Phase change of Ge₂Sb₂Te₅ under terahertz laser illumination

In the last half century, electronic devices have significantly promoted the development of the computing technology following the von Neumann approach.^1,2 Currently, the world generates exponentially increasing amounts of data that need to be processed in a fast and efficient way. The traditional electronic devices with a limited bandwidth and high power consumption cannot satisfy the increasing demand of the modern computing technology. Therefore, it is highly important to find other solutions to further accelerate the computation. Rather than the traditional electronic devices, the deployment of photonic devices with ultrahigh bandwidths, low losses, and high efficiencies³ is one of the best ways to significantly accelerate the processing speed of the next computing systems. For example, on-chip optical interconnects have already replaced electronic wires in connecting central processing units in high performance computing systems to deliver an overall performance for the signal transmission.⁴ Due to the unique characteristics, e.g., fast, scalable, and non-volatile, the phase change material (PCM) is the key component of photonic devices for the all-photonic computing.⁵ Specifically, the large refractive index contrast between different states of the PCM allows its broad applications in novel photonic devices,⁶ such as integrated all-photonic non-volatile memories,⁵ optical color rendering,⁷ and low-power optical switches.⁸ Among different PCMs, Ge₂Sb₂Te₅ (GST) demonstrating fast switching time in a level of a few nanoseconds or even less⁹ and long-term stability for over 10 years¹⁰ is suitable for non-volatile memory devices. The working principle of the conventional GST-based PCM devices relies on the fast phase change between multilevel states under certain temperature conditions employing an electrical-induced or optical-induced heating, i.e., amorphous GST (a-GST) at room temperature, cubic GST (c-GST) at ∼150 °C (423 K), and hexagonal GST (h-GST) at ∼300 °C (573 K).^11–16 

In previous studies, the electrical heating is normally utilized to achieve the phase change of GST because the electrical trigger is easy to implement. These geometries usually consist of top and bottom electrodes or an electrode array to switch the phase states of GST.⁸ For the electronic GST devices, the phase change speed is limited by some issues, such as the impedance mismatch due to the change of the phase state and the uneven distribution of the phase change area resulting from the uneven Joule heat distribution (especially for the planar device where two electrodes are located on the same surface).¹⁷ In principle, the optical implementation is a suitable alternative for fast switching the phase states of GST by employing a laser. The light-induced temperature change can also result in the phase change of GST. The rapid development of the light source in the visible and infrared frequencies has boosted the applications of GST materials in data storage,¹⁸ integrated optical switches,¹⁹ and perfect absorbers.²⁰ In these frequency regimes, the optical reflectivity of GST increases with increasing annealing temperature induced by higher optical power irradiation.²¹ Then, the transmission or reflection of GST can be modulated and detected to verify the phase states.²² 

The applications of GST materials in the terahertz frequency range (roughly defined between 100 GHz and 10 THz) are relatively rare compared to the visible and near-infrared applications due to the lack of high performance terahertz radiation sources. In the past decade, this diversified and rapid development of the terahertz quantum cascade laser (QCL) has led to a bright future for terahertz applications. The electrically pumped and chip-level QCLs characterized by their broadband frequency coverage,²³ high far-field beam quality,^24,25 and high output power²⁶ show abilities for boosting applications of GST materials in the terahertz regime. Up until now, various studies on interactions between the GST materials and terahertz light have been performed, for instance, the near-field imaging²⁷ and spectral characteristics of the various phase states of GST materials.^28–30 However, the phase change of GST materials in the terahertz regime still needs additional electrical^31,32 or optical components, such as femto-second lasers³³ or intense terahertz pulses,³⁴ which makes the system rather complex. Moreover, the intense terahertz pulses will bring about the Zener tunneling breakdown,³⁵ threshold switching,³⁶ and the significant volume expansion on GST surfaces,³⁷ which hinder the practical applications of GST materials. In view of these, investigations of the phase change of GST under terahertz QCL illumination will boost the applications of GST materials and, simultaneously, diversify the application fields of terahertz QCLs.

In this work, we investigate the phase change of GST materials in the terahertz frequency range, employing a QCL emitting around 2.5 THz. A z-scan technique is utilized to gradually change the terahertz power density irradiated onto the GST samples and then to achieve the phase change. The transmission evolution with z or power density is investigated. Numerical simulations based on a three-dimensional (3D) finite-element model with electromagnetic heating are also performed to calculate the change of temperature and transmission of the GST samples during the z-scan process. The simulation results indicate that the phase change of GST can be achieved at the focal point (z = 0), and the transmission of GST is reduced by 30%, which shows good agreement with the experimental results. Moreover, using transmission electron microscopy (TEM), we visually observe the phase change from c-GST to h-GST after higher power terahertz illumination (2.5 mW), further proving the experiment observation of the z-scan measurements. Our work demonstrates a method to directly control and detect the phase state of GST materials under terahertz illumination, without the need of additional electrical or optical components, which paves a way for the applications of GST materials in the terahertz range. Moreover, this work also diversifies the application fields of terahertz QCLs.

The GST samples with a thickness of 100 nm were grown on 4 in. high-resistance double-side polished silicon (Si) wafers (300-μm-thick), employing a physical vapor depositor with a stoichiometric Ge₂Sb₂Te₅ alloy target. The details of the sample growth can be found in Ref. 38. For the spectral characterization of the GST sample in different phase states, the grown samples were cleaved into different pieces for the thermal annealing at various temperatures, i.e., 150, 200, 250, 300, 350, and 400 °C for 30 min. Figure 1(a) shows the transmission spectra of GST samples annealed at different temperatures measured using a Fourier transform infrared (FTIR) spectrometer (Bruker, Vertex 80V). For the FTIR measurements, the diameter of the light spot on the sample surface is around 10.8 mm, which is 1.8 times the selected aperture diameter (6 mm). As a reference, the transmission of the 300-μm-thick Si substrate is also shown. It can be seen that the thin as-deposited amorphous GST (a-GST) film does not contribute too much in the absorption, and therefore, the as-deposited a-GST/Si and the Si substrate show almost identical transmissions in the terahertz frequency range investigated here. However, as the annealing temperature is increased, a drastic decrease in transmission can be clearly observed. For instance, at 150 °C, the phase change starts; and between 200 and 300 °C, the overall transmission is decreased, and an absorption peak around 2.56 THz can be observed, which is a clear sign that phase state of the film is transformed from a-GST to c-GST.^28–30 As the temperature is further increased to 350 °C, the transmission is strongly reduced to 20%, showing that the h-GST is obtained. The significant difference in transmission between c-GST and h-GST is mainly due to the fact that the h-GST (hexagonal crystalline structure) characterizes higher carrier mobilities.³⁹ 

To clearly show the frequency dependence of GST films in different phase states, vertical cuts at different frequencies from Fig. 1(a) are plotted in Fig. 1(b). The three shaded areas indicate the three different phase states of the GST films. It reveals that the transmissions of GST at 2.5 and 2 THz are strongly dependent on the phase states; especially, we can see that the transmission follows a step function (dashed line) when the phase state is changed from a-GST to c-GST. Because of this, it is more suitable to observe the phase change of GST below 3 THz. By considering the spectral coverage of QCLs and the window of minimum water absorption, we finally choose a terahertz QCL emitting around 2.5 THz in this work.

To verify the phase states of the GST samples shown in Figs. 1(a) and 1(b), we perform the transmission electron microscopy (TEM) and x-ray diffraction (XRD) measurements. Figure 1(c) shows the typical TEM images of a-GST and c-GST (obtained by annealing the a-GST film at 200 °C). The insets in Fig. 1(c) are the corresponding fast Fourier transform (FFT) patterns, which clearly show the crystalline and amorphous structures of the c-GST and a-GST, respectively. An analysis of the energy dispersive x-ray (EDX) spectroscopy can further give the ratio of each component, i.e., Ge:Sb:Te ≈ 1.92:1.98:4.19. Figure 1(d) shows the x-ray diffraction (XRD) curves measured for the GST films annealed at different temperatures. For the as-deposited GST film, due to its amorphous feature, except the Si (211) peak, no other peaks from the GST film can be observed. When the GST samples were annealed at 200 or 250 °C, reflection peaks (111), (200), and (220) can be observed, and these peaks are the characteristics of c-GST films. Note that the reflection peaks (002) and (421) are generated because of the oxidation during the annealing process. Regarding the case of 300 °C annealing temperature, the GST sample exhibits both the cubic phase and the hexagonal phase characterized by the additional peaks (013) and (203).^11,29

In Fig. 1, we show the spectral and structural properties of GST samples annealed at different temperatures. The focus of this work is to investigate the phase change of GST triggered by terahertz light irradiation. To study the power dependence of the phase change process of the GST films, a z-scan technique^40,41 is employed using a terahertz QCL emitting around 2.5 THz as the radiation source. As shown in Fig. 2(a), the terahertz light emitted from the QCL is collected by using an off-axis parabolic (OAP) mirror with a diameter of ϕ = 2 in. and a focal length of f = 4 in. Then, it is collimated by using another OAP mirror (ϕ = f = 2 in.) onto the GST/Si sample. The beam waist radius of w₀ ≈ 73 μm at the focal point (z = 0) is directly obtained from the QCL far-field beam pattern measured using a terahertz camera (NEC, IR/V-T0831C).⁴¹ The beam waist is strongly dependent on the waveguide geometry. In this work, we assume that the beam waist does not change with laser power. The GST/Si sample is mounted on a motorized translation stage (Newport, M-ILS200CCL) moving along the optical axis (z), which results in a gradual tuning of the power density irradiated on the sample. The transmitted terahertz light through the GST sample is measured using a Golay cell detector (Tydex, GC-1P). A chopper with a frequency of 24 Hz and a titled wire grid polarizer (PUREWAVE, PRO-1630) are used to modulate and attenuate, respectively, the transmitted terahertz light for accurate power measurements. We measured the series of transmitted terahertz energies by moving the GST sample from z = −6 mm to z = 6 mm with a step of 0.3 mm. The measurement time for each step is 1 s.

Figure 2(b) shows the light–current–voltage (L–I–V) characteristics of the terahertz QCL (2-mm-long and 150-μm-wide) in continuous wave (cw) mode at a heat sink temperature of 15 K. The maximum average power of the QCL is 2.56 mW at a drive current of 950 mA. The inset in Fig. 2(b) shows a typical emission spectrum of the QCL recorded at a drive current of 750 mA in cw mode. The lasing spectrum is centered around 2.54 THz. Figure 2(c) plots the normalized power of the terahertz QCL as a function of z using the z-scan setup [see Fig. 2(a)] without a sample on the translation stage. In principle, if the laser is stable enough, the measured power will not change as z is changed. However, in this work, a closed-cycle liquid helium cryostat with a vibrating compressor is used for cooling the QCL. The vibration of the cryostat then results in the power fluctuations. To evaluate the power stability, we mathematically calculate the root mean square error (RMSE) that is equal to 0.0188, as shown in Fig. 2(c).

Prior to a z-scan measurement of the GST sample, we first perform a three-dimensional (3D) transient simulation based on a finite-element method (COMSOL Multiphysics) to study the phase change of GST films during a z-scan process. Figure 3(a) schematically depicts the sample geometry and the coordinate configuration when the surface of the GST sample is placed at the focal point of the terahertz light. The origin of the 3D coordinates is the central point of the focused beam pattern on the sample surface. In the simulation, the moving of the as-deposited a-GST sample is treated by gradually changing the terahertz beam size [see Eq. (1)]. At each z point, the electromagnetic heating model is adopted to convert the light power on the surface of the sample to a heat source. Then, the heat distribution can be calculated following the heat transfer theory employing the 3D finite-element simulation. Simultaneously, in this model, we calculate the optical transmission to verify the phase states induced by terahertz illumination.

To simplify the simulation model, the terahertz beam pattern on the sample surface at each z point is considered as a Gaussian distribution. The z-scan process can be simulated by gradually changing the beam radius w on the sample. The beam radius w is a function of z, which can be defined as

where z_R = $πw02n$/λ is the Rayleigh length and n is the refractive index.^42,Figure 3(b) shows the thermal hysteresis of the GST film in the heating (black arrows) and cooling (colored arrows) processes. It can be seen that the a-GST can be transformed to c-GST as the temperature T is increased (150 < T < 300 °C), and it can be further transformed to h-GST (300 < T < 650 °C). When T is greater than 650 °C, the GST film will be melted. For the melted GST film, a slow quench will bring the film to the c-state or h-state, and a fast quench will make the film return to the a-state as indicated by the red arrow in Fig. 3(b).⁴³ Note that once the GST film is transformed to the c-state or h-state before reaching the melting point (T < 650 °C), no matter how we lower the temperature, the phase of the GST film will not change [see green and blue arrows in Fig. 3(b)].¹¹ 

In the 3D finite-element simulation, the electric field and power distributions on the surface of the GST sample at different z positions can be written by⁴⁴ 

where P₀ is the optical power of the laser, where we use 1.5 mW in this simulation, c is the speed of light in vacuum, ϵ₀ is the absolute permittivity, and ϕ = kz + $k(x2+y2)2z(1+z02/z2)$ − $tan−1(z2/z02)$ with k being the wave number. For the thermal simulation, the power shown in Eq. (3) is absorbed by the GST film and converted to a heat source on the sample surface. Then, the heat on the surface dissipates in the sample vertically and laterally. The absorption of the GST film is assumed to be 6% in the simulation,²⁸ and the absorption of the Si substrate is neglected due to the too small imaginary part of the refractive index of Si in the terahertz regime.⁴¹ The difference in the optical transmission of the a-GST and c-GST is taken into account by using different refractive index values of 17 and 22 for a-GST and c-GST, respectively, in the electromagnetic simulation.²⁹ The thermal conductivity values of 0.16, 0.56, and 130 W/mK for the a-GST, c-GST, and Si substrate, respectively, are used in the simulation.^11,29 The thermal boundary resistance (TBR) is also an important parameter for the thermal simulation. Here, inspired by the published works reported in Refs. 18 and 45, the TBR in the interfacial form of the Wiedemann–Franz Law of 0.9 m² KW⁻¹ (10⁻⁵ m² KW⁻¹) is assumed between the a-GST (c-GST) and the Si substrate. It is worth noting that most previous studies of the TBR of GST concentrate on the interface between the GST film and the metal or the GST film and silica.^46,47 The TBR can vary largely depending on the deposition technique and types of materials utilized.⁴⁸ Because the TBR between the c-GST and the Si substrate is much smaller than that between the a-GST and the Si, in the c-GST or once the a-GST is transformed to c-GST, the heat will dissipate fast into the Si substrate. To clearly show the phase profile of the GST sample, in the simulation, we assume that the a-GST is transformed to c-GST once the temperature reaches 150 °C. Figure 3(c) shows the phase profile of the GST film when the as-deposited a-GST moves from z = −6 to 0 mm. It can be seen that the central area of the GST sample with a diameter of 70 μm is transformed to c-GST under terahertz illumination. Because the central part of the terahertz beam irradiated on the surface of the GST film carries the highest optical power (hence the heat), we observe the c-GST phase uniformly distributed in the central circle. Once the central part is transformed to c-GST, the heat will fast dissipate into the Si substrate because the TBR between c-GST and the Si substrate is almost 5 orders of magnitude smaller than that between a-GST and the Si substrate. Therefore, the heat cannot transfer further on the GST surface, and we observe that the central phase changed area is much smaller than the focused terahertz beam size. In Fig. S1 of the supplementary material, we show the calculated temperature profiles on the GST sample surface at different z positions. Due to the effect of the fast vertical heat dissipation, at z = 0, the surface temperature at the central area of the sample is even smaller than that measured at outside parts. Because the phase changed area is much smaller than the light beam size, there is still a large amount of heat dissipating laterally on the sample surface. Therefore, at some specific z positions, we can see the ring structures where the c-GST is observed, as shown in Fig. 3(c) and Fig. S1. To show the entire picture of the phase state of the GST sample, in Fig. 3(d), the phase states along the horizontal line in Fig. 3(c) (y = 0) are plotted as the sample position z is changed from −6 to 6 mm. It can be clearly seen that before the sample reaches the focal point (z = 0), the small central area is already phase changed from a-GST to c-GST. Then, a ring area with a diameter of 300 μm on the sample surface is transformed to c-GST. The detailed temperature distributions for the two characteristic phase change points can be found in Fig. S1 of the supplementary material.

The propagation simulation of the terahertz wave through the GST sample is also coupled in the 3D finite-element model. To evaluate the terahertz transmission, we set two optical ports in the model, i.e., port 1 for the injection and port 2 for the power measurement of the transmitted terahertz wave, as shown in Fig. 3(e). Two perfect matched layers (PML) are used to ensure no reflection along the z direction in the upper and lower boundaries. In x or y direction, the perfect electrical conductivity (PEC) or perfect magnetic conductor (PMC) conditions are implemented for the boundaries to ensure that the terahertz wave propagates along the z direction. Figures 3(e) and 3(f) show the calculated electric field distributions on x–z and the GST surface (x–y) planes, respectively, of the GST/Si sample when the sample is positioned at z = 0. The electric field distribution shown in Fig. 3(f) is determined by two effects: first of all, the light reflections from the interfaces of the GST, Si, and air will affect the electric field distribution; on the other hand, in the simulation, the optical transmission and the thermal analysis are coupled to each other; therefore, the phase change effect will affect the refractive index and then the electric field distribution. In view of these effects, although the electric field at the injection (port 1) is homogeneously distributed (see Fig. S2 of the supplementary material), its distribution on the surface of the GST film is inhomogeneous. In Fig. 3(g), the black curve shows the transmission of the GST as the sample position is changed from z = −6 to 6 mm. Similar to the results shown in Fig. 3(d), before the sample surface is moved to the focal point (z = 0), the phase change already happened [marked by the sharp decrease in the transmission; see Fig. 1(a)]. However, when the sample is beyond z = 0 position and goes further away from the focal point, an increase in transmission is observed. This is because the fact that as z is increased, the beam waist of the terahertz wave also increases, and the majority of the light transmits through the a-GST rather than the c-GST. The small area of c-GST does not affect the overall transmission as the sample is far away from the focal point. Therefore, we can only see the low transmission when the sample is near to the focal point. As a reference, we also plot the maximum temperature of the GST sample in Fig. 3(g) (red curve) as a function of z extracted from Fig. S1. Both the thermal and optical simulation results agree well with each other and indicate that the a-GST film can be phase changed to c-GST triggered by terahertz laser illumination at a frequency around 2.5 THz.

To experimentally prove the simulated results shown in Fig. 3, we perform the z-scan experiment. The a-GST and c-GST (annealed at 200 °C) samples were tested with different optical powers (or different laser drive currents), employing the z-scan technique shown in Fig. 2(a). The measured transmissions of the two GST samples at 700, 750, and 950 mA laser drive currents are summarized in Fig. 4. From the L–I–V characteristics of the QCL [see Fig. 2(b)], it can be seen that optical powers measured at 700, 750, and 950 mA drive currents are 1, 1.5, and 2.5 mW, respectively. It should be noted that the transmission of the GST films should not exceed 100%, and the experimental points that locate in the shaded areas in Fig. 4 are artificially introduced by the data normalization. In this work, the thickness of the GST film is only 100 nm, which is far thinner than the Si substrate (300-μm-thick). The normalization is strongly disturbed by the interference peaks of the Si substrate. As a result, at some z points, we observe unexpected transmissions greater than 100%. The transmission measurements for the c-GST sample are used to verify if the a-GST samples under terahertz illumination can phase change to the c-GST because the transmission difference between the two samples at 2.5 THz can vary by 30%. As the a-GST sample is irradiated by lower optical power [1 mW @ 700 mA, Fig. 4(a)], at each z position, the measured transmission never goes down to the transmission level of the c-GST. This indicates that the phase change cannot happen when the a-GST is illuminated by 1 mW terahertz power. As the power is increased to 1.5 mW [see Fig. 4(b)], it can be clearly seen that as the sample is close to the focal point (z = 0), the transmission falls drastically to the level of the c-GST. The sudden decrease in transmission, as shown in Fig. 4(b), is a sign of the phase change from a-GST to c-GST, which agrees well with the simulated results shown in Fig. 3. However, when the optical power is further increased to 2.5 mW, as shown in Fig. 4(c), the transmission of a-GST does not decrease to the level of c-GST. Actually, it was expected that as the optical power was further increased (hence the temperature was increased), the phase change from a-GST to c-GST or a transmission decrease could be observed. This abnormal behavior of the GST sample under large optical power illumination can be explained as follows: as can be seen from Fig. 3(b), when the temperature is increased due to higher power illumination (2.5 mW), around the focal point, the a-GST phase is changed to the c-GST and then h-GST. Finally, it reaches the melting point and returns back to the a-GST. It is worth noting that the crystalline-to-amorphous transition depends on a sufficiently high cooling rate. In most previous studies, the melt-quenching process is initiated by a short pulse to raise the temperature momentarily above the melting temperature. Then, the top/bottom metal quenches the melted GST sample immediately after the optical pulse.^4,18,33 In our experiment, the decrease in the TBR makes the Si substrate act as a heat sink after the a-GST phase is changed to the c-GST or h-GST. Furthermore, it can be seen from Figs. 3 and 5 that the phase change area is much smaller than the focused terahertz beam size. Therefore, during the z-scan process, after the a-GST phase is changed to the melting phase, once the z position is slightly deviated from the focal point, the power density irradiated onto the phase change area of the GST sample decreases, which functions like an optical pulse. Because of the two effects, in Fig. 4(c), we see that the transmission of the a-GST sample under 2.5 mW terahertz irradiation almost does not change during the z-scan process. The transmission electron microscopy (TEM) measurement shown in Fig. S4 of the supplementary material also proves that the a-GST phase state is obtained under high power terahertz irradiation, which infers that the fast quenching rate in the z-scan experiment is achieved.

In Fig. 4, we experimentally show that low-power irradiation cannot phase change the a-GST sample, while high power irradiation can result in a phase change to c-GST at 1.5 mW and a melted phase at 2.5 mW, revealed by an optical z-scan technique. To further prove the phase change of the GST samples, the transmission electron microscopy (TEM) is employed to visually show the crystalline structures of the GST films after terahertz illumination. From the thermal simulation, as shown in Fig. 3, the phase change area resulted from the optical pump is rather small. Therefore, it is difficult to find the phase change area under the TEM when there is no marker on the sample surface. In view of this, to facilitate the location of the phase changed GST, we position the GST sample surface at the focal point of the terahertz beam and move it along x and/or y directions with a step of 0.1 mm. At each position, the sample stops for 2 s. This can create multiple phase change areas on the sample surface and help search for the target under TEM examination.

For the TEM measurements, both a-GST and c-GST samples were first illuminated by the terahertz beam with different power levels, i.e., 1.5 and 2.5 mW. After illumination, we then checked the sample surfaces under the scanning electron microscope (SEM). We found that the samples illuminated at the lower power level (1.5 mW) cannot show any visible “stamp” on the surfaces of either a-GST or c-GST samples, while circular spots can be seen on the sample surfaces of both a-GST and c-GST under high power terahertz illumination [see Fig. 5(a) and Fig. S4(a) of the supplementary material]. Because of this, we only perform the TEM analysis for the samples illuminated by the high power (2.5 mW) terahertz beam. In Fig. 5, we show the results of the c-GST sample illuminated at 2.5 mW terahertz power. Figure 5(a) is the SEM image of the sample surface. Multiple circular dots, marked by numbers from 1 to 6 can be observed on the sample surface. However, the dots are not aligned along a straight line. This is possibly due to the vibration of the cryogenic cooler and a slight shift in the sample position during terahertz irradiation. Figure 5(b) shows the enlarged SEM image of the circular dot marked by “4,” as shown in Fig. 5(a). The circular shape can be clearly seen, and the diameter is measured to be 3.6 μm. To perform the TEM analysis and examine the crystalline structure of the sample, a cross sectional specimen of the GST sample is prepared (see Sec. II for the detailed sample preparation). Figure 5(c) shows the cross sectional TEM image of the GST sample. The circular region on the GST sample surface can be clearly identified.

Figures 5(d)–5(f) show the TEM images of the GST sample taken for the regions in the circle, out of the circle (∼200 nm away from the edge of the circle), and another random location that is far away from the circle and not irradiated by the terahertz beam, respectively. Therefore, the random location is for sure in the c-GST state, and the results shown in Figs. 5(f) and 5(i) are presented here as a reference. The different layers, i.e., Pt, GST, and Si, can be clearly distinguished in the three images. Note that phase segregations appear in all three regions.⁴⁹ Different from Figs. 5(e) and 5(f), for the TEM image taken in the circle, as shown in Fig. 5(d), an oxide layer with the thickness of few nanometers can be observed on top of the GST film. This oxidation can be proved by the EDX measurement. As shown in Fig. S3 of the supplementary material, the EDX mapping clearly indicates that the oxygen is the dominant element in the layer between Pt and GST. The oxidation is resulted from heating terahertz irradiation.³⁹ To obtain the phases of the GST films in different regions, we perform the FFT for the images in Figs. 5(d)–5(f). The corresponding results are plotted in Figs. 5(g)–5(i), respectively. From the comparison of the diffraction patterns, it can be seen that the c-GST in the circular region is phase changed to the h-GST after terahertz irradiation. The hexagonal phase can be proved by the crystal faces 7(003) and 8(002), as shown in Fig. 5(g). However, the region out of circle (∼200 nm away from the edge of the circle) is still in the cubic phase (c-GST), as shown in Fig. 5(h). The TEM measurement of c-GST illuminated by the 2.5 mW terahertz beam explains the reason why there is a slight drop in the transmission of c-GST during the z-scan experiment shown in Fig. 4(c). As the c-GST is irradiated by higher optical power [2.5 mW, Fig. 4(c)], the c-GST is phase changed to the h-GST. Although the FTIR measurements indicate that the transmission of h-GST is reduced to 20%, as shown in Fig. 1(a), the transmission of c-GST illuminated by 2.5 mW terahertz power [Fig. 4(c)] is reduced by 5% due to the phase change area with a diameter of 3.6 μm. It is worth noting from the simulation shown in Fig. 3(c) that there is a phase changed ring structure. However, in the experiment, from SEM or TEM images, we are not able to see this ring feature. We have to emphasize that to clearly observe the ring structure, it should be strongly oxidized. If the ring area is only phase changed and not oxidized, we will not be able to see the change in the surface morphology.

In Fig. S4 of the supplementary material, we show the case of the a-GST sample illuminated by the high power (2.5 mW) terahertz beam. A circular region with a diameter of 5.5 μm is studied [see Fig. S4(b)]. Two regions, i.e., in the circle and out of the circle, are selected for the high resolution TEM analysis. The FFT patterns shown in the insets of Figs. S4(c) and S4(d) of the supplementary material reveal that both regions are in the amorphous phase. This is because after terahertz illumination, the region in the circle first reaches the melting point and then recovers to the amorphous state during the cooling process [see Fig. 3(b)]. Therefore, we observe that both regions in Figs. S4(c) and S4(d) of the supplementary material are a-GST. However, for the case of c-GST, as shown in Fig. 5, we only observe the phase change from c-GST to h-GST after terahertz irradiation. The reason is as follows: as we already explained, the TBR between c-GST and the Si substrate is much smaller than that between the a-GST and the Si substrate. When the c-GST and a-GST are illuminated by the same power of terahertz radiation, the heating effect in the c-GST will be much less than the a-GST because a large part of the heat in the c-GST will dissipate into the Si substrate. Because of this, the c-GST after terahertz irradiation cannot reach the melting point. To further prove that the TBR between the GST film and the substrate is important for the phase change behavior, we performed the z-scan measurements for GST films grown on a 300-μm-thick quartz substrate. The TBR value between the GST film and the quartz substrate is supposed to be 1–100 m² K/GW, which is extremely smaller than that between the GST film and Si.⁵⁰ The main results are shown in Fig. S5 of the supplementary material. It can be clearly seen from Fig. S5 that for all z-scan measurements performed under 1, 1.5, and 2.5 mW terahertz irradiations, the transmission of the a-GST sample never goes down to the transmission level of the c-GST. The results indicate that due to the fast vertical heat dissipation (smaller TBR value), the phase change of the GST film on a quartz substrate cannot happen, which shows good agreement with the explanations in the main paper. Furthermore, the results shown in Fig. S5 of the supplementary material show good agreement with the z-scan measurement, as shown in Fig. 4(c). The transmission of the a-GST after 2.5 mW terahertz illumination does not show a drastic decrease at the focal point (z = 0), which indicates that the GST remains in the amorphous phase after high power terahertz irradiation.

To show the reversibility of the phase change between a-GST and c-GST, and its switching ability, we illuminated an a-GST sample alternately with two different cw terahertz power values, i.e., 1.5 and 2.5 mW. The transmission spectra of the GST sample for 100 period numbers with different period times, t, i.e., 40, 20, and 10 s, are shown in Figs. S6(a)–S6(c). The duty cycle used in this experiment is 50%. Note that the data shown in Fig. S6 are normalized to the results of the Si substrate. Therefore, the results shown in Fig. S6 are purely for the GST film without the contribution of the substrate. From Fig. S6, it can be clearly seen that when the period t = 10 or 20 s, the phase change between a-GST and c-GST can be repeated at least 100 times. However, as t is increased to 40 s, the switching degrades after ten repetitions. The reason for the degradation is probably due to the oxidation and/or damage resulted from long time illumination.³⁷ Note that in Figs. 4 and 5, we show that for a large area c-GST sample (∼2 × 2 cm²), due to the vertical heat dissipation (smaller TBR values), the c-GST cannot be switched back to the a-GST in the power range studied here. However, if we irradiate the a-GST sample with a 1.5 mW power, the phase change to c-GST can be observed, as shown in Fig. 4. However, the phase changed area is much smaller than the terahertz beam size, which means that for the larger part of the GST sample, the TBR is still larger, and only the small phase changed area has a low TBR value. In this situation, if we further apply 2.5 mW terahertz irradiation, we are able to switch the phase back to a-GST, as shown in Fig. S6 of the supplementary material. The experiment also proves that the terahertz induced phase change can be used for the terahertz optical components, e.g., modulators and switches.

In summary, we have observed the phase change of the GST material illuminated by a terahertz quantum cascade laser emitting at 2.5 THz. A 3D finite-element model with electromagnetic heating was used to simulate the transmission and temperature evolutions of GST under terahertz laser irradiation. To experimentally observe the phase change of the GST sample, a z-scan technique was employed to investigate the transmission as a function of power density (z position). Both the simulation and experiment confirmed that the a-GST film can be phase changed to c-GST under terahertz illumination with a power of 1.5 mW. The TEM analysis visually proved that the phase change from c-GST to h-GST can be achieved by higher power terahertz irradiation (2.5 mW). It is worth noting that although the phase change can be observed, both the simulation and TEM analysis revealed that the phase changed area induced by terahertz illumination was much smaller than the terahertz beam size at the focal point (z = 0). Furthermore, it has been demonstrated that under high power illumination, the a-GST sample can reach the melting point and then go back to the amorphous phase during the cooling process.

See the supplementary material for the calculated temperature profiles, EDX spectroscopy mappings, SEM and TEM images of a-GST under terahertz illumination, relative measurement of the GST sample grown on the quartz substrate, and reversibility experiment.

This work was supported by the National Natural Science Foundation of China (Grant Nos. 61875220, 62035005, 61927813, and 61991430), the “From 0 to 1” Innovation Program of the Chinese Academy of Sciences (Grant No. ZDBS-LY-JSC009), the Scientific Instrument and Equipment Development Project of the Chinese Academy of Sciences (Grant No. YJKYYQ20200032), the Major National Development Project of Scientific Instrument and Equipment (Grant No. 2017YFF0106302), the National Science Fund for Excellent Young Scholars (Grant No. 62022084), the Shanghai Outstanding Academic Leaders Plan (Grant No. 20XD1424700), and the Shanghai Youth Top Talent Support Program.

The authors have no conflicts to disclose.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

The active region of the terahertz QCL used in this work is based on bound-to-continuum transitions. The detailed layer structure can be found in Refs. 51 and 52. The QCL active region was grown by a molecular beam epitaxy system on a semi-insulating GaAs (100) substrate. The grown wafer was then processed into a single plasmon waveguide geometry with a ridge width of 150 μm. The processed laser ridges were cleaved into 2-mm-long laser bars, which were then mounted onto copper heat sinks. After the wire bonding, the laser device was mounted onto the cold finger of a cryostat for electrical and optical characterizations.

The cw output power of the terahertz QCL was measured using a terahertz power meter (Ophir, 3A-P THz). For the cw operation, the current driver of the laser was operated in a constant current mode. For the power measurement, two off-axis parabolic mirrors were used for collecting and collimating the terahertz light into the power sensor. In addition, the beam path was purged with dry air to reduce water absorption. Note that the power values shown in Fig. 2(b) are the measured values without considering any calibrations, i.e., window transmission, water absorption, and mirror reflection.

The terahertz laser-irradiated GST samples were observed by scanning electron microscopy (SEM, JEOL 7800F). In the process of fabricating TEM samples, the interested areas of GST films were first deposited and protected with Pt by using a dual-beam focused ion beam (FIB, FEI Helios Nanolab 600), then mounted into a half copper grid with a standard lift-out process, and finally thinned into TEM lamellas. A JEOL JEM-ARM300F microscope was used to analyze the structural and elemental information of these lamellas at 300 kV. An energy dispersive x-ray (EDX) spectroscope was employed to collect the element distribution in the scanning transmission electron microscope (STEM) mode.