Epitaxial growth of Mg_xCa₁₋xO on 4H–SiC(0001) and β-Ga₂O₃$\left( {\bar 201} \right)$ wide band gap semiconductors with atomic layer deposition

Abstract

SiC and Ga₂O₃ are promising wide band gap semiconductors for applications in power electronics because of their high breakdown electric field and normally off operation. However, lack of a suitable dielectric material that can provide high interfacial quality remains a problem. This can potentially lead to high leakage current and conducting loss. In this work, we present a novel atomic layer deposition process to grow epitaxially Mg_xCa_1−xO dielectric layers on 4H-SiC(0001) and β-Ga₂O₃$\left( {\bar 201} \right)$ substrates. By tuning the composition of Mg_xCa_1−xO toward the substrate lattice constant, better interfacial epitaxy can be achieved. The interfacial and epitaxy qualities were investigated and confirmed by cross-sectional transmission electron microscopy and X-ray diffraction studies. Mg_0.72Ca_0.28O film showed the highest epitaxy quality on 4H-SiC(0001) because of its closest lattice match with the substrate. Meanwhile, highly textured Mg_0.25Ca_0.75O films can be grown on β-Ga₂O₃$\left( {\bar 201} \right)$ with a preferred orientation of (111).

Introduction

With the rapid expansion of semiconductor industry, power transistors are widely applied in the fast-growing fields such as electrical vehicles, solar energy, and giant data centers [1, 2]. The most important figure of merit (FOM) for high frequency power transistors is R _on × Q _g that characterizes the energy loss for high frequency operation [3, 4, 5]. Here, R _on is the resistance during “on” state, which is proportional to the conducting loss. Q _g stands for the total gate charge, which represents the loss during switching operations. Silicon-based power devices cannot fulfill the required high FOM in high-power applications because its low breakdown electric field requires a larger gate to drain distance, which would increase the R _on substantially and, thus, increase the total energy loss [6, 7, 8].

Wide band gap semiconductors, including SiC, Ga₂O₃, and GaN, can overcome this challenge because their high breakdown electric field facilitates a smaller gate to drain distance, so that R _on can be reduced proportionally. Furthermore, GaN-based high electron mobility transistors (HEMTs) can further reduce R _on by introducing 2-dimensional electron gas (2DEG). However, the application of GaN-based HEMT arouses safety concerns for power electronic devices because of its intrinsically normally-on feature [9]. The current commercially available normally-off GaN HEMT devices are achieved by adopting the so called “cascode” configuration. This configuration consists of the series connection of a low-voltage normally-off Si-based metal oxide semiconductor field effect transistor (MOSFET) with a high voltage normally-on GaN HEMT [10]. Other strategies to achieve normally-off GaN HEMT, such as p-Gate [11], fluorine gate [12], and gate recess [13, 14], were proposed and received numerous attention from both academia and industry. However, their complicated fabrication process and low reliability prevent them from mass industrial-level production. Therefore, the intrinsically normally-off SiC and Ga₂O₃ power transistors would be more favorable in industrial applications because of their safety nature.

A key challenge for SiC and Ga₂O₃ devices is the lack of a suitable dielectric material that can both reduce the gate leakage and the interfacial defects [15]. High gate leakage current would deteriorate the device reliability and increase the on-state loss. Also, interfacial defects under the gate electrode can substantially increase switching loss by trapping electrons that in turn increase Q _g. Moreover, charged defects can also scatter electrons in the channel so that R _on is increased as well.

SiC can form a native oxide (SiO₂) with one or two order of magnitude higher interfacial defect density than the SiO₂/Si interface because of the incomplete oxidation of carbon [16, 17]. Moreover, the low dielectric constant (3.9) of SiO₂ limits the maximum safe operating electric field, and the high breakdown field of SiC cannot be used fully [18]. As for β-Ga₂O₃ transistors, Schottky gate structure [19, 20] was used to form metal–semiconductor field effect transistor (MESFET). This type of structure has a nature of leaky gate that prevents high efficiency device operation and renders a low on/off ratio. HfO₂ [21], SiO₂ [22], and Al₂O₃ [23, 24] had been proposed as the gate dielectric to reduce the gate leakage currents. However, very few dielectric materials had been proposed to achieve a very high-quality interface along with very low leakage [25].

Both MgO and CaO have relatively high dielectric constants (MgO 9.8, CaO 11.8) and relatively large band gaps (both 7–8 eV) compared with SiC (3.3 eV) and Ga₂O₃ (4.5–4.9 eV) and, thus, might be suitable gate oxides for wide band gap semiconductors. Recently, our group has demonstrated that an epitaxial Mg_xCa_1−xO film can be grown on GaN(0001) surfaces using a suitable ALD process, achieving low interfacial defect density and high electrical performance [26]. We demonstrated a low defect density interface between semiconductor substrates and a lattice matched epitaxial gate dielectric material. Q _g can be effectively minimized with this reduced interfacial defect density, and therefore, the overall device loss was reduced. Moreover, because there are few grain boundaries to serve as the leakage pathways, gate leakage loss was also reduced. 4H SiC has a similar hexagonal symmetry on the (0001) surface with the GaN(0001). With only −4% of lattice mismatch with GaN, SiC is widely used as the epitaxial growth substrate for GaN in LED industry. Bhuiyan et al. also demonstrated that Mg_xCa_1−xO is a promising radiation-tolerant gate dielectric material for GaN-based MOSHEMTs [27]. β-Ga₂O₃ on the other hand has a monolithic structure that does not exactly match the hexagonal symmetry of the rock salt Mg_xCa_1−xO(111) surface. But its $\left( {\bar 201} \right)$ surface has a structure similar to a hexagonal structure, which makes it also suitable as the substrate for GaN epitaxial growth [28, 29]. Therefore, it is highly possible to grow Mg_xCa_1−xO epitaxially on the SiC(0001) and β-Ga₂O₃$\left( {\bar 201} \right)$ surfaces which are close to GaN(0001).

Moreover, as shown in Fig. 1, the SiC lattice sizes lie between those of MgO and CaO lattices. Therefore, by tuning the composition of Mg_xCa_1−xO, a lattice match with SiC substrates can be achieved. According to Vegard’s Law [30], a 76% Mg and 24% Ca oxide mixture can achieve a lattice match with SiC with an epitaxy relation of $\left( {111} \right)_{{\rm{Mg}}_{\it x} {\rm{Ca}}_{1 - {\it x}} {\rm{O}}} //\left( {0001} \right)_{{\rm{SiC}}}$. In the meantime, because β-Ga₂O₃$\left( {\bar 201} \right)$ surface is similar to GaN(0001) surface, it is anticipated to be latticed match with a composition near Mg_0.25Ca_0.75O. Epitaxial Mg_0.75Ca_0.25O grown by molecular beam epitaxy (MBE) on SiC has been reported as an alternative dielectric oxide [31]. Moreover, epitaxial GaN has been reported to be grown on β-Ga₂O₃$\left( {\bar 201} \right)$ surface by MBE [32]. Therefore, MBE-grown dielectrics can be expected for Ga₂O₃ substrates as well. Measurements on HEMTs with MBE-fabricated MgCaO insulators showed improved gate-lag [33]. However, MBE endures problems such as low industrial compatibility due to ultra-high vacuum requirements and high cost. Furthermore, MBE is not applicable to large substrates needed for cost-effective industrial processing. Thus, a more scalable and lower cost method to deposit epitaxial Mg_xCa_1−xO is highly desirable. Atomic layer deposition (ALD) is a well-established method to achieve films with high uniformity and highly controllable precision. Moreover, ALD technique is more cost effective and industrial compatible compared with MBE, providing potential way to achieve precisely controlled and high-quality dielectric films. Thus, in this study, we focused on applying the ALD process to grow epitaxially Mg_xCa_1−xO on SiC(0001) and β-Ga₂O₃$\left( {\bar 201} \right)$ substrates.

Figure 1: (a) 2θ–ω scans of three Mg_xCa_1−xO/SiC samples. (b) Lattice constant of Mg_xCa_1−xO with respect to Ca content in the film. Comparing SiC(0004) lattice with Mg_xCa_1−xO(222) lattice, it shows that the Mg_0.72Ca_0.28O has the smallest lattice mismatch with SiC.

Results and discussions

Epitaxial growth of Mg_xCa_1−xO on SiC

SiC has two primary polytypes for semiconductor applications, 4H and 6H, respectively. Both 4H and 6H are hexagonal, and each has a similar (0001) surface structure and a relatively small lattice mismatch to GaN, −3.8% and −3.5%, respectively. Traditionally, 4–8 degrees of off-axis SiC were grown with liquid or vapor phase epitaxial growth because the higher density of surface steps can improve the growth quality [34]. However, Hijikata et al. had pointed out that the lower angle of off-axis SiC can effectively reduce the interfacial traps [35]. In our case, the surface symmetry of off-angle SiC also does not match with the hexagonal structure of rock salt (111) surface. Therefore, an on-axis SiC 4H is selected as the substrate. To perform a comprehensive study on how lattice mismatch influences epitaxy growth, films with three compositions such as pure MgO, Mg_0.72Ca_0.28O, and Mg_0.51Ca_0.49O were investigated. According to Vegard’s law, as shown in Fig. 1(b), the SiC lattice matches with Mg_0.72Ca_0.28O is expected to have best epitaxy quality. The substrate surface pretreatments and ALD Mg_xCa_1−xO deposition were performed with the conditions described in the Experiments and Methods section. X-ray diffraction (XRD) was conducted to further confirm the lattice matching of the Mg_xCa_1−xO and SiC substrate. Figure 1(a) shows 2θ–ω scans of three Mg_xCa_1−xO/SiC samples. 2θ–ω results show that the lattice mismatch of Mg_0.51Ca_0.49O and Mg_0.72Ca_0.28O are +3.2% and +0.7%, and −3.9%, respectively. It is clear that the Mg_0.72Ca_0.28O has the closest match with SiC. This result is further verified by the corresponding rocking curves shown in Fig. 2. The narrower rocking curve of Mg_0.72Ca_0.28O/SiC (FWHM = 0.1°) than Mg_0.51Ca_0.49O/SiC (FWHM = 0.36°) and MgO/SiC (FWHM = 0.39°) indicates the former has the highest epitaxial quality. The fact that the lattice constants of Mg_xCa_1−xO films are very close to the corresponding Vegard’s law prediction, indicating that there is little stress. The extra sharpness of the Mg_0.51Ca_0.49O and Mg_0.72Ca_0.28O’s rocking curve might be due to the relaxation of the top layer of the film.

Figure 2: Rocking curves of Mg_xCa_1−xO/4H-SiC(0001). (a) Rocking curve of Mg_0.51Ca_0.49O/SiC at 2θ = 72.55°. (b) Rocking curve of Mg_0.72Ca_0.28O/SiC at 2θ = 74.70°. (c) Rocking curve of MgO/SiC at 2θ = 78.96°.

Ren et al. had pointed out that the drawback of this ternary dielectric system is the large difference in the ionic radius between Mg and Ca causing severe immiscibility [31, 36], particularly at high temperature. As a result, solid-solution CaO–MgO would be difficult to synthesize using bulk techniques because the two phases are immiscible over the entire composition range. However, the use of ALD as a film growth technique often allows for the formation of metastable phases, overcoming miscibility issues due to the unique self-terminating nature of growth. According to our research, Mg_xCa_1−xO ternaries are kinetically stable as solid solution films over the entire compositional range when grown at a remarkably low temperature of 300 °C. The absence of CaO and MgO peaks in the XRD pattern also confirmed that there is no phase separation.

Cross-sectional transmission electron microscopy (TEM) was used to study the interfacial region of this epitaxial structure. Like the typical GaN cross-sectional TEM, SiC also shows a three atomic line periodicity pattern in the vertical direction because of its hexagonal structure [26]. Therefore, the interface between the substrate and Mg_xCa_1−xO film can be determined by where this three-line periodicity discontinued. A sharp interface without any interfacial layer can be seen in Figs. 3(a) and 3(c). The traditional oxidized SiO₂ dielectric on SiC is caused by the roughened interface [37]. In the ALD-grown epitaxial Mg_xCa_1−xO on SiC, the pristine interface is preserved. The very well-defined diffraction spots in (b) also confirm the high epitaxial quality of the film. Twinning formation is common in crystalline films, with 180° mirror images having almost identical formation energy as shown in the Mg_0.72Ca_0.28O left-right symmetric diffraction pattern [38, 39]. Moreover, from the diffraction pattern, the zone axis of the SiC substrate in these TEM images is $\left[ {11\bar 20} \right]$, and the film zone axis can be determined as $\left[ {0\bar 11} \right]$. Therefore, the epitaxial relation between the film and substrate is Mg_0.72Ca_0.28O(111) × $\left[ {0\bar 11} \right]$//SiC(0001) × $\left[ {11\bar 20} \right]$. Theoretical studies on Mg_xCa_1−xO ternary systems also predicted a CsCl phase is possible [40]. But the TEM diffraction pattern clearly showed that there is no phase other than rock salt and also no phase separation is observed.

Figure 3: TEM cross section of Mg_0.72Ca_0.28O/SiC(0001). (a) Cross-sectional TEM image. (b) Diffraction pattern of the (a) region. The zone axis of the SiC is $\left[ {11\bar 20} \right]$ and for Mg_0.72Ca_0.28O is $\left[ {0\bar 11} \right]$.

In summary, the XRD and TEM studies confirmed that the ALD Mg_xCa_1−xO can be grown epitaxially on 4H-SiC (0001) substrates. Mg_0.72Ca_0.28O shows the best lattice match with a +0.7% mismatch with the SiC substrate. The rocking curve also confirms that this smaller lattice match has a better epitaxial quality than the samples with larger mismatch.

Epitaxial growth of Mg_xCa_1−xO on Ga₂O₃

Although β-Ga₂O₃ has a monoclinic crystal structure and does not have hexagonal symmetry on any crystal face, the O atoms on β-Ga₂O₃$\left( {\bar 201} \right)$ surface are roughly lined up on the same plane [Fig. 4(a)] and formed a nearly hexagonal structure from the top view [Fig. 4(b)]. Moreover, the distance between these O atoms are close to the distance between the N atoms on GaN(0001). Kitamura et al. had demonstrated that an epitaxial GaN can be grown on β-Ga₂O₃, with a lattice mismatch as small as 2.6%. Therefore, it is possible to create an epitaxial growth of Mg_xCa_1−xO on a β-Ga₂O₃$\left( {\bar 201} \right)$ surface as shown in Fig. 4.

Figure 4: (a) Side view of β-Ga₂O₃$\left( {\bar 201} \right)$ lattice. (b) Top view of Mg_xCa_1−xO(111)/β-Ga₂O₃$\left( {\bar 201} \right)$.

As shown in Fig. 4(b), there are also two different possible orientations denoted as A and B. In both cases, Mg/Ca atoms are always stacking over substrate O atoms. In case A, half of film oxygen atoms are stacking over Ga atoms and the other half are on the vacancies surrounded by three oxygen atoms from the substrate. On the other hand, in case B, the film oxygen atoms are all on top of the vacancies surrounded by three oxygen atoms from the substrate. Thus, a φ scan is used to determine which orientation is achieved with ALD growth. The Mg_xCa_1−xO(200) peak and Ga₂O₃(202) peak are close in both 2θ and ω offset in Mg_xCa_1−xO(111)/β-Ga₂O₃$\left( {\bar 201} \right)$ structure. Therefore, a φ scan is set for scanning based on the position of Ga₂O₃(202) peak at ψ 54° and 2θ 38.62°. Figure 5 demonstrates the φ scan of Mg_0.25Ca_0.75O/β-Ga₂O₃, which has a 3-fold symmetry, confirming that there is only one in-plane orientation existing in the film. Furthermore, the substrate (202) peak does not overlap with the film (200) peaks in φ scan, indicating that the film has a B-type orientation as shown in previously in Fig. 4(b).

Figure 5: φ scan of Mg_0.25Ca_0.75O/Ga₂O₃$\left( {\bar 201} \right)$.

To achieve epitaxial growth of Mg_xCa_1−xO on β-Ga₂O₃ substrates, ultra-clean surfaces of β-Ga₂O₃ substrate are needed. However, β-Ga₂O₃ is unstable under both acidic and basic conditions. Therefore, its surface treatment is very challenging. Some reports show that Piranha and buffered oxide etch (BOE) treatments are effective to prepare the β-Ga₂O₃ surface [41]. In this study, both pretreatments are used for comparison.

The cross-sectional TEM images of Mg_0.25Ca_0.75O grown on both Piranha and BOE treated β-Ga₂O₃$\left( {\bar 201} \right)$ are shown in Figs. 6 and 7, respectively. The Piranha treatment is a common method for β-Ga₂O₃ surface cleaning used in a few Refs. [42, 43]. Although benefiting electrical performance in the literature, the Piranha-treated sample exhibits a 10 nm amorphous-like interfacial layer. This interfacial layer may be from the hydroxylation of β-Ga₂O₃ in the acid pretreatment.

Figure 6: Cross-sectional TEM of (a) Piranha-treated bare Ga₂O₃ surface and (b) Mg_0.25Ca_0.75O on Piranha-treated Ga₂O₃.

Figure 7: TEM of Mg_0.25Ca_0.75O/Ga₂O₃$\left( {\bar 201} \right)$ after BOE surface treatment. (a) High resolution image of Mg_0.25Ca_0.75O and Ga₂O₃ interface, a 2 nm interfacial layer is visible. (b) Diffraction pattern. The zone axis of β-Ga₂O₃ is [010]. The main diffraction spots of Mg_0.25Ca_0.75O indicate that the zone axis of the film is $\left[ {0\bar 11} \right]$. The stray diffraction spots from Mg_0.25Ca_0.75O are indicating that the film is highly textured with some misoriented crystal phases.

In the meantime, as shown in Fig. 7(a), the 30 s BOE-treated sample (1:6 original diluted with 5 times volume of water) has only a 2 nm interfacial layer. This is a substantial improvement compared with the Piranha-treated substrate. However, severe delamination has been observed on the edge of β-Ga₂O₃ substrates, indicating surface damage caused by BOE treatment. In Fig. 7(b), the diffraction pattern of both the film and substrate indicates an epitaxial relationship to be Mg_xCa_1−xO$\left[ {0\bar 11} \right]$ × (111)//β-Ga₂O₃$[010]$ × $\left( {\bar 201} \right)$. In addition, the diffraction pattern in Fig. 7(b) indicates a highly textured film, with some misoriented crystal phases.

XRD was then carried out to further investigate the epitaxial growth quality of Mg_0.25Ca_0.75O/β-Ga₂O₃. The full 2θ–ω scan of Mg_0.25Ca_0.75O/β-Ga₂O₃ is demonstrated in Fig. 8. The existence of (111) and (222) peaks of Mg_0.25Ca_0.75O indicates the film has a preferred (111) orientation. The (222) peak of the film is not as obvious as (111) because it is overlapping with (801) peak from the substrate. The β-Ga₂O₃(111), (601), and (801) peaks indicate that the substrate contains other small crystals with disordered orientations.

Figure 8: 2θ–ω of Mg_0.25Ca_0.75O/Ga₂O₃ and bare Ga₂O₃.

In summary, highly textured Mg_0.25Ca_0.75O can be grown on β-Ga₂O₃$\left( {\bar 201} \right)$ with a preferred orientation of (111). The dielectric films are highly textured rather than epitaxial mainly because β-Ga₂O₃$\left( {\bar 201} \right)$ lattice is not perfectly hexagonal, which can introduce mismatch to the film (111) surface. Another minor cause might be the nonideal surface pretreatments. The β-Ga₂O₃ surface pretreatment has a profound effect on the interfacial property. BOE treatment can effectively reduce the interfacial layer thickness damages of the substrate to some extent.

Conclusion

In this work, we demonstrated for the first time that epitaxial and highly textured Mg_xCa_1−xO can be grown on SiC(0001) and β-Ga₂O₃$\left( {\bar 201} \right)$ with an ALD process. Because SiC(0001) has hexagonal symmetry, Mg_xCa_1−xO can be grown epitaxially on it. The closest lattice matching Mg_0.72Ca_0.28O showed the highest epitaxy quality in an XRD rocking scan. Therefore, it has the greatest potential to achieve superior dielectric properties and should be the best candidate for device applications. In the meantime, highly textured Mg_0.25Ca_0.75O can be grown on β-Ga₂O₃$\left( {\bar 201} \right)$ with a preferred orientation of (111). The Mg_0.25Ca_0.75O film on β-Ga₂O₃$\left( {\bar 201} \right)$ is textured instead of epitaxial because the substrate surface is not perfectly matching the hexagonal symmetry of the film (111) surface. The crystal orientation relationship between Mg_xCa_1−xO and β-Ga₂O₃ was determined by XRD scans. Cross-sectional TEM showed that BOE treatment can effectively reduce the thickness of amorphous β-Ga₂O₃ top layer compared with Piranha treatment.

Experiments and methods

Substrate pretreatments

Both SiC and Ga₂O₃ substrates were purchased from MTI Corp. Surface pretreatments of substrates are crucial to the success of epitaxial growth of Mg_xCa_1−xO films because epitaxy is very sensitive to interfacial conditions. SiC substrates were soaked and rinsed 5 min in isopropanol and acetone and then methanol was conducted subsequently as basic surface cleaning. Then a 5 min UV-ozone and 30 s buffered oxide etch (BOE; 1:6 original diluted with 5 times volume water) were followed to remove the organic contaminants and surface oxide. In the case of Ga₂O₃, same surface pretreatment procedures were carried out for one batch of substrates. As a comparison experiment, the final step of 30 s BOE etch was replaced with 30 s Piranha (a mixture of 3:1 sulfuric acid and hydrogen peroxide) etch for the other batch of substrates.

Mg_xCa_1−xO film growth

ALD of Mg_xCa_1−xO was carried out in a home-built tubular reactor. Mg_xCa_1−xO films were obtained by alternating doses of calcium, magnesium, and oxygen sources, where bis(N,N′-diisopropylacetamidinato)calcium(II) dimer [13], bis(N,N′-di-sec-butylacetamidinato)magnesium, and H₂O were used as precursors, respectively. Structural formulas for Mg and Ca precursors are shown in Fig. 9. The Mg precursor is kept at 110 °C, whereas the Ca precursor is kept at 140 °C to provide enough vapor pressure. H₂O is kept at room temperature. During the deposition, the samples/substrates are placed on the substrate holder kept at 310 °C under vacuum. In each dosing cycle, either the Mg or Ca precursor is dosed into the reaction chamber and then purged for 30 s. Then H₂O vapor is dosed and purged. Four different precursor dosing ratios, namely, Mg only, Mg:Ca = 1:1, Mg:Ca = 1:2, and Mg:Ca = 1:3, were used. The dosing sequence is repeated for the corresponding number of Mg cycles, then Ca cycles. Because of the large size of the ligands, each metal cycle cannot fully cover one atomic layer. Therefore, the alternating Mg and Ca cycles can give a uniform film instead of layered structure. The compositions of the resulting films were determined by RBS (Rutherford Backscattering Spectrometry) to be MgO, Mg_0.72Ca_0.28O, Mg_0.51Ca_0.49O, and Mg_0.25Ca_0.75O, respectively. The film growth rates with different dosing ratios and the corresponding compositions are summarized in the literature published by our group [26]. The total ALD Mg_xCa_1−xO film thickness is 40 nm for further characterization. Because MgO and CaO are both reactive with moisture, a 5 nm of Al₂O₃ is grown in situ on top of Mg_xCa_1−xO for both SiC and β-Ga₂O₃ substrates to prevent water in air from permeation into the epitaxial film during XRD and TEM characterization.

Figure 9: Structures of the metal precursors. (a) bis(N,N′-di-sec-butylacetamidinato)magnesium and (b) bis(N,N′-diisopropylacetamidinato)calcium(II) dimer.

Film epitaxy characterizations

Cross-sectional TEM and diffraction patterns were conducted on Mg_xCa_1−xO/SiC and Mg_xCa_1−xO/Ga₂O₃ samples with the Joel 2100 High resolution TEM in Harvard CNS (Center for Nanoscale Systems). The TEM samples were beforehand prepared with an FEI Helios 660 FIB (focused ion beam) system and then cleaned with 500 eV Ar ion beam with a Fischione NanoMill 1040 in Harvard CNS. XRD measurements were carried out in the Bruker D8 high resolution XRD facility in Harvard CCB (Chemistry & Chemical Biology Department) X-ray Laboratory. The φ scan of Mg_xCa_1−xO is achieved by first determining the 2θ and ω angle of Mg_xCa_1−xO (200) position with asymmetric 2θ–ω scan and rocking scan. Then a φ scan is conducted by collecting the Mg_xCa_1−xO (200) peak while rotating the sample stage.

Author contributions

The manuscript was written through contributions of all authors. X.B.L. and X.G. contributed equally.

Supplementary material

To view supplementary material for this article, please visit https://doi.org/10.1557/jmr.2019.376.