uniformity in the observed area. Figure 2c, d shows the change of binding energies of Sb and Te atoms when Cr dopants are introduced in. It is known by experience that when the bond of L-M turns to be L-N, the binding energy of L will increase if the electronegativity of N is larger than that of M, or conversely.21 The electronegativity of Cr, Sb or Te is 1.66, 2.05, or 2.1, respectively. The binding energy of Sb decreases after the doping of Cr, as well as Te. It tells that some of the Sb and/or Te in Sb2Te3 crystal lattice are probably substituted by Cr, which could also be inferred by XRD results, as has been discussed in ref 20. The compositional uniformity of As state films, which are almost amorphous, is mainly controlled during deposition by selected sputtering conditions, such as the rotation of sample table, suitable gas flow, et al. Theoretically, the high degree disorder of amorphous film leads to an extremely thermodynamic nonequilibrium state, generally companied with a large undercooling and a relatively high solid solubility. Therefore, the compositional uniformity could be easily realized if the film is uniformly deposited, even though the concentration of Cr dopant is as high as 10.5 at %, whereas for crystalline films, cluster or segregation may occur when the concentration of Cr dopant is high, because of the limit of solid solubility in crystals. However, in this paper, the element distributions still keep quite uniform even with 10.5 at % Cr dopants in the crystalline film. Here, the crystalline CST_10.5 film was achieved by annealing the related amorphous film with original wonderful uniformity. It was in polycrystalline state and the grains were relatively small (tens of nanometer).20 The grain boundary stress and defect stress may play a positive role to balance the stress caused by extra Cr dopants in the small crystal grains. On the other hand, XPS results show that Cr dopants probably bond with Sb/Te atoms, which is quite beneficial for maintaining the compositional uniformity during different operations. In conclusion, the film quality in this paper is

To detect the film quality, we used scanning-transmission electron microscopy (STEM) to find out the distributions of different elements in the films of as-deposited states and annealed crystalline states (40 nm, annealed at 300 °C for 5 min). X-ray photoelectron spectroscopy (XPS) was used to explore the effects of doping-Cr on the bonds of different elements in CST_10.5; the films for XPS were 200 nm in thickness and annealed at 300 °C for 5 min. Oxidation was removed by Ar+ etching for 240s. The experimental results show that the studied films are uniform in composition, and Cr dopants probably bond with the matrix elements, which will be discussed in detail in section 3.1. 2.2. Experiment Design. Figure 1 shows the pump−probe system for real-time in situ reflectivity evolution measurement. The pump

Figure 1. Experimental setup for real-time in situ reflectivity measurements during femtosecond laser-induced phase change process. light was used to trigger the phase transition process, with a pulse duration of approximately 130 fs and a wavelength of 800 nm, which could form an irradiated spot of about 1 mm in diameter on the sample, using a convexo-convex lens with focusing length of 150 mm. The frequency for multipulse experiments was 1000 Hz. The probe light with continuous wave of 650 nm was focused at the center of the irradiated spot with a diameter of approximately 0.3 mm, and its intensity was decreased by an attenuation slice in order to eliminate its influence on the sample. The data were collected by a high-speed photodetector connected to a fast digital phosphor oscilloscope. The total time resolution of the pump−probe system was about 2 ns. In this paper, we define the operation from low reflectivity to high reflectivity as SET process, and the reverse operation as RESET process. Experiments have shown that the difference in reflectivity should attribute to different fractions of amorphous state to crystalline state.11 Therefore, SET process relates to the increase of crystalline state fractions, while RESET relates to the increase of amorphous state fractions. We organized totally 7 group experiments. The studied B

DOI: 10.1021/acsami.6b06667 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces Table 2. Experiment Process Designa

a

It is the pretreatment process above the dotted line and the studied process under the dotted line.

Figure 2. (a) STEM spectra and (b) atomic percentages of different spectrum points of crystalline CST_10.5 film; (c, d) XPS results of CST_10.5 compared with Sb2Te3 film.

reliable, which could ensure the repeatability and comparability of the following experiments. 3.2. SET. Figure 3a−c shows the SET processes on different films with As initial states, which are induced by multiple pulses with different fluences. Generally, because of the existence of a

minimum time required for crystalline nucleation to form and grow,22 the crystallization process is obviously not favored by using ultrashort laser pulses.23 Successful crystallization induced by single ultrafast pulse (ps/fs) can be fulfilled only under some well-chosen parameters by careful tailoring of heat flow C

DOI: 10.1021/acsami.6b06667 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces

Figure 3. SET processes on different films with (a−c) As states and (d−f) RE/C states induced by multiple pulses with different fluences (mJ/cm2).

conditions, such as the control of film thickness10 and substrate,10,12 well-primed initial state and selected fluence,11 et al. In this paper, we focused on the comparison of different films, so multiple pulses were used to promote easier crystallization. For crystallization of As state films induced by single ultrafast laser pulse under picosecond, recalescence phenomena, i.e., the melting process before final crystallization, is widely observed in Sb-based growth-dominated phase change materials, such as GeSb8−10 and AgInSbTe,18,19 whereas it is hardly observed in nucleation-dominated GeSbTe films.11−13 In Figure 3a−c, recalescence phenomena can also be observed in multipulse induced crystallization process for Sb2Te3 and CST films (related to stage I), whereas it is not observed for GST film (inset of Figure 3a), which indicates that the crystallization of CST films with suitable Cr concentration should be growthdominated, as same with pure Sb2Te3. On the other hand, we can see that the pulse number and fluence for Sb2Te3 and CST films to reach complete crystallization are much more than those for GST. The main reason should be that the crystallization threshold power and complete-crystallized time for growth-dominated films generally significantly increase with increased size of irradiated spot, while they are not sensitive for nucleation-dominated films;18 and the size of irradiated spot in this paper is about 1 mm in diameter, which is too large compared with the recording mark (tens to hundreds of

nanometer) in phase change optical disk. Therefore, growthdominated phase change films have great advantages in speed when used as storage medium with high density. What’s more, it can be seen in Figure 3a−c, when Cr atoms are doped into Sb2Te3, the reflectivity decrease in stage I becomes much weaker. It has been known that the difference in reflectivity should attribute to different fractions of amorphous state to crystalline state.11 In fact, with quite low crystallization temperature, the As state Sb2Te3 film has more fraction of crystalline state. Therefore, the reflectivity could decrease to much lower value during stage I, which indicates the appearance of a lot fraction of melting/amorphous state. Crdopants have been proved to increase the crystallization temperature of amorphous state,20 so As state CST films are almost in total amorphous state; therefore, the reflectivity decreases in stage I are not obvious as Sb2Te3. Figure 3d−f show the SET processes on different films with RE/C states induced by multiple pulses with different fluences. It can be seen that recalescence phenomena is obviously suppressed for all films, and the crystallization process is greatly promoted compared with Figure 3a−c, which are most probably due to the different surroundings of irradiated spot, as shown in Table 2. Generally for phase change materials, the thermal conductivity of crystalline state is larger than that of amorphous state.24,25 For RE/C state, the surrounding of the D

DOI: 10.1021/acsami.6b06667 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces

Figure 4. Comparison of SET results for films with (a) As states and (b) RE/C states induced with different pulse number; (c) the best values of fluence for SET processes; and (d) the TEM morphologies of a CST_10.5 film in As initial state and S final state. Notice: the best values shown in c relate to the points in green circles in panels a and b.

irradiated spot is in crystalline, instead of amorphous for As state. When the pulse energy is deposited in the spot, it is easier and quicker to be diffused through the crystalline surrounding, which results in the inability of enough temperature rise for possible melting in recalescence process. In particular, for Sb2Te3, the irradiated spot after RESET should have much more fraction of amorphous state than As sate film, which also leads to less reflectivity decrease during recalescence process. On the other hand, the interface between the spot and its crystalline surrounding could greatly decrease the crystallization barrier by turn the nucleation mode from homogeneous to heterogeneous nucleation, which is especially beneficial for growth-dominated material.26 In Figure 4, the SET results for different films with As or RE/ C initial states are detailed compared and analyzed. First, it clearly shows that the necessary pulse number to induce highdegree crystallization (reflectivity contrast >20%) for RE/C state (50−100 pulses) is much less than that for As state (100− 200 pulses), which is due to their different surroundings and has been discussed above. Second, with a certain pulse number, when the fluence increases, the reflectivity contrast will increase at first and then decrease, i.e., there exists a best value of fluence. It is because that high fluence may lead to partial melting, which results in the competition of amorphization to crystallization. Figure 4c shows these best values, which relate

to the points in green circles in Figure 4a, b. At the same initial state, the best value for CST_5.3 is the lowest, whereas that for Sb2Te3 is the highest, which may partly originate from their different RESET power consumption for amorphization as shown in Figure 7 a, whereas with the same composition, the best value for the RE/C state is lower than for the As state. The total energy needed for crystallization depends on the product of fluence and pulse number. Therefore, the needed crystallization energy for RE/C state is much lower than for As state, and the RE/C state is more common in practice. For RE/C state shown in Figure 3d−f and Figure 4b, the needed pulse number for complete crystallization of Sb2Te3 is much less (75), although with a bit higher best value (as shown in Figure 4c). Therefore, the total crystallization energy for Sb2Te3 is the lowest; and the more Cr concentration, the higher crystallization energy is needed. It is welcome because higher needed crystallization energy generally indicates better amorphous thermal stability. What’s more, for RE/C state, the reflectivity contrast of CST film is higher than Sb2Te3 film, which contributes to better data resolution. In addition, as an example of structure transition induced by femtosecond pulses, Figure 4d shows the TEM morphologies of a CST_10.5 film in As initial state and S final state, which further confirms the fact that the change of reflectivity is mainly E

DOI: 10.1021/acsami.6b06667 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces

Figure 5. RESET processes on different films with (a−c) C states, (d−f) S/As states and (g−i) 4m-S/As states induced by multiple pulses with different fluences (mJ/cm2).

Figure 6. RESET processes on CST_5.3 films with different states induced by single pulse with different fluences (mJ/cm2).

we tried a lot of effort to conduct SET-RESET cycles on one spot of Sb2Te3 film with different fluences, and found that only the first RESET operation could lead to high reflectivity contrast (>30%), whereas during the following cycles, reflectivity contrasts are quite low (