1. Introduction

Phase-change materials (PCMs) alloys are now commercially used in both optical and electronic data storage technologies [1–9]. The phase change process is usually modelled as a short heat pulse inducing a fast and reversible amorphous ↔ crystalline phase transition. The chalcogenide materials that can support resonant bonds tend to exhibit a large optical and electrical contrast between their crystalline and amorphous states [10, 11]. The excellent scaling [12, 13] and fast structural transitions [14] makes these materials attractive for new applications in nanophotonics [15, 16]. In particular, the pronounced contrast of the refractive index in the visible and near-infrared are exploited to design plasmonic absorbers, colour filters, metamaterial devices, meta surfaces, and integrated active photonics devices [17–20]. Properly characterising and understanding the phase transition process, optical nonlinearities, and the temperature dependent optical properties will aid the design of PCMs for new functional photonics applications.

The crystal structures of PCMs are inherently unstable and their measured properties are strongly dependent on the measurement procedure. For example, the electrical resistance change in the cubic structure of Ge₂Sb₂Te₅ is several orders of magnitude due to disorder induced localisation [21]. The structural changes from crystalline to amorphous tend to be stochastic due to activated thermal diffusion-driven processes. Hence, the PCM structural state is dependent on the history and any previous switching attempts can significantly impact the measured crystallisation time [22].

In order to compare the switching performance of PCMs developed by different research groups, a standard testing protocol should be established. One of the most common forms of testing switching time and switching energy of phase change materials is by “static testing”. Static testers were originally designed to test the performance of rewritable optical disc materials. The term ‘static’ refers to the fact that the discs are not spinning during the test. In contrast, dynamic disc testers (DDTs) are used to measure the disc writabilities and erasabilities whilst the disc spins.

In recent years phase change material research has moved from developing disc materials to materials for phase change electrical data storage, and now the focus is on designing functional materials for active nanophotonics. Characterising the phase transition speed with the DDT is inconvenient, since a new optical disc is required to test each new composition. Analysis of the phase transition process is also made difficult by the relative movement between the laser heating spot and the disc. Moreover, the DDT discs require homogeneous films of uniform thickness, and this prevents using high throughput design parameter-spread combinatorial techniques to efficiently assess different materials and structures [23]. In contrast, the static tester probes microscale areas on the sample, where the thickness and composition is locally uniform. Yet on a larger scale, the composition can be varied, which in turn allows mapping of the switching performance against composition, or other spatially varying design parameters [24]. For these reasons pump-probe static testers are the systems of choice for characterising phase change materials.

The first static tester was developed by IBM Research Laboratories to analyse reversible phase change optical data storage materials. The system was based on a Kr-ion laser that was pulsed by an acousto-optic modulator. The Kr-ion laser was used for both writing and reading crystalline and amorphous marks. A second laser was used for autofocus by focusing to a spot offset from the written marks [25]. The same group introduced phase transform kinetic (PTK) plots, which are used to study the laser pulse dependent crystallisation and amorphisation of thin films. Essentially, the plots display a matrix of reflectance or transmittance, laser pulse power, and laser pulse duration. The PTK plots are particularly useful for comparing the switching parameters for different material systems and optical structures.

With the ever increasing optical output power of laser diodes, Mansuripur et al., designed a static tester to measure the optical reflectivity of the PCMs [26]. The pump (680 nm) and probe (643 nm) laser diodes were capable of producing pulses with 5 ns rise and fall times with optical powers of 40 mW and 50 mW respectively. The two laser system was used to compare and study the crystallisation and amorphisation PTKs of different PCMs. Moreover, using two lasers in a pump-probe arrangement allowed in-situ reflectivity measurements during the switching process. The system was also able to measure the magneto-optical properties of thin films. However, using a bulk electromagnet in the experimental setup prevented the transmission measurements.

A fully automated static tester with pump (658 nm) and probe (635 nm) lasers was designed by Salinga to study laser-induced phase transitions in PCMs [22,27]. In this system, an autofocus system was used, and the sample’s reflectance was measured at 635 nm before and after heating with the 658 nm pump. This type of pre-pump–post-pump measurement does not provide information about the transient change in optical properties. Simpson also designed a system based on a 658 nm pump laser and a 635 nm probe laser [24,28]. It was capable of measuring changes in sample transmission as a function of time during and after the pump laser pulse. However, the system was limited to transmission measurements.

A slight variation on the static tester design was used by Rude et al., to measure switching of 1550 nm signals through Si ring resonators [29]. A Ge₂Sb₂Te₅ phase change material was patterned onto the surface of the ring resonator waveguide, and laser switching its phase tuned the resonant frequency of the waveguide. The static tester system consisted of two lasers, a 980 nm pump laser which was focussed onto the Ge₂Sb₂Te₅ section of the ring-resonator, and a 1550 nm fibre-coupled laser beam, which was diffracted into the ring resonator waveguide by grating couplers. The 1550 nm signal was switched with a 12 dm modulation depth when the phase of the Ge₂Sb₂Te₅ was switched between its crystalline and amorphous states. Moreover, partial crystallisation of the Ge₂Sb₂Te₅ by the 980 nm pump allowed for continuous tuning of the resonator’s resonant frequency. This was the first experiment where a static laser tester was used to characterise a photonics device. The same group have since used a similar set-up to switch Ge₂Sb₂Te₅ on surface plasmon polariton (SPP) waveguides, thus modulating the SPP transmission, and they have also employed a femtosecond pump-probe set-up to measure the resonance switching time of Ge₂Sb₂Te₅ tuned Au nanohole arrays [16,30]. A rather different method was used to modulate the transmission of light through a Si₃N₄ waveguide. The Ge₂Sb₂Te₅ was patterned on a tapered section of the waveguide. The 1560 nm pump light evanescently coupled from the waveguide at the Ge₂Sb₂Te₅ section, thus causing it to heat and reversibly switch the transmitted 1570 nm signal [31].

We aimed to develop a static tester that is highly stable, easy to align, and capable of simultaneously measuring the transmitted and reflected signals during and after the pump laser pulse. This ability to measure the transient change in reflectivity and transmission is important for studying the phase change process, and for assessing the volatile changes in optical properties for potential application in modulators. The system can also be used to perform pre-pump–post-pump measurements in a similar fashion to Salinga’s system [27].

In this work, we demonstrate a pump-probe static tester that utilises two fast photodetectors (1 GHz) to collect both transmitted and reflected probe signals. To enable easy alignment, the pump and probe laser beams were permanently coupled into a single mode optical fiber (SMF, λ = 600 nm) and aligned into a home-built microscope setup. The single mode propagates along the optic axis and ensures the spatial alignment of laser beams throughout the system. The amorphous ↔ crystalline switching time and energy was calculated by measuring the change in transmission and reflection of the thin films. A 14-bit camera with a resolution of 1294 × 734 pixels was incorporated into the system to image the sample surface. Additionally, the system can be used to study the nonlinear refraction and nonlinear absorption of the thin films without adding any additional optical components.

The system, overcomes the limitations of previously designed static testers by enabling transmission mode measurements, time resolved measurements, and by reducing bulky optical components [24, 26, 27]. The system allows the characterisation of phase transitions, thermally induced changes to the optical properties, and non-linear absorption/refraction of thin films by performing pre-pulse–post-pulse, time-resolved optical switching, and z-scan measurements. The measurements performed by this system facilitate the design of active materials for photonics and data storage technologies. Table 1 compares the static testers developed by different research groups.

Table 1. Comparison of static testers used by different groups

View Table | View all tables in this article

2. Experimental set-up design

A diagram of the laser pump-probe setup is shown in Fig. 1. The higher power 658 nm pump laser was used to induce a phase transition in the PCM film, whilst a 100 μW, 1 μs probe laser was used to measure the change in reflected and transmitted signal. To focus the probe beam within the write mark, a shorter wavelength of 635 nm probe laser was used. The maximum achievable pump laser power at the sample surface was 40 mW.

 

Fig. 1 (a) Schematic diagram of the dual laser pump-probe static tester. SMF, single mode fiber; BD, beam dump; FP, fiber port; CL, collimation lens; BE, beam expander; M1, M2, silver mirrors; DM, dichroic mirror; MOL1, MOL2, magnifying objective lens; CF1, CF2, bandpass line filters; BS, beam splitter; FL1, FL2, FL3, focusing lens; PD1, PD2, photo detectors; LED, light emitting diode. Mirror M1 is movable and only used for imaging the sample surface. (b) The optical microscope image of crystalline marks on an amorphous Ge₂Sb₂Te₅ sample.

Download Full Size | PDF

Both laser beams were coupled into a 90:10 evanescent coupler such that 90% of the pump beam and 10% probe beam powers were coupled into one of the two outputs of the coupler. The other outputs 10% of the pump and 90% of the probe; this was used for monitoring the coupling of the write laser beam and observing the laser’s stability. By coupling both pump and probe beams into a SMF, both beams are guaranteed to be aligned at the sample surface. This design minimises the use of bulky optics, hence improving the system’s stability. The combined beams were transmitted through a dichroic mirror (DM, λ_cut-off = 638 nm) and a microscope objective lens 0.5 NA (MOL1, working distance = 10.6 mm) to focus on the sample. The beam spot size was measured to be 1.08 ± 0.02 μm by the knife edge method [33, 34]

The amplitude and pulse duration of the pump beam was controlled by an Agilent 33250A arbitrary function generator. A bias-tee was used to add the RF signal to the DC signal. The output pulse from the bias-tee was fed through the pump laser to modulate the pump signal. With this set-up, the pump pulse duration can be controlled from 20 ns to 2000 ns and a pulse power up to 40 mW. To avoid continuous heating of sample, the probe laser was switched on for a duration between 1 μs and 1 s with a peak optical power of 100 μW at the sample surface.

The reflected light from the sample surface was collected by MOL1 and further reflected by the DM toward the photo detector (PD1). The dichroic mirror transmits 90% of the pump and reflects 65% of the probe beams. The transmitted light through the sample was collected by the MOL2 and focused onto the photo detector (PD2). Bandpass line filters (CF1, CF2) were used to separate the probe beam from the pump beam before it reached the photodetectors. The reflected and transmitted probe signals were detected simultaneously by the fast silicon photodetectors and the final data was measured by a 10 bit, 1.5 GHz digitising oscilloscope (NI PXIe-5162).

The white light source (LED), the focusing lens (FL3), the beam splitter (BS), movable silver mirror (M1), and the camera were attached to the system to image the sample. Mirror M1 was dismounted from the system during pump-probe measurements to avoid blocking the laser beams. The camera was used to find the focus position. The focused z-position was measured by a displacement probe (KEYENCE, GT-2-71N/71P) with 100 nm accuracy. This position was used in an automated focus locking feedback loop to ensure focussed measurements. The focal position was controlled by a z-axis nano-positioner (Piezosystem Jena, PZ-100) with 100 nm resolution. The laser pump-probe static tester system was fully automated and controlled by a LabVIEW interface.

2.1. Pre-pulse heating effect of probe laser operated in continuous-wave and pulse mode by Finite Element Analysis

A continuous-wave (CW) probe laser can lead to continuous heating of the sample and influence the optical switching measurements. We demonstrate this effect using a Finite Element Analysis (FEA) of a laser heat transfer model. The temperature of the PCM layer was calculated as a function of probe time and power.

A sandwich structure of SiO₂/Ge₂Sb₂Te₅/SiO₂ was used for the simulation. The energy transfer to the Ge₂Sb₂Te₅ layer was simulated by the heat conduction equation [35]:

In this equation, P is the power, R is the reflectivity, α is the absorption coefficient, d is the film thickness, ω is the Gaussian beam waist, r is the distance from the beam centre, and τ is the pulse duration. A laser pulse of 100 μW and 1 μs duration was considered for the simulation. The heat absorbed E_th (r) by the thin film was calculated. The thermal conductivities of Ge₂Sb₂Te₅ and the insulating SiO₂ layer were chosen to be 0.2 and 1.4 Wm⁻¹K⁻¹ respectively [37].

A short probe-pulse is essential to minimise unwanted heating of the sample. The temperature at Fig. 2 is shown as a function of time for a 1 μs, 100 μW pulse (dashed) and continuous wave (solid) laser heating. The 1 μs laser pulse increases the Ge₂Sb₂Te₅ temperature by 45 K after which the temperature drops due to heat dissipation and radiation to the surroundings. In contrast, the CW probe continues to heat the sample and after 2.5 μs, the temperature has increased above 400 K, which is close to the Ge₂Sb₂Te₅ glass transition temperature [38]. Clearly, even low power CW probe lasers are unsuitable for measuring properties of μm areas of phase change materials. For this reason, we used probe laser pulses to measure the material’s optical reflection and transmission.

 

Fig. 2 Two-dimensional FEA simulation of time dependent temperature of Ge₂Sb₂Te₅ layer for a pulse (dashed) and continuous (solid) probe laser.

Download Full Size | PDF

3. Demonstration of switching time measurements

The performance of our static tester is demonstrated by measuring the change in the transmission of a Ge₂Sb₂Te₅ sample as a function of time. 30 nm thick Ge₂Sb₂Te₅ films were deposited on fused silica and silicon (100) substrates by RF magnetron sputtering at room temperature. A SiO₂ film of 140 nm was used as the capping layer to avoid evaporation and oxidation of the Ge₂Sb₂Te₅ film during laser heating. We assume absorption in the SiO₂ capping layer is negligible at 658 nm. Ge₂Sb₂Te₅ films deposited on fused silica were used for the transmission-mode measurements while the Ge₂Sb₂Te₅ films on silicon substrates were used for the reflection-mode measurements. The as-deposited Ge₂Sb₂Te₅ samples were in the amorphous state.

The thickness of the phase change layer may influence the laser crystallisation and recrystallisation phase transformation kinetics. For example, increasing the thickness of the phase change material, Ge₂Sb₂Te₅, increases the crystallisation time [39]. Hence, a higher optical power is required for it to crystallise. In contrast, decreasing the film thickness beyond 10 nm substantially increases the crystallisation temperature due to the interfacial stress [13]. This effect is compounded by the lower absorption of thinner films thus it is difficult to crystallise ultra thin phase change samples. For amorphisation, thicker films are hard to amorphise and a heat sink is usually required to realise the melt-quenched amorphous state. Additionally, an adequate contrast in the reflection/transmission signal is required to detect the phase transition. We found that a Ge₂Sb₂Te₅ thickness of 30 nm was optimal and therefore this was used throughout the measurements reported in this paper.

A laser pulse of 200 ns, 16 mW was applied to the amorphous Ge₂Sb₂Te₅ film by the pump laser and the change in the transmitted light was monitored by the probe laser. The transmission as a function of time is plotted in Fig. 3(a). The pump laser pulse was initiated at 0 ns and the transmission was measured up to 1200 ns. The transmitted light gradually decreases and saturates at a lower level than the pre-pulse state. There is a 40% change in transmission after applying the pump laser pulse. This change in the transmitted signal is due to the amorphous-crystal phase transition of Ge₂Sb₂Te₅.

 

Fig. 3 (a) The change in transmitted signal as a function of time in the crystallisation process of Ge₂Sb₂Te₅. The JMAK fitting (black curve) of the transmitted signal. (b) Distribution of switching time for 250 times laser switching measurements. The gaussian fitting (pink colour) of the switching times.

Download Full Size | PDF

The relative change in transmission (ΔT) between the amorphous and crystalline state was calculated and fit with a Johnson-Mehl-Avrami-Kolmogorov (JMAK)–like crystal growth function f(t) = −A [1 − exp (−(t/τ)^b)] [40]. Where, t is the pulse duration, τ is the crystal growth rate, b is the crystal dimensional factor, and A is the fitting parameter. The fitted curve is shown in Fig. 3(a). Since JMAK theory applies to isothermal crystallisation, we cannot obtain any meaningful information from the fitting parameters for non-isothermal laser heating crystallisation measurements. Others have used a clever combination of laser amorphisation of samples held at a constant elevated temperature and finite element modelling to gain insight into isothermal crystallisation processes [41].

In this manuscript “pulse duration” refers to the time that the pump laser pulse is on. Switching time is the time from when the laser pulse is initiated to the time when the the transmitted or reflected signal has reached 90% of the normalised change in transmission or reflection. Therefore, the crystallisation switching time, t_s, was calculated as f(t=t_s)= 0.9ΔT or 0.9ΔR. We repeated this switching time measurements 250 times on the amorphous Ge₂Sb₂Te₅ sample. Each measurement was performed on a fresh area of the same Ge₂Sb₂Te₅ sample. A histogram of the switching times is shown in Fig. 3(b). For this non-homogeneous nucleation process, the new nuclei are formed in a stochastic way and depend on several random variables such as, activation energy, sample homogeneity on micro scale, and rate of formation of new nuclei. As is usual for a measured values that depends on the values of a group of random variables, the switching time is normally distributed; a consequence of the the central limit theorem. Thus the switching time, t, distribution was fitted with a Gaussian normal distribution (pink curve) with a mean value of 212 ns and standard deviation of 4 ns, i.e. t ~ N(212, 4) ns. Therefore we consider the random repeatability error in switching time measurement for our system to be 4 ns.

3.1. Transient PTK measurements

To study the crystallisation and amorphisation kinetics we use a transient PTK measurement. Similar to the procedure introduced by IBM, we generate a matrix of laser power and reflectivity as a function of time [25]. For these switching experiments, the pulse duration was kept constant at 500 ns while the pump power was increased in steps of 0.5 mW from 0.1 mW to 40 mW. For a given laser power and laser pulse duration, an average of five measurements were considered to determine the change in transmission. After each laser pulse measurement, the sample was translated by 10 μm using an x-y stepper motor stage. Therefore, each measurement was performed on a fresh area of the same sample. A time interval of 30 seconds was kept between two consecutive laser pulses to prevent the heat accumulation and to readjust the focus position.

The measured transmitted probe signal contains high frequency background noise from dark current of the photodetector as shown in Fig. 4. A finite impulse response (FIR) digital low-pass filter was used to remove the high frequency background noise [42]. The reflection and transmission measurements were smoothed using a band pass filter with a frequency range of 0.01 to 0.3 GHz, which rejects frequencies above 0.3 GHz. Thus, using this filter the temporal resolution of the transmission and reflection measurement is 1.7 ns, which is less than the repeatability error for the switching time measurement. The raw data and the corresponding filtered data for a crystallisation and amorphisation switching process are shown in Fig. 4(b) and 4(e) respectively.

 

Fig. 4 Power-time-transmission image plot of the Ge₂Sb₂Te₅ sample during crystallisation and amorphisation. (a) and (d) shows the experimental raw data. (b), and (e) using of short pass FIR filter to smoothen the raw data and (c) and (f) represents the filtered transmitted data.

Download Full Size | PDF

A matrix of the laser power, time, and transmission was used to plot the PTK diagrams shown in Fig. 4. The colour represents the change in transmission due to a phase transition. Green colour represents the amorphous state while red represents the crystalline state of Ge₂Sb₂Te₅. The crystallisation power-time-transmission plot shows four distinct regions: (1) no change in transmission, the Ge₂Sb₂Te₅ remains in as-deposited amorphous state, (2) a reduction in transmission, which indicates a structural transition from amorphous to crystalline state, (3) an increase in transmission, indicating the melting of the PCM, and (4) a large increase in transmission, indicative of melting and then ablation of the PCM from the sample surface. Thus the red area of plots (a) and (c) can be used to determine the switching power, time, and therefore switching energy, of the PCM. These plots are, therefore, especially useful for comparing the switching performance of different phase change materials and structures.

To study amorphisation of Ge₂Sb₂Te₅, the sample was annealed above its crystallisation temperatures at 200 °C in an argon atmosphere for 30 minutes and then cooled to room temperature. A 50 ns pump pulse, initiated at 0 ns was used for amorphisation. The raw and filtered power-time-transmission plots are shown in Fig. 4(d) and 4(f). The increase in transmission indicates the Ge₂Sb₂Te₅ is reversibly switched back to an amorphous state. The reversible switching is a faster process than crystallisation. Here we see that Ge₂Sb₂Te₅ can be amorphised with a laser pulse of 27 ± 4 ns with power of 32 mW.

3.2. Comparison of methods used to measure switching times

To compare the transient and pre-pulse–post-pulse crystallisation time measurements for Ge₂Sb₂Te₅, a stacked Si/Al (60 nm)/Ge₂Sb₂Te₅ (28 nm) structure was used with an additional metal layer to improve the reflected signal. The Al layer acts as a mirror and a small change in the reflected signal induced by PCM structural change can be easily detected. For the pre-pulse–post-pulse measurements, the pulse duration was varied from 100 ns to 2000 ns with steps of 100 ns. For each pulse duration, the pump power was gradually increased from 0.1 mW to 40 mW with steps of 0.5 mW, and the reflected signal was measured. The probe laser reflected signal was measured before and after the pump pulse. The reflectance change was calculated and plotted as a function of pump laser power and pulse duration. This is shown in Fig. 5(a). Each colour pixel represents a laser pulse of particular power and pulse duration. For the transient measurement, a fixed pulse duration of 200 ns was used and the change in reflected signal was continuously measured before, during, and after the pump pulse and the results are shown in Fig. 5(b). The reflected signal increased abruptly at 50 ns, indicating the phase transition from the amorphous state to the crystalline state. A volatile change in the reflected signal was observed during the 300 ns pump pulse. This is marked as region (1) in the Fig. 5(b). This volatile change is attributed to thermal excitation of carriers [43–46]. Post-pulse measurements can not observe these transient changes during pumping. The transient modulation is an important characteristic of these materials and can be used to design ultrafast optical modulators and sub-wavelength apertures [14, 28].

 

Fig. 5 Power-time-reflection image plot of the Ge₂Sb₂Te₅ in crystallisation process by (a) pre-pulse–post-pulse mode and (b) transient mode measurements.

Download Full Size | PDF

The crystallisation and amorphisation switching times can also be measured by analysing images from the sample after laser excitation. A 10×10 array of crystalline marks was written to the amorphous Ge₂Sb₂Te₅ film in the structure Si/SiO₂ (30 nm)/Ge₂Sb₂Te₅ (28 nm)/SiO₂ (140 nm) by varying the pulse laser power and duration. The reflected image of PTK matrix pattern is shown in Fig. 6. In this laser crystallisation experiment, the pulse duration was varied from 100 ns to 1000 ns with steps of 100 ns, while the pump power was increased from 0 mW to 40 mW with a step of 4 mW. For the crystalline marks matrix pattern, the columns have identical pulse lengths, and the rows have the same pump power. The crystallisation process was studied by analysing the reflected intensity of the marks on the sample surface. Figure 6 shows that a laser pulse duration greater than 100 ns can induce a structural transition in Ge₂Sb₂Te₅. For this sample structure, the pump laser power was below 16 mW and the Ge₂Sb₂Te₅ remained in a low reflectivity, amorphous state. Increasing the laser pump power increases the probability of crystallisation and the resultant mark size. We see here that crystallisation is possible in 100 ns for pump powers greater than 20 mW. In both cases, the laser pulse duration and power was varied to trigger the structural change in PCM and the corresponding switching time and power was calculated. Therefore, this image analysis is equivalent to the pre-pulse–post-pulse measurement. The area of the crystalline mark in the image plot is proportional to the reflected intensity, therefore each crystalline mark in Fig. 6 has a parallel mapping to the single colour pixel in Fig. 5(a). This optical microscope image with crystalline marks can be used to determine the threshold value of laser pump power and pulse duration in the switching measurement. However, a major limitation of the image analysis approach is that the whole power-time array must fit within the field of view of the static tester microscope. In our case, the field of view is 120 μm by 120 μm, and the minimum reliable x-y translation is 5 μm, hence the total number of laser marks can be written into the film is 576. This restriction limits the power and pulse duration resolution of the measurement. Initially, a low resolution power-time-reflection experiment was performed with larger pulse power and time step sizes. This lower resolution power-time image was used to find the power-time range, where the transition occurs. Then, the measurement was repeated with a smaller laser power and pulse duration step size over the power-time range the of interest. Thus, providing a higher resolution measurement of the crystallisation time dependence on the laser pulse duration and power.

 

Fig. 6 The optical microscope image of a crystalline write marks matrix on the amorphous Ge₂Sb₂Te₅ surface. The red colour represents high reflective (crystalline) state and the blue colour represents the low reflective (amorphous) state.

Download Full Size | PDF

The final method that we use to measure crystallisation uses a moving sample. A schematic ray diagram of the dynamic disc tester (DDU-1000, Pulsetec Co.) is shown in Fig. 7 [47]. It consists of a 680 nm diode laser source which is focussed on to the rotating disc. The laser light which is reflected from the disc is passed through a beam splitter (BS1). Half of the light is used to measure the reflectivity of the discs surface, whilst the other half of the light is sent to a polarising beam splitter (PBS1). One polarisation state is directed to a split detector and used to maintain alignment on the disc track, and the other polarisation state is directed to a quadrant detector for auto-focussing. The crystallisation time, t_c, is measured by erasing amorphous marks of a known length at a known disc linear velocity. By plotting the reflectivity of the track as a function of disc linear velocity, the maximum linear velocity that can be used to erase the disc can be found. Erasability is calculated by measuring the maximum change in signal-to-noise ratio of an amorphous mark before and after applying the write laser [48]. The erasability for the phase change materials along the GeTe − Sb₂Te₃ pseudo-binary line are calculated and shown in Fig. 8. A single laser is used to measure the crystallisation and amorphisation rate of PCMs. The sample linear velocity, v, laser power, P, laser pulse frequency, f, and duty cycle, d, is controlled. Hence, amorphous or crystalline marks of length, can be written to the sample according to $l=\frac{vd}{100f}$. Where the laser duty cycle, d, is a percentage.

 

Fig. 7 Schematic diagram of the dynamic disc tester.

Download Full Size | PDF

 

Fig. 8 The erasability as a function of disc linear velocity of GeTe − Sb₂Te₃ based phase change material alloy. The green colour curve represents the Ge₂Sb₂Te₅ alloy.

Download Full Size | PDF

Figure 8 shows a typical measurement for phase change materials along the GeTe − Sb₂Te₃ pseudo-binary line. For Ge₂Sb₂Te₅, we see the erasability abruptly drops when the sample velocity is increased above 11 ms⁻¹. For this particular sample, the amorphous mark length was 600 nm, hence the minimum time for complete crystallisation is $t_c=\frac{6\times {10}^{-7}m}{11m{s}^{-1}}=55ns$. This method is known as a dynamic disc tester (DDT) and was commercially used to characterise phase change DVD-RAM.

It is interesting to compare the crystallisation time measured using the DDT with the crystallisation time measured using a static tester. The re-crystallisation time of 600 nm long amorphous Ge₂Sb₂Te₅ marks was measured by the DDT to be 55 ns with a laser erasure power of 8 mW. For the as-deposited amorphous Ge₂Sb₂Te₅ films, which were measured by the transient static tester measurement, shown in Fig. 4(a), the complete crystallisation time was found to be 54 ns with an incident laser power of 32 mW, and a laser spot size was 1 μm. In contrast, for the same laser power of 32 mW, the pre-pulse–post-pulse procedure required a 100 ns pulse duration to crystallise the amorphous Ge₂Sb₂Te₅ film, see Fig. 5(a). Similarly, Fig. 6 shows that a 100 ns pump pulse is sufficient to crystallise Ge₂Sb₂Te₅ with a laser power of 32 mW.

The 50 ns to 100 ns spread of crystallisation times measurements for Ge₂Sb₂Te₅ clearly demonstrates a measurement method and procedure dependence and justifies standardisation of phase change measurement procedures. Such standardisation will clearly minimise cross-lab measurement variability. We now highlight the key difference between the different procedures and methods. The DDT system measures the linear crystal growth velocity of re-amorphised marks written into the crystalline film. If the crystal growth front keeps up with the scanning laser, then the amorphous mark will be completely crystallised. The DDT is, therefore especially well suited for measuring crystal growth rates. In contrast, the static tester measures the change in reflection/transmission as a function of time at a constant sample position. This crystallisation time measured by the DDT and the transient methods are similar, 55 ns and 50 ns respectively, but the laser pulse energy used by the static tester is nearly a factor of four higher. This is due to the smaller amorphous marks and thermally optimised disc structures, which were used in the DDT measurements. The disc structure employed a ZnS–SiO₂ (80:20) insulations layers, which enable recrystallisation with a low laser power. The longer crystallisation times measured using the pre-pulse–post-pulse measurement and the image analysis suffer from precision problems. For the image analysis and the pre-pulse–post-pulse experiment reported here, measurements were collected every 100 ns. Thus the shortest crystallisation time measured was 100 ns and the minimum power used to crystallise the material was also similar at 24 and 20 mW respectively. However, by comparing the crystallisation time and power at known image analysis and pre-pulse–post-pulse measurement points, we see that there is good agreement with the transient measurements. For example from the transient measurement shown in Fig. 4, we see that a 20 mW pulse causes complete crystallisation 100 ns, which agree with the other methods.

Since the nucleation time for Ge₂Sb₂Te₅ is shorter than most phase change materials [49], we see very similar crystallisation times for static tester measurements of as-deposited films, and DDT measurements on re-amorphised marks. For this reason we expect a minimum crystallisation time of as-deposited amorphous, growth dominated sample by static tester measurement to be substantially longer than crystallisation time measurements of melt-quenched amorphous marks in the DDT measurements.

The re-crystallisation time is technologically relevant, however, we refrain to discuss the re-crystallisation kinetics in details because it has been extensively studied previously by Abel-son [50, 51]. The re-crystallisation time is dependent on the formation of amorphous mark. Amorphous Ge₂Sb₂Te₅ has a spectrum of amorphous states, which can be prepared by both thermal and non-thermal melting [52, 53].

The main advantage of the transient measurement is the 1 ns temporal resolution, and the main disadvantages are the level of noise in the laser intensity measurements and the system complexity. The main advantage of the image analysis and pre-pulse–post-pulse systems is the set-up simplicity. The pre-pulse–post-pulse measurements can give an insight into the crystallisation behaviour of PCMs but they are unable to measure transient volatile changes to the reflected/transmitted signals, which are observed in the transient measurements.

It is also noteworthy that the crystallisation of phase change materials is dependent on a number of factors including the phase change material thickness, history of heating, the interfacial materials, strain, and the capping layer [13, 54, 55]. Therefore, we need to establish a standard test structure of testing phase change materials such that cross-lab variability can be minimised.

Table 2 compares the dynamic and static testers.

Table 2. Comparison between the dynamic and static testers

View Table | View all tables in this article

3.3. Nonlinear optical properties measurement

In addition to phase transformation measurements, we have also used the static tester to measure the nonlinear absorption and refraction using the z-scan method [56]. We demonstrate this measurement on Bi₂Se₃, a chalcogenide material and a well known topological insulator [57]. The sample was exposed to the 658 nm pump laser in CW mode. The incident laser intensity was below the optical damaged threshold level and kept at a low power of 1 mW/μm². The pump beam was focused to a spot size of ω₀ ≈ 1 μm, using a microscope objective lens, which resulted in a Rayleigh length, $z_R=\frac{\pi \omega \frac{2}{0}}{\lambda }$, of 4.77 μm. The sample was translated along the z-axis through the focus by a distance of 80 μm. The step size was 100 nm. The transmitted power was measured in an open aperture (OA) and in a closed aperture (CA) configurations using a photodetector.

Figure 9(a) shows the normalised transmittance for Bi₂Se₃ samples in an OA configuration at an incident on-axis intensity of 0.39 GW/m². The OA z-scan measurements show a saturable absorption (SA) behaviour. This indicates that the SA dominates over the other effects like two photon absorption or reverse-saturable absorption. More importantly, due to its narrow band gap (0.3 eV), there is a high probability to photo excite an electron from the valence band to the conduction band [58–60]. Close to the laser focus z-position, the photon flux is high which depletes the irradiated Bi₂Se₃ valence band. Consequently, the photon absorption probability is lowered and the transmitted laser intensity increases [61]. This effect is observed as a sharp peak in the transmitted intensity as the sample is translated across the focus position, see Fig. 9(a).

 

Fig. 9 Normalised change in transmission in a z-scan measurements of Bi₂Se₃. (a) Open aperture (OA) (b) Ratio between CA and OA.

Download Full Size | PDF

Figure 9(b) shows a typical z-scan curve by taking the ratio between the CA and OA normalised transmittance. The pre-focal minima and post-focal maxima is a sign of self-focusing effect with a positive value for nonlinear refractive index (n₂).

To calculate the nonlinear optical parameter, we fit the normalised transmittance curve of OA and CA to the well known formula [62]:

4. Summary

To conclude, we have characterised the performance of our in-house built dual beam pump-probe static tester and compared it with other optical systems that are commonly use to measure the phase transformation kinetics of chalcogenide phase change materials. In contrast to measurements of crystallisation using electrical devices, optical measurements can be made on patternless thin films without the need for expensive and time consuming lithography. There are essentially, three different measurement techniques that are used for measuring the switching times of phase change materials, (1) dynamic measurements, where there is a relative movement between the laser beam and the sample. The crystallisation time is then computed by the maximum scanning speed at which the amorphous mark can be erased, (2) transient static measurements, where the both the sample and laser are static, and a probe laser is used to measure reflection and transmission transients due to a pump pulse, and (3) pre-pulse–post-pulse measurements of the change in reflection and/or transmission. In all cases it is critical that the probe laser, which is used to measure the sample optical properties, does not heat the chalcogenide film. Our models show that a laser intensity less than 100 μW/μm² for 1 μs will heat a Ge₂Sb₂Te₅ film by approximately 40 K, and therefore we recommend using lower laser probe intensities and short pulse durations.

We have performed switching time measurements on Ge₂Sb₂Te₅, which is a nucleation dominated phase change material, using the three different measurement procedures. The measured minimum phase change time seems to be consistent for all three methods. However, the pre-pulse–post-pulse switching time measurements are less precise and do not provide information on the transient behaviour of the material. The precision of these measurements can be improved by using a finer pulse duration measurement grid. However, the pre-pulse–post-pulse method provides information about the switching power and time, and the change in the samples transmission and reflectivity. Moreover, the single laser set-up is comparatively simple, and therefore we recommend this type of set-up for assessing materials to be used in optical logic, switches, and reconfigurable photonic circuits [29, 31, 63], where the transient optical behaviour of the material is not important. A further advantage of this method is that low speed detectors and amplifiers can be used. These generally have a higher signal to noise ratio than those used in the transient measurement setups and this leads to more accurate measurement of the chalcogenides optical property change. The “Transient mode” measurement procedure is suited to characterising materials for active photonics devices that exploit the chalcogenides transient optical changes. For example, devices that used to form transient apertures or photonic signals [16, 30, 64].

Our in-house built pump-probe system can perform pre-pulse–post-pulse measurements, transient mode measurements, and measurements of non-linear refraction. At the present time the system is limited to characterising thin films that have high absorption at the pump laser wavelength (λ = 658 nm). To characterise chalcogenide material tuned meta devices and meta surfaces, which typically use thicker films, a pump laser wavelength must be used that is less strongly absorbed such that the temperature distribution through the film thickness is relatively uniform. Clearly adding a tuneable laser source to our system will extend it’s capability.