Cryo-mediated exfoliation and fracturing of layered materials into 2D quantum dots

See allHide authors and affiliations

Science Advances  15 Dec 2017:
Vol. 3, no. 12, e1701500
DOI: 10.1126/sciadv.1701500


Atomically thin quantum dots from layered materials promise new science and applications, but their scalable synthesis and separation have been challenging. We demonstrate a universal approach for the preparation of quantum dots from a series of materials, such as graphite, MoS2, WS2, h-BN, TiS2, NbS2, Bi2Se3, MoTe2, Sb2Te3, etc., using a cryo-mediated liquid-phase exfoliation and fracturing process. The method relies on liquid nitrogen pretreatment of bulk layered materials before exfoliation and breakdown into atomically thin two-dimensional quantum dots of few-nanometer lateral dimensions, exhibiting size-confined optical properties. This process is efficient for a variety of common solvents with a wide range of surface tension parameters and eliminates the use of surfactants, resulting in pristine quantum dots without surfactant covering or chemical modification.


Quantum dots (QDs) with diameters in the range of 2 to 10 nm exhibit unique electronic and optical properties due to confinement of charge carriers in all spatial dimensions (15). Because of the tunability of optoelectronic properties and band structures, QDs are of interest in many applications, such as photovoltaics (6), bioimaging (7, 8), catalysis (9), light-emitting devices (10), photodetectors (11), etc. Two-dimensional (2D) materials, including typical transition metal dichalcogenides, graphene, and hexagonal boron nitride (h-BN), intrinsically exhibit exceptional physiochemical properties when the vertical dimension is reduced into the thickness of a monolayer or few layers (1219). The reduction of lateral size into nanometer scale of a 2D monolayer or few-layer nanosheets to form 2D QDs can result in synergistic effects, such as a combination of the distinctive properties associated with both 2D nanosheets and QDs. Before exploring promising physiochemical properties of 2D QDs from a range of available compositions, it is essential to develop effective, scalable, and generic processes to produce these structures. Some recent reports showed top-down cutting methodologies for preparing graphene and MoS2 QDs via hydrothermal, electrochemical, microwave, nanolithography, or ultrasonic shearing processes but these processes involved multiple steps and required harsh conditions (concentrated acids, strong oxidizing agents, energy intensive grinding, or high temperature, etc.) (2023). Here, we report a simple, universal, and scalable synthetic approach based on cryo-mediated liquid-phase exfoliation (LPE) to prepare 2D QDs directly from bulk-layered material powders in a series of solvents within a short time.

In our approach, we introduce a cryo-mediated pretreatment step in which the bulk-layered material powders are soaked in liquid nitrogen for a period of time. Then, the bulk powders are dispersed in a solvent, such as isopropanol (IPA)/H2O mixture (for example, IPA/H2O = 1:1 by volume), followed by LPE with the assistance of bath ultrasonication (24, 25). First, to the best of our knowledge, the cryogenic pretreatment by soaking the raw powders in liquid nitrogen was adopted for the first time to prepare 2D QDs. The cryo-pretreatment facilitates the efficiency of peeling off few-layer (comprising mostly monolayer) sheets from the bulk powders in the following bath ultrasonication-assisted LPE while simultaneously mediates the fracture of these few-layer sheets into high-quality pristine QDs within short time. This is because the interlayer van der Waals force between layers is weakened, and some small cracks in intralayer are formed after cryo-pretreatment process. Second, the cryo-pretreatment process only needs to adopt the pristine raw powders as the precursors without introducing any other impurity elements, and thus the pristine QDs can be directly achieved without any surfactant and other chemicals contamination (20, 26). Third, the cryo-pretreatment process exhibits less dependency on solvent properties, which means various common low–boiling point solvents with a wide range of surface tensions and even pure water can be selected for LPE of layered materials into QDs, showing great improvements compared with the existing LPE technique for graphene and 2D nanosheets in typical high–boiling point organic solvents, such as N-methyl-2-pyrrolidone (NMP), N,N-dimethylformamide (DMF), and dimethyl sulfoxide (24, 27, 28). Fourth, the as-developed cryo-mediated LPE process can be applied to a wide range of layered materials, and a series of 2D QDs can be fabricated with pristine state, demonstrating a universal approach for the production of 2D QDs. Last, but not the least important, the quality of the as-prepared pristine QDs can be controlled to have predominant mono- and bilayer thickness and a uniform lateral size of 2 to 3 nm.


Figure 1A illustrates the procedure as exemplified by preparing 2D MoS2 QDs. In the early ultrasonication (delamination) stage, the bulk powders are easily exfoliated into few-layer or even monolayer nanosheets in a relatively high yield [~8 weight % (wt %)]. These nanosheets have lateral sizes in a wide range of 100 to 5000 nm, as shown in the representative case of MoS2 in Fig. 1B. The origin of facilitation of LPE efficiency may lie in crack formation in the layered powder and weakened interlayer van der Waals force between 2D layers. Our process involves two steps: An initial soaking of room temperature bulk layered powder in liquid nitrogen (quench step), and then dispersion of the soaked raw powder into room temperature solvent following by sonication (recovery step). This procedure results in a marked temperature change of 2D powder in a cycle, room temperature/liquid nitrogen temperature/room temperature. The fast thermal relaxation in the quenching step would form small cracks in the 2D powder, which is a similar feature to the effect of the quenching treatment for the preparation of graphene nanosheets (29). These cracks, serving as capillaries, allow permeation of solvent, which facilitates the LPE by bath sonication (30). Moreover, the thermal expansion in the recovery step helps weaken the interlayer van der Waals force of 2D materials. With the extension of ultrasonication treatment time, nanosheets are further disintegrated into QDs by ultrasonic wave–induced lattice breakup as evidenced by the emergence of lots of nanoscaled holes on the basal planes of nanosheets (Fig. 1C). The liquid nitrogen pretreatment of the raw powders will affect the mechanical properties, especially the intralayer molecular bond embrittlement induced by the ultralow temperature treatment (31), which subsequently promotes the fracturing of few-/monolayer nanosheets into QDs. The QDs are finally separated from relatively larger nanosheets by centrifugation and vacuum filtration resulting in a QD yield of ~1 wt % (Fig. 1D). The whole process is free of concentrated acid and surfactant. Without the cryo-pretreament, only nanosheets (thickness, ~1.6 nm; and lateral size, ~100 nm) can be obtained by identical LPE conditions (fig. S1).

Fig. 1 Procedure of cryo-exfoliation for preparing pristine 2D QDs (represented by MoS2).

(A) Schematic illustration of the steps for preparing 2D QDs. The outlined step in a red rectangle is the cryo-pretreatment. (B to D) AFM results showing the evolution of nanosheets to QDs, (B) ultrathin MoS2 nanosheets at the delamination stage, (C) small MoS2 nanosheets with tiny holes (the red arrows point to the holes.), and (D) pure MoS2 2D QDs after vacuum filtration.

The thicknesses of the 2D QDs are in the range of 0.5 to 1.2 nm, corresponding to one to two layers, as analyzed by atomic force microscopy (AFM) (Fig. 2B, showing an example of MoS2 QDs). The transmission electron microscopy (TEM) image reveals that MoS2 QDs are nearly circular in shape, with a relatively narrow diameter distribution ranging from 0.78 to 2.4 nm (Fig. 2E). The high-resolution TEM (HRTEM) image shows the lattice fringes of MoS2 QDs with a spacing of ~0.27 nm corresponding to the (100) faces of MoS2 crystals (Fig. 2H), suggesting the high crystallinity of the MoS2 QDs. This cryo-exfoliation strategy toward fabricating 2D QDs can be universally applied to other 2D materials, including graphite, h-BN, TiS2, NbS2, WS2, Bi2Se3, MoSe2, MoTe2, and Sb2Te3. The quality of these QDs in terms of thickness, diameter, and crystallinity is similar to that of MoS2 QDs. For examples, an average thickness of 0.6 and 1.0 nm is observed for graphene QDs (GQDs) and WS2 QDs, respectively (Fig. 2, A and C), whereas the diameter ranges from 0.77 to 3.20 nm (mean value of 1.71 ± 0.32 nm) for GQDs (Fig. 2D) and 2.18 to 5.99 nm for WS2 QDs (mean value of 3.73 ± 0.56 nm) (Fig. 2F). Both GQDs and WS2 QDs are crystalline with exposed (100) facets, respectively (Fig. 2, G and I). The other QDs made from bulk powders of h-BN, TiS2, NbS2, Bi2Se3, MoTe2, and Sb2Te3 are also composed of mono- or bilayers and exhibit mean lateral size of 2 to 3 nm (Fig. 3, A to F).

Fig. 2 Morphological and crystalline structure of the 2D QDs.

(A to C) AFM images and their statistic thickness distribution (insets). (A) GQDs, (B) MoS2 QDs, and (C) WS2 QDs. (D to F) Low-resolution TEM images of (D) GQDs, (E) MoS2 QDs, and (F) WS2 QDs, respectively (insets showing the lateral size distribution). (G to I) High-resolution TEM images of (G) GQDs, (H) MoS2 QDs, and (I) WS2 QDs, respectively.

Fig. 3 Cryo-exfoliation applied to fabricate 2D QDs from different bulk-layered material powders in various solvents.

AFM images of QDs for (A) h-BN, (B) TiS2, (C) NbS2, (D) Bi2Se3, (E) MoTe2, and (F) Sb2Te3. (G) Statistic average thickness of MoS2 QDs as a function of cryo-pretreatment duration (the blue curve, with fixed 1-hour ultrasonication time) and sonication time (the red curve, with fixed 1-hour cryo-pretreatment time). The average thickness of MoS2 QDs as-exfoliated in different solvents with varied polar/dispersive ratio (H) and surface tension (I).

The crystalline structure of the QDs, taking graphene, MoS2, and WS2 QDs as examples, was studied by x-ray diffraction (XRD). Compared to the coexistence of a variety of XRD peaks for interlayer and intralayer crystal faces in the bulk powders, only one peak for the interplanar face (002) appears in the XRD patterns of all QDs samples prepared by drop casting, indicating a random restacking of QDs along <001> direction (fig. S2). In addition, the (002) peaks for QDs samples become broader due to the reduced size in both thickness and lateral size (32, 33). The x-ray photoelectron spectroscopy (XPS) results reveal that these QDs exhibit a pristine chemical state with no obvious oxidation or functionalization (figs. S3 and S4), which is distinctly different and unique compared to QDs prepared by vigorous chemical treatments (34).

Two key process parameters, the cryo-pretreatment duration and the subsequent ultrasonication time, potentially determine the quality of as-prepared 2D QDs. When the sonication time is fixed to 1 hour, the resulting QDs gradually become thinner with the increase of cryo-pretreatment time and finally become steady when the pretreatment lasting for 1 hour or longer (Fig. 3G, blue curve). With the fixed cryo-pretreatment time of 1 hour, the thickness gradually decreases with the increase of ultrasonication time until 2 hours (Fig. 3G, red curve). A combined cryo-pretreatment and sonication for 1 hour each could result in 2D QDs with a thickness <1 nm corresponding to mono-/bilayer of 2D materials from normal AFM measurement. To further improve the yield and concentration of QDs, we can cycle the cryo-pretreatment and the LPE steps, during which the residual mono- or few-layer nanosheets could be further disintegrated into QDs (figs. S5 and S6).

From previous work on the controlled LPE of 2D materials, matching of surface tension parameters of solvents to that of 2D materials plays a critical role in producing nanosheets from bulk powders (25, 35). We sonicated the cryo-treated MoS2 powders in various solvents with a wide range of surface tensions and surface tension component polar/dispersive ratios to explore the solvent effect on the production of QDs. The results show that high-quality MoS2 QDs could be produced in all of these solvents (Fig. 3, H and I; table S1; and fig. S7), demonstrating that the cryo-pretreatment could effectively reduce the dependence of exfoliation efficiency on solvent. The obtained results also apply to other layered materials such as graphite (fig. S8). This is of great significance for the future development of 2D QDs in a variety of solvent systems for practical applications.

The optical properties of the QDs were explored by ultraviolet–visible (UV-vis) absorption and photoluminescence (PL) spectroscopy. Under irradiation with UV light at 365 nm, the GQDs solution emits blue light (inset in Fig. 4A). Both MoS2 QDs and WS2 QDs solution exibit blue fluorescence as well but with higher brightness (fig. S9). All QDs solutions exhibit an excitation-dependent PL behavior (Fig. 4A and fig. S9). The PL peak redshifts upon increasing the excitation wavelength, similar to that of previously reported 2D materials–based QDs (3638). Visible (3.05 eV) pump–midinfrared (0.25 eV) probe transient absorption experiments were performed to compare the ultrafast dynamics of monolayer graphene and GQDs. Figure 4B shows the free photoexcited carrier recombination lifetime of monolayer graphene and GQDs. Monolayer graphene exhibits a biexponential decay containing a fast process of 0.18 ps and a slow process of 2.0 ps, whereas the recombination rate for GQDs slows down by several times to 2.7 ps. The fluorescence quantum yield (QY) of these 2D QDs is typically below 10% (Fig. 4C). The 2D QDs (for example, MoS2 QDs) can uniformly disperse in an epoxy matrix to form a photoluminescent composite (inset in Fig. 4C). The 2D QDs present larger optical bandgaps between 3.0 and 6.0 eV, as compared to their bulk and monolayer counterparts, resulting from the quantum confinement effect (Fig. 4D and fig. S10) (39).

Fig. 4 Optical properties of the pristine 2D QDs.

(A) PL spectra excited at different wavelengths of GQDs. Insets show digital images of the corresponding QDs under day and UV light. a.u., arbitrary units. (B) Pump-probe transient absorption spectra of GQDs and monolayer graphene. (C) The QY and (D) collective optical bandgap for selected 2D QDs. The inset in (C) shows epoxy filled with MoS2 QDs exhibiting blue fluorescence under UV irradiation.

The cryo-mediated LPE serves as a powerful and universal approach to produce 2D QDs directly from bulk-layered material powders. With the introduction of cryo-pretreatment, the efficiency of LPE is enhanced and exhibits less dependency on solvent properties. These exfoliated mono-/few-layer sheets are more amenable to fracturing into QDs compared to nanosheets from LPE procedure alone. This methodology applies to a variety of layered materials processed in a broad range of common solvents without harsh chemical modifications and with no use of surfactants or additives, which enables the synthesis of pristine QDs.


Cryo-mediated exfoliation and fracturing process for QDs

Commercially available h-BN (average, 45 μm; 99.5%) and Bi2Se3 (99.999%) were purchased from Alfa Aesar. Graphite (<45 μm, 99.99%), MoS2 (average, 2 μm; 99%), and WS2 (average, 2 μm; 99%); and MoSe2 (~325 mesh, 99.9%), TiS2 (~325 mesh, 99.9%), and Sb2Te3 (~325 mesh, 99.96%) were purchased from Sigma-Aldrich. All materials were used as supplied. All solvents were purchased from Sigma-Aldrich and used as supplied without purification.

MoTe2 powder was synthesized using a tube furnace. A porcelain boat containing mixed Mo (99.9%; Sigma-Aldrich) and Te (99.999%; Sigma-Aldrich) powder was placed in the center of the furnace. The mole ratio of Mo/Te was about 1:2. Ar and H2 gases were used as the carrier gas and were kept flowing during the whole process. The furnace was heated to 600°C and was kept at the reaction temperature for 20 min. Then, the furnace was cooled down naturally. The phase of MoTe2 powder was further confirmed by XRD (fig. S11).

The cryo-mediated exfoliation and fracturing process mainly included the following two steps: First was the cryo-pretreatment, performed by soaking the raw 2D materials bulk powder in liquid nitrogen for a period of time (10 min to several hours), then the bulk powders were immediately dispersed into the solvent IPA/H2O (the volume ratio of IPA/H2O is 1:1), and finally liquid-exfoliated with the assistance of bath ultrasonication. The initial concentration of the dispersions was 3 mg ml−1 for all 2D materials. After the cryo-mediated exfoliation and fracturing treatment, the resulting dispersions were centrifuged at 6000 rpm (rounds per minute) for 30 min to remove nonexfoliated materials. To further separate the QDs from the mixed dispersions (containing QDs and the exfoliated nanosheets), the supernatants were collected by pipette followed by vacuum filtration through ultrafine membranes with pore size of 25 nm (VSWP02500, Millipore). For the AFM characterization, the as-obtained QDs solution was diluted and then directly dropped onto a new mica substrate, which would be first dried naturally at room temperature and then further treated in a vacuum oven at 80°C for 4 hours to completely eliminate the influence of residual solvents. The as-filtered QDs samples were dropped onto ultrathin carbon type A (3 to 4 nm) with removable formvar grid (400 mesh) and then vacuum-dried for TEM and HRTEM test. The XRD and XPS samples were prepared by dropping the QDs samples onto glass substrates and drying in a vacuum oven. The yield of nanosheet and QDs were determined by inductively coupled plasma mass spectrometry (ICP-MS) measurement.

Solvents with different surface tensions and polar/dispersive ratio parameters usually have crucial influence on the LPE for the preparation of mono- or few-layer nanosheets of 2D materials. It is desirable for the LPE process to optimize the experimental results by matching the surface tension of solvents to that of 2D materials (24, 25, 35). To explore and determine the effect of solvent types on the production process and morphology of QDs, some other typical organic and inorganic solvents (including NMP, DMF, acetone, and ethanol) with varied surface tension and polar/dispersive ratio parameters were chosen to disperse the cryo-pretreated MoS2 powders for the following sonication treatment, and the corresponding produced QDs were characterized by AFM (fig. S7 and table S1).

We also conducted a comparative study on the effect of the liquid nitrogen pretreatment. Without cryo-pretreatment, we can only obtain nanosheets of 2D materials with lateral sizes ranging from tens of nanometers to a few hundred nanometers (fig. S1) after identical LPE (for example, for 4 hours in IPA/H2O = 1:1 mixture). In our normal cryo-mediated exfoliation and fracturing process, we immediately dispersed the cryo-pretreated powders into solvents followed by bath sonication. There was no time gap between the cryo-pretreatment and sonication processes. However, we had also preliminarily studied the effect of gap on the morphology of the as-prepared sample. With a 1-hour gap, it was possible to obtain the 2D QDs but with lower yield. There existed some incompletely fractured nanosheets with lateral size >10 nm (fig. S12).

We tried to collect the WS2 QDs by vacuum filtration with a membrane pore size of 25 nm. As the ICP-MS test show, the concentration of WS2 QDs will be greatly improved with the extension of ultrasonic time (fig. S13) under the same cryo-pretreatment duration (1 hour), indicating the increase of yield. After 4 hours of sonication treatment (for example, WS2-1h-4h), the concentration of WS2 QDs could achieve 28.7 parts per million (28.7 mg liter−1). The yield of the QDs could be calculated up to 1 wt % of the raw powders. In actuality, the obtained QDs concentration was underestimated because many QDs settled on the remaining nanosheets and were filtered out. If the as-prepared QDs could be directly used together with the as-exfoliated nanosheets in a dispersion (for example, a photocatalytic water splitting application), the superiority of our high-concentration QDs in low–boiling point solvents would be much more remarkable.

Characterization and calculation of QDs

In all cases, the ultrasonication was performed using a Branson 5510 model bath sonicator with a frequency of 40 kHz. The resulting dispersions were centrifuged by the Thermo Fisher Scientific Sorvall Legend X1 centrifuge at 6000 rpm for 30 min to remove nonexfoliated materials. The topographic height/thickness of the as-obtained QDs of 2D materials was characterized by an AFM (Bruker NanoScope V MultiMode 8.0) on mica substrates. TEM images were recorded on a JEM-2100F at an accelerating voltage of 200 kV. XRD patterns of the samples were performed on D/Max 2500 V with Cu Kα radiation (λ = 1.54056 Å). The chemical valences of the sample elements were measured by XPS using an ESCALAB 250 with a monochromatic Al Kα x-ray beam (1486.60 eV). All the binding energies of the elements were calibrated to the C 1s binding energy of 284.8 eV. PL spectroscopy experiments were conducted using a HORIBA Jobin Yvon NanoLog spectrofluorometer with varied excitation wavelengths in the range of 300 to 400 nm. A Shimadzu UV-2450 UV-vis spectrophotometer was adopted for the UV-vis absorption spectra characterizations.

We calculated the thickness of the as-obtained QDs mainly from the AFM images, as shown in Fig. 2 and fig. S14. As we can see from the fig. S14A, when we draw a horizontal line in the NanoScope Analysis software (shown in fig. S14A, blue line), we can obtain the thickness distribution data, from which we can see that there are some negative values (fig. S14B). We guess that this kind of negative value should be generated by the instrument itself or the mica substrate. Then, we characterized the pure mica substrate by AFM (fig. S14C). When we draw a horizontal line on the mica substrate, we can find some periodic thickness fluctuations ranging from several tens to 300 pm (fig. S14 D). We suppose that this kind of fluctuation would affect the thickness distribution statistics of the as-obtained QDs.

The optical bandgaps of various 2D QDs were extracted according to the Tauc plots (40)Embedded Imagewhere α is the diffuse absorption coefficient at wavelength λ, h is the Planck’s constant, v is the light frequency, n is 2 or 0.5 for direct and indirect transitions, respectively, C is constant, and Eg is optical bandgap of materials.

The QYs of 2D QDs were measured at an excitation wavelength of 370 nm using [Ru(bpy)3]Cl2 (QY = 0.04) as a reference. The QY was calculated by (41)Embedded Imagewhere Φ is the QY of sample with respect to reference material, Int is the area under the emission peak at a certain wavelength, A is absorbance at the excitation wavelength, and n is the refractive index of the solvent.

For the preparation of QDs/epoxy composite, a liquid epoxy resin system 2000 based on diglycidyl ether of bisphenol A and a curing agent 2020 based on ethylenediamine were supplied by the Fibre Glast Developments Corporation and used as received. The QDs solution was added to the epoxy resin and magnetically stirred at 80°C under vacuum for 48 hours. The final mixture was cooled down to room temperature and the curing agent was added (Phr 23). The viscous mixed composite was placed in silicone molds, cured at ambient temperature, and further postcured at 100°C for 2 hours. Then, the as-prepared QDs/epoxy composite was placed under day light and UV light for the optical characterization.


Supplementary material for this article is available at

fig. S1. The AFM images of the as-exfoliated MoS2 nanosheets with only sonication-assisted LPE and without cryo-pretreatment.

fig. S2. XRD patterns of the as-obtained QDs and raw powder of layered materials.

fig. S3. XPS spectra of the as-obtained QDs.

fig. S4. XPS spectra of the corresponding raw bulk powders.

fig. S5. UV-Vis absorbance spectra of MoS2 samples with a series of cycles of cryo-pretreatment and liquid exfoliation treatment.

fig. S6. Digital image of MoS2 supernatants without vacuum filtration with series cycles of cryo-pretreatment and liquid exfoliation.

fig. S7. AFM images of MoS2 QDs produced in different solvents.

fig. S8. The thickness statistics of the as-exfoliated GQDs with different cryo-pretreatment duration and solvents.

fig. S9. Photoluminescence spectra excited at different wavelengths for MoS2 QDs and WS2 QDs.

fig. S10. Tauc plots used to determine the optical bandgaps of various 2D QDs derived from UV-vis spectra.

fig. S11. XRD pattern of the as-synthesized MoTe2 powders.

fig. S12. The AFM image of the MoS2 sample obtained in a process with 1-hour gap between the cryo-pretreatment and the ultrasonication.

fig. S13. ICP-MS results of the WS2 QDs concentration with different sonication time at a fixed cryo-pretreatment duration of 1 hour.

fig. S14. Thickness statistics of the as-exfoliated MoS2 QDs and the background thickness signal of the new mica substrate.

table S1. Surface tension and polar/dispersive ratio of the solvents adopted for the cryo-exfoliation of MoS2.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We appreciate A.A. Martí (Department of Chemistry, Rice University) and A.Aliyan for the UV-vis and PL test and meaningful discussion. Funding: This research is financially supported by SABIC (Saudi Basic Industries Corporation) and National Natural Science Foundation of China (51772072). Y. Wang also would like to acknowledge the financial support received from the China Scholarship Council during his visit to Rice University. Author contributions: Y. Wang and J.W. conceived the idea and designed the experiments. Y. Wu and P.M.A. supervised the project. Y. Wang and Y.L. performed the synthesis experiment and the physical characterization in collaboration with J.Z. and J.W. Y. Wang and J.W. wrote the manuscript. Y. Wang is responsible for the Figs. 1, 2 (A to C), 3; and figs. S1, S5 to S8, and S12 to 14. J.Z. claims responsibility for the Fig. 2 (D to I) and figs. S2 to S4. Y.L. and X.W. were responsible for the Fig. 4 and figs. S9 and S10. X.Z. takes responsibility for fig. S11. All authors discussed the results, analyzed the data, and commented on the manuscript. Competing interests: Y. Wang, P.M.A., Y.L., J.W., H.X., N.C., and I.N.O. are authors on a patent application related to this work filed by Rice University and SABIC (no. 62/380,902, filed 29 August 2016). All other authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article