Introduction

Ferroelectric materials, which possess the capability of polarization reversal in response to electric fields, exhibit vast potential in various technological applications1,2,3,4, such as nonvolatile memories and neuromorphic devices. Owing to the unique feature of the polarization and charge distribution discontinuity5, the emergent topological textures in ferroelectric materials further provide a plethora of exotic physical phenomena and device applications6,7,8,9,10,11,12,13,14,15. The delicate balance among the electric, elastic, and gradient energies is critical for the formation of topological domains16. Diverse strategies like stain and interface engineering are utilized for meticulously tailoring the intricate interplay between the lattice and charge degree of freedom17. Due to the intricate nature of the prerequisite, the observation of topological polar domains is predominantly achieved within complex superlattices or substrate-specific heterostructures15,18. Reducing the dimensions or increasing the confinement of the ferroelectric materials is proved to be a promising strategy for facilitating the formation of topological texture and beneficial for controllability19,20,21. However, the critical thickness issue poses a significant constraint in efforts to further reduce the thickness of perovskite oxide films22,23,24, especially for ultrathin and high-density storage applications. In addition, practical applications of topological structures require the ability to manipulate them via external stimuli25,26,27, which is promising but still challenging in the perovskite oxides28.

The recently discovered sliding ferroelectrics provide an intriguing platform for exploring topological polar domains29,30,31,32. The interfacial charge transfer nature inherent in this material confines polarization to the interface33,34, thereby naturally introducing polarization discontinuity and fostering a robust coupling between polarization and the lattice35. The van der Waals (vdW) superlattice in sliding ferroelectrics could be facilely modulated by electrically or mechanically controlling crystallographic alignment36, which is expected to provide a potential fascinated route to generate and manipulate the topological structures dictated by polarization. Direct visualization of electromechanical response upon such 2D topological polar textures via Piezoresponse Force Microscopy (PFM) is promising for exploring the fundamental physics and potential device application21,37, as PFM provides a feasible way to read and write the polar textures, via locally applying electric and strain stimuli31. However, due to the relatively weak polarization of the interfacial ferroelectrics, as well as many imaging artifacts and cross-talks encountered, precise PFM characterization of 2D polar textures, even 2D ferroelectrics, still remains challenging38. Here, utilizing the high-resolution angle-resolved Lateral Piezoresponse Force Microscopy (LPFM), we experimentally observed the distribution of the in-plane (IP) electric polarization induced by the interfacial charge redistribution. Coupling of the IP and out-of-plane (OP) component of the polarization induced the formation of the topologically nontrivial spatial textures, which was further proved by the Scanning Transmission Electron Microscopy (STEM) characterization with atomic-scale resolution, and in good agreement with previous theoretical predictions39,40. The size of the polar texture can be controlled by the twist angle, as well as by means of applying a mechanical stress through an Atomic Force Microscope (AFM) probe. The facility of achieving high controllability makes the polar nanoregions featured with a network of topological textures in sliding ferroelectrics a promising platform for fundamental physics research and next-generation electric device application.

Result

Moiré patterns formation with an interlayer twist

By parallel stacking two monolayer boron nitride (BN) with a small twist angle, moiré patterns are formed, consisting of different stacking configuration networks (Fig. 1a). In the AB/BA stacking configuration with C3 symmetry, the transfer of electrons from N atoms to the neighboring B atoms occurs, thus leading to the interlayer charge redistribution and the generation of sizable OP polarization29. The IP component of polarization vanishes since electric dipoles are oriented only along the OP direction due to the exact horizontal overlapping between B and N (Fig. 1b). In the AA stacking configuration, both the OP and IP electric polarization vanishes due to the C3 and IP mirror symmetries. While in the intermediate (IM) stacking configuration between AB/BA and AA, only IP electric polarization could exist owing to its π-rotational (C2) symmetry about an IP axis41. For an arbitrary stacking configuration that falls outside the above high-symmetry stacking orders, both the OP and IP components of the induced electric field are expected to be detected. The appearance of both OP and IP polarization enables the construction of the polar domain networks in such a system40. To explore the topological texture in the moiré pattern, taking BN as an example, we constructed the almost parallelly stacked BN heterostructure, by using the “tear and stack” method. Kelvin Probe Force Microscopy (KPFM) characterization shows connected triangular patterns with clear contrast between the adjacent AB and BA triangle domains, indicating the formation of the moiré pattern41 (Fig. 1c).

Fig. 1: Atomic stacking and the surface potential of t-BN layers with a slight twist angle.
figure 1

a Sketch of moiré patterns through the parallel stacking of two monolayers BN with a slight twist angle, which comprises complex networks characterized by diverse stacking configurations. The light blue dots represent B atoms while the light pink dots signify N atoms. b Lateral perspectives of the lattice arrangement in an R-type bilayer h-BN, illustrating various stacking configurations including AA, AB/BA, IM and arbitrary stackings. The arrows depict the local electric dipole moment pointing from a B atom in one layer towards the nearest N atom in the adjacent layer, stemming from the interlayer charge redistribution. c The surface potential characterized by KPFM in a local region of t-BN layers with near zero-degree angle.

Full size image

Observation of the polarization textures in moiré patterns

The angle-resolved LPFM characterization could provide the information regarding the polarization orientation and magnitude21,42, affording a new perspective to explore the IP polarization distribution associated with the polar textures. It is worth noting that in the domain patterns acquired by PFM in 2D materials with weak polarization, especially the twisted 2D materials, a phase difference of less than 180° between regions of opposite polarities was commonly documented in current literatures (Table S1 in Supplementary Note 1), contradicting the expectations based on classical theory of PFM. The phase contrast of less than 180° is believed to be indicative of the existence of some crosstalk signals43, such as electrostatic effects and capacitive coupling, hindering the reasonable disclosure of the domain configuration in the twisted 2D materials due to the artefacts easily encountered in the corresponding PFM measurements. In this work regarding the PFM measurement, a valid method focusing on the enhanced signal-to-noise ratio (SNR) through meticulously selecting the operating frequency and driving alternating current (AC) voltage was proposed and implemented (Supplementary Note 1). This approach enables us to address the above issues, affording us the chance to obtain more reliable PFM results regarding domain configuration and thus unveil the intriguing 2D topological polar texture confined in moiré interface. To delve into the features relating to IP polarization in these connected triangular patterns, the angle-resolved LPFM phase and amplitude images acquired by the proposed method were collected through rotating the sample (Fig. 2a–d, Supplementary Fig. S23). Such domain patterns in the twisted BN (t-BN) sample could be immune to the thermal disturbance (Supplementary Fig. S24). The LPFM phase images are in accord with the corresponding LPFM amplitude images. When altering the rotational angle, the magnitude of the LFPM amplitude in the regions adjacent to triangular sides changes whereas the reversal between 90° and −90° in LFPM phase could be observed in the regions adjacent to triangular sides. By contrast, the relatively low magnitude level of amplitude and near 0° phase degree are always observed in the inner regions and vortexes of triangular patterns regardless of the rotational angle. It is worth noting that when evaluating IP polarization based on the LPFM, several factors especially the direction of the PFM cantilever should be considered44,45. Only the response of the projection of the IP polarization components perpendicular to the cantilever can be captured by the LPFM46,47, as they can result in the torsional bending of the cantilever leading to a lateral deflection on the photodiode detector (Supplementary Fig. S25). Thus, when the orientation of the IP component of the polarization is aligned parallel to the direction of cantilever, the absence of an effective IP polarization component precludes a lateral signal on the photodiode. The cantilever deforms stemming from the IP surface deformation cannot occur, thereby rendering the observation of the corresponding lateral PFM signal improbable. For the perpendicular case, the largest LPFM contrast should be acquired as there exists the largest magnitude of the effective IP polarization component, which is identical to the magnitude of the IP component of polarization (noted as LP). For the case where there is a deviation angle φ between the direction of cantilever and the orientation of the IP component of the polarization, the magnitude of the effective component of the polarization (LPeffective) is regulated by the formula LPeffective = LP·sin(φ). Thus, a trigonometric correlation is expected to be established between the LPFM response and the deviation angle. In this work, the cantilever was maintained to be perpendicular to the scanning direction during the whole test. When the sample was anticlockwise rotated at an angle from 0° to 180° relative to the initial state, the amplitude magnitude in the regions adjacent to triangular sides almost reached their maximum level when the sides were perpendicular to the cantilever detection. By contrast, the minimum value level of amplitude occurred when the sides were parallel to the cantilever detection. The relationship between the rotational angle α and the LPFM response around sides could be described by the trigonometric function, which is consistent with the preceding analysis (Fig. 2e and Supplementary Fig. S26). The corresponding LPFM phase around the sides reversed its position between the positive and negative value ranges when the inclination of the triangular sides alternates between left and right orientations relative to the cantilever’s direction. This indicates the direction of the LPeffective distributing along the regions adjacent to these sides reversed since the effective IP components is the projection of the IP polarization components. Based on the above observations, it can be concluded that the direction of the corresponding IP polarization in the position adjacent to triangular sides is approximately parallel to the triangular sides. Since the analogous evolution in LPFM contrast of each side could be observed during the rotating of the sample, the direction of the IP polarization along each side should be the same when the side was rotated to an identical position. In other words, the orientation of the IP polarization distributing in the regions adjacent to the three triangular sides was approximately parallel to its corresponding side and had the identical rotational trend with respect to the inter triangular regions, which is an indication of Bloch-type merons.

Fig. 2: LPFM mapping in t-BN layers and the construction of an IP piezoresponse vector map.
figure 2

The exemplified LPFM amplitude (a, b) and phase (c, d) images of the t-BN layers obtained before (a, c) and after (b, d) the sample is rotated anticlockwise at different angles (namely, 0° and 90°) relative to the initial state. The angles in the figures denote the rotational angle relative to the initial orientation of the sample and the different. It should be noted that the direction of the cantilever is maintained unchanged during the whole experiment. e The relationship between the LPFM response intensity and the rotational angle α, which is expected to be expressed by a trigonometric function. f The IP piezoresponse vector distribution of a local region in the t-BN layers corresponding to the region in (a, b). The gray arrows in (c, d) denote the orientation of the cantilever and the double-headed arrows indicate the scan direction.

Full size image

To further verify the above consideration, the map of local IP piezoresponse vectors was further constructed according to the high-resolution LPFM amplitude and phase images of the quasi-identical region with the sample oriented along both 0° and 90° orientations relative to the initial state48,49,50,51,52. As elucidated before, only the component of IP polarization perpendicular to the cantilever can be detected by the LPFM. Thus, two vertical components x and y of the PFM response due to IP polarization can be acquired, where x and y component is orthogonal with respect to the original crystallographic alignment before the sample is rotated, respectively. The magnitude and direction of an IP piezoresponse vector can be determined by the PFM piezoresponse of the x component and y component at each local position (Supplementary Note 2). Based on the assumption that there is a linear correlation between piezoresponse vector and electric polarization, we can gain insight into the intricate positional distribution of the IP polarization. From the obtained IP piezoresponse vector map (Fig. 2f), it is observed that the piezoresponse vectors exhibit an anticlockwise curling pattern in an upward triangular region, whereas they curl in the clockwise direction in a downward triangular region. This arrangement gives rise to a configuration resembling Bloch-type merons. Such a pattern formed by the IP piezoresponse vectors exhibits a strong concordance with the PFM contrasts acquired under any rotational angle (Supplementary Fig. S23).

In addition to LPFM, the Vertical Piezoresponse Force Microscopy (VPFM) is also a powerful tool to investigate the OP polarization distribution as the piezoelectric response is generally proportional to the OP polarization39. Similar connected triangular patterns could be observed in both the VPFM phase and amplitude images (Supplementary Fig. S27), which is corresponding to the connected triangular patterns characterized by KPFM. It is worth noting that obtaining ideal 180° phase difference between the regions of AB and BA stacking is challenging due to the unavoidable electrostatic effects associated with the fluctuation of electric potential between these regions (Supplementary Note 1). The largest VPFM signal contrast along the triangular sides could be observed. However, the lowest OP polarization is assumed to lie in these sides as indicated by the theoretical prediction that these sides are the regions featured with the IM stacking where only lateral polarization could exist. The IP flexural crosstalk could be a plausible explanation for the above inconsistency (40, 41), which is corroborated by the observation of abrupt changes in the VPFM signal contrast after rotating the sample (Supplementary Fig. S27). This phenomenon also indicates that a sizable IP polarization distributes along these triangular sides as it is one of the prerequisites for the occurrence of IP cantilever flexure47,53. To gain further insight into the OP polarization distribution, the VPFM with quadrature phase differential interferometry (QPDI) was employed to obtain the OP electromechanical response signal of t-BN, where the observed parallel strip-like domains along the triangular contour could possibly be associated with the strain and the strain gradient adjacent to the triangular sides (Supplementary Note 3).

The fabricated t-BN samples were investigated using atomic-resolution Scanning Transmission Electron Microscopy annular dark field (STEM-ADF) imaging to provide rational evidences for the presence of Bloch-type merons. From the diffraction contrast in low-magnification STEM-ADF images (Fig. 3a, b), large reconstructed hexagonal arrangements with moiré periodicity are clearly observed. These domains often maintain good periodicity over hundreds of nanometers, and were occasionally distorted by the global strain unavoidable in suspended 2D crystals. Adjacent triangular domains are confirmed to be AB and BA stacking configurations by their distinguishable atomic structures (Supplementary Fig. S28). They were separated by wide lines of brighter contrast, consistent with the KPFM results. Figure 3c presents high-resolution images of various stacking configurations including AA, AB/BA, IM and arbitrary stackings in the moiré rhombus supercell, which are confirmed by their unique atomic patterns with well agreement with the simulations (Supplementary Fig. S29). Characteristic structural units were selected as templates, and similarity matching was applied across large-scale atomic-resolution images. The matching results are highlighted with shaded regions in Fig. 3b, clearly revealing the presence of extensive arbitrary patterns between the adjacent AB and BA domains. The atomic-scale ADF images of the transitioning domain wall between AB and BA patterns are presented in Fig. 3d. By extracting the characteristic triangular structural units from the AB and BA domains on either side as reference frames, we can map the relative in-plane lattice displacements within the rotated BN atomic layers (Fig. 3e). This displacement vector field provides a clear representation of the AB\BA domain wall width and the interlayer sliding between the layers down to single atom scale associated with the rotation. In this case, where the domain size is 49.2 × 54.3 nm, the width of the AB/BA domain wall is estimated to reach 38.1 nm. The results from STEM suggest that the sliding in-plane ferroelectric signal is originated from the atom displacement induced by the interlayer stacking transited gradually from the center to the edge of the moiré supercell. This is consistent with the results from LPFM characterizations shown in Figs. 1 and 2, where the dipoles are oriented perpendicular to the interface of t-BN at the center region of AB/BA domains, and the dipoles are parallel to the surface at the outer domain walls of the AB/BA domains. The coincidence between STEM and PFM characterization provides strong evidence to further confirms that the polarization texture within t-BN shows the theoretically predicated Bloch-type merons characteristics39,40.

Fig. 3: Atomic STEM observation of the polar texture in t-BN sample.
figure 3

a, b Low-magnification annular dark-field (ADF) STEM image of the periodic moiré domain array in t-BN layers with twist angles of 0.28°, showing clear reconstructed domain contrast variation, consistent with the PFM measurement. The dashed rhombus indicated the supercell size of the polar texture. AA, AB, BA, IM and arbitrary regions are shaded in red, blue, green, yellow, and pink, respectively. c A series of high-resolution atomic-scale images of various stacking configurations found in the polar texture in (b), including AA, AB/BA, IM and arbitrary stackings, with structure models overlaid on the upper portion. d Atomic scale STEM-ADF images of domain wall between the AB and BA pattern as highlighted by white rectangle in (b). e Atomic displacement mapping of the reconstructed boundary showing the displacement variation between the layers when transforming from AB to BA domains. The arrows indicate the displacements of atomic columns within the two triangular reference frames as illustrated by the atomic models below. The purple dashed lines delineate the width of the domain wall.

Full size image

Ex-situ and in-situ modulation of polarization textures

To evaluate the changing behavior of the polar textures with the twist angle, θt-BN, the angle-resolved LPFM phase and amplitude images of the t-BN samples with different θt-BN were collected (Supplementary Fig. S30). Based on the theory of moiré superlattices, the relationship between the periodicity Λm and the twist angle θt-BN is governed by the equation Λm = (a/2)/sin(θt-BN/2) if the lattice mismatch could be neglected54, where a denotes the h-BN lattice constant (~0.25 nm) and the periodicity Λm is determined by the distance between the two regions of AA stacking according to the PFM images. This formula is considered to be reasonable to determine the twist angle if there is a negligible strain condition in the layers with a rigid lattice55,56. Given the strong bonding of B-N and the fact that the residual strain could be hugely suppressed after the annealing processing, the above approach is thought to be applicable to the calculation of the twist angle θt-BN of the t-BN samples in this work. These t-BN samples exhibit a periodicity Λm ranging from ~160 to ~505 nm, where the accompanied θt-BN varies from ~0.09° to ~0.028°. Following the same method for the reconstruction of an IP piezoresponse vector map (Section II in supporting information), the IP piezoresponse vector maps of the t-BN samples with several different θt-BN were determined (Fig. 4). Similar patterns accommodating a network of Bloch-type merons could be observed in the t-BN samples with these different θt-BN ranging from ~0.09° to ~0.028°. Besides, the t-BN sample, possessing a greater θt-BN, displays a conspicuous enlargement in the area of topological textures. Besides, it is worth noting that a reduction in θt-BN of the t-BN samples is accompanied by an increase in both the stacking domain size and the width of the domain wall (Supplementary Fig. S31). The variation of domain wall width with domain size is similar to the theoretical calculated value in twisted graphene system57.

Fig. 4: The evolution of IP piezoresponse vector map with different θt-BN.
figure 4

IP piezoresponse vector map of the t-BN samples with different θt-BN, namely, ~ 0.09° (a), ~ 0.055° (b), ~ 0.041° (c) and ~ 0.028°(d).

Full size image

As an exemplary case to evaluate switchable features of the polar nanoregions after their formation, the evolution of the domain patterns consisting of a network of Bloch-type merons in t-BN after the application of the strain applied by PFM probe was imaged using LPFM mode. After the sample was subjected to stress conditions induced by the PFM probe (an applied force of 250 nN), the expansion of the inner regions of triangular patterns accommodating AB/BA stacking occurred (Fig. 5). According to the mapping of IP piezoresponse vectors determined by the PFM images obtained before and after the sample was rotated at 90° (Figs. 5d, f), the expanded rhombic patterns forming by these IP piezoresponse vectors after the stress manipulation exhibited winding numbers, divergences and curls corresponding to Bloch-type merons analogous to that of the initial state. To put it another way, only the area of the topological textures was enlarged while the underlying topological state/feature was maintained unchanged when the sample was subjected to the applied stress. It’s obvious that the polarization distribution can be modulated through the local mechanical force applied to the sample using the PFM probe, which is anomalous to the variation in the area of topological textures modulated by the twist angle. Notably, the t-BN samples exhibit a subtle broadening of the domain walls after the application of stress (Supplementary Fig. S32). While the electric field can regulate polarization transitions in twisted van der Waals structures, the pinning effect occurring at AA stacking sites with considerable energy barriers in twisted samples hinders the mobility of domain walls, thereby constraining the capacity of the electric field to alter the topological texture. By contrast, stress can directly induce shear-driven domain motions, thus circumventing the domain-wall pinning effect of the AA sites and effectively switching polarity states58,59. These phenomena highlight the promising tunable and non-volatile characteristics in response to the stress of topological textures formed by 2D vdW twisted layers, which might open a new way to design ferroelectric memories.

Fig. 5: The LPFM mapping in t-BN layers subjected to the stress and the corresponding IP piezoresponse vector map.
figure 5

The surface morphology (a, e), LPFM amplitude images (b, f), LPFM phase images (c, g), and the corresponding IP piezoresponse vector maps (d, h) in the region denoted by red squares in (c, g) in BN layers before (a–d) and after (e–h) the sample was subjected to stress conditions applied by PFM probe.

Full size image

Discussion

In summary, the electronic OP and IP of the polarization were unambiguously illustrated in sliding ferroelectrics fabricated by twisting BN layers across varying strain scenarios, experimentally demonstrating the presence of the Bloch-type meron. The topological textures were found to be facilely tunable by the twist angle and the application of a strain, thereby demonstrating promising controllable characteristics. Such 2D topological polar texture has the unique advantages. The topological structure originates from the relative position of atoms with different electron affinities in the lateral and vertical directions. The simple lattice-polarization coupling at the 2D limit, distinct from the interplay of elastic, electrostatic and gradient energies in oxide, would make such a system ideal for symmetry and strain engineering to form diverse topological nontrivial phases. The direct PFM observation of such topological polar texture presented here provides a feasible way for further study of the electromechanical coupling and engineering of topological structures. As the topological polar structures are fully confined at the moiré interface, 3D stacking of different topological layers with coupled/decoupled vertical interaction is very straightforward for potential high-density storage and coupling between different topological textures, i.e. polar-spin/valley texture coupling. This work delves into the topological textures in ultrathin 2D vdW sliding ferroelectrics, revealing its potential for the advancement of next-generation electronic devices.

Methods

Fabrication of artificially stacked BN

The BN crystals were purchased from HQ Graphene. Stacked BN is produced by the “tear and stack” method, where half of the single BN is picked up and stacked on top of the remaining half. The BN was exfoliated on SiO2 (285 nm)/Si substrate, and its thickness was determined by an optical microscope. After picking up half a piece of BN through the PC film, we tried to stack the remaining half piece while maintaining nearly zero degree of twist. Subsequently, the t-BN was placed on top of the graphene and grounded via an evaporated electrode.

Kelvin Probe Microscopy measurement

The surface potential of t-BN was probed using KPFM, which was performed on an Asylum Research Cypher S system with HQ:NSC14/Cr–Au conductive probe. KPFM is a non-contact scanning microscopy technology that detects the potential difference between the probe tip and the sample. KPFM is a standard two-pass method (called nap mode at Asylum Research). First, the surface topography is obtained using a tapping method similar to standard AC mode (without direct current voltage). The feedback loop of the AFM controls the probe tip to lift a fixed distance in the second scan line, and then records the surface potential of t-BN based on the electrostatic interaction between the sample and the probe.

Piezoelectric Force Microscopy measurement

Both the Asylum Research Cypher S system and the Asylum Research Vero system from Oxford Instruments were used to conduct the PFM observation. If not specially stated, the PFM data were acquired using the Asylum Research Cypher S system. VPFM was performed on the Asylum Research Cypher S system employing a Arrew-EFM conductive probe with a spring constant of ~2.8 N m−1 and a free resonance frequency of ~75 kHz. In VPFM imaging, the OP polarization is measured by recording the cantilever perpendicular deflection signal (Supplementary Fig. S25). LPFM was performed on the Asylum Research Cypher S system using a Arrew-EFM conductive probe with a resonance frequency of ~75 kHz and a spring constant of ~2.8 N m−1. LPFM is a close relative of VPFM, but it is performed in a scanning direction orthogonal to the probe cantilever. LPFM detects the IP component of polarization by tracking the lateral deflection of the probe cantilever, which is caused by the IP electric field components induced surface shearing. The schematic diagrams of VPFM and LPFM are shown in Supplementary Fig. S25, where the lateral torsion signal of the cantilever beam is recorded by a four-quadrant detector. To obtain a better signal-to-noise ratio, LPFM worked in Dual AC Resonance-Tracking mode during the test. When obtaining LPFM data on t-BN at different rotational angles, the AFM sample stage was kept inside the instrument to ensure consistent testing conditions. The probe was lifted a small distance away from the sample surface, and a custom-made rotational holder was used to rotate the sample to the target angle. This approach guaranteed that the testing parameters stayed consistent among measurements. To ensure the sample is rotated at the target angle, we initially identify a prominent edge of the sample using the CCD camera integrated with the AFM system, followed by tracing the outline of this edge on the screen. Once the sample is rotated, we verify the angle between the edge and the initial outline displayed on the screen and then make adjustments by further rotating sample until the desired angle is achieved. For the PFM observation using the QPDI technique, the Asylum Research Vero system (or the Vero AFM) and an HQ:NSC18/Pt conductive probe with a free resonance frequency of 75 kHz and a spring constant of 2.8 N/m are used. The details regarding the Vero AFM integrated with the QPDI technique are depicted in Supplementary Note 3. All the tests were performed at room temperature.

Scanning Transmission Electron Microscopy characterization

STEM imaging was performed by using an ultrahigh vacuum aberration-corrected JEOL-ARM200, equipped with a cold field-emission gun operating at 200 kV. The probe current was approximately 15 pA. The convergence semi-angle and the inner acquisition semi-angle were set to 27 and 110 mrad. The STEM images were captured using a frame size of 2048 × 2048 pixels, with an exposure time of 24-36 μs. All STEM experiments were conducted at room temperature.